2009 Ncta Technical Papers PDF

THE COMPLETE
TECHNICAL PAPER PROCEEDINGS

FROM:
“My (TV) Space” –
Using Subscriber Management Systems to refine and redefine video-based social networking
Andrew Poole
Joe Matarese
ARRIS Group, Inc.
Abstract The TV is King.
Social networking sites like MySpace and A further aspect of MySpace and Facebook
Facebook have enabled on-demand user- video-sharing is that the display mechanism is
generated content to millions of viewers. In on the PC. Cable technology has the easiest
this paper we ask whether it is possible to consumer access to the most preferred
leverage social-networking web technology, viewing device for video: the television.
user-generated video, and cable According to a 2008 CTAM study, 96% of
infrastructure to provide a “best of all adults who subscribe to cable or satellite
worlds” experience for these millions of services prefer to watch television on
viewers. traditional TV sets. [Reference: CTAM]
While it is possible for viewers to hook up a
PC to the television, the process remains more
difficult than accessing content via a cable set
INTRODUCTION top box. Given Cable’s widely deployed VOD
infrastructure, can Cable leverage its
customer-friendly on-demand access to the
Glass Houses television to provide user-generated content
via the television?
Social networking sites have enabled, for
better or sometimes worse, unfettered access This paper investigates three scenarios for
to as much personal information as selective entitlement of subscriber-generated
individuals dare make public. When Warhol video content via cable: local sharing, global
uttered the line about everyone being famous sharing within an MSO, and global sharing
for 15 minutes, he appropriately conjugated across MSOs, all of which may in fact
the verb with a “will” rather than a “may”. overlap. In all of these scenarios, the solution
One feels that it is fast becoming impossible acts as a trust broker, allowing subscribers to
not to be famous for 15 minutes. In this issue entitlements to other subscribers –
“glass house” environment, cable potentially friends, family and/or acquaintances. The
has a role in allowing subscribers to cable operator stores the subscriber-generated
selectively draw the shades. Subscriber video content, uploaded via a web portal over
Management Systems (SMSs) have long an MSO’s high speed data service, and
served as a means to provision specific maintains the associated entitlements on the
services to subscribers. Such services subscriber’s behalf. The solution provides a
typically include high speed data tiers and mechanism for the subscriber to allow or
subscription VOD packages. Why not take revoke entitlements on any subset of the
this a step further and allow subscribers to act subscriber’s content.
upon service and account information so as to
offer personal content in a more selective
manner?
2009 NCTA Technical Papers - page 71

THE GROWTH OF USER-GENERATED site, an existing MSO web portal, or a new
CONTENT web site.
The question could be asked, “Who cares Now, envision a connection between the
about user-generated content anyway?” and social-networking web site and an MSO site.
the answer is a resounding “Users do.” The social-networking solution would
understand the connection between web users
In terms of user-generated video and MSO subscriber accounts, deliver user-
specifically, in 2005 there were 3.3 billion generated content to the MSO site, and inform
user-generated video views, growing to 34 the MSO site of what subscribers have access
billion in 2008. [Reference: MediaPost] to what content.
User‐Generated Video Views
(billions)
40
30
20
10
0
2005 2006 2007 2008
Figure 1. User-Generated Video Views
Cable has an opportunity to tap into this

phenomenon to provide a value-add service
that brings this wealth of user-generated
content in a user-friendly manner all the way
to the television set.
THE VISION
The “My (TV) Space” vision starts with a Figure 2. High-level System Overview
social networking web site providing all the
features of the current “social sites” of today:
account management, profile management, Subscribers at the MSO site can subscribe
“friend” management, and content to the user-generated content subscription
management. Content management today package in order to gain access to the user-
includes text, photos, and video. So, on the generated content that their friends have
social networking web site users can upload shared with them. As in subscription Video-
user-generated content and share it with one on-Demand (SVOD) packages, the user may
or more of their friends. Note that this web subscribe to the package via an interactive
site could be an existing social networking screen on the EPG or through a web portal.

(Unlike SVOD packages, the user would not Think Local, Act Global
necessarily subscribe via a call to a CSR.)
In the global sharing case within an MSO,
The MSO’s VOD system would be a subscriber might post video to extended
extended to understand the per-subscriber, family, to dispersed college friends, or to
per-content entitlements to conditionally remote colleagues. In this situation a cable
allow subscribers to access the user-generated operator with a national footprint must be able
content. Titles will automatically appear in the to exchange entitlement information and
subscriber’s VOD portal—either a special content across its markets. So in our little
social-networking VOD client application or a league example, the Malibu-based
special category in the existing VOD client. grandparents of the little league player in St.
Though outside the scope of this paper, there Louis could watch the video of the game. Of
are many possible ways in which cable course, even cable operators without national
subscribers viewing user-generated content footprints can participate in this use case
may provide feedback to the content author or through integration with a web portal
other viewers about the video. incorporating streaming video.
Think Global, Act Global
“MY (TV) SPACE, YOUR (TV) SPACE” – An enhancement of the above case extends
THREE SCENARIOS the solution to share video content across
markets that span multiple cable operators. In
At least three basic scenarios exist for this solution friends can share user-generated
selective entitlement of subscriber-generated content via Cable television regardless of the
video content via cable: local sharing, global city or cable operator. This service could be
sharing within an MSO, and global sharing provided only as long as the viewer’s home
across MSOs. market is part of a coordinated effort among
the cable operators.
Think Local, Act Local
A SOLUTION – THE UNITED
In the local sharing scenario, a subscriber FEDERATION OF CABLE SYSTEMS
might share content with schoolmates or with
members of a neighborhood association. One solution option to implement the “My
Another example is a subscriber sharing video (TV) Space” vision by necessity involves
content of little league games with the coordination of subscriber information across
families of the other little league players. MSO sites, coordination of content across
These groups are geographically local by MSO sites, and coordination of entitlements
nature and thus could be served by a single across MSO sites.
MSO site (e.g. a single city). The cable
operator can easily manage entitlements and While a “united federation of cable
video content within the confines of the local systems” with coordinated entitlements,
VOD system. content, and subscriber information may seem
to be somewhere between the implausible and
the impossible, the introduction of social-
networking solutions which can leverage SMS
interactions may simplify some aspects of the
solution.

Figure 3. Cross-MSO System Overview
involve using web development technologies

or perhaps partnering with existing social
TECHNICAL ASPECTS OF THE networking sites. Note that this site could be
SOLUTION an MSO-branded site or a social-networking
site which is independent of MSO
Now let’s explore some technical aspects interactions. Also note that multiple social-
of the solution. networking sites could be supported by the
overall solution.
The Social-Networking Web Sites
One new function of the “My (TV) Space”
The social-networking web sites will web solution will be back-end transcoding of
provide a web-based control interface to users video to a format conducive to display on
of the system. Capabilities to manage profiles, cable set top boxes. Existing CableLabs
connect friends, upload videos, and assign content encoding standards and existing
viewing access to videos are key aspects of MPEG transcoding technologies may be
this component. Technical considerations here leveraged for this purpose.

Federated Cable Subscriber and Cable Service Federated User-generated Content Session
Information Feedback
In order to enable content access to only As cable viewers interact with user-
those cable viewers who have subscribed to generated content there exists an opportunity
the user-generated content package, the for feedback to the social network. This
solution as a whole will need to understand feedback could be as simple as the number of
cable subscriber information (and package views for a given asset. Furthermore, the
entitlement) across cable operator sites and cable viewer could send messages back to the
even across the cable operators themselves. user who generated the content such as “Great
This will require interaction with existing Video!” or “Very funny!” Again, this aspect
Cable Subscriber Management Systems of the system is a subject unto itself, but the
(SMS’s) to coordinate the information. basic feedback mechanism must be
Several industry-standard or existing SMS- acknowledged.
vendor interfaces may be leveraged for this
interaction. Social Networking VOD Applications
Federated Assets (Metadata and Content) The cable operator’s VOD Application
Server and VOD Client may or may not
A key element of the solution is to allow change depending on the approach taken to
user-generated content to flow to cable the solution.
operator networks. This will include both the
user-generated videos and the metadata One approach is to leverage the existing
associated with the videos such as title, VOD Application. By enhancing the existing
description, length, and the user who created VOD application server and client, the VOD
the content. This data will need to move system could display a “User-Generated
across MSO sites and across multiple MSOs Content” category in the existing VOD Guide.
as well. One approach to getting this content This category would contain the titles of the
and metadata to the service provider site is to user-generated content that the subscriber is
leverage existing CableLabs Asset entitled to view.
Distribution and Metadata interfaces.
Another approach is to develop a new
Federated User-specific Asset Entitlements VOD Application. There are advantages to
providing a new VOD Application (server and
Once the correct assets are on the MSO client), for example, leveraging tru2way, to
site and the VOD solution recognizes that the form the solution. The viewer could be
account has access to the user-generated provided with a branded custom “User-
content package, the solution still needs the Generated Content” user interface with
entitlement information of which users are features specifically designed for interacting
entitled to which specific assets. This with social networking user-generated
information will need to be provided across content. The VOD client experience then
the cable operator sites and potentially across becomes an extension of the web-based
the cable operators. experience. This approach might further
benefit cable operators by providing a means
to “upsell” set-top devices capable of

supporting more sophisticated user • Multi-screen – TV, PC, Mobile content
interaction. availability
Such services amount to a natural offering,

NON-TECHNICAL CHALLENGES in keeping with cable operators’ position as a
trusted provider.
There are some non-technical challenges
that need to be overcome. Among these are
potential legal challenges. Most social CONCLUSION
networking web sites have site guidelines
prohibiting upload of copyrighted or illegal With the ever-increasing ways for people
material. Will this be sufficient in the domain to consume video, Cable operators face ever-
of cable networks? Will the social-networking increasing competition for the eyes viewing
web site need to provide moderation of video that video. The “My (TV) Space” concept, as
uploads? These challenges will need to be described here, explores some of the technical
overcome prior to adoption of user-generated aspects of providing user-generated content in
content solutions on cable networks. a user-friendly manner to cable-enabled
televisions across the globe. The challenge of
cable operators will be to further leverage the
WHAT THE FUTURE HOLDS efforts of those individual content creators
and to move those user-generated content
Given the infrastructure described above viewers from small-screen PCs back into the
the possibilities for new functionality are big-screen living room.
endless.
REFERENCES
Cable operators can extend the approach to
create appealing and easy-to-use new services, [CTAM] CTAM (6/30/08) “TV Web Sites
including: Grow More Popular, but Viewers Still Prefer
• Video albums containing personal content Their TV Sets”
and links to entertainment content to be [MediaPost] MediaPost (1/31/08) “User
shared among a limited audience. Generated Video Expects 34 Billion Views in
2008”
• Personal TV channels, programmed by the
subscriber
• “Internet TV” – bringing web-based series
programming to the TV. Today, these
series can be found at web sites such as
blip.tv and Crackle.com.
• Video mail – either multicast or unicast
• Real-time subscriber interaction across
cable systems – chat with your friend in a
different city via the TV screen – even if
your friend is serviced by a different MSO

A NEW CONCEPT FOR ROBUST VIDEO MARKING
Niels Thorwirth
Verimatrix, Inc.
Abstract are often unable to read the mark after small

geometric content transformations.
Digital watermarking is the concept of
permanent and imperceptible tagging of This paper outlines the research results
video data. User-specific watermarking that enable human readout and the
embeds tracking information that accurately challenges to make the mark robust yet
identifies the last legal recipient of a video. invisible. The result allows for wide content
This type of watermark does not enforce distribution, ensures the mark is secure and
content use restrictions but it enables can establish strong evidence linking the
identification of sources of content abuse, content to a consumer while maintaining
acting as a deterrent and encouraging privacy.
responsible consumer behavior. These types
of techniques represent a way to maintain
control over video content while permitting
convenient content consumption that can DIGITAL WATERMARKING 101
compete with free but inconvenient and
illegal content distribution. Over the last few years, the
representation, storage and distribution of
Required is a secure, robust and digital video have grown massively in
imperceptible marking technology that does diversity and popularity, and are now
not degrade consumer experience and yet approaching the level of digital music in its
establishes a reliable trace that survives ease of use and ubiquity. The advantages of
attacks or passages through the analog hole. this type of distribution and consumption,
This paper presents our implementation just as with music, also fuel the growth of
approach, optimized to create a large scale piracy. The rightful owners of
commercially significant capability video content are deprived of their revenues
incorporating new concepts tailored to and could be discouraged from investing in
digital video. the creation of content in the future.
A key element of the approach may Several approaches have been applied to
appear to be counter intuitive, in that the secure digital media:
embedded information is not designed to be Digital encryption technology is
machine readable but it can be extracted effective to enable secure delivery.
into a human readable form. While it However, once decrypted and presented in a
requires human interaction, it allows use of human visible format, it can be re-recorded
well honed human perception processing to to obtain and distribute an illegal, unsecured
aid the extraction task - which is simply copy. No secure technology currently exists
superior to machine recognition and thereby to prevent re-recording using a camcorder.
solves the problem of content
synchronization. It overcomes the Amongst other applications, the marking
challenges of content misalignment that of media can help investigation to identify
fools machine readable approaches, which individuals that are responsible for abuse,

either by embedding recipient information in rotation or conversion to another file format,
the media or ownership information that including lossy compression, during which
indicates copy restriction. perceptually insignificant information is
eliminated to reduce the size of a digital
One way of marking media is done by media file.
adding information to the digital media file
that is ignored during normal playback, but Critical transformations that occur during
can be extracted and used by more piracy re-recording include:
specialized tools. Apple iTunes is widely • Camcorder capture
assumed to be using this method to tag • Digital-to-Analogy (DA) conversions to
DRM-free file distribution (1). This kind of S-video or VHS
tagging enables identification of the • Re-compression to low bitrates using
unmodified file, but are easily removed and popular formats like H.264, DivX,
destroyed when the file is re-recorded or MPEG2
converted to another format. • Geometric distortions like heavy scaling,
rotation, AR change, random bending,
For more robust marking, visible text or cropping
dots have been used to carry identifying • Color conversion, e.g. grayscale
information in video, and while this • Filtering like blur, sharpen, contrast
information does survive re-recording, it can modification, noise reduction and de-
easily be identified and removed in order to flicker
disable the ability to track the content. In
• Frame rate conversion
order to maintain robustness, these markings
are also visible, a compromise that degrades Relative misplacement of the mark and
the consumer experience. underlying content can also be created
intentionally by imperceptible, slight or
Digital watermarking is another marking
combined modifications, such as shifts,
approach that has been suggested in many
rations and time jitter. Publicly-available
variations. Common digital watermarking
tools (3, 4) apply these modifications, also
schemes embed information, by introducing
called attacks, in an automated fashion.
manipulations at a certain positions in space
Since current image processing algorithms
or time. These watermarks are detected by
are not very successful in recognizing
specialist software (2). In other words, it is a
misplacements in distorted content (a
process for modifying media content that
process also called registration) these
embeds a machine-readable code into the
modifications render these types of
data content.
machine-readable digital watermarks
ineffective.
When interpreting the manipulations
during machine processing of the content,
Digital still images have been the early
processing can be greatly reduced through
focus of watermarking research and
pre-knowledge of the insertion positioning.
commercial exploitation (5). Video
When the positions are modified, i.e.
watermark approaches often have been
misplaced or weakened, the readout may
based on the application of the image
become difficult or impossible. Such
watermark applied to video frames. While
modifications do routinely occur during
this is the natural progression and allows the
simple media treatment such as cropping,
embedding of much data, this approach does

not efficiently use the time domain for transformations thereof, unlike bitstream
gathering embedded information because manipulation and encryption. Unlike visible
detection is only successful if some marking it is unnoticeably hidden in the
information in individual frames can be media. The difference to previous
recovered. This approach fails when none of watermarking approaches is the possibility
the watermarks can be read at least in part of using the human perceptual system for
due to a failure in registration or destruction actual recognition, which is superior to
of relevant areas. current machine reading capabilities, in
particular for degraded, noisy and
Digital watermarking typically involves transformed content.
a complex transformation of the original
image and of the message to be embedded in The message is emphasized by a
order to allow invisible embedding. In the computer, but the actual interpretation of the
forensic, user specific application, mark can be performed by a human, making
watermarks are used to embed information it possible to detect the mark in degraded
about an individual playback device to allow content and independent of registration.
tracing of the last legal recipient. It this case, While it is very difficult for a watermark
it acts as a serial number or license plate and technology to interpret a misplaced mark, a
requires execution of the embedding in a message can be easily read by a human even
playback device, which typically has little if it is rotated, made smaller, stretched and
processing power to spare for additional on a noisy background.
functionality like digital watermarking.
During the embedding process, only
HUMAN-READABLE some areas are modified, while all video
WATERMARKING data is used for detection. The areas that are
modified are chosen by a perceptual model
To overcome the challenges outlined that takes into account where in the content
above, a new approach is required that takes the modification is able to stay just below
the specific requirements of forensic any noticeable difference and thereby make
watermarking into account. The approach the modification invisible to the consumer.
starts with the realization that the Another component of user-specific forensic
recognition of distorted content is a task that watermarking is that it ensures security by
humans can still perform better than distributing the embedding locations in a
machine. This fact is used for example in so random fashion. Consequently, the precise
called CAPTCHA images (6) that blocks locations of the alterations can not be
Web site robots and identify human users. observed or removed, even if the mark can
So instead of embedding machine-readable be recognized.
information, a human recognizable image is
invisible embedded in the video, distributed The mark is spread over the entire frame
over time. During extraction, that area and combined with the media using the
information is aggregated in order to derive basic content, such that the embedding
a human-readable image containing the portion can not be generally removed
embedded information. without destroying the video.
The result is a robust mark that exists Additionally the mark is spread over
with the actual media and survives time, and while it generally can not be

recovered from a single frame, each frame The specifics of the technology do not
contributes to the detection result. The allow for automatic measurement of the
detection process accumulates results from actual extracted signal strength as it is part
several frames over time, contributing to of the design that the information is read out
invisibility and robustness of the mark. by a human. While the mark can be
measured for verification by correlation to
The basic principle of combining frames the known embedded information or OCR
in time is that during the process the actual applied for reading, this will not be adequate
movie content of the frames will average out to determine the level of human readability.
to be a more or less smooth surface with The detection is therefore somewhat
pixel values close to the mean –with a faster asymmetric. For one, the readout is done by
convergence for high-motion content. The a human and the embedding by a machine.
information embedded in the content in In addition the extraction filter is mostly
contrast is at constant locations and independent of the embedding and does not
therefore will increase in relative signal follow the specifics used during embedding.
strength as the frame content diminishes. A
simple averaging though is not sufficient to For instance, the embedding locations
expose a readable mark. To actually archive are spread in a pseudo random fashion that
robust detection, additional filtering is does not have to be known during
applied to enhance the mark in every frame. extraction. Both levels of asymmetric
The filter is aimed to reduce the effect of embedding and reading increase the security
underlying content to be significant in the against attacks that aim to understand the
combined image while at the same time embedding algorithm and to invert them. It
improve the effect of individual frames. also provides robustness against the so
Over the last years of researching this called oracle attack that degrades the
approach, we have evaluated a variety of watermarked content in an automated loop
different filters and often found surprising, to find the least amount of degradation that
counter intuitive results on their individual causes the detector to fail to recognize the
effectiveness. The best achievement so far to mark.
highlight the remaining bits of information
is a combination of filters applied in Unlike machine-readable digital
sequence. watermarking, the detection process displays
an obvious and unambiguous human
Our initial research however already understandable outcome, e.g. a serial
revealed that the method of combining number. When uncovering the mark, the
frames over time shows remarkable embedded graphic slowly appears from the
robustness, even in the presence of marked video, showing that the mark is
modifications that would suggest a removal derived from the content. The extraction can
of the minute, unnoticeable variations. be interpreted or read by a layman. The
These degradations include compression of result is easy to understand and can be used
the color domain or analog transformation. as persuasive evidence of wrongdoing
The reason for the approach being effective during an investigation into content misuse.
in these scenarios is that a fraction of the
modification will survive any transformation HOW TO USE ROBUST
that leaves the content in a reasonable WATERMARKING
quality.

User-specific digital watermarking is actually pay-per-view for premium content
ideally suited for forensic applications that compared to trawling for “free” peer to peer
aim to identify the source of piracy. It acts downloads. Effective content protection can
as a deterrent because it registers the make the difference between few
individual video to its owner. The subscribers that distribute movies to a large
embedding task is performed in each set-top Internet community and a growing
box (STB) or DVR, allowing individually- consumer base that values premium content.
marked copies, without the requirement to
process individual videos for each receiver References
at the head-end. The extraction is performed 1.http://www.theinquirer.net/inquirer/news/3
on copies that are in public distribution and 57/1050357/drm-free-itunes-files-have-
originate from a single piracy source. To your-number (retrieved on 1/29/09)
identify the embedded information, the
copies are read in a central location avoiding 2. Jian Zhao, "Look, It's Not There", in:
the distribution of the extraction service BYTE Magazine, January, 1997
preventing a possible attacker to verify the
success of modification applied to the video 3. Fabien A. P. Petitcolas, “Watermarking
with the intention to remove the mark. schemes evaluation”, I.E.E.E. Signal
Processing, vol. 17, no. 5, pp. 58–64,
September 2000
CONCLUSION
4. Fabien A. P. Petitcolas, Ross J. Anderson,
With the obvious need for better Markus G. Kuhn, “Attacks on copyright
protection and infrastructure in the growing marking systems,” in David Aucsmith (Ed),
business of digital video distribution, a Information Hiding, Second International
layered protection approach that Workshop, IH’98,, Proceedings, LNCS
incorporates digital watermarking can help 1525, April 1998
alleviate the revenue loss content owners are
facing through current piracy. Digital video 5. Saraju P. Mohanty, “Digital
delivered through an STB provides a unique Watermarking: A Tutorial Review,” Dept.
environment for content protection using of Comp Science and Eng, University of
user-specific marking, potentially providing South Florida, Tampa, 1999
a superior distribution channel and a
preferred form of distribution by movie 6. Luis von Ahn, Manuel Blum and John
studios that are looking to secure their Langford, “Telling Humans and Computers
sensitive early release windows. As a Apart Automatically,” In Communications
business stimulus, such early release of the ACM, vol. 47, pp. 56 – 60, February
windows translate into more demand for 2004
pay-TV operators and they are a crucial
component to increase the momentum and Niels Thorwirth
Director Advanced Technology Implementations
customer base of this distribution channel. Verimatrix, Inc.
6825 Flanders Drive
Transparency of consumption enabled by San Diego, CA 92121
watermarking may be an additional +1 (858) 677-7800 x3003
important factor for the bottom line - +1 (858) 677-7804 Fax
maintaining the consumer motivation to +1 (858) 357-1529 Cell
[email protected]

AN OVERVIEW OF TV WHITE SPACE FOR THE CABLE INDUSTRY
Tom du Breuil, Dave Gurney, Mark Kolber
Motorola, Inc.
Abstract After the DTV transition, all of the full

power analog TV transmitters will go off the
After years of study, public comment and air, freeing up additional spectrum compared
testing, the FCC has adopted rules for to today. This is often termed the “Digital
unlicensed operation of TV Band Devices Dividend” by regulators. The amount and
(TVBDs). The TV spectrum’s desirable exact frequencies of unused spectrum, i.e.,
propagation properties and the promise of unused channels not assigned to any TV
significant amounts of usable spectrum make broadcast stations, varies from location to
this band attractive. TVBDs are expected to location. These unused segments of spectrum
support a wide range of applications, from are referred to as TV White Space (TVWS).
providing broadband wireless internet
connectivity in rural areas to consumer-based TV: 402 MHz of Spectrum (VHF & UHF)
whole house automation and networking TV CHS VHF 2-13 and UHF 14-36, 38-51 TV CHS 52-69
solutions. This paper provides an overview of 698 MHz 806 MHz
the TVBD rules and their impact, including Analog TV Clears by June 12, 2009
TV: 294 MHz 108 MHz
some potential TVBD applications and key
700 MHz Band
cable operator concerns. Allocation Done
Public Safety: 24 MHz
Commercial: 84 MHz
TV: 294 MHz 108 MHz

INTRODUCTION TV White space (Proposed): Spectrum Amount & Location
Varies By Geography
In the United States, 402 MHz of

spectrum in the range of 54 MHz to 806 MHz Figure 1. US TV Spectrum Before and
After the DTV Transition
has been allocated for TV broadcasting as
shown in Figure 1. Coincident with the DTV Since each market has different TV
transition, now scheduled for completion June channel assignments, the potential "white
12, 2009, the uppermost 108 MHz of that space" in each market is different. For
spectrum, referred to as the 700 MHz band, example, in Baltimore, Channel 25 is
will be vacated from all TV broadcasting and available for use as white space after the DTV
the spectrum has been reallocated for transition while in neighboring Washington
commercial and public safety systems. The DC, it will not be available as a low power
auctions for this spectrum have brought analog station, WZDC, will still be using that
almost $20B[1] to the government with one spectrum. Unlicensed TVBDs will be
block still unsold. The remaining 294 MHz of regulated under the FCC Part 15 rules and are
this prime radio spectrum will remain for allowed to operate at relatively low power
over-the-air (OTA) TV and other licensed levels.
operations. However, throughout the United
States, portions of this 294 MHz resource Making this white space spectrum
remain unused. available is intended to address the need for
more spectrum for Wireless Internet Service

Providers (WISPs), home WiFi-type wireless to traditional WiFi spectrum potentially allow
devices and other wireless services. wireless Internet Service Providers (WISPs)
to service rural communities with fewer
This white space spectrum should not be transmission towers. Such wireless
confused with the spectrum above 700 MHz broadband is not limited to rural areas,
that is going to become available after the however, and could also be offered in certain
digital TV transition. As noted, the spectrum limited suburban areas as an alternative or
above 700 MHz has been re-allocated to supplement to other licensed mobile data
licensed commercial services and public services or unlicensed (WiFi) wireless data
safety services and will no longer be used for services.
TV broadcasts. TV stations currently
broadcasting in the 700 MHz band will move TVBDs may also potentially be used for
to lower parts of the TV band. wireless home networking, both for home
automation applications as well as high speed
APPLICATIONS wireless HD video streaming. Thermostats,
lights, and the alarm system could all be
Many potential TVWS applications have networked with TVBD technology in order to
been identified. Wireless rural broadband allow the homeowner seamless wireless
service (Figure 2) is an application that has integration which permits operation from
received significant attention and could anywhere in the home. TVBDs could further
potentially be a good fit in the TVWS bands be incorporated into a smart power grid
since rural areas tend to have fewer TV solution where home appliances communicate
stations, and thus have more spectrum in real-time with the local utility company in
available. In addition, these rural areas are order to more efficiently utilize both
less likely to have wired broadband options generation and distribution capacity by
due to the low population densities. balancing the real-time price against the
timing of when major appliances are operated
in order to provide better efficiencies for the
utility and lower power costs for the
Wireless consumer. TVBD networking applications
need not be mutually exclusive of other
network technologies; a home automation
application might connect through a wired
broadband service permitting the homeowner
to securely peer into the TVBD-enabled
wireless camera at the front door. Smart grid
4W applications would also most likely include a
combination of networking technologies.
Coverage circles not drawn
TVBDs may be used to provide the

networking infrastructure for a whole home
video entertainment network, and can also be
used to augment existing WiFi data networks.
Figure 2. Wireless Consumer Broadband Where there is sufficient white space
Service
available, multiple vacant TV channels can be
used to concurrently and wirelessly stream
Similarly, the favorable propagation
multiple HD videos around the home from a
characteristics of the TV spectrum compared
digital video recorder in one room to multiple

TVs in other rooms. TVBDs could also be including Motorola and Google [3] has been
embedded in consumer camcorders and recently formed to address the need for
enable wireless streaming from an HD industry standard geo-location databases.
camcorder directly to the home entertainment
system for easy wireless viewing or to a PC There are also provisions in the rules for
for wireless file transfer. sensing-only TVBDs, but such devices must
go through an additional testing and
Another potential TVBD application area certification process that is not yet fully
is data networking for enterprise and public defined. Therefore, one can expect that geo-
safety applications. Such applications already location enabled TVBDs will be the first to
use licensed spectrum and public safety will reach the market, once third party database
get additional spectrum in the 700 MHz providers are established.
spectrum, so these TVBD applications may
allow supplemental services. A university or The FCC order allows for operation of two
large business might use TVBDs to add high classes of unlicensed TVBDs: fixed and
quality video conferencing between campus personal/portable devices. Fixed TVBDs may
buildings without impacting existing LAN output up to 4 W EIRP, and operate on OTA
infrastructure. channels 2, 5-36, and 38-51, while
personal/portable TVBDs may output up to
FCC TVBD RULES OVERVIEW 100 mW EIRP, and operate on channels 21-36
and 38-51. Sensing-only TVBDs would be
In November, 2008, the FCC released the limited to 50 mW EIRP. Table 1 summarizes
first draft of the Second Report and Order the operating restrictions for the different
detailing Unlicensed Operation in the TV devices. There are two classes of personal
Broadcast Bands [2]. The order covers the portable TVBDs: Mode I client devices, and
initial operating rules for unlicensed TV Band Mode II master devices. Mode II master
Devices. The order provides for unlicensed portable TVBDs utilize geo-location
operation of TVBDs using geo-location databases to determine usable channels in
database as the primary method of incumbent their area, as do fixed TVBDs. Mode I client
protection mechanism, along with sensing as a portable TVBDs must rely upon a master
secondary method of incumbent protection. device (fixed or portable) to acquire a list of
usable channels for a geographic area.
Geo-location database enabled TVBDs Portable TVBDs are permitted to operate at
must determine their location to ≤ 50 m locations inside of an adjacent (TV) channel
accuracy and access a geo-location database service contour at a maximum power level of
via the internet to determine open channels in 40mW. Fixed devices are presently not
that location. The geo-location database allowed to operate on adjacent channels.
contains protection information for analog and Note that the FCC has left the record open
digital OTA TV broadcast services, as well as regarding fixed TVBD operation on adjacent
for cable headend and TV translator receive channels, and higher powered rural fixed
sites, radiotelephone services, and certain TVBDs. The FCC prohibited TVBD use of
(e.g., FCC Part 74) fixed wireless microphone Channels 3 and 4 because of the widespread
deployments. Generally, broadcast TV use of home RF remodulator interconnects on
stations are protected from any co-channel these channels between set top boxes, VCRs,
TVBD operations in their grade A and Grade and TVs.
B service areas to avoid harmful interference.
A white space database coalition consisting of
several major technology companies

Specification Personal/ Fixed TVBD and protections are discussed in detail in the
Portable next section.
TVBDs
Tx EIRP max 100 mW 4 Watts
40 mW adj ch
There are no FCC restrictions as to the
Tx TPO max 100 mW 1 Watt types of modulation or bandwidth that
40 mW adj ch TVBDs may utilize. Vendors are free to
Geo-location yes, from host yes innovate in this area and some IEEE groups
or self have expressed interest in developing
Sensing yes yes standards for operation. The IEEE 802.22
Prohibited 2-20, 37 3, 4, 37
group has been developing a standard for
channels
TVBD database no yes wide area networking (i.e., broadband
registration wireless internet service) for well over two
Tx ID no yes years.
Tx power control yes yes
FCC yes yes POTENTIAL CABLE OPERATOR ISSUES
certification
Tx antenna n/a ≤ 30 m
height
The deployment of TVBDs can impact
Rx antenna n/a ≥ 10 m cable systems in several ways. There are
height possible interference issues as well as possible
Tx Spectral -55 dBr (measured in 100 kHz BW) business opportunities as cable operators may
Mask want to extend their wired networks to
Sensing -114 dBm with 0 dBi antenna (ATSC, wirelessly support additional applications.
threshold NTSC and wireless microphones)
The interference issues fall into two broad
Border 32 km from Canada
restrictions 40 km from Mexico UHF categories:
60 km from Mexico VHF 1) Interference to the reception of weak
Database check at power up and once per day
period relocation
OTA TV signals at the headend , and,
Table 1. Summary of FCC TVBD Rules 2) Interference due to direct pickup
(DPU) at the subscriber’s residence.
While, the database is the primary method Interference in the cable distribution plant
of protecting OTA TV broadcasts, sensing between the headend and the subscriber taps
must also be utilized by all TVBDs to detect of the cable system is not expected to be an
the presence of incumbents. TVBDs sensing issue. This is because the distribution portion
ATSC or NTSC signals above the -114 dBm operates at higher RF levels and employs well
threshold must report the detection to the shielded hardware. Shielding weakness in
equipment operator and give the operator the this portion of the system would cause a cable
option to vacate the channel (i.e., TVBDs are operator to fail their Cumulative Leakage
not required to vacate a channel when sensing Index (CLI) requirements and require
detects a TV signal if the database indicates immediate remediation. The fiber portion of
that the channel is open). Wireless a Hybrid Fiber Coax (HFC) system is also
microphones must also be sensed and TVBDs immune to RF interference.
must vacate the channel when a wireless
microphone transmission is detected Cable Headend Protection
(regardless of the database). In general,
TVBDs must sense a channel for a minimum The newly adopted FCC TVBD rules
of 30 seconds before using it, and sense at afford protection for OTA cable headend
least every minute during use to determine the receive sites. As noted previously, a TVBD
presence of incumbents. Cable system issues device is not allowed to operate co-channel to

a TV station within the TV station’s Grade A Cable operators will need to register
and B service contour. This provides information about these cable headend
protection to cable headend receive sites and receiver sites including their location and
TV viewers alike. Cable headend receivers affected channel(s) with the TV white space
that operate outside of OTA TV protected database provider.
service contours are afforded specific
protection by the geo-location database, as There is no specific protection given to
requested of the FCC by both NCTA and cable headend receivers that operate inside of
Motorola [4][5]. A TVBD is allowed to the grade B channel contours from adjacent
operate beyond the grade B contour, but not channel TVBD interference. Recall that
within specified distances of a headend that is portable TVBDs are allowed to transmit up to
registered in the geo-location database as 40 mW EIRP within the grade B contour on
detailed in Table 2 and Figure 3. Note that the adjacent channel of a TV station. Cable
both co-channel and adjacent channel headend receivers as well as TV receivers
operation is prohibited within the “keyhole could potentially be adversely affected by
zones" although the area of protection for adjacent channel interference from TVBDs.
adjacent channel is much smaller than that for Therefore, it is important that cable headend
co-channel as shown in Figure 3. OTA receivers have excellent adjacent
channel rejection, and that the OTA antennas
Within Beyond be as high and directional as possible.
Grade B Grade B
Contour Contour Direct Pickup (DPU) Interference at the
Co- TV station Headend in data base Subscriber’s Residence
channel in data base establishes "keyhole
protection protection zone" As stated above, the newly adopted TVBD
Adj no Headend in data base operating rules allow up to 100mW EIRP
channel protection establishes "keyhole levels for personal/portable devices. It was
protection protection zone" shown during several FCC tests that direct
Table 2. Geo-location Protection pick-up (DPU) to cable systems can occur at
Mechanisms for Headend ATSC Reception levels well below 100 mW [6][7]. Both
Motorola and the NCTA recommended
limiting in-home devices to much lower
power levels (i.e., 10 mW) to avoid the DPU
interference issue [5][4]. However, the FCC
chose to allow higher power levels in the
TVWS ruling. Remediation of DPU
interference will become more critical in cable
systems when TVBDs are deployed.
DPU Analysis
Interference from TVBDs to the cable

system headend is addressed by the geo-
location database and sensing. These are the
Figure 3. Distant
Database Protection
Cable Headend same mechanisms designed to protect OTA
TV reception. However, interference can also
occur at the other end of the cable system, at
the subscriber’s home as shown in Figure 4.

This is the result of co-channel ingress or will become less troublesome, as the full
leakage into the cable system. Ingress can power analog TV transmitters go off the air.
enter the cable system anywhere there is a In the VHF range, the OTA and cable TV
weakness in the system shield including video carriers were often on the same channel
poorly shielded or damaged coaxial cable, and DPU caused severe beat interference and
poor quality, loose or corroded F-connectors, ghosting. After the transition when the off air
open system ports on splitters and wall plates, analog signals are replaced with 8-VSB
and poor shielding in TVs, PC tuner cards, digital signals, the high energy video carriers
VCRs, set top boxes, cable modems, or will no longer be present and the visually
multimedia terminal adapters (MTAs). severe DPU interference will be replaced by a
lower level of snow-like DPU interference for
on-channel analog cable services. Offsetting
Indoors this, however, the wide-spread deployment of
Outdoors
700 MHz 4G mobile handsets as noted above,
along with nearby lower powered TVBDs
DPU into STB
DPU into cable
STB
DPU into
drop cable Ch 3 / 4 and
have the potential to significantly increase the

DPU
into cable video wiring
Hard line Trunk Cable
TV not susceptible
Distribution Amp
instances of DPU becoming a problem.

drop cable
Distribution not susceptible 4-way
DPU into splitter
due to high RF levels and
Sufficiently strong DPU can completely

DPU into TV
excellent shielding DPU into cable
disrupt services on the affected channel,

Cable
Ready TV
DPU into
Open cable tap
DPU into
Cable modem which may include multiple SD and HD
Cable Modem
video, high speed data, or telephony services.
Some operators have chosen to stop using
Figure 4. Home Cable System Ingress
Points some cable channels in the 800 MHz band
that were suffering from widespread severe
The interference due to DPU from TVBDs DPU, caused by mobile handsets.
is similar to DPU interference that cable
operators have experienced in the past from The fundamental interference protection
over the air signals including those directly mechanism used by TVBDs to protect OTA
from TV stations, ham radio operators, land reception is the selection of an unused
mobile police and fire radios etc. More frequency or white space. Unused OTA
recently, as some cable system operation has frequencies, however, have little relationship
extended above 800 MHz, DPU interference to unused cable channels. There are very few,
from cell-phones has also occurred. After the if any, white spaces on a cable plant,
digital TV transition, the 700 MHz to 800 especially with the growing demands for more
MHz band will become occupied with high HD, on-demand and data services. This
powered 4G wireless transmitters (e.g., 700 means that every TVBD has the potential to
MHz base stations EIRP levels can exceed 1 interfere with one or more services on the
kW) as well as additional mobile handsets cable plant via DPU. The remaining
(e.g., with typical 200 mW transmit power interference, or DPU, protection mechanism is
levels). The possibility of DPU interference the "shielding effectiveness" (SE) of the cable
due to these services will also need to be system, which includes the entire residential
addressed through the same remediation cabling system including all the devices
techniques. attached to it such as TVs, set tops, and
modems. Once an interfering signal has
It is interesting to note, that after the digital entered the cable system, this undesired
transition, what was previously one of the interferer travels throughout the home along
most troublesome forms of cable system with the desired cable signals. As such, a
DPU, pickup from analog TV transmissions, TVBD causing interference may be located at

the opposite end of the home well removed Radiated E Field of TVWS Devices
at Various Power Levels
from where the interference is observed. 10.000
4 W (+36 dBm) EIRP
100 mW (+20 dBm) EIRP
The FCC has established the RF DPU 40 mW (+16 dBm) EIRP
E Field V/m Far Field

1.000 FCC Cable Ready 0.1 V/m
standard for analog cable ready TV receivers

in 47 CFR 15.118 as 0.1 V/m. The rules do
not explicitly refer to digital cable ready 0.100
digital TV's but we can assume the same

requirement applies. The FCC acknowledged 0.010
the widespread use of Channels 3 and 4 as the 10 100

Distance Feet
1000
in-home interconnect frequency and has

Figure 5. Susceptibility as a
prohibited TVBDs from operating on those Function of Radiated Power and
channels. Subscribers using a set-top box Distance
(STB) with a Ch3/4 connection from the STB
to the TV will be protected from interference Even when a cable ready TV is not being
in this interconnect. Likewise systems using a used, the in-home cable wiring itself can be
baseband audio / video (e.g., composite, s- susceptible to DPU [5]. Motorola conducted
video, component, DVI, or HDMI) tests on various home wiring components
connection should not suffer from interference including a selection of retail purchased
to this connection. However, for those cables and splitters and found a wide range of
subscribers not using a STB but instead using shielding effectiveness. For the purposes of
a cable ready TV connected directly to the these tests, the shielding effectiveness (SE) of
cable, the TV itself may be susceptible to a sample of cable is the ratio between the
DPU as shown in tests conducted by the FCC external electric field measured in V/m and
[4]. the induced voltage on a 75 Ω load connected
to the cable. For example, if a cable exposed
Figure 5 shows the free space far field to 1V/m (60 dBmV/m) induces 0 dBmV into
strength at various distances radiated by the load, the shielding effectiveness is 60 dB.
TVBDs operating at 40 mW, 100mW and 4
Watts. The dashed horizontal line at 0.1 V/m For comparison, the shielding
represents the minimum FCC specification 47 effectiveness of an FCC 47 CFR 15.118
CFR 15.118 for cable ready TV set immunity. compliant TV set was calculated. FCC 47
The lines cross at about 40, 60 and 400 feet CFR 15.118 specifies that with a desired
respectively. Thus we can expect the signal level of 0 dBmV and a 0.1V/m
possibility of DPU interference directly to a interfering field, the DPU interference must
cable ready TV that just meets minimum FCC be better than -45 dBc. Thus a 40 dBmV/m
DPU specifications at these distances in free interference field must induce an equivalent
space. interference signal below -45 dBmV.
Therefore the minimum required shielding
effectiveness is 85 dB.
The specified minimum tap level for 256

QAM is usually -12 dBmV. The C/I ratio for
Threshold of Visibility (TOV) for 256 QAM
is about -27 dBc. Therefore any interfering
DPU signal above -39 dBmV will disrupt the
QAM signal. Note that in the case of an
analog signal, the DPU interference becomes

visible above -45 dBmV but will not Sample Typ SE at
completely degrade the picture or sound. In 500 MHz
the case of 256 QAM digital signals, Noise floor of test system 130 dB
interference above -39 dBmV will not just High Quality Cable with 110 dB
become visible but will completely destroy barrel and termination
the picture and sound. This is due to the "cliff High Quality Cable with 90 dB
effect" of the digital video systems. This barrel, NO termination
difference needs to be kept in mind when High Quality Cable, open F 75 dB
considering the ramifications of various levels connector
of interference. Subscribers may be expected
Splitter with soldered back, 125 dB
to tolerate some level of visible DPU
all ports terminated
interference to analog signals but cannot be
Splitter with glued back, all 80 dB
expected to tolerate a total loss of picture and
ports terminated
sound due to DPU interference to digital
signals. Splitter with soldered back, 95 dB
all ports open
Motorola tested the shielding effectiveness Retail cable, RG6 Quad 130 dB
(SE) of various coaxial cables and wiring Shield, Snap ‘n Seal
components and provided this information to Retail cable, RG59, Molded 85 dB
the FCC [5]. An example of the SE test data connector assemblies
Table 3. Shielding Effectiveness of
is shown in Figure 6. Various Cables and Terminations
The best cables had a SE measured around Figure 7 shows the effect of the common
130 dB which was the limit of the test practice of leaving an unused cable end
equipment. The worst cables were in the exposed (i.e., unterminated) which causes
range of 70 to 80 dB. The type of termination large degradation to the SE. This is likely to
Reference Cable with Various terminations
be typical of pre-wired homes in rooms
140 without TVs or other devices connected to the
130
System Noise cable system.
Shielding Effectiveness dB
120 Floor
with screw on
110 Termination
with push on System with open ports
100 termination
(same cables, various splitters)
90 with Barrel but
no termination
80 open cable end 140
130
Shielding Effectiveness dB
70
120
60
0 200 400 600 800 110
Frequency MHz 100
90
Figure 6. Shielding Effectiveness of 80
Various Cables 70
60
on the end of the cable was also found to have

0 200 400 600 800
Frequency MHz
a large impact on SE, and hence DPU Figure 7. The Effects of

interference susceptibility. Unterminated Ports (Open Wall Plates)
Table 3 summarizes the mid-band 500 MHz

value of shielding effectiveness we found for
various samples of cables, terminations and
splitters.

DPU Mitigation cable ready TVs should be set to levels that
are well above minimum system and
There are several steps a cable operator can regulatory requirements. Every extra dB of
take to mitigate interference from TVBDs, the signal level here provides another dB of DPU
new 700 MHz services, and the existing 800 immunity. As analog signals are replaced
MHz mobile services. In the case of TVBD with digital signals, the need to maintain very
interference to the headend, the operator must low levels of composite second-order (CSO)
register headends receiving OTA signals and composite triple beat (CTB) are reduced.
beyond the grade B contour in the TVBD geo- Amplifiers can be used at the residential
location database and keep these records point-of-entry to increase the signal level on
accurate. the in-home wiring without having to raise the
signal levels throughout the entire distribution
For all headends receiving OTA signals, plant.
the OTA antenna(s) should be highly
directional, properly pointed, and mounted as Figure 8 shows the E Field DPU
high and as far as possible from likely susceptibility for 256 QAM against a range of
locations for portable/handheld TVBDs. cable shielding effectiveness for several levels
ATSC headend receivers should also have of QAM operating level.
good adjacent channel rejection characteristics
and quality low-loss cabling between the E Field DPU Suseptability for 256 QAM at various levels
antennas and receivers. 10.000
A combination of remediation approaches

E Field V/m Far Field
may be needed to resolve DPU interference in

1.000 +6 dBmV
0 dBmV
subscriber homes. One of the most important -6 dBmV

-12 dBmV
techniques is to properly terminate open ports. 0.100
Cable drops, splitters and terminations should

all be of high quality and properly installed.
TVBDs and cell phones should be located 0.010
70 75 80 85 90 95 100 105 110 115 120
away from the cable system and TVs. Old or Shielding Effectivness dB
unused cabling/splitters in the home should be Figure 8. DPU Susceptibility as a

disconnected from the system. And in Function of SE
extreme cases, it may be necessary to
disconnect poorly shielded cable ready TVs For example, using a cable with 100 dB SE
and install well shielded set top boxes in front and a 256 QAM operating level of -12 dBmV
of them. Cable operators may also want to yields a DPU susceptibility of 1 V/meter. We
work with local retailers to insure that cable can use Figure 5 and Figure 8 together to
components with high shielding effectiveness examine the relationship between desired
are readily available in retail outlets. And signal level, and system SE vs. TVBD power
cable operators may want to prepare and level and separation distance. Using the 1
distribute information to their subscribers V/meter level we obtained from Figure 8, and
explaining self-help remediation procedures. referring to Figure 5, we should not expect
DPU problems from a 4 Watt TVBD at a
There is another technical step the cable distance of about 35 feet or greater free space.
operator can consider. Increasing the signal Note that any intervening walls or structures
levels on the in-home wiring directly reduces between the TVBD and the affected cable
the effect of DPU interference. Signal levels device would decrease the required separation
delivered to the inputs of set top boxes and distance.

Such ingress, referred to as DPU, is most
CONCLUSION likely to occur in subscriber residences both
due to the lower signal levels used within
TV White Space presents interesting subscriber drops and homes versus the main
opportunities to utilize the TV broadcast cable plant as well as the wide variety of
spectrum for a wide range of unlicensed cabling, device types, quality and installation
applications. The FCC rules currently provide of the home wiring systems. Additionally,
for two classes of TVWS devices: fixed and many of the potential TVBD applications are
personal/portable TVBDs; each are permitted consumer focused, and such TVBDs are likely
different maximum power levels. The to be in close proximity to subscriber cable
primary protection mechanism for licensed systems.
users of the TV spectrum is the mandatory use
of a geo-location database. Sensing is defined The TVBD rules that the FCC decided to
as a secondary protection mechanism. adopt, allow higher in-home power levels than
Additionally, the FCC rules allow for a class either NCTA or Motorola recommended in
of sensing-only products at a lower power FCC comments [4][5], as well as power levels
level, but such devices will have to undergo higher than the levels at which the FCC found
additional testing and qualification by the caused cable system DPU interference in its
FCC before such devices may be sold. own field tests [6]. As a result, many
subscriber homes may experience DPU
For cable operators, TVWS presents both interference resulting from the deployment of
the opportunities for additional wireless TVBDs and remediation would be required to
applications and services as well as risks to resolve these issues. These interference issues
their current cable system operations. The are likely to be similar to those that some
risks include the potential for TVBDs to cause operators have already experienced with
interference in cable systems. existing mobile phones in the 800 MHz band.
With the reallocation of the 700 MHz band
The FCC acknowledged the NCTA and including significant spectrum licensed for
Motorola comments on the potential of TVBD high powered 4G mobile services, cable
interference to headend OTA reception of services on those frequencies are likely to see
distant ATSC TV signals and the FCC additional interference as well. Fortunately,
provided a specific protection mechanism in the same remediation techniques can resolve
the TVBD rules. Such headends are permitted all these potential interference issues.
to register in the geo-location database and are
granted a keyhole protection zone to facilitate Consumer demand for bandwidth
these distant OTA reception requirements. continues to grow, and this demand provides
continuing growth prospects for cable
Cable systems today typically operate up operators. TVBDs provide unlicensed access
to 750 MHz, 860 MHz, or 1GHz and often to significant spectrum, thereby enabling
have no white space due to the high additional cable operator wireless applications
subscriber demand for more HD, on-demand expansion. Cable operators will have to lead
and data services. As such, all TVBDs DPU remediation efforts with both their
operating in broadcast TV spectrum “white subscribers and local retailers in order to
space” are, by definition, transmitting co- maintain and grow the capacity of their
channel to one or more cable services. As current cable systems, paving the way for
such, any ingress of a TVBD into the cable these additional wireless applications.
system will cause interference and possibly
even loss of service on the affected channels.

REFERENCES
[1] FCC Auction 73, 700 MHz Band.

[2] FCC Report and Order, Docket 04-186,
11/14/08.
[3] White Spaces Database Group, 2/4/09,
http://www.google.com/intl/en/press/press
rel/20090204_whitespaces.html
[4] Comments of NCTA, Docket 04-186,
1/31/07
[5] Ex Parte of Motorola, Docket 04-186,
12/7/07
[6] FCC OET DPU Test Report, Docket 04-
186, 7/31/07.
[7] FCC OET Phase II Test Report, Docket
04-186, 10/15/08.

Cloud Computing for Cable TV: Expanding Choice, Control & Content for a New
Generation of Television Viewer
John W. Callahan, Chief Technology Officer, ActiveVideo Networks
Jeremy Edmonds, Director, Product Management, ActiveVideo Networks
INTRODUCTION: streams, i.e., bundling, “tiering” and 30-

second ad insertion.
The multichannel television industry is
caught between two significant realities: And second, the inexorable growth in IP
Growing consumer demand for increased access network bandwidth, exemplified by
choice and control of TV content, and an the advent of DOCSIS 3.0, serves to enable
installed base of technologically fragmented, further disaggregation of the bundled model
resource-limited digital set-tops and CE for video consumers.
devices.
As the operators’ service delivery
Consumers have become acclimated to platform continues to change, however, so
the media richness, immediacy and dynamic does the software applications platform. The
nature of Internet video. It follows that availability of cable-delivered EBIF
consumers – particularly those under 30 [Enhanced Binary Interchange Format],
years of age -- are increasingly seeking to tru2way and unicast streaming applications
avoid the usability limitations associated such as NDVR [network-based digital video
with traditional broadcast multichannel recording], will allow operators to more
video delivery systems. easily introduce new video services –
subscription, transactional, and advertising
As a consequence, content owners, supported. The debut of a new generation of
service providers, and advertisers alike face “broadband-connected” TV sets will
major challenges to traditional business catalyze this as consumers will find it easier
models as video consumers shift to the one- to “watch television” without subscribing to
to-one world of unicast, on-demand video the traditional operators’ services.
streams that characterize VOD, DVR, and
Internet video. In the same way that the Internet is
shifting toward a “cloud” model -- where
The notion of “just plug a broadband computing resources are provided as a
connection into your HDTV” – a service, over the Internet -- so can
predominant trend at the January 2009 multichannel video providers leverage their
Consumer Electronics Show -- is the latest local, regional and national footprints to be
step in the process of moving from the the “TV cloud” for more and better content.
traditional operator business model of
bundling access and content, to a world Specifically, this approach could deliver
where the consumer chooses “what, when, a stream of personalized content as a single,
and where” to “watch television” [as our standard MPEG stream -- to any digital set-
parents called it]. The implications for top box or web-connected CE device.
cable are clear. First, this movement
damages the operator’s traditional revenue

This paper provides an overview of • Faster software application lifecycle
current impediments to TV-based and lower total cost of ownership
interactivity, and describes the high-level (development, integration,
requirements and capabilities of the “cloud deployment, content management).
computing for the TV” architecture that
could improve the outlook. • The ability to deploy rich media
services (i.e., complex video,
THE CASE FOR THE CLOUD graphics, and interactivity)to all
CPE, eliminating “lowest common
The language of the computing “cloud” denominator set-top issues” and
is typically associated with the Internet, increasing scale for operators,
even though the term itself pre-dates the programmers and advertisers.
Internet by at least two decades. Before it
was the “cloud,” computer scientists • Consistent presentation of content
recognized the need for “grid computing,” and advertising opportunities (which
“distributed processing,” and other minimizes unpredictable nature of
variations on the theme of sharing a the user experience on different CPE
processing workload over clustered and optimizes the ability to
computers. accurately place advertising
elements).
The benefits of cloud computing apply to
interactive television, as well. The “TV There are certainly capital investment
cloud,” for instance, could remove workflow attributes to the argument above. Generally
gaps and bridge the application and media speaking, investment in server-based
processing requirements between head-ends computing and network bandwidth leads to
and set-tops. A cloud approach also would less expensive CPE being able to provide an
enable developers of programming and array of video services (i.e., simple MPEG
advanced advertising to use familiar, web- decoders). That said, CPE with more
based development tools that are similar to capabilities can be equally supported. With
or the same as those used to provide respect to the investment in software
interactivity within a web site (e.g., technologies, the greatest expense often is in
DHTML, JavaScript, etc.). targeted development, integration and
regression testing across dozens of different
As in all approaches, there are pros and CPE platforms -- each with its own
cons to the notion of locating computing and performance characteristics, graphics
media processing power in the operator display capabilities, and consequent impact
network, versus investing in generations of on the viewer experience. While tru2way,
more powerful Customer Premise EBIF, and other standard client software
Equipment (e.g., set-top boxes). Noting that platforms are certainly an improvement,
each user is given an individual, they do not eliminate the cost of integration
personalized video stream can sum up the and regression testing.
con of the network approach. Obviously,
this uses the operator’s network bandwidth As cable operators and multichannel
to carry the unicast stream. However, the video providers have deployed more and
pros of this approach include: more sophisticated applications, the costs of
client development have increased greatly –
measured both in the time to market (24 to

36 months not being at all uncommon) and In addition, the total cost of ownership
direct technology expense for complex labs for applications lifecycle is a phenomenon
and software expertise. well understood in the enterprise and
personal computing domains, but only
The server-centric approach allows an recently becoming an issue for multichannel
application to be developed, integrated, and video providers. In this sense, “total cost of
regression tested on a single software ownership” includes not only initial
platform. The time to realize a new or applications development, but also
modified application can be measured in customization, integration, bug fixing, and
months (if not weeks). The server-centric regression testing. That is, a “fat client
approach takes advantage of powerful solution,” to the exclusion of all else, is an
computing resources to enable sophisticated, extraordinarily difficult and expensive
media rich applications that can meet the business proposition.
operators’ collective need for a rapid and
repeatable way to “change the category,” by One of the features of the Web is that it is
creating must-have features that are unique essentially a "cloud computing" model in the
to the video platform. "software as a service" sense. It allows the
application lifecycle to be managed on a
Given the growth in IP bandwidth – both small and well-behaved domain of common
the access and backbone networks – devices – network servers and common
delivering “TV Quality IP Video” is quickly client devices, i.e., PCs. The fact that PCs all
becoming a viable model. The server- have enormous computing capabilities
centric approach is well suited to deliver means that media-rich applications depend
video services over the IP network on client-based media processing
(somewhat obvious when one considers the technologies for application execution (e.g.,
growing number of such providers). Flash).
Questions for operators relate to Multichannel video operators can share
development and deployment of new video the benefit of the Web approach and utilize
services that maximize their investments in network-based servers to support
their robust, high QoS “last mile networks” complicated, media-rich applications -- but
for both traditional HFC and IP connected they cannot depend on the client to have the
devices. What is the appropriate application necessary capabilities. Unlike the ubiquity
software model for video applications? In of just one or two operating systems and
short, via “the server cloud or the CPE [set- common APIs, the multichannel video
top] client?” provider has dozens, if not hundreds, of
client [set-top] platforms - all differing in
Addressing the industry’s set-top
the details of capabilities and performance.
fragmentation problem – meaning the
hundreds of permutations of differing Standardization of the client side will
hardware and software combinations that take many years. By analogy, it took more
exist amongst the 37+ million digital cable than a decade for Microsoft Windows® to
boxes currently installed – will go far in become a de facto standard for desktop PCs.
assuaging the “critical timing factor” Application deployment to PCs is a
necessary to quickly remove applications demonstrably slow and expensive
that aren’t attracting consumers, and to proposition as compared to the “web model”
quickly add those that are. of software as a service. Web applications

are iterated frequently and at low cost. Cable Immersive interactive video applications
operators all have one common “standard” will bring the desired content “to the
in their client devices, i.e., MPEG decoders. viewer,” not make the viewer search to find
This, combined with server-based a “destination” in an unnatural way. This
applications in the cloud will bring a more “surfacing the content to the viewer” (versus
“web like” application life cycle model to “destination-based interactivity”) can be
the television viewing platform. found on many video streaming websites,
e.g., YouTube, where the activity of viewing
This Web model approach would any given video stream is augmented by
combine the concept of “cloud computing” meta-data links to several other video assets
with some aspects of the traditional set-top (as well as non-video applets).
approach. In essence, it binds together the
best features of cable, cloud, and client. Transposing this experience to a full-
screen video monitor, viewed at distance,
CURRENT IMPEDIMENTS TO TV-
and eliminating the computer mouse and
BASED INTERACTIVITY
keyboard bring the “VOD” and “interactive
The storied past of TV-based interactivity TV” models together as natural features of
goes back as many as three decades, yet “watching television.”
“interactive TV,” as a category, seems
always to be on the horizon. The word THE WORKFLOW ISSUE
“interactivity” itself is part of the problem:
As soon as an “interactive application” gains Several other impediments to TV-based
consumer traction, it takes on a descriptor interactivity exist. One is an overall lack of
other than “interactive.” Consequently, the an automated systems infrastructure to
application exits the perceived realm of connect the “sales order process” with the
“interactivity,” becomes part of the “creative process” to the “content
“normal” viewing experience, and ceases to management and provisioning process” and
exist as an example of “interactive TV.” finally to the “delivery process”. The overall
Examples include the electronic program word to describe this is “workflow.”
guide (EPG) and video on demand (VOD).
Both provide interactivity, yet neither is For the traditional multichannel video
considered an “interactive application” as subscription business, this workflow is well
they have passed into the realm of the established. In its simplest form, movies and
“normal” viewing experience. TV shows are produced, licensed to an
aggregator (e.g., NBCU), wholesaled to an
The historical predominance of operator (e.g., Comcast) for distribution, and
“destination-based” interactivity is also a then retailed to the consumer. The
factor. To access VOD, consumers were advertising and subscription models are well
taught to “go to” a place to find and order established for this process.
titles. Likewise for the electronic program
guide, which exists as a separate menu The important point is that there are
destination. A strong argument can be made automated systems (encoding, content
for interactivity to occur as a natural part of protection, “billing systems,” trafficking
the viewing experience that it enhances. systems, royalty payment and settlement)
that support this model so these businesses
can scale.

That workflow scaffolding includes the
With respect to interactive applications, following: A known, easy and repeatable
this “workflow” does not exist in any method for creating applications and
uniform, scalable way. The current applying any QA [quality assurance]
ecosystem of extant and desired interactive mechanisms to ensure applications behave at
video applications and services relies on a their best; data collection, to fulfill the
patchwork of business systems and creative application’s intent, and to feed any primary
tools, all of which are delivered to a or third-party billing mechanisms; and the
heterogeneous population of operators with links to those billing systems.
no “billable event tracking” except by
sneaker-net and swivel chair operations. Consider an advanced advertising
Without the “back-end” tied to the “front- application that allows the viewer to click on
end” via an automated workflow that a widget associated with an ad to receive
generates invoices and tracks payments and more information on the product. From a
respects copyrights, it will be very hard to workflow perspective, gaps emerge
build a scalable business around a popular immediately:
interactive application.
1) Creative: What should the widget
A specific example of the workflow look like? Who builds the creative
conundrum is the notion of the “bound” for the campaign – and to what
application, meaning an application that template, using what authoring
executes synchronously with the program or tool(s)?
advertisement within which it runs. 2) Application provisioning:
Operationally, the interactive
EBIF, the Enhanced Binary Interchange application must be provisioned on
Format, is the Cable Television Laboratories to the network – its widget assets
specification developed to establish bound transferred for playout, its
applications over two-way video plant. availability parameters fed into the
EBIF’s strength is its overall reach – traffic/billing system.
essentially the entire installed base of digital 3) Stewardship: All ad campaigns
cable set-tops. However, EBIF defines only follow general and specific rule sets
a portion of how to execute bound – competing products may not be
[program-synchronous] and unbound shown within the same ad pod; time
applications. parameters to protect children from
inappropriate content, etc.
Specifically, EBIF defines only the 4) Data Collection: After playout, data
delivery chain of the “trigger” or “widget” associated with the spot needs a
that enables a consumer to “click” from the method to flow into the aggregation
remote control and to engage with the TV engines feeding national and local
and the program or advertisement at hand. campaigns.
Such definition is critical and necessary, but 5) Billing: Any additional revenue
for EBIF-based “bound” applications to associated with the interactive spot
become mainstream, a necessary scaffolding needs a feed into operator billing
of workflow must emerge. systems.
6) Reporting and Settlement: automated
mechanisms must be available to

operators and advertising and without the benefit of network server
constituencies, etc. to create reports resources means the wealth of capabilities in
both for advertising effectiveness the newest units is eclipsed by the care-and-
and contract fulfillment purposes. feeding needs of the oldest units.
While many efforts are underway, the

authors do not know of any available THE CHALLENGE OF
solutions that will connect the traditional THE INSTALL-ING BASE
day-to-day business of advertising sales to
the operators’ broadcast and unicast Equipment fragmentation problems are
streaming platforms. Individually and not contained to the set-tops of multichannel
combined, workflow gaps prevent the video providers. While the “TV Widgets” so
business from scaling and the ability for prevalent at the 2009 Consumer Electronics
multichannel video providers to build both Show achieved high marks for “cool factor,”
local and national advertising revenues. they will ultimately face similar challenges.
Assert: Consumers will tolerate poor
THE CHALLENGE OF quality of service when using an “interactive
THE INSTALLED BASE application” on the Internet versus on the
TV. Consider: Millions of dollars have been
Digital cable set-tops, as a category, are spent trying to make channel changes take
approaching their 15th anniversary. Until half a second less time, because consumers
fairly recently, they’ve existed as the “thin dislike having to wait two seconds instead of
clients” that lag the Moore’s Law trend of one and a half to change from one video
computing devices. Compared to PCs, stream to another.
digital set-tops have long been dismissed as
Consider the developer, though, seeking
not including enough processing power or
to get a new “Yahoo Widget” into a
memory to enable immersive, media rich
Samsung television set. First, the widget
applications. In short, what’s thick today is
gets submitted to Yahoo’s TV widget
thin tomorrow, and, for digital cable boxes -
working group for a testing cycle (with
- compared to PCs -- it’s always tomorrow.
associated costs) that may be lengthy. After
approval, the widget is presented to
The installed base of digital boxes
Samsung, likely requiring a convincing
presents a “lowest common denominator”
business model agreement to justify
problem for application development and
association with the Samsung brand. These
software version control. Building
steps are repeated for inclusion on other
applications only for high-end boxes reduces
“widget-ready” devices, and likely would
potential reach; building applications for all
not be possible for an independent widget
set-top variations reduces the application’s
developer to perform.
attractiveness to the lowest common
denominator of graphics chips, processing Another popular client environment,
power, and memory. Adobe Flash®, typically releases an entirely
new software stack on an annual basis. Each
Put another way, operating interactive release assumes a Moore’s Law-like
applications solely upon the limited increase in available computational power
capabilities of the aggregate set-top base, and memory in the underlying platform –

yet that assumption likely will not fit the Requirements to enable server-centric
razor-thin margins and cost justifications applications, using content cached locally as
required in the CE manufacturing well as pulling content from Internet
environment. sources, include:
Indeed, making a Yahoo Widget or Requirement 1: Provisioning of backbone
Adobe Flash engine run exactly the same, on network interconnections to 100 Mbps or
all consumer display devices, and so that the greater between headend servers and
application author is not exposed to a maze relevant Content Distribution Network
of confusing, conflicting specifications is a locations. This is obviously sensitive to the
very complex task. The result, as has been characteristics of individual applications. In
the case with set-top boxes, is the likely effect, operators need to enable a terrestrial
pruning of platform features to the lowest caching point at the edge of set-top network
common denominator. (as is done for IP services). The required
bandwidth will be a calculated value derived
REQUIREMENTS TO ENABLE SERVER-
from customary traffic analyses.
CENTRIC APPLICATIONS ON THE
ACCESS NETWORK
Requirement 2: The provisioning of
The “new new thing,” vis-à-vis using adequate access network bandwidth to
server-centric applications to deliver deliver unicast video streams to a meet a
interactive video applications to thin client specified QoS. That is, the operator will
set-tops and/or CE client devices, is estimate “peak simultaneous stream usage”
connecting the population of client devices and ensure the network Session Resource
to the server cloud with sufficient Management components are provisioned
bandwidth. This seems obvious yet, in accordingly.
practice, it more than likely is not. Why:
Multichannel video operators generally Requirement 3: The transcoding of
maintain copious amounts of bandwidth to content to formats supported by the client
connect their IP access networks to the population. While there is a diversity of
Internet, yet the traditional headend is not formats on the Internet (and a diversity of
likely to have more than a data connection plugins to process them), the video
for File Transfer Protocol (FTP) support. streaming format that ALL CPE devices can
This is a result of the fact that the process is MPEG. In addition, for cable
subscription video content has traditionally operators, the ability to multiplex EBIF
been served by point-to-multipoint satellite applications and/or deliver Java applets is
distribution, or by low bandwidth links for required.
IPG data or other non-video applications.

Requirement 4: Workflow support for DOWNSIDES OF A “PURE CLOUD”
integrating the application platform with TOPOLOGY:
“business systems” and “content
management systems” back ends. This issue To be fair, an entirely cloud-based
was outlined in the workflow discussion. approach to interactive TV isn’t the right
answer, either. Doing so wouldn’t take
Requirement 5: The availability of advantage of client device capabilities, like
common Web authoring. That is, tools that local graphics blending, and a location
use DHTML as well as common scripting execution environment for “overlay
and video object toolsets. As applications applications.” An easy example of this is the
are transcoded to the specific client ability to detect embedded data “triggers” on
capability (as above), the application broadcast video streams (think “EBIF
developer does not need to be schooled in Trigger”). This allows a local application to
the various client capabilities and
idiosyncrasies.
Requirement 6: The guarantee that

applications will present the user experience
per the designer’s intent, consistently, on all
client devices

“overlay” graphics on the video stream (and provide associated overlay applications, as
execute application logic) only for those well as detect embedded video triggers that
viewers who choose to do so; it creates a can then be locally processed (on the client)
locally interactive individual experience to further enhance the viewer’s experience
with a broadcast video service. There is no by offering hyperlinks to related video or
need for such a mechanism with unicast other applications.
video streams as the viewer is already
engaged in an individual, one-to-one, An example of this could be a TV
experience. As the bandwidth intensive commercial with embedded EBIF triggers
video is shared (broadcast), this approach that enable telescoping into an ad microsite.
can be more efficient than unicast streams. The user reacts to the trigger, initiating a
local application overlay. The overlay
Likewise, certain applications are almost application displays several options,
entirely client-based. Chat and social including the ability to see special how-to
networking applications are good examples. videos. The user selects a video of interest
They typically require a user to log on and and the local EBIF application signals a
the network to maintain knowledge of their network server, which initiates a unicast
location, a.k.a., “presence-based” video stream in concert with the local
applications. These applications are application. The user is now seamlessly
typically comprised of low-bandwidth text viewing a linked video. This video stream is
and graphics and are not associated, or inherently personal (as it is unicast) and may
bound, to a unicast video stream. Simple contain further embedded triggers. As the
text and graphics can usually be processed server can process user requests for complex
by client devices using graphics overlay and media types and encode them to client
are not directly associated to a unicast video compatible media formats in real time, the
stream (think “chat room associated to a live application may bring all the depth of a
sporting event”). typical multimedia web site application to
the viewer.
The ideal environment for immersive,
video-linked interactivity, which infuses Meanwhile, a scaled version of the
Web-like characteristics into TV shows and broadcast stream can be included in every
ads, is a combination of cloud computing scene, enabling the viewer to continue to
and a traditional client-server applications view the original programming channel.
architecture approach. Network servers have Given appropriate “TV production values,”
the processing power and resources to create the net effect on the viewer is no obvious
media-rich, video-intensive applications. “interactive” or “VOD” application, simply
Client-side application execution can viewing a video with all the richness

of the Web and the immediacy of the classic not completely eliminate) the problem of
television presentation experience. lowest common denominator set-top
fragmentation. This is a serious issue when
CONCLUSION considering “scale.” The server and client
hybrid – one that maximizes the inherent
The optimal solution to provide the best strengths of a high-bandwidth, real-time,
viewing experience that combines the media two-way connection between the headend
richness of the modern Web with the quality and home, and that takes advantage of
of service and ease of use of traditional simultaneous MPEG and IP transport paths -
television uses a software applications - is the optimal approach to maximize the
platform that maximizes both server-centric return on investment in software delivered
computing and media processing power with multimedia applications.
client overlay and local application logic.
The Web model, if used as a guide for
modernizing and realizing the promise of the
cable broadband television platform, would
argue to use authoring tools and publishing
workflow for applications targeted at both
mainstream video and advertising
applications (the latter, taking advantage of
the oft cited “targeting” capabilities inherent
in the Web model and when using unicast
video applications).
The judicious use of server-centric

software applications will ameliorate (but

COMMON ADVANCED ADVERTISING SYSTEM
Arthur Orduña
Canoe Ventures LLC
Abstract
To operate within the CAAS, it is generally
The cable industry has a long history of acknowledged that the Participating MSOs
working together to advance our services for must adhere to common guidelines. MSOs
the benefit of consumers and our commercial must provide assurances that each of their
customers, including the advertising participating systems will:
community. CableLabs, National Cable • Support uniformity of product and ad
Communications (NCC) and the numerous formats
interconnects that dot the map are examples • Support the defined business processes
of this cooperative spirit. It is a natural • Provide common system interfaces
evolution that cable would work together to • Provide consistent data per the agreed
develop a standardized national platform for upon interfaces, semantics and policies
advanced digital advertising. That is the Given the difficulty of building the CAAS to
prime development task for Canoe Ventures. support multiple heterogeneous MSO delivery
systems, the scope will initially be limited to
At the heart of Canoe’s ecosystem will be MSO video services. However, this will be
the Common Advanced Advertising System expanded to include other platforms based on
and its over-arching business process market demand (e.g. mobile, HSD portals,
management system. This implies a massive other service providers).
data record keeping system and associated
business intelligence capable of aggregating, Reference Architecture
storing, mining and reporting on extremely
high volumes of data. Advanced analytics The reference architecture presents the
and inventory management systems are also envisioned end-to-end system and highlights
crucial to reaching the full potential for the portion of the system for Canoe. The
addressability and adaptive campaign underlying MSO distribution systems will use
management. a common set of interfaces, however the
specific design may vary by MSO or systems
within an MSO.
COMMON ADVANCED ADVERTISING
SYSTEM
Overview
The Common Advanced Advertising System

(CAAS) consists of the national Advertising
Stewardship business systems and enabling
infrastructure that interfaces with advertisers
and advertising agencies, media owners and
MSO distribution systems to deliver a range
of advanced digital advertising products. The
CAAS will be a centralized control center and
clearinghouse, and will communicate with
many different MSO delivery systems.

MSO Ad Sales Cable Networks Buyers
Common Ad Sales & Operations
& Operations & Broadcasters ( Advertisers /Agencies /Programmers /
Networks /MSO Marketing )
Private UI Web -based Master Control Interfaces

MSO Inventory Aggregated Inventory
Master Workflow M
o
n Common
Content it
Pre-Planning Campaign Reporting & o Advanced
Delivery & ri Advertising
& Ordering Delivery Billing n
Fulfillment g Systems
Reference DB Campaign Planner Data

Warehouse
Standard
Interface to MSO Systems I/F &
data
Campaign Asset Service Viewer

Manager Mgmt Measurement Response M
on
it
o MSO
Addressability Interactive Platform ri
Subscriber Billing n Specific
g
DB
Delivery System
Mobile
STB Client PC
Phone
Figure 1 - Reference System Architecture

The following reference diagram deconstructs the CAAS into its primary elements and business
processes.
Customer Assurance
eBusiness Systems
Ops Support
Ops Support
Finance
Master Workflow
Ops Support
Infrastructure & Services

Ops Support
Figure 2 - CAAS Business Processes
Advertising Stewardship enables a national Ad Network, and could possibly support

sales organization to manage the complete ad future business verticals. The data within the
sales lifecycle. Web based interfaces will SIS will support a wide range of data mining
provide the necessary access for both ad sales and analytics process, one of the most
stakeholders and customers for sales and important being audience segmentation to
campaign development, in-flight campaign support targeting. Through the use of bonded
management and post-flight reporting and 3rd party data services providers, blind-
financials. These systems will rely heavily on matching can be performed against MSO and
the underlying infrastructure services as well advertiser customer lists while maintaining
as interfaces to MSO systems. full consumer privacy.
The Subscriber Information Service (SIS) is The Distribution Services are intended to
the primary data warehouse aggregating simplify the process of creating, delivering
audience and service measurement data from and fulfilling digital advertising across
all of the distribution systems. It will supply multiple MSOs. While customers could
the necessary storage, processing, and interface directly with the MSO interfaces,
business intelligence functions to support the this service greatly reduces the complexity

and customer overhead by providing a single
organization to handle content preparation and Tying all these systems together is the Master
delivery across many different systems, and Workflow. Workflow automation will
potentially different target platforms such as connect the many business processes and
wireless or broadband. And, while it is the stakeholders with automated and traceable
long-term objective for all MSOs to use processes. Because of the many stakeholders
common content formats and delivery and complexity of the business processes, a
standards, in practice it will take time for all highly sophisticated, flexible, and secure
systems to reach a consistent level of solution based on proven technology is
deployment, and this service enables a more mandatory for successful operation of the
manageable migration path. system.
Workflow
The following diagram illustrates a high-level workflow between sellers, buyers, the CAAS and
MSO systems.
Proposal
Inventory
Settlements
Figure 3 - Example Workflow

modifications or enhancements, the complete
Design Objectives set of CAAS requirements.
Time to market is of paramount importance

for the Common Advanced Advertising
System. Therefore, it is highly desirable to
minimize custom development, and maximize
the use of existing or off-the-shelf systems.
However, it is important that the
recommended systems meet, or will meet with

Within the above overarching guidelines, our Information Systems
design approach has been to:
A robust data infrastructure coupled with the
• Evaluate best of breed products that business intelligence necessary to extract
provide functionality in line with the maximum value is the cornerstone of this
CAAS requirements and long-term effort. As the business objectives for the
objectives CAAS evolve, and as Internet advertising
• Minimize unnecessary custom work continues to experience strong growth, the
• Enable procurement flexibility and data required to enable this effort has also
product innovation increased in magnitude and importance. The
• Limit unnecessary dependencies and ability to accurately target and adapt
integrations – balance the benefits of campaigns, and thus the required Information
off-the-shelf products against vendor Systems to effectively manage, analyze and
management complexity securely govern very large amounts of data
are critical to the success of the Common
The resulting design centers on a loosely Advanced Advertising System.
coupled architecture framework with open
interfaces enabling multi-vendor solutions and
independent operation of distinct elements of
the system. For example, the Subscriber
Information Service is a complex information
system that might benefit from different
technologies, physical distribution and/or
hosting models than the Advertising Network.
Part of the design exercise weighed the
different models and present trade-offs in
terms of reliability, extensibility, scalability,
operability, cost and time to market.
It is also crucial to define and automate the

business processes using proven workflow
automation technology. Because advanced
digital advertising is still a nascent business, a
flexible rules engine and defined roles and
permissions are mandatory in order for the
system to keep pace with changing business
objectives. The workflow analysis should
place equal emphasis on core business
processes and operations monitoring and
quality control functions.
Design Considerations
There are critical system elements that are

expected to require significant new design.

Table 1 presents the range of data that passes
through the system. The design of the CAAS
and MSO system interfaces will cover these
as well as anticipated future needs.
Table 1 - Common Advanced Advertising System Data Needs

Reference & Stakeholder Order & Execution of Order Operational Reporting
Agencies Order Details Service Level Management Report Types

Networks Products Capabilities Delivery Confirmation
System Configuration Price Availability Affidavit
SYS Code Rules Monitoring Operational Reports
STB Financials System W ide Technical Operations Reporting
MSO Subscriber DB Content Reference Regional SLA Management
HSD (Broadband) Expected Info Full Footprint Sales Pipeline
W ireless Reporting Complete Financials
Historical Orders Affidavit HR People Reporting
Verification MSO Report Types
Planning Enhancem ent Inventory Utilization
Proposal data Approval Aggregate Financials
Inventory Order Expressions Dashboards
Dynam ic Confirmation Historical
Capacity Management Category
Product Revisions Sector
Pricing Change orders Performance
Business Rules Cancellations Com parative reports
Permissions Make Goods Co-op affiliate reporting
Reach RFI Fulfillm ent
Polling Results
Performance
Capture more than will report
Clicks associated with Views
Service Delivery
VOD
STB
Raw Data Sets
Im pressions
Initial Sources for service measurement from profiles and identify targeting groups, deliver
MSO subscribers includes but is not limited specific messages to each targeting group, and
to: determine the efficacy of their messages.
• Linear ad insertion The primary components of the targeting
• Settop box usage, including linear capability include:
viewing, DVR and time-shifted
viewing, and Interactive usage • Definition of targeting profiles to
• On demand usage match the advertiser’s segments of
interest using data from multiple
As a service provider, adherence to applicable sources.
consumer privacy laws and policies are • Identification of the targeting groups
always of critical importance. Any design in an iterative process of refining the
must include strict data governance rules and profiles to maximize the reach (e.g.
auditing capabilities. how many households) – through
blind matching and generating counts.
Addressability
• Creation of advertising packages
consisting of message, creative, and
National, Regional, and Local Advertisers
metadata that are then trafficked and
want to segment customers, develop targeting
delivered.
Segmentation can be applied to the
subscribers’ profile for household (and
eventually set-top box level) level targeting.

The efficacy of targeting is measured through owners and product lines. To accomplish this,
key metrics that are specific to the advanced the design must include billing and
advertising product (e.g. linear addressable, accounting solutions focused on a secure
RFI, etc.). infrastructure that supports accounting
standards and provides proven transactional
The targeting capability is largely a part of the
integrity. The billing design will
Campaign Development step in the
accommodate the creation of customer
advertising lifecycle. It is applicable to all the
invoices across companies and products as
products being contemplated within the
well as the distribution and tracking of these
CAAS advanced advertising tracks (e.g.
invoices. Any associated accounting
targeted overlays, linear addressable, etc.).
transactions will be tracked in concert with
It is envisioned that the MSO may require a GAAP Accounting standards and support
3rd party data services provider as an customer collections as well as vendor
infrastructure partner for meeting Personally payments. Finally, the design will allow for
Identifiable Information (PII) requirements, periodic reconciliation and settlement of
and to provide support for advertisers in financial transactions between participating
understanding and applying targeting to their companies.
advertising campaigns. In most cases, a
“blind” data match will be required for Workflow Automation
qualitative profiling to protect the privacy of
the specific subscriber file and information. As illustrated in the CAAS reference diagram,
Clarifying the relationship between MSO data all of the business processes will be
systems, the CAAS and 3rd party partners and orchestrated through automated workflow.
enumerating the trade-offs for structuring the The workflow engine will support flexible
underlying data systems is a critical design and extensible rules definition and will
objective. provide multiple views into the current state
Inventory Modeling and Placement Decisions of any campaign based on stakeholder access
rights. Due to the highly distributed nature of
the problem, it is very likely for campaigns to
The inventory management requirements for encounter errors somewhere in the process.
the CAAS are extremely complex and For example, content may not be successfully
dynamic given the number of distribution received at all sites, or rights may not have
systems and the different types of inventory been cleared by the required date. Therefore,
represented by the envisioned set of products. the solution should also provide active
It is the objective for distribution systems and monitoring and alerting capabilities for
media owners to communicate inventory defined error conditions.
status to the CAAS through well-defined
interfaces. The CAAS will correlate this Interfaces with MSO Systems
information and use it to support inventory
modeling during campaign development and An abstraction level – called the MSO System
to drive trafficking and placement decisions. Interfaces – will be designed to allow data and
message exchange between the CAAS and the
Invoicing and Accounting specific systems and equipment at the MSO
delivery sites.
In order for the CAAS to offer a cohesive
buying and tracking mechanism for customers A critical activity during the design phase is
across the Cable footprint, it must provide to perform a detailed analysis of the data and
customers with a single invoice across media messages that are used by the MSO delivery

systems shown in Figure 1. Canoe is working
with each Participating MSO in order to There is a strong desire that the interchange
collect a full set of requirements to describe between the CAAS and the MSO delivery
the data and messages that must be generated system be governed by standards-based
for and collected from the MSO delivery implementation. This covers the current suite
system in order to ensure proper operation of relevant standards and specifications from
within their systems. This may require ANSI, SCTE, CableLabs, IETF, and other
analyzing the capabilities and interfaces of involved entities. Of particular interest are
deployed traffic and billing, VOD, and SCTE-30, SCTE-35, SCTE-104, SCTE-118,
interactive systems to fully understand the SCTE-130, DVS-766, CableLabs ETV,
how these systems will interoperate within the CableLabs Metrics, and several subprojects of
end-to-end solution. Individual MSOs will the CableLabs VOD Metadata and OpenCable
make independent technology and initiatives. These are key to the development
procurement decisions regarding their of the MSO delivery system interfaces.
delivery systems, including the portions of
those systems that interface to the CAAS.
Standard Specification
SCTE 30 Digital Program Insertion Splicing API, SCTE 30, 2006
SCTE 35 Digital Program Insertion Cueing Message for Cable, SCTE 35, 2004
SCTE 104 Automation System to Compression System Communications Applications Program
Interface (API), ANSI/SCTE 104 2004
SCTE 118 Program-Specific Ad Insertion - Data Field Definitions, Functional Overview and
Application Guidelines, ANSI/SCTE 118-1 2006
SCTE 130 Digital Program Insertion – Advertising Systems Interfaces, SCTE 130
SCTE 138 Stream Conditioning for Switching of Addressable Content in Digital Television
Receivers, SCTE 138 2009
OC-ETV-AM OpenCable Enhanced TV Application Messaging Specification, OC-SP-ETV-
AM1.0-I03-060714
OC-ETV-BIF OpenCable ETV – Binary Interchange Format 1.0, OC-SP-ETV-BIF1.0-I03-060714
OC-SP-METRICS OpenCable Receiver Metrics Gathering Specification, OC-SP-Metrics-I02-070416
Table 2 - Standards Reference
constraints of the capabilities of legacy
Interfaces with Agency and Advertiser stewardship products in use today.
Systems
This effort will involve not only defining the
Canoe will also define an approach for data system components and communication
exchange with the Advertiser and Agency standards referenced within this document,
community. Again, there is a strong desire but also integrating the new products into the
that the approach between CAAS and its document trafficking workflow, protocols and
customers move toward a standards based formats currently in place with Legacy
approach leveraging and driving the creation Stewardship systems for various documents
of standard specifications within the including proposals, orders, revisions and
Advertising community (AAAA). However, it invoices.
also must be recognized that there will be an
ongoing need to electronically transact This demands familiarity with existing AAAA
business with this community within the common documents and protocols, work with
MSO and National Sales, as well as with the

lead software vendors in this space, in order to
define an approach for document exchange
that can support both current and future
capabilities of these platforms.
Conclusion
This paper described at a high-level the

functional architecture of Canoe’s Common
Advanced Advertising System. It is our intent,
in publishing this paper, to educate the
technical community on what we are
designing and building for our various
stakeholders.
It is our goal that the successful launch of the

CAAS will mark the first iteration of a
standardized national platform for advanced
digital advertising, resulting in additional and
ongoing innovation from multiple parties –
not just Canoe and the MSOs - at the services
and application levels.

CONNECTING CABLE’S NETWORK AND APPLICATIONS WITH ITS
INDUSTRY PARTNERS WHILE MAINTAINING OUR INDIVIDUALITY
Mitch Weinraub
Thomas Engdahl
Joshua Seiden
Comcast Media Center (CMC)
Abstract will also want to expand the connectivity

initiatives that involve seemingly competitive
The cable industry is working industries, such as mobile media providers
aggressively to expand the range of services and internet advertising structures.
that meet consumers demand for
personalized programming. In order to
deliver these new services efficiently, the
industry is changing its video delivery INTRODUCTION
infrastructure towards architecture that is
open and network –based, connecting Throughout the history of cable
headends to backbone networks and using television in the United States, the business
data to drive events. has been based on local providers providing
a unique service in each market. Besides
From its position at the intersection of providing an acceptable mode of operation,
programmers, advertisers, application it has also proven valuable in competing
developers, MSOs and hardware against satellite and other providers that, due
manufacturers, the Comcast Media Center to their technological differences, were
(CMC) has been able to develop a unique forced to deliver a homogenous service to
perspective on the current state of systems the entire country.
development and the gaps that are yet to be
resolved. From this intersection, it has Even as some cable MSOs grew larger
become apparent that the cable systems and began to align their services, a number
serving standalone markets will want to join of traditional cable aspects had remained
larger MSOs in expanding their level of unchanged. More recently, cable MSOs
connectivity, both vertically and have begun to standardize their navigation,
horizontally, in order to the opportunities packaging and other offers across their
that are emerging succeed in today’s market footprint, but these standardizations often
place. remain within a particular MSO.
The connections that the cable industry is However, there is less standardization
making fall into two primary categories. between MSOs. Examples include major
First, the industry will want to create tighter market DMAs served by two or more cable
connections to existing partners and other systems that haven’t created ad
MSOs. Second, the industry is extending its interconnects. It is becoming easier to make
connections to key partners in the media and these types of connections thanks to
entertainment industry and to advertisers. standardization among cable operators, as
Perhaps more importantly, cable systems evidenced by CableLabs’ recently issued

specifications for Modular Headend Things Changed When Consumers
Architecture. Became Connected
Another example of movement toward Some business historians may trace the
greater standardization is the multi-MSO beginning of a connected marketplace back
funded organization known as Canoe to the use of the U.S. Highway system, a
Ventures. With this organization in place, national telephone system, railroads, or even
some major, positive steps are being taken to the Wells Fargo Wagon. Similar to these
make sure that interactive and advanced past significant events, the ability for
advertising efforts are as standardized as consumers to go online has served as a
possible. This, when successful, will allow major catalyst for the connected
for the widest possible footprint for new and marketplace. Thanks to the Internet’s World
exciting forms of advertising. Wide Web, services now had the ability to
aggregate communities and consumers in
The industry’s move toward greater ways not possible before. In a significant
interconnectivity will support the steps way, this added a new dimension to business
being taking by initiatives such as Canoe in and created brand new opportunities.
order to offer highly competitive advanced Before this change, it was possible to be
advertising and other new business successful by providing a good product or
opportunities in today’s environment. As service to a limited (in both size and
this paper will demonstrate, one of the major geographic diversity) market. Local stores
challenges facing the cable industry in the flourished and providers that best serviced a
near future is the ability to maintain the local focused marketplace won the day.
and individual aspects that have defined
cable throughout its history while making When consumers went online, this all
these new connections.. began to change, but it took a while. At first,
online services looked a lot like the off-line
The significant and meaningful services that preceded them. Compuserve,
connections that are emerging include using Prodigy, and America Online became the
SIP (session initiated protocol) and other dominant players. Each provided a suite of
key technologies to connect service services that was individual, unique and
platforms between MSOs, connecting to catered to the specific users that they were
content providers, broadcasters and major trying to attract. And for a while this worked
studios, connecting with interactive well. Even as the web began to emerge,
television authors and platforms, and these services lived on as the best way to
connecting with prominent advertisers and connect to the web as well as to access
their agencies. These connections allow the additional “walled-garden” services.
cable industry to continue to grow and However, as web based services began to
remain foremost in the minds of consumers add features, the walled-garden approach
as the primary service provider for became less and less valuable. Web based
entertainment and telecommunications services - those that allowed access by
services. millions of consumers regardless of their
internet service provider (ISP) or means of
access - quickly became more valuable than
those (within the walled garden) that only
served members of the private club.

The recent Internet success stories have without the mass or scale of the connections
been services that reach out to the widest being made. These services could not
possible audience. Whether you look at it survive in small or isolated environments. In
from the perspective of a “tipping point” or fact, it could be argued that they are
a “network effect”, the volume of users successful primarily because they didn’t
doesn’t just demonstrate linear growth for isolate themselves but opened and connected
these services, it is also responsible for the their services to a bigger market.
basic success and is the only way the service
could survive. To paraphrase the famous This is the model that the cable industry
line from Marshall McLuhan, in the new is beginning to follow with its deployment
business environment, it could be said that of advanced video interactive services. In
the mass is the message. today’s marketplace there are multiple
competitors aiming to deliver service, many
One example of the mass of a service in a non-traditional fashion without even
being critical to its success is eBay. eBay owning their own networks. In response,
began its life as Auctionweb, a component cable operators are opening their networks
of the online services offered by Echo Bay, to connect a wider range of services to their
a consulting firm that served the San customers, markets and vendors, which
Francisco bay area. The Auctionweb portion allows the industry’s new advanced services
of the site quickly grew to become the entire to be successful. In this environment, the
Echo Bay site, which was renamed eBay and whole will be greater than the sum of the
broadened its scope to serve all comers on parts. This requires having a (big) slice of an
the web. i Once available to the world, the extremely large pie as opposed being limited
volume of buyers, sellers and items to a single cupcake. The following sections
available grew exponentially and the rest, as of this paper outline specific areas of the
they say, is history. In this case, the mass of business where the potential for these
the offering has completely transformed the opportunities may be realized through
service. building connections.
Likewise, YouTube.com and social MSOs NEED TO CONNECT

networking sites like Facebook provide MORE THAN EVER BEFORE
excellent examples of mass or scale. With
the entire world of on-line users able to In the past, cable was a parallel sort of
participate, these services have been wildly service where each head-end could handle
successful (from a usage if not yet a and process programming independently,
financial perspective). For these services, without any connection or knowledge of one
the network effect or mass is they key to another. In the case of traditional linear
their success. Their features may not be services (like CNN and HBO) it didn’t
extraordinary or cutting edge, but the sheer matter if there wasn’t an interactive
number of users and connections that can be connection. Conversely, the new advanced
made has been integral to their success. services that the industry and its partners are
striving to create are often the kinds of
A number of additional examples could services that can’t succeed or perhaps even
be cited, but the critical point here is that survive without extensive connections. From
these services represent a new kind of gaming to advertising to shopping to new
business success. The success of these kinds ways for viewing content, the cable industry
of services wouldn’t have been possible is poised to play a big role in the interactive,

connected and personalized service offerings will need to work everywhere -on every
of the next decade. However, many of these cable system regardless of size or location.
services fit the new business models
outlined above. This means that they will Additionally, new cable interactive
require mass and scale as opposed to closed services will need to connect to other
systems in order to succeed. systems that may not be based on the same
middleware. For instance, cell phone text
Canoe Ventures is already driving some messaging systems, 1-800 numbers and
of the key connections that will be required websites can already cast votes and
to enable advanced advertising related aggregate polls. Applications riding on the
services. These efforts are expected to prove cable industry’s iTV platforms will need to
extremely valuable and successful for all connect to these existing platforms (and
participants. This section of this paper others yet to be invented) in order to gain
focuses on the connections between the mass and scale necessary to be
providers that will be necessary to support successful.
and drive the growth of services outside of
those directly related to advertising, which Like the games and voting/polling
Canoe is already addressing. services described above, both social
interaction services (living room versions of
Some of the most basic interactive Facebook, etc.,) and content sharing services
services being developed today are in the (such as YouTube) will also require the
area of gaming. While these games are fun interaction of extremely large number of
and entertaining on an individual level, they users in order to succeed. As a result, cable
really become interesting when individuals and other last-mile service providers will
can play and measure their performance want to find ways to connect their offerings
against others. Although larger MSOs may and share or aggregate their users and
be able to aggregate enough of a user base interactions across the video universe, as
within their own footprint, this type of mass well as extending to online, mobile and
market just won’t be possible MSOs serving other electronic platforms.
smaller markets or single headends. In
addition, game developers will have the While applications or content services
opportunity to create truly “game changing” providers require open systems that support
gaming opportunities when they can draw the seamless delivery of their services, it
upon and rely upon a large and diverse user will be important for the connected
base, such the one available via the Internet. ecosystem to support each cable system’s or
MSOs requirements for certain forms of
Likewise, voting and polling applications autonomy. In addition to differences with
are often sited as early opportunities for local navigation, each system or MSO will
cable’s new interactive platforms. Like want to maintain the look and feel that
gaming, these applications thrive only when supports its brand identity as the provider of
the base of users is large enough. One of the services that are delivered over the local
reasons that American Idol works so well network. Fortunately, the connections
and attracts such a large audience is that required on the back-end that allow for mass
anyone and everyone can cast their vote and and scale to be aggregated don’t need to
be part of the process. In order for EBIF affect the front end where user interfaces
and tru2way™ applications to become the and branding opportunities are presented.
new standard for consumer interaction, they Using structures defined later in this paper,

we can accomplish one without impacting 1. Establishing a high level committee
the other. that would agree (on behalf of the
interested MSOs) on the definition
In the case of cable system operators and of centralized services for data
other last-mile service providers, the exchange.
question is “How can these connections be
made?” The initial plans for most of the 2. Creating an RFP to allow interested
services described above focus primarily on vendors to describe the details of
the local servers necessary to deliver the systems necessary to support
applications to customers. However, the the exchange.
connections required to allow these services
to work well and really grow are as 3. Establishing a timeline for the
important as the core functionality. deployment of a beta system that
Examples like cell phone call processing or application developers would use to
SABRE in the airline industry represent test against.
forms of “coop-etition” where competitors
in the same marketplace have come together Making these connections creates
to create common systems allowing an numerous service opportunities. The biggest
entire industry to grow. For example, the opportunities typically lie in the services that
SABRE airline reservations tool is used by will be invented in order to take advantage
more than 90 airlines across the globe. This of the connected platform. However, some
has enabled single interfaces and on-line of the services that can quickly emerge
booking services, but it wasn’t always the include:
case. Likewise, it wasn’t always so easy to
get your cell phone to work across different 1. Television- based social
phone networks. These industries have networking – Whether more like
found a way to share data and transactions. Facebook, Dating on Demand or
This is the model that the cable industry some combination of the two,
could follow in order to achieve the full meeting and sharing with others
potential for advanced services while across the country and the world will
maintaining differentiation. undoubtedly be significant.
For the cable industry, this involves 2. Television program related

establishing standard and trusted points of sharing – Examples include real-
data exchange. In this scenario, each cable time on screen fantasy sports
system operator would need to trust a matchups; or what to watch next -
centralized service that will allow all players recommendations from friends or
(cable and non-cable alike) to publish high just real time program chats with
scores and aggregate votes and user social connections. Interactive
generated content in a way that allows the applications that directly relate to
entire pie to grow. Historically, there are television content are also likely to
examples of cable MSOs that banded proliferate.
together to address this need and Canoe
3. Meaningful Real-time Voting and
serves as the latest example. Achieving
Polling – When most if not all
these requirements would involve:
viewers can be connected and

aggregated, voting and polling with offered to customers, but to also begin to
immediate results becomes possible. leverage ubiquity to generate new forms of
revenue. The current approach for video
4. Real-time competition on a mass delivery in the cable industry focuses on
scale – ideas in this space include “every MSO as its own entity”. The
massively multi-player quiz shows or resulting effect is that some MSOs find
prediction contests. When the themselves able to create new revenue
industry is able to match the quality streams while others struggle to follow on
and presence of television with real- the coattails. As history demonstrates,
time, two way connectivity on a innovation occurs at a faster level when the
large scale, the kinds of games cable industry can leverage its existing
customers are likely to play will capabilities in other platforms and
evolve and expand past what can be implement them across the majority of cable
envisioned today. operators
As mentioned above, the best ideas are As organizations such as Canoe focus on
almost always those that can’t yet even be deploying interactive ads nationally, they
imagined. Innovations that creative cannot be as successful without ubiquity.
developers using tru-2-way can come up Ubiquity is achieved when a single
with in their proverbial garages can’t even application can be distributed and function
be imagined until the world that provides across the entire cable footprint, in every
these kinds of connections comes into environment. The most straightforward
existence. The successes of the internet are a approach for accomplishing this objective
great reminder that vast, new capabilities are would be to construct applications
a prerequisite for game-changing successes generically, using generic interfaces, thus
such as eBay and YouTube. History has preventing the need to create new versions
also shown that replicating walled garden of applications for every single deployment
environments, such as Prodigy or in the country. This is obviously a daunting
Compuserve, on a big a screen in the living challenge, but with a few changes to how
room may not be enough take to take the the industry does business today, it can be
industry where it wants to go. accomplished.
The Elements of Application Creation

and Distribution
TECHNOLOGICAL REQUIREMENTS
FOR THE CONNECTED ECOSYSTEM EBIF (Enhanced Television-Binary
Interchange Format) and tru2way represent
Offering the next generation of services set-top abstraction layers. While these two
described above will require video services abstraction layers work in different ways,
to become IP-based. This will allow for they accomplish a similar task. Deploying
new applications to utilize the Internet to applications that run in EBIF or tru2way,
provide enhanced services as well as make it possible to have the applications be
connect to existing Internet-based services portable from one platform to another
and applications. without modifying the application’s code.
The primary goal of the industry is to not Application success is typically defined
only increase the choice of applications by both reach and use. It will be easier to

achieve national distribution if only one accelerate the deployment of national
version of the application exists. One of the applications in rapid fashion. Once this
challenges to overcome is allowing each cross-MSO CDN exists, an advertiser or
MSO to retain its own look and feel while application developer can gain access to a
all MSOs are using the same application. nationwide footprint at a fraction of today’s
This can be accomplished more easily if costs.
first, the application is broken up into
several components, second, it relies heavily It is conceivable to implement intelligent
on using library-type functions, and third features into this application delivery CDN,
templates are used for allowing the such as the ability to customize applications
application to adapt to the unique elements on the fly. Creating this abstraction layering
of a local environment without touching the capability can allow an application
software. The nature of how the software is developer to ingest a single application into
structured can be expanded from its current this CDN. An automated workflow engine
functionality to achieve these goals. will attach or bind the application to any
number of individual templates. The
As application developers innovate these results of such a process work for both the
new products, implementation and adoption MSO and the content provider. Each MSO
will be easier for those that implement can receive a custom version of the
proper design and technique. Applications application that is branded and behaves in a
developed with industry-approved Software way that maintains the common look and
Development Kits (SDKs) keep developers feel of their network. Multiple content
within the proper guardrails for the EBIF providers and advertisers can use the same
client or tru2way stacks deployed in the application and achieve similar
field. In addition, applications that functionality, but with a look and feel that
implement Templates will allow for quick responds to the content.
customization for new skins, new behaviors,
new look and feel, without modifying the Set-top Abstraction Layers
core code.
Supporting these new interactive
Once well-designed applications are applications on cable plants will require
created, they will need to be deployed leveraging new open architecture
quickly and efficiently. Understanding that technologies such as EBIF and tru2way to
MSOs may deploy a wide variety of support widely deployed applications across
carousels and infrastructure in their a wide variety of headend networks. EBIF
networks underscores the need and benefits and tru2way provide abstraction layers to
of using standard data formats such as both the network and the set-top box in
CODF (OpenCable Content Definition different ways. EBIF is a data format that
Format). An important goal is to avoid the allows an EBIF client on a set-top box to act
implementation of proprietary interfaces in as a presentation engine for EBIF
the application distribution, data gathering applications. Tru2way, conversely,
and reporting chain. A “publish and enables an entirely new platform and
distribute” architecture available to all infrastructure. Tru2way will also allow
content providers and application developers and advertisers to create widgets
developers, and subscribed to by multiple that would work with video-based
cable operators, can help to dramatically applications similar to the role of widgets
with Internet-based applications.

Due to the existing footprint of well over set-top box will be a simpler application
60 million EBIF-capable set-tops, EBIF is than the Internet-based applications they can
becoming the technology of choice for be utilized to initiate this cross-platform
advertisers and application developers experience and provide customers with the
looking to immediately deploy their type of global connectivity and interactivity
applications. In addition, cable MSOs are they are seeking.
also deploying infrastructure to support
tru2way applications. Maintaining MSO Identity - Application
Customization
Interactive Application Sources
As more and more applications become
Interactive applications will originate available, maintaining a common look and
from a variety of sources, including feel on the MSO’s network that is also
interactive content authors who are new to consistent with their navigation paradigm
the cable industry.. As previously outlined, will become an important part of success-
the open, global and interactive fully deploying applications. Unlike
characteristics of the Internet have allowed it applications on a PC, iTV applications are
to benefit from the development of new and presented as a service offered by the MSO.
innovative applications. By creating While application vendors strive to have
translation engines and other types of their own look and feel. A customer who
interfaces to these applications, the cable interacts with many applications from
industry can accelerate the deployment of multiple vendors may find a different style
complex, cross-platform applications on its each time they launch a different
video platform. application. This could be unsettling if an
application loads or navigates differently
Abstracting Cross-Platform from the other applications on the same
Communication cable plant.
Because EBIF applications have the Allowing each MSO to maintain

ability to communicate with application consistency with their navigation paradigm
servers on the Internet, those application involves building applications that can be
servers can achieve a cross-platform customized to “fit” an application to the
presence. An example of this would be an MSO’s network. However, application
EBIF application that communicates over customization will impede progress unless it
the Internet to a Facebook application is done in a way that reduces, or even
through a proxy on the cable headend eliminates, re-testing the application on each
network. Once the connection to the MSO’s network. Rapid deployment and
Facebook application is made, that wide-scale adoption cannot occur when
application can leverage the cross-platform there are multiple testing and certification
footprint of Facebook and carry the processes. Instead, the most successful
customer experience onto iPhones, applications are likely to be those that
Blackberrys and other mobile devices. “white-label” their applications, allowing
Leveraging applications such as Facebook in each MSO to “skin” the application to their
this manner allows an MSO to use the cross- standards. Options include allowing each
platform connectivity of Internet-based MSO to use its own colors, logos, and
application as an abstraction layer. While messaging. Expanded functionality would
the actual EBIF applications running on the include capabilities to drive customers to

other services like VOD, or launch other cable systems will be able to deploy
interactive applications. For example, an applications developed by multiple
MSO or content provider may deploy an application vendors across the same
application that is bound to a specific infrastructure.
television show (i.e. voting and polling, etc.)
In addition, the MSO or content provider SDKs
may choose to add a button on the
application that can launch VOD sessions of Widely used for online interactive
older episodes of the program. applications software development kits
(SDKs) are typically a set of development
News and weather applications will also tools that allows developers to create
require a level of customization. These apps applications for a particular software,
will require the ability to utilize localized middleware or hardware. The iPhone is a
feeds as a source of data.. Traffic reports, great example of how an SDK can be
local news and sports, and local weather leveraged to successfully build a variety of
may come from existing RSS feeds or other applications. Application developers need
data feeds. A single application must be to know little about the iPhone if they have
able to define all of these data sources, the proper SDK to build applications. The
skins, and behaviors without modifying the iPhone SDK keeps developers confined to
source code of the application. the capabilities and limitations of the device.
Standardized Infrastructure The cable industry currently has several

requirements choices when it comes to EBIF and tru2way
SDKs. As the technologies mature, the
A key success requirement for quickly choices of SDKs are likely to consolidate,
deploying applications on a cable plant is to thus providing a consistent and common
limit the amount of required testing while toolkit for developing interactive
ensuring that the application will execute applications. Common toolkits allow
safely and perform well. Accomplishing developers to leverage a common
this will require both the industry’s framework, and help to ensure that
continued embrace of open standards and applications will exhibit similar behaviors.
focusing on building a standardized The benefits to this approach are that the
architecture. resulting applications will tend to have a
common look and feel, they will behave
A variety of technologies may be used similarly, and they tend to be more
for developing this standardized predictable. This approach will also serve to
architecture. First, by standardizing on a shorten the test time.
few software development kits (SDKs),
application developers can be assured that Moreover, applications developed with an
they are developing their applications within EBIF or tru2way toolkits have many
a certified and tested framework. Second, advantages over applications developed by
the use of templates will allow for proprietary tools. SDKs provide an
advertisers and application developers to environment where a developer can simulate
quickly customize and deploy applications. the application’s behavior before loading it
Finally, and as previously described, by on a set-top box. This allows for a certain
standardizing the back-office infrastructure level of testing of an application and the
required to support interactive applications, middleware before the code is loaded on a

set-top. Developer toolkits also help to keep applications are examples of content-driven
software developers within the guardrails templates. An example of a content-driven
allowed by the EBIF client or the tru2way template is a voting and polling application.
stack. As a result, applications are A Voting/Polling template allows the
developed with “safe and tested” methods, content provider to modify the look of the
thereby making the code less intrusive than application and to also define the data that it
a native application. will present, such as a question on which
contestant to vote for or a set of trivia
Templates questions.
What if an organization wants to develop The second form of template is an

interactive ads or applications but it doesn’t application-driven template. Application-
have the luxury to hire its own software driven templates tend to be more complex
development staff? For example, an applications that were originally developed
advertiser may want to implement an as stand-alone applications, like electronic
interactive EBIF ad and distribute it into a program guides (EPGs) and fantasy football
national broadcast. If the advertiser doesn’t applications. Application-driven templates
have a software development staff, or will allow MSOs to customize the
doesn’t want to hire a third party to develop application to achieve a similar look and
a customized application, it could still feel to other key navigational functions and
implement the application by utilizing to provide a consistent experience to the
“templates”. consumer.
The concept of templating is used Implementing an application-driven

throughout the Internet advertising industry template will allow an application developer
today. Advertisers can purchase a template to develop a single application and
and quickly customize it for their advertising customize it for every customer quickly and
needs. Their main benefit is that the safely.
deployed ad is a fully tested application
without the advertiser having to invest in Standardizing back-office requirements
software development. For example,
advertisers can purchase Google AdWords SDKs and Templates simplify the work
templates to create Internet-based ads for on the client-side. However, the back-office
their business. The templates require little and infrastructure required to run these
software development experience and allow applications also needs to be standardized,
advertisers to instantly deploy ads that dis- as described earlier. CableLabs and the
play during internet searches or navigation cable industry are currently working on
to topical websites. ways to standardize back-office
requirements. This will be essential to
Interactive application templates will allowing cable video customers to
come in two forms: content-driven templates experience e-commerce and other
and application-driven templates. Content- transactional applications with the same ease
driven templates implement a single of use that they enjoy with online shopping
function, but they can be configured to or ordering an On Demand movie.
present entirely different sets of data.
Request for Information (RFI), VOD
Telescoping, and Voting/Polling

APIs CableLabs, and centralized services
providers such as HITS AxIS can provide
The complex testing of applications “sandboxes” to application developers to
occurs today because it is necessary. Most develop and test new applications on a
of the software deployed on current cable representative cable plant. This approach
plants is hardware-specific, so any changes can not only reduce the development time,
or variations can cause problems in the but also accelerate the test cycle, enabling
network. However, implementing new applications developers and MSOs to deploy
services with abstraction layers and new services much faster.
standards not only allows for portability and
broader support, but also limits the risk Providing these sandboxes also enables
when the new services are introduced or the cable industry to incubate new
changed. As the industry experienced with application developers and encourages
DOCSIS and PacketCable, open standards innovation on cable networks. By reducing
and abstraction layers work. Hardware and the cost of building an environment to
software vendors in the online ecosystem do innovate new applications, sandboxes also
not have to regression test every iteration of help to foster the next generation of
hardware and software on their platforms application developers. Organizations like
because applications are developed using OEDN, which encourage college students to
open or published (sometimes licensed) begin developing applications for the cable
APIs to the operating system. Similarly, industry as they do today for iPhone and the
web browser developers are not required to Internet, also provide the infrastructure for
test their browsers against all possible students to integrate their applications.
websites because the browser is acting as a Enabling these new application developers
presentation layer for a standardized format helps to grow the ecosystem that will
of data – not unlike EBIF. provide MSOs with more choices of
applications to offer their customers.
Sandboxes
On-boarding
In addition to supporting open standards,
the cable industry’s connected ecosystem Organizations that provide an operational
can benefit from faster methods of environment for application developers
integrating new services. One of the should be prepared to offer developers an
approaches being used to shorten the easy, accessible on-boarding process that
development cycle is to allow developers allows for rapid introduction of new
access to a live, MSO Grade, test cable plant applications into the market. Mini-mobile
that is representative of a real cable network. applications and the iPhone “App Store”
Providing application developers access to illustrate that the connected world is more
this test network, or “sandbox” during their about small, inexpensive and often
development process gives them the ability temporary applications rather than complex,
to learn, and identify issues that cannot be expensive ones. Accordingly, good on-
identified in a simulation environment. To boarding organizations help to breed
encourage innovation from a wide cross- innovation and competition by encouraging
section of developers, access to a sandbox a wide universe of application developers to
may be the only way a developer may be develop, test and deploy an even wider
able to afford to test their application. To range of applications to run on cable
accomplish this, organizations such as networks.

A good on-boarding process also helps in they can provide more information and
the building of a knowledge base of best transactional opportunities. The connected
practices, known infrastructure problems cable ecosystem will allow these interactive
and typical pitfalls that can be shared elements to occur over the video platform
amongst all developers. In addition, it offers itself.
application developers the privacy and
protection required for keeping their Personalizing Content Through
intellectual property safe. Changeable Elements
The connected ecosystem can support the

advertising and content communities’
A CONNECTED ECOSYSTEM interests for optimizing the value of their
FOR ADVERTISERS content by expanding the methods for
reaching potential customers who are the
By connecting MSOs and back offices, most interested in their ads or content.
local cable systems will be able to present Currently, one of the advantages of placing
advertisers and content providers with the ads by node is the ability to sell that
opportunity to communicate to their inventory to neighborhood advertisers. The
customers with the most powerful medium connected cable ecosystem would expand
(high definition, large screen, surround the market for these hyper-local ad avails by
sound television) while using the web- providing the capability for national
proven tactics of personalization and advertisers to reach specific neighborhoods
interactivity. However, some additional or other demographics anywhere in the
connections will be required in order for the country.
advertising and content community to take
full advantage of these advanced Supporting the growing demand for
opportunities. segmented advertising requires an expanded
approach towards managing, tracking, and
In addition to the connections discussed delivering content. It is relatively simple to
earlier, the capabilities of a more robust manage one of a few pieces of content that
platform for seamlessly delivering and are distributed nationwide, or to a few
managing advertising and other forms of regions of the country. It becomes
content will help to increase effectiveness. exponentially harder to manage content that
Moreover, these enhanced capabilities may have multiple versions, with different
would allow content placement interactive applications attached to them,
opportunities on the cable video platform to with different text overlays, linking to
be more competitive with other content different actions, and all designed to reach
platforms, largely by leveraging their different communities.
capabilities.
Combining demographic data concerning
The connected ecosystem described in a neighborhood or customer with the cable
this paper provides advertisers with a system’s ability to deliver content by node,
platform that combines television’s powerful opens up the possibility for personalized
ability to engage audiences with the content. However, delivering the message
interactive capabilities and reach offered by that will be the most relevant to that
the Internet. Today, advertisers use on-air neighborhood can involve managing as
campaigns to drive traffic to web sites where many as 20 different elements that

customize the message for maximum information requests content and
impact. An example would be a cruise ship instructions to the RFI system. All of this
ad that features older adults when it appears content and the associated metadata that
in neighborhoods with that demographic describes it need to be delivered to all of the
while replacing those scenes with images of devices, and verified to ensure the consumer
family activities in neighborhoods that skew experiences the desired result. Ads
to that demographic. Stitching those delivered via tape are physically incapable
elements together requires pairing the ad of containing the data required for
server with an ad-decision engine that will interactive capabilities. The open standards
ensure the ad containing the right elements architecture described earlier in this paper is
appears in front of the right viewer. intended to support interactive advertising
applications by providing the interface to
Delivering this type of content to the remote carousels or ad servers and the asset
hosting and serving devices will require the and scheduling metadata to insert an asset
ability to automate management of the into a digital video stream. Advertisers
bidirectional metadata. This is to ensure that interested in taking advantage of the
the multiple elements of the ad delivered to connected ecosystem’s iTV capabilities will
the serving environment are assembled the also require the interactive application
correct way for that particular node. The templates that were previously described.
ability to take advantage of advanced
capabilities such as these is result of both the Accelerating time to market
technology platforms being available, and
accessible at large-enough scale for the Another requirement of the advertising
content to be available to a large audience. community for delivering personally
The connected cable ecosystem plays the relevant content that can be addressed by a
important role of being able to mange the connected ecosystem is accelerated delivery.
delivery of the content and the metadata Rapid, timely delivery of all of the content
across a national footprint, but to also and its associated elements across the entire
aggregate engagement data across all footprint is an important element of the
markets and by each demo, while observing video system operating in a comparable
strict privacy guidelines. manner to the other (i.e. Web, mobile, etc.,)
platforms. Examples of ads that are heavily
Interactive advertising dependent upon time-to-market for their
effectiveness include political campaign ads
In order to fully support interactive where a candidate will want to create and
advertising – all of the content and its place an ad to respond to a new attack ad
associated elements should be managed and from his or her opponent. Another example
tracked across multiple delivery platforms. would be advertisements that encourage
As an illustrative example, assume an greater audience engagement by mixing
interactive ad may have links to telescope to images from the program with ad content,
a long-form video on VOD, and also to such as a carmaker’s ad that features “the
support a prospective consumer’s ability to play of the game.”
receive more information about a particular
product or service. To achieve the desired To achieve rapid delivery requirements,
result, the linear ad needs to be delivered to the ability to accelerate the collaboration and
the Ad insertion system, the long-form approval processes can also be improved.
content to the VOD server, and the Advertising content is typically created in a

collaborative process between one or more quality throughout the content distribution
production houses. Ad agencies approve process. Incorporating these design,
this content before it’s distributed to the end operational and measurement elements into
destinations. As the number of versions of the asset delivery system will improve the
the content, and associated elements grow, likelihood of adoption and probability of
the processes that track and support the success for advertisers and other enterprise
approval will need to take on the same level market stakeholders that use connected
of sophistication as the delivery processes. cable architecture to meet the needs of their
customers and grow their businesses.
Meeting the demand for accelerating
time-to-market is well beyond the As with the connections described in
capabilities of tape dubbing and overnight previous sections of this paper, meeting the
delivery systems. These ads will need to interest of the advertising and content
travel at the speed of light, not FedEx. A communities will also involve extending
network-based architecture appears to be the beyond the connected cable video network.
most appropriate means to support these In order to provide access to the features and
content delivery needs. Understanding that flexibility found on the web, connected
the participants in this ecosystem will be cable networks will also need to connect to
comprised of both large and small the systems and services that enable these
companies located around the globe web capabilities.
suggests that the content management and
delivery system will need to overcome such One example of the systems cable may
issues as packet loss, corrupted files, longer want to connect with are advertising
transport session times, lost productivity, decision engines. These web based systems
manual monitoring and the lack of real-time, that decide which ad to provide to a
item-detail reporting of file transfer activity. particular customer, at a particular time are
These asset delivery systems will need to becoming extremely advanced and valuable.
provide fast, fault-tolerant, secure, and Today, these systems primarily support
audited digital delivery that will represent a banner ads but they are expanding into video
core element in the supply chain for ads as well. These types of decision engines
addressable advertising. Delivering in real will be advantageous for expanding the
time leaves no time for mistakes. capabilities that cable’s new platform can
offer to advertisers and content providers.
Meeting the Advertiser’s Quality In addition, leveraging the mass and scope
Requirements of the data collected and analyzed by these
web-based systems will enhance the value of
Ensuring quality throughout the asset the cable video platform to companies that
delivery process is also increasingly require this information in order to optimize
important to the advertising community. the effectiveness of their content delivery
Their ads are built and tailored for activities.
customers who will be viewing them on
large, high-resolution display devices. As
such quality of experience becomes an
increased focal point. The ecosystem will SUMMARY
also need to provide content providers and
others stakeholders in the distribution chain With all of the recent advances in
with the ability to monitor and measure technology, the cable industry is quickly

moving past a wired world into a connected The decisions that all cable MSOs make
world. In fact, consumers aren’t just over the next few years will help to realize
connected occasionally, they are always the potential for cable’s video platform to
connected. They can get e-mail, connect to serve an increasingly larger and more
the internet and text or twitter anytime of dynamic environment . Realizing the
day or night, regardless of location. fullness of that potential will require the
connections that allow consumers to access
For cable to lead the development of the personally relevant services they want
advanced services from advertising to game more easily.
playing to social networking, it will want to
embrace the concepts of connectivity. By connecting to other wire providers
Historically, cable systems were being and cable systems, the industry could ensure
limited to a single medium that provided that its services can deliver the mass and
“last mile” connectivity for video scale required for success in today’s world.
consumers. However, a connected view of This entails connecting technologies in ways
the world can’t end there. Limiting the view that allow for localization and
to ways to connect to the consumer would personalization without requiring custom
squander the opportunity for being a major development and extended deployment
player in the products and services that cycles. It also involves connecting
consumers access through those pipes. advertising platforms in ways that allow
advertisers to enjoy the power and presence
Consumers of today and tomorrow have of the high definition, large screen, surround
an expectation that a single device or sound television with the interactive,
application is connected to the rest of their dynamic and personalized capabilities that
world. Evidence that this is necessary have evolved on (the far less impactful)
becomes obvious in situations like looking Internet platforms.
around the plane when the flight lands.
Almost everyone immediately reaches for The time is right for evolving to meet the
his or her cell phone. In many cases the user connected society of the future. Connections
doesn’t immediately make a voice call, but to both its existing base of friends and
rather spends the taxi time reconnecting to partners, as well as connections to other,
their world, checking e-mail, answering somewhat competitive, electronic industries
texts or checking news and scores. This is will be critical for success in a connected
the environment in which cable will play a world. The cable industry is in a unique
greater role by continuing to build upon its position to provide both consumer and back
commitment to an open, connected office services connections in ways that can
ecosystem. take advantage of a world where often, the
mass is the message.
i
How did eBay start?, A brief history of eBay, By Aron Hsiao, about.com

Content Management Systems versus Content Delivery Networks
Michael Adams
TANDBERG Television
Abstract advantages of today’s “push” delivery model,

namely that very popular titles do not
In today’s on-demand deployments, generate unmanageable spikes in network
content management and content delivery are utilization and that customers are never
intertwined into a store-and-forward, offered “hot” assets until they have been
multicast delivery system that relies on successfully provisioned at the video-server.
satellite transmission for secure asset
delivery. The mode of delivery is a “push” Nevertheless, there are excellent reasons
from the content provider to the on-demand to migrate to terrestrial networks for content
site, such that all assets must be provisioned distribution including the ability to turn-
in the video-server memory before they are around assets much faster when required; an
offered to the customer. This same system example would be a political campaign
ensures that titles are offered in a consistent advertisement. In addition, the explosion in
and flexible way according to the operator’s content choices, coupled with the increase in
marketing requirements. the number of high definition titles, makes the
“push” content delivery model impractical at
In contrast, cable’s high-speed data some point. Therefore, mechanisms to identify
services use content delivery networks (CDN) certain, less popular titles as “library
to deliver web assets, as a series of file- content” become essential, as does selectively
transfers over terrestrial networks, such as employing a “pull” content delivery model.
leased interconnects and backbone networks.
Moreover, the mode of delivery is a “pull” Finally, this paper will contrast the
from the subscriber, based on HTML requests requirements for a CMS with those for a
from the website. Websites are essentially CDN, and show how the two technologies can
unmanaged by the operator, except for be used together in certain circumstances to
caching frequently served pages to manage blend the advantages of both types of
backbone bandwidth. approach.
There is considerable interest among cable

operators in migration towards using a CDN
approach for on-demand content. However INTRODUCTION
the transition is not trivial because of cable’s
unique requirements related to scale, Video-on-demand systems have now been
management, and security for on-demand deployed by all major cable operators across
assets. their entire footprint and these systems work
extremely well, serving millions of customers
This paper explains in detail the on a daily basis.
requirements for a Content Management
System (CMS) to manage end-to-end delivery Cable on-demand services are based on a
of on-demand assets in a hybrid network QoS guaranteed, connection-oriented model
environment using satellite and terrestrial from the streaming service to the subscriber.
links. We show how a migration towards This model relies upon session resource
terrestrial content distribution is possible management which allocates a guaranteed
without giving up some of the fundamental slice of bandwidth between the server and the

set-top for the duration of an on-demand developed – this is essentially a collection of
session. This is possible because the path assets and metadata files that completely
from the streaming server to the set-top is describe the on-demand title. Each package is
over a relatively simple access network transformed into a single file using a UNIX
topology (typically via a Gigabit network to a utility called tar. These tar files are
QAM modulator). The only possibility of transmitted over satellite, using multicast to
blocking is in the HFC network itself, but this greatly reduce bandwidth requirements.
well-managed by allocating constant bit-rate Thousands of hours of content are distributed
sessions using standardized QAM resource to systems via satellite from multiple content
management. Cable operators have become providers. Although an hour of standard
adept at increasing the total bandwidth per definition content requires about 1.7 gigabytes
subscriber by reducing the size of the service to be transferred, the actual transmission rate
group as on-demand peak usage increases. from each content provider is relatively low,
in the order of 10 Mbps.
However, this means that video servers
have to be placed at the edge of the network Satellite distribution provides an extremely
and they need to be pre-provisioned with all efficient, “push” mechanism to get the
the assets that the customer could possibly required content to a very broadly scattered
want. This leads to some challenges to solve set of cable systems. This so-called “pitcher-
in deployed systems: catcher” approach uses a reliable multicast
algorithm to make efficient use of satellite
1) Scaling the number of assets from capacity to push assets to a large population
10,000 to over 100,000 to support long- of headend systems simultaneously. Also
tail content. included is robust encryption to prevent
2) Effectively managing a very large unauthorized access to the content files.
number of assets across multiple sites.
3) Ability to dynamically change metadata Unfortunately bundling the content files
independently of content propagation. with the metadata means that the metadata is
also multicast, leading to a one-size-fits-all
In this paper we will examine the role of a approach. This forces the operator to make
Content Management System (CMS) and a metadata changes at each system, after
Content Delivery Network (CDN) in solving content distribution rather than centrally. As a
these problems and show how they address result, asset management has become
different aspects of these challenges. complicated and time consuming because
each system is managed separately. Worse
ON-DEMAND CONTENT DISTRIBUTION still, there is a lack of transparency to
corporate marketing and central operations
Content distribution products were first groups. If delivery of title to a site is
developed to meet the specific needs created unsuccessful, a manual “re-pitch” is required,
by cable operators as they deployed on- which is both costly and inefficient.
demand systems on a headend-by-headend
basis. The title metadata includes fields on
pricing, categorization, and availability
In the earliest days, content files were window. A content provider distributes
shipped around on tape and metadata was packages with these values set according to its
manually entered into the VOD system at business rules. The operator may need to
each headend. In order to scale on-demand override some of them such as different
services, the concept of a package was pricing on a system-wide or system-by-system

basis. Some assets may need to be filtered The situation is similar for on-demand
because they are not required at the particular dynamic ad placement – in this case it is
system, or because there is insufficient storage especially important to verify that the ad
capacity for them. Because of the need to content is actually on the server before it is
change the metadata for each operator, and scheduled to be inserted into an on-demand
even down to the specific site, “Asset stream. The ad decision is usually made at
Management Systems” were introduced. session start-up time, and at this point a play
Initially asset management systems were list is presented to the server.
deployed on a site-by-site basis but they have
evolved to support multiple sites in a regional Since operators offer the majority of on-
environment. demand titles across their entire footprint, it
makes sense to manage them centrally. Since
CONTENT MANAGEMENT SYSTEMS content distribution and delivery is spread
geographically across a large number of
On-demand systems rely on thousands of systems, a distributed content management
hours of content distributed via satellite from solution is required with centralized control.
multiple content providers. With so many files Since local divisions and regions also want to
in so many locations, it becomes a complex offer local on-demand titles, the solution must
task to keep track of everything and today’s support local management of those titles.
manual systems are struggling to keep up.
There is often a significant cost in on-demand A central control panel to manage content
operations due to so-called “content metadata provides a long list of advantages
propagation errors”. These occur when part for the operator, including the potential to
of the content distribution chain fails during enable powerful and timely promotional
the provisioning process. In most cases, the campaigns along with pricing discounts, to
harm is done because there is no mechanism provide targeted advertising support, to adjust
for the operator to automatically check that all the viewing window, to remedy errors, or to
the necessary assets have been placed on all enrich metadata after the title has been
the appropriate servers. So the provisioning provisioned (for example to support extended
error is often discovered first by the customer, descriptions, advanced search, or
who requests a title and gets back an error recommendations).
code. This impacts the session success rate.
Worse, it discourages the customer from using Let’s explore some example scenarios that
the on-demand service and reduces operator are problematic for operators today and I’ll
revenues. explain how a content management solution
can help to make these scenarios much easier
In practice it is very difficult to make the to manage effectively. I’m going to focus on
content distribution chain completely “bullet- scenarios that are operationally intensive or
proof” and doing so would come at a impossible to implement today:
significant cost, so it is better to automatically
check the asset status on a regular basis. If a The corporate marketing and programming
particular asset is missing from a server, it can department believes that a title has truly
be automatically rescheduled for delivery. underperformed and decides that a
Since assets are typically propagated before promotional campaign along with a price
they are made available (before the viewing reduction is justified in the final week of
window), it is possible to repair any missing availability. With centralized control of the
assets before they have any customer metadata and the resulting ability to easily
impacting effect. lower the price for the title, the marketing and

programming department now has the ability maintain more intelligence, awareness and
to pull together a cross-channel advertising flexibility in this critical linkage between
campaign in partnership with the content assets and metadata.
provider. Further, they can use their content
processing solution to embed ETV triggers in For example, for certain distribution needs,
that advertising, to direct viewers directly to the delivery of metadata might be unbundled
the “buy” screen. from the delivery of the associated heavy
asset. This is the case for many operators that
A set of titles with a common element or do not have the backbone capacity to handle
theme can be listed in a special category – for distribution of their content asset. By
example, all the movies featuring “Ben separating the delivery of these two assets, the
Kingsley” or all the “Bond” movies. By cross- operator can deliver very large content assets
linking the metadata of titles in this way, this to local systems over a very cost effective
also creates a foundation for automatically satellite network, and yet still deliver
generated recommendations for new titles. metadata via their intranet. Thus metadata and
rules for the management of the metadata can
An unpopular title can be removed from be organized centrally and delivered to the
the system before the end of its availability local systems whenever content is distributed
window to free up storage space for new or updated.
titles. As new titles are introduced, a set of
rules is used to distribute them to each market Finally, operators will need to be able to
according to pre-defined priorities so that the provision content so that it can be played on
on-demand storage in each market is any device. This provides the operator with a
optimized for maximum revenue generation. competitive response to “over-the-top” video
providers by providing a seamless extension
A set of titles can be ingested in the of on-demand cable services to additional
traditional on-demand infrastructure, and then devices within the home; for example, to TVs
automatically processed according to a set of with broadband connection, to PCs, or to an
rights and rules housed in a seamlessly ultra low cost IP set-top box. To make multi-
connected solution to the appropriate content platform services manageable and scalable,
compression and metadata formats for the content management solution must enable
broadband and mobile video service the operator to provision a single title with
platforms. Depending on the rights and rules, multiple playout options across multiple
in some cases, only the trailers are sent to devices.
these additional platforms, and in other cases,
the movies can be purchased and viewed on CONTENT DELIVERY NETWORKS
them.
Content Delivery Networks have grown up
In order to do all of the above effectively, a to support web-based delivery of information,
fundamental change is required in the way including arbitrary content formats for
that the content management solution graphics, music and video. Wikipedia defines
structures and handles the relationship a content delivery network as “a system of
between a so-called “heavy” asset and its computers and storage networked together
associated metadata. In short, the processing, across the Internet that co-operate
managing and delivering of metadata needs to transparently to deliver content most often for
be separate from, but not disconnected from, the purpose of improving performance,
the associated assets, and vice versa. The scalability and cost efficiency to end users”.
content management solution needs to

Essentially a CDN is a cache-based system metadata that ties certain files together. The
where files are moved around the network reason for this is that the hyper-linked web
based on the pattern of usage. Considerable pages themselves provide the framework and
ingenuity has been applied to the algorithms the navigation for the website. Essentially the
and protocols to do this. Peer-to-peer (P2P) “metadata” equivalent is embedded in the
protocols are one example of an approach to HTML markup of the web pages. Web
building a distributed, collaborative network technology improvements including dynamic
of content stores. HTML, applets, and servlets make web-sites
much more dynamic and flexible than before
A typical CDN implementation is to and allow support for localization and
populate files into an origin server (which personalization, but this principle remains
itself may be a distributed entity). This server unchanged.
is connected to many cache servers by
network links. The cache servers directly Although the CDN automatically responds
serve the clients, which are web-browsers. to client requests to make ensure that the most
When the user navigates to a web-page, the recently used files are held in the cache
embedded markup language (HTML) causes servers to satisfy overall demand, this may not
the browser to request all the necessary files provide the best indication of asset popularity.
for that page using HTTP requests. In the first This is because many cached files are related
fetch, the cache server must retrieve them to navigation requests versus asset requests.
from the origin server and pass them along to Moreover, there is no intelligence in the CDN
the client, keeping a copy in the cache. On to pre-position files in anticipation of a
subsequent fetches, the local cached copy can demand for them which may cause
be used to save bandwidth on the network link unpredictable spikes in bandwidth as new,
back to the origin server and to save processor popular assets are discovered by the clients.
bandwidth on the origin server. [1]
The CDN architecture allows for multiple CMS VS CDN

tiers of cache servers for better scalability and
allows the edge cache server to be pushed Table 1 summarizes the functions
closer to the client, allowing for faster supported by CMS and CDN technologies.
response times and higher bandwidth across a
given link. However, because each client Function CMS CDN
request is independent of every other, each Metadata management Y N
fetch is a unicast session and considerable Transcoding control Y N
effort has been applied to cache management Workflow Y N
algorithms, load-balancing, and so on to make Content Distribution N Y
best use of each fetch operation. Multicast Distribution N N
Cache Management N Y
Although this is a simplified overview of Load Balancing N Y
how a CDN works, it captures the important
principles of operation. It is fair to state that Therefore, we conclude that CMS and
CDN technology has been developed for the CDN functions are complementary to each
general case of managing a large number of other; a CMS allows for functions related to
small files with little regard to what function the semantics of the content while a CDN
those files actually represent at the system provides the mechanism to distribute the
level. As such, the model lacks a higher level various files across the network of servers.
abstraction such as the asset grouping and its

The next step in this analysis is to discuss Therefore the best solution is to use the CMS
how the two technologies can best be to implement distribution rules based on
combined in a practical way. popularity indicators embedded in the
metadata in order to select which delivery
CMS AND CDN mechanism to use on an asset by asset basis.
In this section we will discuss a potential CONCLUSIONS

migration strategy to add CMS and CDN
technologies to an existing on-demand In this paper we have reviewed the
deployment. As we have already discussed, capabilities of CMS and CDN technologies in
CMS can provide a powerful management on-demand video deployments and conclude
layer that will work with the existing satellite that:
content delivery model to provide operational
advantages and additional control and 1) They are complementary technologies
monitoring capabilities. and there are advantages to using the
two in conjunction when the number
Therefore, the first logical step is to add of assets is greater than 10,000.
CMS to the existing on-demand deployment 2) CMS technology is a useful addition to
to reap the benefits of this technology. existing on-demand deployments in
conjunction with todays “push”
As demand for long-tail content increases, content distribution systems.
the system may reach a limit on the number of 3) CDN technology may be added as an
assets that can be “pushed” to the on-demand additional distribution technology for
systems in the available time. Ultimately this long-tail assets, essentially “pulling”
is limited by the cost of satellite bandwidth these assets from a central content
available to content providers and the cost of library only when they are required on
storage in each on-demand system. Titles the first customer play.
towards the end of the long-tail distribution 4) Extensions to support multi-platform
will be so rarely viewed that it does not make delivery of assets will require a CMS
economic sense to distribute them and store platform to manage the additional
many copies of them locally. complexities of content transcoding
and multiple content files formats per
At this point, CDN technology provides a asset.
complementary addition to the existing
“push” content distribution model. It should
be noted that the latest generation of “pitcher- [1] THE BENEFITS AND CHALLENGES
catcher” based solutions will operate across a OF DEPLOYING LARGE REGIONAL
mix of satellite and terrestrial facilities. VOD ASSET LIBRARIES by Michael W.
Nevertheless, for popular titles that are Pasquinilli, Sunil Nakrani, Jaya
expected to generate significantly more than Devabhaktuni, 2008 NCTA Technical
one play on average per system per month, the Papers™
“push” model is still superior from a cost
perspective because of its use of a multicast
distribution protocol.

DEVELOPING A GRADING SYSTEM FOR DIGITAL VIDEO QUALITY
Dave Higgins
Chuck Wester
Comcast Media Center
Abstract acquisition of new customers as well as

dramatically influence whether a customer is
This paper describes an “End to End” retained. In fact, HDTV can be seen as a
measurement and analysis philosophy for clearly defined competitive battleground
transport of digital video signals from a based almost entirely on the proposition that
centralized distribution point to the the recovered video “Quality” must exceed a
consumer. A key component of the effort has consumer’s current service. According to
been to develop specific and repeatable recent industry studies, quality issues are
metrics that describe both the integrity of responsible for 40% or more of all customer
the signal delivery, as well as its relative churn, second only to cost. i In addition,
subjective quality. industry analysts have observed that cable
system operators risk losing the entire RGU
Given that there are a number of (revenue generating unit) when bundled
approaches and technologies used to deliver customers move to an alternative
linear video content to households, this multichannel video programming provider
paper examines the use of multiple transport (MVPD). ii
options commonly deployed in the
marketplace today and how metrics can be With so much at stake, cable system
created to measure the efficacy of the end operators must actively participate in the
product as delivered to the consumer. video content delivery chain which includes
Specifically, the delivery plant considered is the creation, management and distribution of
based on MPEG 2 transport to the HD content. This paper identifies as many as
consumer, however the model can easily be eight critical touch-points in the content
conformed to support an MPEG 4 approach. distribution process that can directly impact
video quality performance and affect the
Central to the measurement efforts customers’ perception of quality and
employed is the adoption of common satisfaction with their video services
terminology to describe known digital video provider. In addition, it is based on the end-
impairments. Further, the ability to describe user HD experience to provide a more
these impairments in a manner that can be accurate depiction of the impact on quality
easily understood, communicated, and perception from multiple factors including
taught is a key outcome of the effort. advanced compression techniques, the
nature of source content acquisition, and
EXECUTIVE SUMMARY satellite vs. fiber transport delivery.
As multichannel video providers achieve While most cable operators focus their
parity in the quantity and subscriptions fees quality monitoring efforts from the output of
for popular television channels and their headend to the consumer set top,
programming, the quality of service (QoS) achieving optimal and competitive quality
associated with the delivery platform has assurance requires establishing and applying
become a key differentiator that can a consistent system for measuring the
significantly impact the successful quality of digital video and audio from the

source to the display device. To achieve this critical and needing to be maintained. iii
objective, the use of a repeatable grading Cable operators indicated that service
system that summarizes both the quality issues are one of the main reasons
characteristics typically noted by expert or for customer support calls, resulting in a
“Golden Eye” viewers as well as significant reason for customer churn – as
impairments the average consumer objects much as 40%. With millions of customers
to is key. The challenges associated with across the country, end-user video quality
creating such a program include the lack of monitoring is an integral part of a cable
industry accepted standards and practices or operator's business,” according to the report.
inconsistent application of complex
technical concepts in an operational Extrapolating from the report’s findings,
environment. the level of churn attributed to QoS
represents millions of digital video
Implementing an end-to-end system for customers, hundreds of millions in annual
quality assessment requires tools such as revenue, and billions in asset value. Given
probes for measuring the quality and the importance associated with customer
reliability of HD content at each touch- retention, several system operators are
point. In addition to monitoring the probes launching significant efforts to focus on
and applying the grading system to QoS. MRG’s analysis found that some cable
anomalies that are detected and measured, MSOs are spending as much as $2 – 5 per
data will need to be exchanged between subscriber per year for QoS.
television programming networks, cable
system operators, and others involved in the “With competitive pressures increasing,
origination and distribution of HD content in cable operators need a comprehensive video
order to address and correct reported monitoring solution to ensure they meet
shortcomings. customer expectations, or face possible
increases in churn and operational costs,"
This paper also includes a summary of a the study concluded.
grading system that is being used today and
examples of the ways that the system is Developing an accepted and customer-
helping to minimize impairments and driven quality grading system will enable
improve the HD customer’s experience. It cable system operators to address
will conclude with recommendations for misperceptions that hinder the use of
steps that the industry can take in order to advanced compression techniques, such as
facilitate the implementation of a quality HD3:1 using MPEG-2. If left to applying
grading system, such as greater automation only the simple math behind 3:1, or the
of the quality assessment process, additional greater compression offered by MPEG-4,
independent research, and consumer potential HD customers would assume that
education. the quality of HD signals cannot be as good
as an HD signal using compression levels of
BENEFITS 2:1 or lower. However, consumer research
conducted by the CMC (Comcast Media
A recent MRG research report found that Center) and others has demonstrated that by
90 percent of cable operators consider video using best case practices for digital video
quality monitoring was either “crucial” or encoding, stat-muxing, and transport, HD3:1
“very important”, and 58 percent viewed is highly competitive with MPEG-4 via DBS
end-user quality of experience (QoE) as or HD2:1.

This research has helped to demonstrate degradation, which occurs as a video service
that HD3:1 can allow cable operators to is created or acquired, processed or re-
launch more HD content using less encoded, delivered through various
bandwidth and without sacrificing the networks, groomers, commercial ad
customer QoE. It also underscores that there insertion equipment, and finally to the
is a tremendous opportunity - and need - for customer’s set tops, is an emerging area of
consumer education concerning all of the opportunity to improve.
factors that affect their viewing experience.
Prior to digital video, the quality of video
INTRODUCTION delivery systems was measured using
vertical interval test signals (VITS), Vertical
Existing industry standards describe Interval Reference Signals (VIRS), and
video quality comparison of the video performance metrics outlined in documents
measured at any point along the delivery such as ANSI/EIA/TIA-250-C. Operations
path to a source reference; a referenced engineers could use these test signals at
video quality measurement which numerous points in the distribution chain
determines how much a system or process without impact to the service to perform
has degraded a given video service. measurement of the quality of the delivery
However, they do not address a significant of analog video. Today, transmission
issue which is the true quality of the video operations engineers are more likely to use
as measured at all critical touch points in the MPEG analyzers to determine the quality of
delivery path as well as at the consumer the delivery path. Typically these analyzers
display. Regardless of the Herculean efforts are used to measure known impairments as
to improve the video source and delivery part of ETSI TR 101-290 compliance such
systems, the quality of the video delivered to as continuity counter errors, missing PIDs
the cable subscriber can be no better than the and Jitter performance. While this approach
worse case video quality at any touch point. is effective for measuring transport stream
errors such as packet loss, MPEG analyzers
Source providers, equipment do not provide any ability to measure
manufacturers, and MSOs have made customer perceived video quality
tremendous improvements to their video particularly as it relates to subjective
transport and delivery systems. The components (i.e. sharpness, noise, macro-
challenge is to find further significant video blocking, etc.)
quality improvements to the end video
product as delivered to the cable consumer. Further, the consumer marketplace has
Given that technical advances have allowed overwhelmingly adopted larger display
for greater density of channels in a multiplex screens that provide enhanced noise
while achieving approximate parity with reduction, filtering, improved contrast ratios,
earlier efforts has resulted in a tangible and materially greater resolution which
increase in plant capacity. The balance result in the ability to see enhanced details in
associated with “Quality vs Quantity” all video content. Unfortunately this ability
continues to be a complex and challenging also facilitates more critical viewing of
equation to solve. That said, the digital impairments or “artifacts” inherent to
improvements associated with upgrading digital video compression and transmission.
one area of the Content Distribution Chain The obvious result is that these impairments
can be marginalized by other impact points whether they are associated with HDTV or
and in some cases made worse. Cumulative SDTV, are becoming much more evident to

customers and there is considerable mis- providing commentary. The transport path,
information present in the marketplace. One whether satellite or fiber, is very likely being
example is channels branded as “HD” which compressed, particularly if the feed is native
show a majority of their content as standard HD. This initial compression causes digital
definition 480i video “upconverted” to HD. impairments that will never completely be
Another example is larger consumer removed from the “video” and as you will
displays which are often broadly lumped see, can become materially worse as they
together and referred to as “HD Displays” travel thru the distribution chain to the
while the viewing characteristics between consumer.
the different types can be dramatically
different, particularly under varying lighting The studio camera is likely the very best
and “off-axis” viewing conditions. source of video quality we will consider
here as the conditions are controlled, the
The confusion in the consumer lighting is optimal, and there is very little
marketplace is understandable, given the opportunity to induce impairments beyond
complexity of the topic in general, as well as those introduced by the camera optics.
the technical appreciation required to Typically, material is mixed between the
acknowledge all of the possible reasons Venue, the camera, and server playback of
behind a given video quality impairment. pre-recorded media in Master Control.
Differences in the video quality displayed by Given the cost associated with media
various brands and types of monitors and storage, particularly HD native content,
increasingly rapid video monitor response Video Servers will ingest content utilizing
rates add to the confusion. While video compression and induce another set of
engineers evaluate video quality using impairments. Again, the deeper the
defined parameters such as viewing angles compression the more likely video
and distance from the monitors, there are impairments will result.
differences in ambient conditions of monitor
locations consumers likely do not T he next step in the process is associated
appreciate. with transporting the video signal from
Master Control to the Outbound
VIDEO QUALITY TOUCH-POINTS Transmission system. In many cases this is
where a significant impact can occur in the
The examples of video quality overall delivered video quality product. All
measurement provided in this paper are major programming providers utilize
based upon analysis that is being conducted compression, whether it is MPEG 2 or
by the Quality Assurance team for the CMC, MPEG 4 based, to maximize the use of
which is currently placing nearly 2Gbps of satellite or terrestrial fiber bandwidth.
MPEG2 video onto the an IP network for Traditional approaches for MPEG 2
delivery to cable markets. Drawing 1 is a statistically multiplexed HD bandwidth
simplified view showing a typical service usage here are in the range of 12 to 15 Mbps
path of an HD signal broadcast out to cable per channel, but there are exceptions in both
subscribers. directions. Whatever the approach, the
outcome can be seen at the baseband video
As an example, assume that there is a output of the satellite receiver or
sporting event being produced at the Venue receiver/transcoder and in all cases there are
which needs to be transported back to the video impairments.
studio where an on-air announcer is

The CMC’s Quality Assurance Team is signals into a single multiplex while
also monitoring content that is distributed sustaining competitive video quality.
over Direct Fiber. The connections, which
will serve all of the major programming As Drawing 1 shows, the output of the
providers, will operate at a higher bandwidth Imagine Encoding system is then
usage model to minimize the impairments transmitted onto the Fiber Backbone for
associated with the Outbound Transmission. distribution.
Once the signal arrives at the CMC, it is While not specifically addressed in this
either converted to MPEG 2 if required or paper, downstream impairments can
presented to the Imagine Encoder as a obviously be caused by local re-encoding,
Gigabit interface. The efficiency associated rate-shaping or grooming as well as local Ad
with removing the need to take the receiver Insertion systems. Further, the set-top box,
output down to its baseband video the consumer display device, and the
components and then re-encoding has viewing environment all play a contributory
allowed the opportunity to integrate 3 HD role in recovered video quality.
Drawing 1

The working assumption is that if the MPEG transport stream probes placed
video source is of superior quality, the re- strategically across its network to monitor
packaging encoders are performing the availability of each of the video services
optimally, and the network is functioning as transported. It is important to note that only
it should, then the video from the source service availability (i.e. Packet Loss as an
should be delivered to the consumer with example) is monitored through automated
little degradation and the availability of the means. CMC has been using the MPEG
service should not be an issue. However, it probe system since January, 2008 to
is critical that as a service provider, the measure its delivery performance, optimize
CMC tracks not only the video quality but the network, and to accumulate fault data for
the performance of the delivery platforms in statistical and alerting purposes. Table 1
order to determine the quality of each shows an example of several weeks of
service, as best as possible. performance data collected on a per program
basis facilitating analysis and review; note
To measure the uptime performance of the anomaly that occurred at Site 11, Red,
its HD video delivery over the fiber yellow and green color coding indicates the
network, CMC has installed a system of severity of the impact.
)
ut)
Programs by
(in
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
1
2
3
4
5
6
7
8
9
C (o
Imparied
rce
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Site
Sou
CM
Seconds
Channel A 2 16 11 10 11 11 11 14 11 18 11 11 168 11 11 11 12 11 20 11 11 10 14 11 11 14 16 11 11 11 11
Channel B 1 1 0 0 0 0 0 0 0 6 0 0 160 0 0 0 0 0 4 3 2 2 2 1 1 2 7 1 0 2 0
Channel C 9 2 1 1 1 1 1 1 1 5 1 1 159 1 1 1 1 1 5 2 1 1 2 1 1 2 6 1 1 1 1
Channel D 0 12 11 10 11 11 11 11 11 15 8 11 168 11 7 10 10 10 15 8 10 10 12 7 11 12 16 11 11 11 10
Channel E 1 7 2 2 2 2 2 2 2 6 2 2 160 2 2 2 3 2 6 2 2 2 2 3 3 6 5 3 2 2 2
Channel F 9 19 9 9 9 9 10 9 9 13 9 9 167 9 8 9 9 8 14 8 9 8 9 14 14 17 12 15 8 10 9
Table 1
SUBJECTIVE ANALYSIS measurements. There are considerable

differences noted between content that is
Because the probe system only measures aired (i.e. SD infomercials on the Overnight
the transport stream delivery quality data, block, legacy or older material that is
CMC developed and implemented a Golden upconverted during Day-part, and first run
Eye program to ascertain a subjective video Native HD material in Primetime.) It is
quality rating of the sources of services it important to note that these channels are
delivers. This data is accumulated over time marketed as “High Definition”, “HD” or
and used to determine average video quality “High Def” and the consumer expectation is
of each service placed on the backbone. not a variable.
This program is a subjective video The Golden Eye observer assigns a level
quality assessment method devised by the of quality to the channel under test based on
CMC using trained observers to perform their observations of any impairment during
subjective quality measurements of the the test cycle. A sample of the Golden Eye
source video, which is then processed and observer subjective test results is shown in
placed onto an IP network. Each service is Table 2 and is based on the test criteria
viewed for 10 minutes on a predetermined which are listed in Table 3.
schedule, which ensures random quality

At the heart of video service quality Understanding there will always be some
grading is the use of common industry level of each of these impairments in every
terms, definitions, and tolerances to known digital video service, and that some are more
digital video impairments such as irritating to the cable subscribers than
contouring, haloing, macroblocking, noise, others, weighting of each impairment
smearing, and pumping. While these terms measurement type is required. After each
are commonly used for describing artifacts impairment type is graded, the impairments
that impair the customer’s QoE, it is very that are more irritating to the consumer must
important that Golden Eye analysis apply be assigned a greater weighting. The total of
them in a consistent manner. Table 3 was the six impairment types combine with these
developed in order to assure a common weightings to determine an overall score.
language to describe the nature of the digital This provides a view of program picture
artifact presence in the video being quality and the probable offending
analyzed. impairment types contributing to lower
quality scores.
Date Time Channel Program CONTOUR HALOI MACRO NOISE SMEARPUMP Channel Entry Total Channel Average
12/15/2008 12:15 A Program 1 2 2 2 2 3 3 44 44
12/19/2008 23:40:00 A Program 2 4 4 4 4 5 4 82 82
12/24/2008 04:32:00 A Program 3 3 4 5 4 4 5 86 86
12/15/2008 12:20 B Program 1 2 2 3 2 3 2 48 48
12/19/2008 23:50:00 B Program 2 3 4 3 3 4 3 65 65
12/24/2008 04:39:00 B Program 3 3 3 4 4 3 3 71 71
12/15/2008 12:25 C Program 1 2 2 2 2 4 4 48 48
12/19/2008 23:56:00 C Program 2 2 3 3 3 3 3 58 58
12/24/2008 04:45:00 C Program 3 3 4 3 3 2 4 63 63
12/16/2008 11:20:00 D Program 1 2 3 2 5 5 2 64 64
12/19/2008 24:05:00 D Program 2 5 4 5 4 5 4 90 90
12/24/2008 05:00:00 D Program 3 3 4 2 3 3 4 59 59
12/16/2008 11:25:00 E Program 1 2 2 4 4 2 4 66 66
12/24/2008 05:52:00 E Program 2 5 3 4 4 5 4 81 81
12/16/2008 11:30:00 F Program 1 2 2 2 4 3 4 56 56
12/24/2008 06:01:00 F Program 2 4 3 4 3 2 4 68 68
Table 2

Compression Artifact Definitions Compression Artifact Ratings
CONTOURING 1 2 3 4 5
Contouring is a defect where abrupt changes between shades of Present in all Present in most Present, but not noticeable, but not no contouring.
the same color create color bands instead of a gradual change. scenes. scenes. distracting to consistently seen.
This can be seen during scenes with a large amount of smooth viewing.
color or in scenes where this is a color contrast change like in wide-
angle sunsets, dawns, or clear blue skies.
HALOING 1 2 3 4 5
Haloing is typically seen around areas of high contrast, such as Heavy blocking Blocking along the Localized blocking Blocking occurring no haloing.
sharp lines, text edges, and graphics. On close inspection, part of that occurs on the edges of most noticeable along at the edges of
the graphic appears to extend into the background. edges of all objects. the edges of logos only.
objects on the objects; always
Halling can manifest as smaller details in graphics appearing to screen. present at the
soften or loose edge resolution resulting in apparent blocking edges of all text
along graphic edges. and graphics.
MACROBLOCKING 1 2 3 4 5
Macroblocking is a defect where the edges of blocks or rows of The whole screen 10 - 15 seconds of 5 - 10 seconds of 3 - 5 seconds of no blocks.
blocks, are typically seen as a grid-like-pattern. This defect often Blocks up, blocking covering blocking covering blocking around a
occurs during dissolves from one scene to another or during action regardless of half of the of at least half the small portion of
scenes involving a great deal of complex movement. scene content. screen. screen. screen.
Another way this artifact is presented is with small to large pixels Note: This is a e.g. Consistent e. g. Blocks occur e.g. Scene
and/or blocks containing corrupted or green pictures. This is very rare event. blocking occurs at the focus of the transitions /
caused by transmission or transport anomalies. during every scene screen; often dissolves / short
transition, during motion. action scenes
fade/dissolves, (explosions).
action scenes, etc.
NOISE 1 2 3 4 5
Noise appears as random speckles on an otherwise smooth Black speckle Scenes presenting Scenes with black Black dots that No noise.
surface and can significantly degrade video quality. clusters that a raining effect of speckles that randomly pop into
manifest clearly black speckles appear to be any portion of the
Although noise often detracts from an image, it is sometimes defined block that cluster and moving on static scene.
desirable since it can add a grainy look that is reminiscent of film. edges. move into the backgrounds.
Noise can also increase the apparent sharpness of an image. foreground.
Closely resembles e.g. walls,
traditional e.g. Noise that curtains, or sky
Macroblocking, manifests itself that appear to
though not and is a distraction have movement in
associated with within the scene. the background.
motion.
SMEARING 1 2 3 4 5
Smearing is a defect where part of the image remains fixed in Regular loss of Losing object Intermittent loss of Individual or no smearing.
space while the adjacent parts of the image moves leaving a trail. object detail that detail on faces or object detail on unrelated
Smearing may also be observed in faces or across large areas of manifests as materials that are faces or materials. occurrences of
a similar type that have fine detail (e.g. grass fields). localized blocking the focus of object detail loss
on a face or a scene.
Smearing commonly affects facial color tones causing video to material.
take on an unnatural look.
PUMPING 1 2 3 4 5
Pumping is a defect where the video or parts of the video appear All scenes with Noticeable Scenes that have Low motion No pumping.
to pulse at a regular interval. static backgrounds movement on minor regular scenes that have
begin to block and static backgrounds movement on intermittent but not
This is typically seen in areas of smooth neutral colors. bleed to the during both motion static constant
foreground. and still scenes. backgrounds. movement on
e.g. Noticeable on static
an overhead view backgrounds.
of a golf course or
a wall of wood
paneling.
Table 3
SOURCE PROCESSING QUALITY perceives various video characteristics is

built into the ICE-Q® measurement system,
In addition to the Probe data and the such as sensitivity to analog and digital
Subjective Video Quality Analysis, the noise, spatial and temporal frequency, and
CMC also relies on Imagine’s ICE-Q® factors such as luminance, color, texture and
technique to measure its HD delivery edges. Imagine’s research has found that the
performance. How the human visual system accuracy of an objective measurement

system is primarily dependent on how frequently than with JND, but the quality is
closely its results correlate with subjective still excellent. Furthermore, the system can
test results from a pool of expert and be designed such that “97” is the target
ordinary viewers. In addition, since different average grade over time, while “95” is the
video coding standards exhibit different target minimum quality. It is also important
types of artifacts, a good objective to design the grading system such that the
measurement system should be tuned and numerical increments are reasonably linear
optimized for a particular video coding with respect to subjective video results. In
environment, e.g., MPEG-2 or MPEG-4 other words, the subjective quality
AVC. difference between “97” and “95” should be
similar to the subjective gap between “95”
The ICE-Q system processes every and “93.”
macroblock of every frame, using variable
bit rate coding to achieve and maintain The CMC has created a summary report
constant video quality. The system selects structure that allows the organization to
the optimal macroblocks and frames by monitor the incoming video quality and
using the objective video quality availability, its delivery performance, and
measurement system to preserve the highest the performance of the fiber network
video quality at the lowest possible bit rate. delivery. The level of quality is calculated
The ICE Broadcast System continually from the CMC Golden Eye subjective video
measures its ability to process the input quality measurements, the video processor
signal as accurately as possible (i.e. with quality grading, and the delivery
minimal added impairments). performance of the backbone as measured
by the MPEG2 probe system. Each of these
The actual grading system relies on a metrics is prioritized to determine actions on
convenient approach using a numerical scale channels of impaired delivery quality.
of 1 to 100, with 100 representing the Channel uptime and events of packet loss
compressed source quality. For example, a are most critical followed by the Imagine
score of “97” may be defined as Just quality scoring, then the CMC Golden Eye
Noticeable Difference (JND), in which subjective scores which have the lowest
expert viewers can rarely discern the priority. The results displayed in Table 4 are
difference between the compressed source examples of actual video services placed on
and the re-processed signal. A score of “95” the fiber network at the CMC.
may be defined as the point of No Material
Degradation (NMD), in which differences
from the source can be perceived more
Dec 1 - 8 Dec 8 - 15 Dec 15 - 22 Dec 22 - 29 Dec 29 - Jan 5

Source Source Source Source Source
Average CMC Average CMC Average CMC Average CMC Average CMC
Programs Quality Quality Quality Quality Quality
Grade Delivery Grade Delivery Grade Delivery Grade Delivery Grade Delivery
(OMC) (OMC) (OMC) (OMC) (OMC)
Channel A 83 98.16 5 83 98.10 80 69 98.05 35 71 97.90 16 71 97.83 8
Channel B 72 97.51 4 72 97.59 14 65 97.56 6 65 97.61 13 70 97.60 9
Channel C 81 97.90 4 80 97.98 4 77 97.95 14 67 97.99 5 67 97.99 16
Channel D 74 95.28 5 73 95.12 3 70 95.28 3 76 95.27 2 76 94.92 1
Channel E 92 95.12 3 92 95.17 80 63 94.95 8 63 94.87 4 70 95.01 7
Channel F 79 96.14 3 86 96.19 7 75 95.99 9 73 95.91 4 73 96.05 7
Table 4

QUALITY TOUCH POINTS under test to the source video providing a
measurement of difference between the two;
To maximize video quality across the a level of degradation. A non-referenced
delivery environment there are eight key video quality measurement is made without
“Touch Points” that begin at source a comparison to the source video, which is
acquisition and end with the consumer not available at most touch points, and is the
display. Video quality measurements at each more telling measurement of the two.
touch point should be made using a non- Drawing 2 shows the video delivery path
referenced approach though a referenced with the Touch Points indicating critical
approach is acceptable where a reference measurement points and opportunities to
source is available. A referenced improve video quality.
measurement is a comparison of the video
Drawing 2
Touch Point 1 is a measurement point at the Touch Point 2 at the Master Control facility
source, such as a live event venue or a requires non-referenced measurements
dubbing house. This is a critical point because there are many video sources
requiring non-referenced video quality required to create a video service. Service
measurements. Video artifacts are caused by degradation causes include varying levels of
analog to digital conversion and tape to video quality of programs and commercials
digital storage conversion requiring and failures of automation resulting in loss
compression of the video. of video or the wrong video played out. This

is also the point where the CMC will be Touch Point 7 is a measurement point
placing mezzanine encoders for direct fiber within the set top box in the customer’s
delivery. home. This requires non-referenced
measurements, since there is no reference
Touch Point 3 can be a referenced source available. Video degradations can be
measurement assuming the service provider caused by HFC delivery and conditions
is also the owner of the MPEG encoders and within the customer’s home.
multiplexors and that a reference,
uncompressed source is available. Service Touch Point 8 is a non-referenced
challenges include poor video encoding and measurement point within consumer video
multiplexor bit starvation causing the six displays measuring the performance on the
video impairments described in Table 1. very end of the delivery path. Clearly, this is
the point where all upstream events and
Touch Point 4 is the point of entry into the conditions affect the perceived video quality
CMC facilities and requires a non- as well as transmission availability as
referenced measurement. The service is left presented to the cable customer.
intact as MPEG encapsulated onto IP. The
affects to video are typically due to packet CONCLUSION
loss on terrestrial delivery systems and
packet loss on satellite delivery systems that With the increasing importance of
are due to RF conditions. However, because Quality of Service as a key marketplace
this is the first point at which CMC has advantage, the ability to measure, report on,
access to the video service, effectively and improve video quality delivered to the
making it the point of demarcation, the consumer is an essential tool for improving
video service must be validated. competitive strategy. Replicating both ideal
and imperfect viewing conditions within the
Touch Point 5 is interface to and from the software algorithms of Picture Quality
CMC IP delivery system and requires Analysis tools is critical to automating non-
referenced measurements assuming referenced video quality measurements.
reference from the delivery source is
available. This is the first point in the In addition, accumulated video quality
service path where the affects to the video scores provide MSOs reliable and repeatable
service caused by processing can be metrics. This data can be used for engaging
measured. Degradations of the video service content providers (and all parties involved in
at this point are often caused by over content management and distribution) in a
processing and bit starvation of the collaborative effort to achieve the best
multiplexor. possible viewing experience for the
customer. Additionally, industry accepted
Touch Point 6 is a measurement point from video quality measurements and grading
a monitor tap of the QAM devices. This enables MSOs to perform regular proof of
point requires non-referenced measurements performance maintenance of cable systems.
because there is no reference source
available. Video degradation is caused by In the long term, digital video quality
local conditions of delivery, grooming, and grading must take place in all devices in the
commercial add insertion. path, from source to consumer, which can
impact the quality of video. Further,
instantaneous scoring is necessary for best

determining devices and systems in the • Display viewing distance and “off-axis”
video service path that are contributing to viewing can impact perceived video
the degradation of video quality and quality.
accumulated scores will enable trending and
isolation of issues that lead to quality • Display size, type, native resolution
impacts and impairments. capability, contrast ratio, and pixel size
and shape can have impacts on
It is the authors’ hope that this paper recovered video.
helps to stimulate further discussion and
movement toward a common system and • Deeper compression approaches such as
automated processes for measuring and HD 3:1 prevent the use of downstream
scoring digital video quality from an “Rate Shaping” or “Grooming” as the
unreferenced source based on industry impacts to the recovered QoE are
accepted tolerances. dramatic.
Summary of Video Quality Measurement Acknowledgements

“Lessons Learned”
Video Layer Quality of Service:
• Linear video is subject to varying video Unprecedented Control and the Best Video
quality from one channel to another as Quality at any Given Bit Rate, NCTA
well as from one program to another. It Technical Papers, May 18, 2008, The Cable
is also evident that video quality varies Show ’08, Ron Gutman and Marc Tayer,
from one source provider to the next due Imagine Communications, Inc.
to source acquisition and delivery
approaches.
i
March 31, 2008, “Cable Operator Video
• Time of day has a direct impact on type Quality Study,” Multimedia Research
of content being aired and whether it is Group, Inc. (MRG, Inc.),
ii
high quality native HD or legacy SD January 1, 2008, “HD Monitoring -
upconverted material. Raising the Quality Bar,” Craig Kuhl,
Communications Technology magazine
• MPEG 4 transcodes to MPEG 2 cause iii
IBID, MRG’s “Cable Operator Video
new “mini” macro-block impairments. Quality Study”
• Older film product and marginal quality

associated with film to video transfers
also have a significant impact on QoE.
• Live sports that is over compressed

(whether it occurs on the path from the
Venue, the Outbound Transmission or as
part of a downstream re-encode process)
can and will cause significant and visible
impairments.

DVB-C2 – THE SECOND GENERATION TRANSMISSION TECHNOLOGY
FOR BROADBAND CABLE
Philipp Hasse, Dirk Jaeger, Joerg Robert

Institut fuer Nachrichtentechnik at Technische Universitaet Braunschweig
Abstract broadcasting services via cable networks. The

technology was also incorporated in the
Competition in the media and communication European version of DOCSIS [2]. Since its
sector is increasing. MSOs have to align their development more than 10 years ago, DVB-C
business models with new emerging require- has been commercially implemented in
ments on a permanent basis. Although millions of products such as (Edge)QAMs, set
requirements vary in different national tops, and cable modems all over the world.
markets in Europe and worldwide, the cost
per transmitted bit is an important factor for During the recent 10 years, research in
staying competitive. Thus one of the key communications techniques has progressed.
requirements for modern communication State of the art algorithms have improved
technologies defined by European MSOs is to significantly in terms of performance and
increase the spectral efficiency of downstream flexibility and Moore’s law has continued to
transmissions by moving the efficiency as happen. These trends have facilitated the
close as possible to the Shannon limit – the commercialization of advanced systems for
theoretical optimum. wireless and wire-line communication.
The second generation of the DVB system for Forced by the increasing competition in the
cable – called DVB-C2 – is an innovative media and communications sector, European
approach making use of state of the art MSOs considered utilizing the benefits pro-
communication technologies which have vided by such modern communication tech-
never been implemented in broadband cable nologies in broadband cable networks. In
networks before. An OFDM (Orthogonal 2006, the DVB [3] Project was approached to
Frequency Division Multiplex) based launch a work item aiming at the development
modulation scheme in combination with a of a second generation DVB system for cable
two-dimensional interleaving (in both called DVB-C2.
frequency and time) and an LDPC (Low
Parity Density Code) error protection
mechanism provide a spectral efficiency
fractions of a dB below the Shannon limit. Figure 1: Logos of DVB and ReDeSign
Will there ever be any reason to develop a
new PHY for cable after DVB-C2 is widely This article describes the DVB-internal
implemented? development process of the DVB-C2 specifi-
cation. It focuses on the work carried out by
the DVB Technical Module and its TM-C2
INTRODUCTION technical experts group launched to prepare
the technical specification. The DVB work is
The Digital Video Broadcasting (DVB) trans- supported by the research project ReDeSign.
mission system for cable (DVB-C) [1] defines The Institut fuer Nachrichtentechnik (IfN) of
physical layer and low-layer signaling Technische Universitaet Braunschweig has
techniques for the transmission of digital taken key positions in both organizations,

DVB and ReDeSign. Furthermore IfN staff derived during the work of IEEE 802.14 [6].
has been in charge of DVB work on dedicated The non-linear behavior of digital signals,
DVB-C2 techniques such as signaling and however, was researched in ReDeSign.
synchronization and has made important Software models were developed incorpo-
contributions incorporated in the final rating the results of this work. As explained
specification. later, intensive investigations were necessary
to generate the knowledge required for the
creation of software tools modeling the
DVB AND REDESIGN statistical behavior of second and third order
intermodulation products generated by
The development of DVB-C2 has involved multiple QAM signals.
some 20 companies including MSOs, IC
manufacturers and equipment vendors. The
chairmanship is provided by Kabel Deutsch- COMMERCIAL REQUIREMENTS
land, one of the major European MSOs.
Based on a liaison agreement, the work is The DVB Commercial Requirements [7]
supported by the European research project constitute the fundament for the development
ReDeSign [4] which has allocated consider- of the technical solution. The major require-
able resources to the work. One of the tasks of ments are summarized as follows. DVB-C2
the ReDeSign team is for instance to support needs to provide:
the standardization work by means of • A toolkit solution for optimal imple-
computer simulations. mentation under all conceivable
transmission conditions of various
An initial Technology Study Mission cable networks
executed among DVB member companies by • Increased spectral efficiency by at least
the Technical Module confirmed the potential 30 % compared with the highest
improvements achievable with latest com- efficiency provided by DVB-C
munication technologies in cable. Prior to the • Advantageous operational flexibility
technical work, use cases were studied and • Support of removal of fixed cable
business requirements were defined by DVB’s channel rasters
Commercial Module. Major requirements are • A low latency mode
summarized further down in the article. The • Adoption of state of the art
technical work was kicked off by means of a technologies
Call for Technologies (CfT). The responses
• An integral solution of the DVB
created the basis of the subsequent develop-
Family of Standards approach: reuse
ment of the DVB-C2 specification.
of elements of existing DVB standards
(e.g. of DVB-S2 and DVB-T2),
An important tool for the development of
wherever appropriate
DVB-C2 has been the software simulation
• Support for possible incorporation in
platform. The simulations focused on two
DOCSIS
goals: First to compile the final DVB-C2
• Cost efficient production of cable
system from individual elements of the
devices such as set-tops but also of
responses to the CfT and secondly to execute
performance simulations of the system. A trans-modulators used in headends of
cable channel model was adopted for this SMATVs
purpose [5]. The model was developed by the
ReDeSign consortium. To a large extent, the It is worth mentioning that the issue of
channel model takes account of the results backwards compatibility with DVB-C was

sacrificed to the requirement of optimizing the constellations which eventually will be
spectral efficiency and thus the throughput of assigned to the individual sub-carriers of the
the new technology. 26 requirements and 5 OFDM symbol. A number of QAM constel-
exemplary use cases were defined in total. lations are defined ranging from 16-QAM up
to 4096-QAM. For each QAM constellation a
dedicated set of LDPC code rates is selected.
THE DVB-C2 TECHNOLOGY A combination of the modulation and coding
parameters is called ModCod. The ModCods
The DVB-C2 specification (still in final draft defined for DVB-C2 are listed in table 1.
stage) defines the signal processing at the Adding a PLP header completes the creation
transmitting end. The high-level block dia- of the PLP container.
gram is given in figure 2. Basically the signal
processing comprises an input processing 16- 64- 256- 1k- 4k-
block, the multiple stages FEC unit and the QAM QAM QAM QAM QAM
OFDM generator.
9/10 X X X X X
Physical Layer Pipe (PLP) 5/6 X X X X X
The DVB-C2 system is intended to provide a 3/4 X X X X

transparent data link for all kinds of digital 2/3 X
information. Data formats such as MPEG
Table 1: ModCods of DVB-C2
Transport Stream (TS) [8] and DVB Generic
Stream Encapsulation (GSE) [9] specially
Data Slice
designed to support the transport of IP data
can be interfaced to the system. For enabling
Several PLPs may be combined to form a data
the transparent transmission, a PLP concept
slice whereas the protection power of the FEC
was defined. A PLP takes the role of a basic
assigned to each single container can be
container utilized for the transport of the data.
optimized individually by using an appro-
A PLP container can for instance comprise a
priate ModCod referred to above. The Data
multiple program Transport Stream, a single
Slicer has the task to move the PLP container
program, a single application, or any kind of
and group of containers, respectively, to a
IP based data. The data being inserted in a
defined location within the bit stream in such
PLP container are converted into the DVB-C2
a way that they will be transmitted by
internal frame structure by the Data Input
dedicated sub-carriers of the OFDM symbol
Processing Unit and subsequently processed
and thus to appear at dedicated frequency sub-
by the FEC (Forward Error Correction)
bands of the spectrum occupied by the OFDM
encoder. The FEC is composed of a Bose-
symbol.
Chaudhuri-Hocquenghem (BCH) outer Code
[10] and a Low-Density-Parity-Check
A two-dimensional interleaving (in time and
(LDPC) inner code [11]. The LDPC is a very
frequency domain) is applied to each indivi-
powerful tool for correcting transmission
dual data slice in order to enable the receiver
errors whereas the BCH’s task is to eliminate
to eliminate impacts of burst impairments as
the error floor which may be produced by the
well as of frequency selective interference
LDPC decoder in the receiver under certain
such as single frequency ingress or possibly
transmission conditions. The BCH code rates
even by fading.
add less than 1 % redundancy. The LDPC
code rates available range from 2/3 to 9/10.
The FEC encoded data are mapped onto QAM

OFDM frame builder Also the equalization process is easy for an
OFDM system since a simple multiplication
The OFDM frame builder combines several of each sub-carrier with an almost time-
data slices together with auxiliary information invariant factor can be applied rather than a
and additional pilot sub-carriers. The pilot complex convolution process needed in case
structure chosen is composed of continuous of a single carrier solution. The number of
and scattered pilots. The continuous pilots are scattered pilots is aligned with the length of
allocated at dedicated sub-carrier locations in the guard interval. 96 and 48 parts of all sub-
each OFDM symbol, which are constant and carriers are used as scattered pilots in case the
do not vary between the symbols. The relative guard interval length equals to 1/128
amplitudes of the pilots are boosted by a and 1/64, respectively, of the symbol duration.
factor of 7/3 for improved robustness. A
DVB-C2 receiver makes use of the 30 conti- The auxiliary data referred to above contain
nuous pilots allocated per 6 MHz bandwidth mainly the so-called L1 (Layer 1) signaling
to perform fine synchronization of the signal information which is put in front of each
in frequency and time direction. By an OFDM frame in terms of a preamble. The
averaging calculation of the continuous pilots, preamble uses the complete set of sub-carriers
the common phase error caused by phase of an OFDM symbol and provides the
noise (induced e.g. in the RF frontend of the receiver with means required to access the
receiver) can be detected and compensated. PLPs. For instance, information about the
After the synchronization is completed at the start positions of the data slices within the
receiving end, the scattered pilots are used for OFDM frame is transmitted as well as the
channel equalization purposes. The inverse ModCod information of the PLP headers.
channel impulse response can easily be Due to their importance, the PLP headers
determined from the differing values between themselves are highly protected by means of
the received scattered pilots and their equiva- robust modulation (e.g. QPSK or 16-QAM).
lent values transmitted known to the receiver.
Figure 2: High-level block diagram of the DVB-C2 encoder

MAC layer. In contrast, DVB-C2 is capable
OFDM generator of combining various adjacent channels to a
single wide-band channel at physical layer.
The OFDM symbols are generated by mean of For combining 4 channels, the 4K-IFFT unit
an inverse Fast Fourier Transformation of the OFDM generator at the transmitting
(IFFT). A 4K-IFFT algorithm is applied end needs to be replaced by a 16K-IFFT
generating a total of 4096 sub-carriers, 3409 algorithm. In case 8 channels are to be
of which are actively used for the trans- combined, a 32K-IFFT units needs to be
mission of data and pilots within a frequency installed in the headend equipment. It is
band of 6 MHz. These 3409 active sub- noted that higher order IFFTs can be utilized
carriers occupy some 5.7 MHz. The sub- for backwards compatible transmissions in
carrier spacing calculates to 1,672 Hz which 6 MHz channels, for instance, requiring a 4K-
corresponds with an OFDM symbol duration IFFT algorithm. The feature to combine
of 598 μs. For markets employing 8 MHz channels at physical layer has some advan-
channels such as the one in Europe, the entire tages compared to the conventional bonding
system can easily be scaled up by a factor of approach at MAC layer. For example, the
4/3 which results in a sub-carrier spacing of method allows making use of the frequency
2,232 Hz and a symbol duration of 448 μs. bands located at the edges of the individual
The OFDM signal injected in an 8 MHz 6 MHz cable channels traditionally occupied
channel occupies a bandwidth of 7.6 MHz. by filter slopes of the DVB-C and DOCSIS
The guard interval used between the OFDM signals (see figure 3). This feature further
symbols has a relative length of either 1/128 enhances the efficiency of DVB-C2 since, for
or 1/64 compared to the symbol length itself instance, a combination of 4 cable channels
and thus durations of 4.7 μs in the US results in an increase of the total bit rate
respective 3.5 μs in Europe. which is higher than this factor 4.
Support of 6 MHz receiver bandwidth

DVB-C2 FEATURES
Another example of a benefit is the possibility
The OFDM based concept implemented in to map individual data slices to dedicated
DVB-C2 provides a couple of benefits in OFDM sub-carriers and thus to dedicated
terms of flexibility as well as of efficiency frequency sub-bands of the OFDM spectrum.
advantages against single carrier solutions. In addition, DVB-C2 ensures that the L1
Some of these benefits are introduced in the signaling has a periodic structure with a
following sections. repetition period equal to 6 MHz (in the US –
8 MHz in Europe). As the wanted data slice
Channel assembly at physical layer has an equal or smaller bandwidth, it is
possible for a receiver devices operating at a
The DVB-C2 system supports the feature of bandwidth of 6 MHz to receive a 6 MHz
flexible and dynamic bandwidth allocation to frequency window of the DVB-C2 spectrum
individual data slices as well as to the entire although the entire bandwidth of the DVB-C2
signal. It is for instance feasible to combine signal is 18 MHz, for instance. In fact, the
various adjacent channels at physical layer in receiver decodes only those sub-carriers
a very effective manner. Note that the term transmitted within this 6 MHz window, as
channel bonding was deliberately not used in illustrated by figure 3.
this context in order to prevent confusion with
the DOCSIS 3.0 feature of channel bonding at

Figure 3: Illustration of features of DVB-C2s OFDM based transmission techniques
single SNR value above which perfect

Prevention of narrow-band interference reception is achieved whereas below that
value no reception is possible (see related
Figure 3 also shows a further useful feature of marks for DVB-C and DVB-C2 in figure 4).
DVB-C2. The signal power of subcarriers can The spectral efficiency adds information
be reduced to selectively adapt the power about the available capacity in terms of bit-
density distribution of the signal for preven- rate per Hz bandwidth.
tion purposes of co-channel interference. This
feature can be helpful in case a safety radio Simulation parameters
service transmitted at a co-channel frequency
needs to be protected by regulation. While The spectral efficiency figures presented in
occupying the remaining part of the channel, this article were obtained by means of
DVB-C2 reduces only the power of those sub- computer simulations carried out at Institut
carriers transmitted at the critical frequencies. fuer Nachrichtentechnik at Technische
Universitaet Braunschweig for the research
project ReDeSign. The results represent a
PERFORMANCE FIGURES quasi error-free DVB-C2 transmission in a
Gaussian noise environment.
This section provides information about the
performance of DVB-C2. In many technical For the simulation, 4 channels, each of a
articles, the bit-error rate (BER) as a function bandwidth of 6 MHz, are combined to form a
of signal-to-noise ratio (SNR) is used to wide-band channel of 24 MHz. A guard
demonstrate the performance of a transmis- interval length of 1/128 of the symbol dura-
sion system. This article chooses the spectral tion was chosen as well as the corresponding
efficiency as an appropriate performance indi- pilot density of 1/96. Each of the 2 spectral
cator. Digital transmission systems equipped guard bands at the wide-band channel edges
with powerful error correction mechanisms has a width of 200 kHz.
produce BER curves with very steep slopes.
In practice this means that we can refer to a

Figure 4: Spectral efficiency as a function of SNR for DVB-C2 and DVB-C
For determining the general performance of DVB-C2, the latter using a code rate of 5/6. It
DVB-C2, an additive white Gaussian noise is observed that the robustness of DVB-C2
(AWGN) channel model was implemented. increases by some 7 dB compared to DVB-C.
This rather simple channel model allows a The SNR required to receive the corres-
better comparison with equivalent perfor- ponding DVB-C signal equals approximately
mance figures of DVB-C and DOCSIS which 24 dB whereas the simulation of the
both represent digital cable transmission equivalent DVB-C2 signal resulted in a
systems in operation today. In addition, the required SNR of some 17 dB.
results can be better compared with the
Shannon limit [12], [13] defining the Increased spectral efficiency
theoretical channel capacity limit for the
transmission of digital information through Comparing the spectral efficiencies of both
channels impaired by Gaussian noise. A more technologies at an SNR of 24 dB( e.g. 64-
sophisticated cable channel model was QAM for DVB-C and 1k-QAM for DVB-C2),
implemented in the further simulations it is observed that the efficiency increases
explained later. from 4.8 bps per Hz to 7 bps per Hz which is
an increase of almost 50 %.
Increased robustness
First estimations of efficiency gains based on
The diagram in figure 4 clearly shows the an introduction of DVB-C2 in combination
advancement of DVB-C2 compared to DVB- with further technologies such as statistical
C (and DOCSIS, respectively) in terms of multiplexing and advance audiovisual coding
both spectral efficiency and reduction of SNR (e.g. MPEG-4) resulted in a total efficiency
required for reception of a signal with a increase of more than 100 %. This figure
dedicated spectral efficiency. In the follow- corresponds with a bandwidths saving
ing, the 64-QAM mode is used by means of achievable in cable networks of more than
an example. This modulation scheme 50 %.
provides a spectral efficiency of almost 4.8
bps per Hz when applied in both DVB-C and

FURTHER SIMULATIONS are not completed yet. First preliminary
conclusions show that intermodulation
Further system simulations were carried out products of digital signals have an impulsive
using a more sophisticated channel model. noise-like behavior with a varying statistical
The model was developed within ReDeSign. occurrence probability of their amplitudes
It compiles information generated in earlier depending on the network load and on the
research work (e.g. by IEEE 802.14 [7], signal levels respectively. Related investi-
DigiSMATV [14]) as well as by ReDeSign gations performed mainly by the Dutch
internal studies and measurement campaigns. research institution TNO (member of the
Different parameter sets of the channel model ReDeSign consortium) are ongoing. [5] It is
were defined to address the varying planned to submit the results to CENELEC
transmission conditions of real networks as [14] and IEC [15].
well as to reflect short-term and longer-term
network scenarios. For the short-term
scenario, a mixed transmission of analogue SUMMARY
and digital signals was considered whereas
the longer-term scenarios reflected the Launched by a request from European cable
situation after analogue switch-off. operators, the DVB Project has developed a
second generation downstream transmission
The complex channel model includes typical technique for cable networks called DVB-C2.
impairments which a signal suffers when The specification defines a toolbox providing
travelling through a cable network such as a set of parameters utilizable to adapt the
phase noise, echoes, impulse noise, etc. system to varying transmission conditions of
While the statistical or deterministic nature the networks in Europe and worldwide. The
and quantity of several impairments is well spectral efficiency compared to DVB-C and
known to the cable industry, one of the most DOCSIS has been increased considerably and
challenging investigations for completing the allows transporting more than 60 Mbps
channel model was the study of effects through a regular cable channel of 6 MHz.
created by interference which are generated Many advantages are supported by means of
by non-linear transfer functions of active the flexible channel bandwidth agility of the
cable components. In particular, the system. During its development the DVB TM-
contribution from digital signals in both a C2 experts have taken account of the require-
mixed analogue-digital and a digital-only ment for a low latency mode allowing to
environment was analyzed. Measurement incorporate the DVB-C2 physical layer
campaigns were organized at premises of technology in the existing DOCSIS system.
equipment manufacturers Kathrein and DVB-C2 therefore has the potential to
VECTOR (the latter also being a member of become the physical layer and lower layer
the ReDeSign consortium). At the labs of the signalling technology for the future
two companies, a cable network emulator was generations of cable based downstream
set up compiling a chain of broadband transmission systems.
amplifiers and an optical link as well as in-
home network components. The network
emulator was fed with 94 DVB-C signals. ACKNOWLEDGEMENT
Alternatively, a mixed signal scenario with
adequate signal numbers was established. The authors appreciate the European Commis-
Intermodulation products were captured into a sion’s funding and support given to the
large memory for statistical investigations ReDeSign project. We also thank the member
performed afterwards. These investigations companies of the ReDeSign consortium,

namely Alcatel-Lucent, ANGA, BLANKOM [6] Kolze, T.: HFC Channel Model
Digital, Telenet, TNO, VECTOR, ZON TV Submission, IEEE Project 802.14, Cable
Cabo who all have contributed to the findings TV Protocol working Group, 26 May
presented in this article. Without the impor- 1998
tant technical contributions to the develop- [7] Advanced Coding and Modulation
ment of the DVB-C2 specification made by Schemes for CATV systems (DVB-C2) –
the members of the DVB Technical Module Use Cases and Commercial
and its experts group TM-C2 represented by Requirements. DVB internal document
its chairman Christoph Schaaf of Kabel no CM-C20018R4, January 2008
Deutschland it would have not been possible [8] Information technology – Generic coding
to write this article. We thank all of them for of moving pictures and associated audio
the devotion to the DVB-C2 development. In information: Systems. International
particular we thank the representatives of Standard ISO/IEC 13818-1, December
Sony for the excellent cooperation in creating 2002
the DVB-C2 encoder and decoder modules of [9] Digital Video Broadcasting (DVB);
the software platform. Also the opportunity Generic Stream Encapsulation (GSE)
provided by Kathrein to analyze the effects of Protocol. ETSI TS 102 606 V1.1.1,
intermodulation products generated by digital October 2007
cable signals at the labs at Kathrein premises [10] Lin, S.; Castello, D. J. Jr.: Error Control
is appreciated by the authors. Evenually, the Coding. 2nd Ed. Pearson Prentice Hall,
cooperation with CENELEC made possible New Jersey, 2004
through a liaison with ReDeSign is greatly [11] MacKay, D. J. C.: Good Error-Correcting
acknowledged. Codes Based on Very Sparse Matrices.
IEEE Transactions on Information
Theory, Vol. 45, No. 2, March 1999
REFERENCES [12] Shannon, C. E.: A mathematical theory of
communication, Bell System Technical
[1] Reimers, U.: DVB – The Family of Journal, vol. 27, pp. 379-423 and 623-
international standards for Digital Video 656, July and October, 1948
Broadcasting, 2nd Ed. Springer, Berlin [13] Shannon, C. E.: Communication in the
Heidelberg New York, 2005 Presence of Noise. Proceedings of the
[2] Data-Over-Cable Service Interface IEEE, VOL. 86, NO. 2, February 1998
Specifications – Downstream RF (reprint)
Interface Specification. CableLabs, [14] http://www.cenelec.eu/Cenelec/Eur
September 2008 opean Committee for Electrotechnical
[3] Digital Video Broadcasting project, Standardization, CENELEC,
DVB, www.dvb.org (status 28.01.2009) www.cenelec.eu
[4] ReDeSign project, www.ict-redesign.eu (status 28.01.2009)
(status 28.01.2008) [15] International Electrotechnical
[5] HFC Channel Model. ReDeSign Commission, IEC, www.iec.ch
Deliverable 08, December 2008, , (status 28.01.2009)
www.ict-redesign.eu (status 28.01.2009)

Getting to 4G via Wi-Fi in the Real World
David Park
Belair Networks
Abstract system, going to the extent of putting fiber

optic connections to individual properties, in
As the economy shrinks and the order to compete with MSOs. The next
competition for profitable revenue generating frontier in bundling is the addition of wireless
subscribers increases, operators increasingly services to the package.
use bundled packages of services to capture
and retain customers. One key tool for that is
the inclusion of wireless services. This paper Mobile wireless systems offer particular
describes how wireless services can be challenges to MSOs, notably;
effectively deployed on an MSO’s existing
HFC infrastructure. It describes both licensed • Spectrum
and unlicensed technologies and how they • Mobility
compare. It also describes how mobility can
be deployed on HFC and the difference
between macrocellular and microcellular In this paper I describe the difference
systems and their performance. between macrocellular and microcellular
techniques for providing wireless coverage
and capacity. I also describe how the HFC
INTRODUCTION can be leveraged as a mounting location for
wireless basestations and some of the
technologies that are required to deliver this.
Traditionally, cable operators have looked In conclusion I show that MSOs can deploy
on wireless options associated with licensed Wi-Fi today and achieve capacities and user
and unlicensed spectrum as either/or experiences that are better than all of today’s
propositions. But, with the emergence of 3G cellular standards and the future 4G
fully mobile cable-optimized basestations, standards.
Wi-Fi can be used as a complement or an
alternative to next-gen mobile standards that
rely on licensed spectrum. Multi-radio Figure 1 shows that this approach is reality
platforms can operate in both licensed and today, with numerous MSO deployments of
unlicensed bands and connect directly to the HFC based microcells in progress.
HFC (Hybrid Fiber Coax) network.
One of the key business strategies for

acquiring and retaining customers of the past
few years has been the use of bundled service
packages. In this MSOs have been uniquely
positioned as the HFC offers a very cost
effective platform to deliver TV, data and
voice services to a customer premise. In
contrast traditional Telcos has been forced to Figure 1 Wi-Fi Microcell being installed on HFC
put in place extensive upgrades to their

MACROCELLS, MICROCELLS & Microcells
CAPACITY
Macrocells The alternative approach has been to

deploy microcells. These smaller cells offer
the following characteristics;
The traditional method of deploying
wireless services used by cellular carriers has • Many times the capacity of macrocells
been the use of roof top macrocells. These • Go around obstructions by terrain and
types of cells were quite suitable for buildings
deploying services that were primarily voice No cold/dead spots
orientated. A macrocell deployment has the • Signal levels are higher due to small cell
following characteristics. sizes
Higher signal strengths = higher
throughputs
• Prone to obstructions by terrain and • Frequency re-use simple
buildings Reuse channels frequently
Cold spots in coverage common • Pluses
Building shadowing causes dead Capacity and user experience
zones • Challenges
• Signal levels are constrained at distance Need more cells
by diffraction losses over buildings and more mobility events
into streets More challenging to provide in-
Lower signal strengths = lower building coverage for tall
throughputs buildings (3D)
• Frequency and spectrum re-use limits Backhaul
capacity at central point
Can’t keep adding channels
• Pluses
Provides overlay coverage
Large cells mean fewer mobility
events
Provides in-building coverage for
tall buildings (3D)
• Challenges
Capacity and user experience
Figure 3 Microcell approach
Microcells have been used infrequently by

cellular operators to date because of the
challenge of getting wired backhaul to them.
Figure 2 Macrocell approach The high cost of deploying a fractional T1
connection into the streets has resulted in
microcells being used for either cold spot
(lack of coverage) of hot spot (capacity) fill-in
rather than ubiquitous coverage. This has been

the case for voice systems, however, as will What this means to a cellular provider is
be demonstrated, 4G systems need microcells that only a small percentage of the cell is
to fulfill their capacity and throughput claims. covered by the highest order modulations.
Fortunately the MSO industry has the Given that cell site locations are driven by
mounting assets required. real estate concerns, and the customers are
individuals who are mobile, it is reasonable to
assume that end users are uniformly
4G systems need microcells distributed throughout the cell. This means
that the average throughput of a cell which
One of the key differentiators of 3rd and 4th also represents its capacity is degraded due to
generation wireless systems as compared to the distribution of modulation rates.
the traditional cellular systems has been the
use of higher order modulation schemes to
deliver enhanced capacities and throughputs. This has a profound impact on the
MSOs are no strangers to this, with the deliverable performance from a cellular
DOCSIS® system using high order QAM to network. While most commentators describe
deliver bandwidth to their customers. wireless systems in terms of the peak
However, the HFC is delivered to the end throughput that they can deliver, or sometimes
customer via coaxial cable, with amplifiers even the raw modulation rate, it is more
that keep the signal to noise high enough for instructive and useful to look at 2 aspects of
all modems that are connected to the system the system in more detail.
to operate at the best modulations. A wireless
system in contrast, propagates its signal over • The peak user performance that can be
the air, and the signal that a user gets is highly expected
dependent on their distance from the cell site. • The average capacity across the covered
This is especially true once higher order area.
modulations are used. This is illustrated in
Figure 4. To determine the average capacity of a cell we
need to estimate the percentage of the cell that
can operate at each modulation An example
calculation is shown in Figure 5 for an LTE
[1] macrocell using a 5MHz channel in a tri-
sectored cell. This assumes the availability of
2 blocks of 15MHz of licensed spectrum.
Figure 4 Modulation versus Distance

Modulation QPSK 16QAM 64QAM
Base Bits per Hz 2 4 6 EVDO, capacities in the order of 4Mbps per
Code rate 0.67 0.83 0.83 sector are achieved. Results can also be
Derated bits per Hz 1.33 3.33 5
Bit Rate 6.67 16.67 25 calculated for WiMAX and for use of larger
Thoughput 5.33 13.33 20
SNR 5 15 25 amounts of spectrum. For example a blended
Uplink Sensitivity -97.99 -87.99 -77.99
Downlink Sensitivity -95.99 -85.99 -75.99 WiMAX macrocell with 10MHz of spectrum
BTS TX pwr
BTS ant gain
38
17
38
17
38
17
yields 13.27 Mbps of capacity (note this is
MS ant gain -2 -2 -2 effectively less than the case shown for LTE
Uplink link budget
Downlink link budget 149.0 139.0 129.0 as it is using twice as much spectrum). It
Fade margin 9 9 9
overall DL link budget 140.0 130.0 120.0 should also be noted that these calculations
are for the downlinks only, as that tends to
Distance (km) 0.65 0.35 0.20
drive the user experience of an end user.
Distance (mile) 0.40 0.22 0.12
Cell Area (km^2) 1.33 1.33 0.38 0.13
Ring Area (km^2) 0.94 0.26 0.13
% 71.01% 19.53% 9.47%
Blended Data Rate (Mbps) 8.28 The propagation model used in this case is
the Empirical COST-Walfisch-Ikegami Model
[2] which is representative of an “above the
LTE Rx BW 5 MHz
subcarriers 15 kHz
rooftop” macrocell type deployment.
subcarriers 301 The loss versus distance and parameters used
uplink resource blocks 25.083
for this model is shown in Figure 6.
occupied BW 4.515 MHz
LTE BTS Noise Figure 4 dB
LTE UE Noise Figure 6 dB
LTE BTS Rx Noise Floor -103.0 dBm
160
LTE MS Rx Noise Floor -101.0 dBm
150
140
Figure 5 Example LTE link budget (macrocell)
130
loss (dB)
120
110
This shows that while the 5MHz channel 100
being used can in theory deliver a 25Mbps 90
80
connection, the top modulation rate is only 0 0.2 0.4 0.6 0.8 1
achieved across 9.5% of the cell. The blended distance (km)
capacity across the cell is 8.3Mbps. As can be Frequency f 1990 MHz

Base Station Height h_1 27 m
seen this is far below the marketing numbers Mobile station Height h_2 1m
Distance d 0.7 km
of 100Mbps download and 50Mbps upload. Height of Building Roofs hr 20 m
This capacity is shared across the entire Width of Road w 20 m
Building Separation b 40 m
cell/sector of the BTS. For a cell radius of Road Orientation p 45 degrees
650m the circular area is 0.5 square miles. If City Type 0 1 = medium/suburban or 0= large metro
the cell site is tri-sectored then the sector area

is 0.17 square miles and the capacity per Figure 6 Example LTE link loss (macrocell)
square mile is 48Mbps. This can be contrasted
with the performance expected from a Wi-Fi
based microcell approach where capacities of MSOs have generally prided themselves
200Mbps per square mile are deliverable on delivering the fastest user experiences, in
today with growth to 0.5Gbps per square mile. contrast to the typical wireless system. In
order to improve theses capacities, a
microcellular approach is required.
When similar calculations are done for
today’s 3G standards such as HSPA and

Wi-Fi microcells as a complement Loss
130.000
125.000
120.000
115.000
In the preceding sections it has been
loss (dB)
110.000
demonstrated that macrocells deliver 105.000

100.000
relatively low capacities and user throughputs, 95.000

90.000
but MSOs have another tool that can be used 85.000

80.000
– the microcell. MSOs also have the ability to 0 0.2 0.4 0.6
distance (km)
0.8 1
use the unlicensed frequency bands and the

Figure 8 Example Microcell Loss
802.11 Wi-Fi standard. Taking the same
technical approach to a Wi-Fi based system
and assessing the percentage of each cell that
Using this number of 8.3Mbps average per
operates at the various modulation rates we
Wi-Fi microcell, and using 30 microcells per
get an average capacity of 8.3 Mbps in a
square mile, we arrive at a capacity of
20MHz band when considering just the
249Mbps per square mile, which is 5 times
current 802.11b/g standards.
that capacity of the LTE macrocell. Using
Modulation
Base Bits per Hz
cell edge BPSK
1
QPSK
2
16QAM
4
64QAM
6
802.11n, the evolution of 802.11g, capacities
raw rate
Code rate
6.89
0.50
13.78
0.75
27.57
0.75
41.35
0.75
of in excess of 500Mbps per square mile can
Derated bits per Hz
Bit Rate (Mbps)
0.50
3.446
1.50
10.338
3.00
20.677
4.5
31.015
be achieved.
Thoughput 2.76 8.27 16.54 24.81
SNR required (dB) 0 8.5 15 21
Uplink Sensitivity (dBm) -99.97 -91.47 -84.97 -78.97
Downlink Sensitivity (dBm) -95.97 -87.47 -80.97 -74.97
BTS TX pwr 27 27 27 27
BTS ant gain
MS ant gain
4
-2
4
-2
4
-2
4
-2 Comparison of 3G/4G/Wi-Fi systems
MS Tx pwr 20 20 20 20
Uplink link budget 134.0 125.5 119.0 113.0
Downlink link budget 125.0 116.5 110.0 104.0
Fade margin 9 9 6 6
overall DL link budget 116.0 107.5 104.0 98.0
Figure 9 shows a comparison of the

Distance (km)
Distance (mile)
0.30
0.19
0.47
0.29
0.23
0.14
0.17
0.11
0.10
0.06
throughputs expected from various standards
Cell Area (km^2) 0.28 0.28 0.69 0.17 0.09 0.03
Ring Area (km^2)
% of whole cell
0.12
40.74%
0.53
76.05%
0.08
10.86%
0.06 0.03
8.56% 4.53%
in the real world.
% with cell edge limit 41.22% 26.67% 21.00% 11.11%
Blended Data Rate [whole
cell] (Mbps) 8.30
Figure 7 Example Wi-Fi link budget (microcell)
This model uses a microcell power loss

that was derived from empirical
measurements of basestation deployed at
strand height using Wi-Fi at 2.4GHz. The best
fit to the data was a propagation power law of
2.8 resulting in the loss profile shown in
Figure 8.
Figure 9 Comparison of systems
A recent study by Ken Biba of Novarum

[3] confirms the real world effectiveness of
Wi-Fi based systems for delivering a good
user experience in comparison to the 2G and
3G cellular standards that are deployed. In the

study the Wi-Fi networks delivered 3 times HSPA was 3.4Mbps, LTE was 11Mbps, both
the throughput of the WiMAX and Cellular on a 500m cell.
networks with the same coverage.
Overall this demonstrates that wireless
capacities and user experiences are generally
A study by Nortel Networks [4] also drew lower than that experienced by home users,
similar conclusions to the above calculations. but that Wi-Fi based microcells offer MSOs
That is that the expected performance of an important tool that can deliver superior
performance.

STRAND MOUNTED WIRLESS BASE
STATIONS
As mentioned earlier, MSOs have a

unique asset that enables them to play a large
role in the 4G wireless world. This asset, the
HFC, offers the three elements necessary for
deploying microcells.
• Power – Plant power, quasi square wave

low voltage ac
• Backhaul – DOCSIS or fiber
• Mounting – the messenger wire Figure 11 Strand Mount Basestation Details
A typical strand mounted wireless BTS is The architecture of such a strand mounted
shown in Figure 10. basestation is shown in Figure 12. The
combined DOCSIS® and power enter the unit
via power protection module that splits the
power and the DOCSIS® apart. This module
provides filtering to prevent emissions from
entering the plant and lightning and power
surge protection. A power supply takes the
plant power quasi square wave and converts it
to usable dc voltages for the rest of the BTS.
Figure 10 Deployed Strand Mount Basestation
The BTS has 2 universal radio slots, which

The basestation uses standard strand in the example shown contain a LTE (3GPP
hangers for mounting and has a single Long Term Evolution) radio paired with an
connection to the HFC for both DOCSIS® 802.11 Wi-Fi radio.
and power. Various antenna options are
required depending on the RF coverage that is
being deployed. In order to ensure that omni
directional coverage is provided, a bracket is
used to offset the omni directional antennas
from the basestation to avoid shadowing.
Directional antennas are also possible for
coverage of tall buildings or for making
backhaul connections. The DOCSIS®/Power
connection is made via a dc splitter and/or a
power passing tap from the main coax.
Figure 12 Inside a Strand Mount Basestation

• IP address of the end device shall not
Central to the Basestation is a control change.
board that contains Software that enables • Customer shall not be required to
integration into the HFC. This processor login multiple times.
provides the packet processing to apply QoS • Specialized Software shall not be
as well as the mobility features described in required on the user’s device.
the next sections.
MOBILTY IN THE HFC
One key difference between the delivery of

wireless service and fixed services is the
aspect of mobility. Today’s HFC is designed
to provide fixed subscribers with a high speed
connection, and contains the assumption that
the user device does not move. In most
Figure 13 Tunneling approach for mobility
implementations there is an IP address limit
per connection of 2 to 5 addresses and a MAC
address limit of 8-16 devices. The network This level of seamless mobility can be
assumes that a user has a NAT (Network scaled to systems in the order of 10,000
Address Translation) router device deployed contiguous basestations. We also need to
on premise that hands out addresses to local address larger scale systems and mobility
computers. In a wireless system deployed on between standards. To do this, a mobile IP
the HFC, the mobile devices can move infrastructure can be deployed as shown in
between basestations at will. In fact, even a Figure 14.
static/fixed wireless device will be mobile as
variability in RF propagation and the One might ask why not deploy a full
environment means that a so called fixed mobile IP infrastructure to handle the mobility
device will connect to several basestations day one. The reason for this comes back to
over time. one of the challenges of microcells which is
that there are lots of cells. This translates into
a lot of mobility events in the system as the
One of the main requirements is to deliver cell boundaries are frequent and users
this mobility without requiring wholesale transition though cells quickly, generating a
changes to the existing HFC components such lot of mobility events. With a hybrid tunneled
as the CMTS and Fiber nodes. In order to and mobile IP architecture, the transitions
deliver such mobility on the HFC a tunneled between microcells are handled at layer 2
architecture is proposed as shown in Figure whereas the transitions between groups of
13. The tunnels are IP constructs that can microcells and between different systems are
traverse a layer 2 and or layer 3 network, handled at layer3.
including a NAT device if necessary.
In order for mobility to be a seamless

experience for the customer, there are some
critical requirements.

concept of centralized subscriber
management.
In the classic HFC case, the modems

represent the customer and the CMTS and
Network can be provisioned to authorize a
particular fixed device, whether that is a set
top box, PVR or Cable modem. In the case of
the mobile device, the device can exist
anywhere on the network, and the user is
attached to the device rather than a fixed
location such as a home. This necessitates a
Figure 14 Mobile IP between networks centralized approach to login. This can be
transparent to a user such as a SIM
A more detailed view of the entire network (Subscriber Identity Module) based
can be seen in Figure 15. This shows another authentication or MAC address based, or a
fundamental difference between a mobile web based user login.
network and a fixed network. That is the
Figure 15 HFC with tunneled mobility

INDOOR WIRLESS
Indoor wireless is another part of the story

that needs to be addressed. In this case,
cellular technologies have struggled to
deliver a consistent user experience and the
experience of dropped calls within the home
has been an experience that has frustrated
many. In fact some cellular providers have
deployed Wi-Fi already as a solution for in
the home wireless using the UMA [5]
technology. Others are exploring the use of
Femto-cells which are miniature basestations Figure 16 Pedestal mounted BTS
for delivery of cellular service. This trend
fully supports the argument for strand based
microcells in the outdoor world, as
Femtocells are becoming necessary to deliver
even 3G data speeds to end users.
Fortunately the MSO/HFC friendly

architectures described above can also be
used to deliver wireless services into the
home or small business. With the hybrid
tunneled architecture seamless mobility can
be provided between outdoor and indoor
locations.
WHAT HAPPENS WHEN THE PLANT IS

UNDERGROUND?
Of course, as ubiquitous as HFC is, not all

of it is overhead strand. A reasonable
percentage is underground, and is accessed
from vaults and pedestals. Figure 16 shows a
typical residential area pedestal with a HFC
fed base station embedded.

SUMMARY By using the HFC to deploy Wi-Fi today,
an MSO can deploy a wireless experience
that will be unmatched until 4G licensed
This paper has demonstrated that the HFC band microcells become ubiquitous.
is an ideal location for hosting wireless
basestations. Broadband wireless systems
need microcells in order to deliver the REFERENCES
broadband experience they promise, and
microcells have traditionally been limited to
locations where backhaul and power are [1] 3GPP Long Term Evolution
accessible. The HFC can provide the 3 http://www.3gpp.org/article/lte
necessary ingredients (location, power and
backhaul) for a microcell deployment.
[2] Cost 231 Walfisch-Ikegami Model
http://www.ee.bilkent.edu.tr/~microwave/pro
Full mobility can be provided over the
grams/wireless/prop/costWI.htm
HFC using a tunneled architecture together
with a mobile IP overlay offering a user a
[3] Ken Biba, Novarum
seamless mobile solution, while not requiring
http://www.muniwireless.com/2009/02/03/m
an upgrade to the HFC.
uni-wifi-outperforms-cellular-and-wimax/
[4] The Business Case for LTE, Gerry

If MSOs wish to deliver the superior user
Collins, Nortel
experience that their customers are used to
http://www.umts-
from their cable modems in the wireless
forum.org/component/option,com_docman/ta
world then they should consider the HFC as a
sk,doc_download/gid,1758/Itemid,12/
key asset. Today’s Wi-Fi systems offer
capacities and user throughputs that are far
[5] Un-licensed Mobile Access
superior to today’s 3G cellular systems and
to next generation 4G macrocells.
http://www.umatechnology.org/overview/
Interestingly, the 4G systems have

migrated to an all IP transport system,
something that Wi-Fi was designed to do
from the start. This is another advantage that
MSOs have, with their IP friendly networks.
A Wi-Fi system deployed on the HFC can be
used to deliver broadband IP wireless
services to customers and as a bridge to a 4G
microcellular system in an extremely cost
effective manner. While not covered here, THE END
such HFC based wireless networks can also
deliver a full multi-media experience, a topic
for a future paper.

INCREASING MSO ADVERTISING REVENUES
THROUGH MANAGEMENT OF AD SKIPPING
Dan Holden
Comcast Media Center
Abstract Industry trends and consumer sensitivities

drive advertisers to be creative when placing
New capabilities that allow customers to ads. For example, digital signage has gained
enhance their viewing experience, such as the great penetration in retail outlets and movie
ability to skip advertisements, can be theaters thus presenting consumers with
disruptive to traditional business models. campaign material regardless of their location
However, these new functions also provide or activity. In the same way, consumers
an occasion for the industry to seize market viewing habits can be leveraged to present
opportunities that are created by shifts in alternative and creative means by which linear
consumer behavior. Rather than seeking advertising content can be presented.
to defeat the customer’s desire to fast-forward Specifically, time-shifted non-linear
through content that doesn’t respond to his or programming can be enhanced to add value to
her personal interests, it is more effective to the traditional thirty-second spot. For
acknowledge that desire and present example, it is possible to place advertising
alternative content. Proactively responding “on top” of media, much like animated bugs.
to the consumers actions changes the This and other techniques, which we will
traditional spot advertising approach and collectively call “trick-file advertising”, offer
benefits consumers and advertisers alike. great revenue opportunities by replacing or
enhancing advertisements while consumers
This paper outlines a revolutionary approach engage trick modes such as fast forward,
for using new technology to present pause or rewind.
advertising to consumers. The innovative
solution leverages the action of ad skipping as The main goal of trick-file advertising is to
a catalyst for allowing advertisers, cable leverage consumer behavior for the benefit of
programming networks, and cable system consumers and advertisers alike. As
operators to generate additional revenue that consumers opt to skip advertising, they are
is currently lost as a result of ad skipping. becoming conditioned to reach for the remote
at the beginning of every break. This behavior
demonstrates that:
INTRODUCTION • Consumers are actively interacting
with the television and the displayed
Many consumers actively choose to skip programming
ads. They purchase or lease technology that • They have the remote in their hand
helps them with this endeavor.
• They are not interested in the
Advancements in technology provide cable
commercial content being presented
system operators with the ability to support
the customer’s preference to skip through ads,
This is truly interaction with the television
while presenting an alternate paid campaign
platform. During the instances where viewers
that runs in the foreground while viewers are
choose to fast forward through content, it is
skipping the original ads or content in the
possible to insert alternative content in the
background.
foreground while the original content appears
to be fast forwarding in the background. At

this point in time, there are two clear
objectives: to get more impactful content in
front of the consumer and to log the event. If The statistics indicate that use of time-
the alternative content is unique and shifting equipment and functions to skip ads
compelling, it will be more valuable and will continue to grow. DVR penetration is
relevant to the consumer. As an example, in projected to increase from its current 23
2005 Burger King created a viral buzz and percent to 37 percent of all TV households, by
therefore drawing increased viewership to 2012. In addition, 25 percent of primetime
their linear commercials as viewers flocked to content will be time-shifted by the 18-49
see alternative content hidden in the 30- demographic (Magna iv ).
second spot.
Least-skipped ads
AD SKIPPING
However, viewers with the capability to
DVR penetration skip ads don’t skip through all ads
indiscriminately. For example, TiVo
Customers skip ads for various reasons commissioned a study (Jan 2008) on
including burnout and lack of interest in the advertising and determined not all advertising
advertised product. The potential for is skipped. The least-skipped ads include
consumer ad skipping is increasing on a theatrical film releases, which accounted for
yearly basis. According to data collected in six of the ten most-viewed commercials in
2008 on DVR penetration, twenty-nine DVR playback mode. In addition, eight of the
percent of U.S. television households are ten most-watched time-shifted cable spots
equipped with DVRs, placing the potential for occurred while viewers were watching the
ad-skipping in front of approximately 33 show “Psych” on USA Network. TiVo
million households (Nielsen data) i . attributed this statistic to integration of
characters from the TV show into the
Moreover, nearly two thirds of U.S. advertising, making it difficult for viewers to
households with $100,000-plus incomes distinguish between commercial and program
subscribe to time shifting services or own a content.
time-shifting device, which translates to at
least one DVR in at least 15 million affluent
households (Ipsos Mendelson ii ) Earlier research by TiVo (covering May
2008) supported the idea that viewers skip
In the first four weeks of the Fall 2008 TV fewer ads that they see as relevant to their
season research from IPG’s Magna, indicated circumstances and interests. TiVo’s
DVR time shifting accounted for 11 percent PowerWatch study specifically evaluated
of all prime-time ratings on the five broadcast spots skipped by demographic. The data
networks. They found DVR playbacks now indicated that viewing for children's skincare
equal 16 percent of all prime time viewing by products in homes with children under 12 was
consumers age 18-49. That's more than 37 percent greater than in homes with adults
double the impact VCRs had, which was at 50-plus. Advertising for toys and games had
the 90-percent penetration level. In DVR- 22 percent more viewing in homes with
only homes, nearly 40 percent of household children under 12. Contrast this to political
prime-time viewing (and 50 percent of prime- ads, which had 15 percent more viewing in
time viewing in the 18-49 age range) is now the homes with adults 50-plus. Finally, ads
time shifted content. iii

for hair restoration products and wigs had 10 viewers interested in the linear scheduled
percent more viewing in those same homes. v program but not interested in the inserted
commercial is a very specific demographic.
Identifying and presenting alternative content
– related to the inserted 30-second spot or
Additionally, Fox TV has found through unrelated (alternative campaign) adds value to
testing that viewer attention levels to the inventory.
commercials are higher for shows with
shortened commercial pods. Biometric
engagement research company Innerscope
found that viewers watching shorter TRICK-MODES AND TRICK-FILES
commercial pods had ad attention levels 31
percent higher; with ad engagement levels 21 Trick modes or trick play
percent higher. Unaided recall levels were
recorded 250 percent higher and ad likeability VCR-like functions such as pause, rewind,
levels 61 percent higher than viewers who fast-forward, replay and skip, are collectively
watched the standard-length ad pods. vi known as ‘trick modes’ or ‘trick play.’ Trick
modes are currently available on DVRs,
Yet another study from Innerscope used VOD, PCs and other CE devices. A trick
eye tracking to measure gaze location and mode is essentially a command that controls
duration and smart vests that measure arousal, the playback of video. Specifically, when
respiration, heart and motion responses. consumers press the fast forward button, this
During their testing, prescreened DVR users interaction indicates their choice to skip that
were given remote control devices that material. This action also represents the
allowed them to fast-forward, pause and play opportunity for inserting an alternative
at their discretion. The viewers whose message rather than the traditional mapping to
programming had shorter commercial pods, fast forward. Caution must be exercised when
and 50 percent less commercials in the implementing this new type of functionality,
program, fast-forwarded 136 percent less than as the consumer could become easily
viewers who watched the standard-length confused. The proposed solution resolves this
commercial pods. vii issue by inserting alternative content that only
covers a portion of the real estate.
It is possible to reduce ad skipping by Additionally, clear delineation between ads
disabling the fast-forward function. However, can be created by a spatialecial separation
in light of the data presented this type of (lines or frames between the new
solution is likely to have a negative impact. message)which appears in the foreground and
An alternative approach would be to leverage the original content, which appears as
the ad skipping behavior of subscribers for the background The ability to still see the original
benefit of advertisers and consumers. This content in fast forward mode will allow the
approach can benefit customers by presenting consumer to see that an advertising trick mode
them with an abbreviated, impactful message has been activated
that would appear during the same time
interval as the skipped ads. Trick-files
Trick-file advertising also benefits A trick-file is the video a consumer

advertisers. During fast-forward, the experiences when they activate a trick mode,
advertiser can be certain the viewer is e.g. fast forward. When trick-file video is
engaged with the television. Captured streamed, it gives the illusion of a VCR

running in faster than normal mode. One possible to insert into the viewer’s screen
method for the creation of a trick-file is to alternative content that is different from the
extract all I-Frames then create a new file advertisement or other content that had
comprised of fewer frames than the original triggered the ad-skipping response. While the
video. For instance, a thirty second spot alternative message must be presented in a
encoded to VOD specifications would contain unique and compelling way, it can be from the
exactly 900 frames of video. In order to play same advertiser or content source For
the spot in five seconds 750 frames of video example a traditional 30-second car ad could
would be removed in order to reach a total be replaced by a shorter message from the
frame count of 150. local dealer containing a special offer.
There are three general methods deployed Since customers will choose to view
today for VOD trick file displays: files, content that matches their interests, this new
indexes, and dynamic video generation. The approach represents a win / win for the
file method creates a new trick file (video file) advertiser and consumer. User behavior is the
that is played when a trick mode is activated. key. Once the customer has the remote in his
Indexing is a similar approach. However, or her hand, the content presents an
rather than creating a new video file, indexes opportunity to turn the original commercial
are assigned to frames of the original video. pod into an interactive experience. This
These indexes are utilized by the video pump principle also applies to other content that
to select the correct frames to display in order viewers may seek to skip. For example, a
to achieve a fast forward effect. The dynamic viewer that fast forwards through the first
video generation approach uses tools to several moments of a TV series that contains
calculate how many frames to skip in real- clips from previous episodes could be
time as the video is played. exposed to a short promotional segment about
one of the leading characters or a product
Each method outlined above may be used endorsement.
to create a trick file advertisement. While
each method described has its own As previously noted, trick-file advertising
complexities associated with the creation of is consistent with the direction that consumers
the trick file ad, the true complexity surrounds and advertisers are driving technology.
the successful mapping and tracking of these Advertisers are open with their plans to move
enhanced ads. To meet the need of the advertising dollars to platforms that offer
advertising community for detailed data, it is greater levels of audience engagement.
key that methods be developed to track Meanwhile, studies such as the ones cited
original advertising as well as the replacement earlier in this paper demonstrate that viewers
ad, while exposing the consumer to equipped with ad-skipping technology will
replacement spot. choose to watch ads that interest them.
The new paradigm represented by trick-file

Trick-file Advertising advertising is comprised of a process that will
present, when prompted by the consumer,
Trick-file advertising is essentially the new advertising in the foreground while the
blending of two distinct concepts: ad-skipping original advertising that is running in a trick
and paid campaigns. Today, when customers mode, such as fast forward, continues to
press the fast forward button, they are appear in the background. The customer
indicating they wish to jump through the views an alternative message during the same
current programming. At this instant, it is amount of time that they are fast-forwarding

through the original content. For example, a CREATING TRICK-FILE
thirty-second advertisement may be replaced ADVERTISING
with a shorter, five-second message. This type
of advertising is restricted to time-shifted The implementation of technology to
content (VOD, DVR, StartOver, etc), since it support trick-files builds upon other advanced
is not possible to fast-forward through real video developments, including VOD and the
time content, such as a live sports cast. tru2way™ platform. Three approaches
discussed in this paper include are file based,
In addition, trick-file advertising represents VOD based, and DVR based creation
the opportunity to create new forms of
personalized content. For example, another File based
way to expand further the potential of trick
file advertising would involve providing The first step for developing file-based
unique messages for multiple buttons on the trick files is to create a file, which will move
remote, by using trick modes with each button the current advertising into the background
that are mapped to a different message. This and implements the fast-forward effect. For
capability will allow selling more avails per example, a normal two minute break has 3600
placement opportunity. frames of video. To run the break in the 20
seconds that simulate a fast-forward mode, the
To radically change the fast forward original 3600 frames of advertising would be
experience would not be advisable, as it is cut to 600 frames of video. This decrease in
sure to confuse the subscriber. A better frames is suitable since the background video
approach is to place the new advertising video will only be utilized to track original
in the foreground, and distinguishing it from programming. This process involves creating
the original content, which appears in the a file that uses only a fraction of the frames
background through the use of a spatial contained in the original content. Once
separator. This “picture-in-picture” affect will extracted from the original ads, the 600
allow the user to track the timeline of the frames are re-constituted as a new file that
original programming, while being presented creates the illusion of a background content
with the alternative content ( (figure 1). running in fast forward mode.
Next, the replacement content is placed in

the foreground. Each and every frame of this
alternative content will be squeezed and
appear inside a spatial separation via a black
border, so that it may be easily viewed as
different content from the background video.
Once these steps are completed, the file must

be indexed, tied to specific content and
prepared for distribution to a playback device
(VOD, DVR etc) When the new trick-file
enhanced advertisement is experienced the
customer views the four thirty-second spots
Figure 1 running in the background while one twenty-
second spot is running in the foreground at
normal speed. This approach creates an
alternative experience during trick mode but

in no way modifies the experience during
normal play All of the advertising being The next step in the process is to overlay
displayed is tracked via VOD server logs and the new advertising on the newly created
can provide metrics down to the unique trick-file. This process involves encoding the
subscriber level. replacement spot to exactly 150 I-Frames.
For example, these frames may be labeled
A more detailed look at the process to “A1” through “A150”. Each of these I-
create trick-file advertisements begins with Frames will have dimensions slightly smaller
the following assumptions: than the I-Frames from the trick-file
background video. The following diagram is
• One thirty second spot representative of the first five frames in the
• Video is encoded to VOD specification newly assembled trick-file advertisement.
(MPEG-2 transport stream, long GOP
set to fifteen, thirty frames per second)
• CMC standards for advertising (900
frames of video with bookend twelve
frames of black)
Based on the specifications above, there Once the video has been correctly
are approx 60 I-frames of content after assembled, the next step is to add the audio
removing the black that appear before and track. Then, the newly created elementary
after ads, resulting in approx. 60 frames of stream is wrapped with an MPEG-2 transport
background video. In this example, the I- stream component. Now it is ready for
frames are labeled “I1” to “I60.” As all of the distribution to VOD pumps, DVRs, or other
black frames are equivalent, they are CE devices. This algorithm can be adapted to
collectively referred to as “B”. Creating a support a wide range of durations.
five second trick-file ad will require a total of
150 frames of video (5 seconds * 30 VOD based
frames/second). A simple algorithm for
reconstruction would be 3(I1) + 3(I2) + 2(I3) There are two options for creating trick
+ … + 2(In), where n=location of the frame in files as part of the VOD process. First, with a
the original asset. This would result in a few enhancements to the video pump, it
trick-file of approximately 120 frames, thirty should be possible to create the picture-in-
frames short of the 150-frame objective. picture video effect dynamically. This would
Next, additional I-Frames are added at the allow late binding of the advertising
beginning and end of the file to achieve the replacement spot. As the pump plays trick-
150 frame objective, and to bookend the file file video, the replacement advertisement and
with frames of black in order to provide appropriate spatial separation would be
smooth video transition. This process results spliced into the reserved section of the trick-
in having the original 912 frames of video play video.
reduced to 156 frames of video that will run in
five seconds. This new file will then be used Alternately, it is possible to orchestrate the
to provide the background video of the VOD pump, which in essence would make a
original spot running in fast forward mode. copy of the trick file and provide the
appropriate enhancements to it. The output of
at “trick file generation device” could be
attached to a feed that’s capable of splicing
portions of the video.

changes to the pre-encryption distribution
DVR based process should suffice for content delivery. In
the base case, the VOD server would simply
DVR implementation of trick-file play the trick-files as it does today. VOD
advertising is essentially the same as VOD. server logs may be used for collecting
The biggest obstacle to overcome is getting advertising metrics in a manner that upholds
the trick-files on the DVR and associating strict privacy guidelines.
them with content correctly. Within the
tru2way specification, the ability to push When compared with other advanced
content to and from web servers has been advertising activities, trick file advertising can
defined. These features have yet to be offer compelling opportunities for revenue
implemented within the tru2way 1.0 generation using technology that is largely
specification, but should be available in the available today. These types of advanced
future. Work is currently underway at advertising opportunities include:
CableLabs to enhance the metadata
specifications thus allowing advertising
triggers to be inserted into video and read by Telescoping – The process of launching long
DVRs. Advertising engines on DVRs will form advertising from triggers that are
then be able to dynamically insert advertising embedded in linear or non-linear content.
on top of trick-file playback, provided the
DVR is either equipped with multiple Addressable – The process of placing unique
rendering devices or splicing capabilities. advertising to match viewer interests.
Request for Information (RFI) – Is the

COMPARISON TO OTHER ADVANCED process of gathering customer interest to
ADVERTISING TECHNOLOGIES create mailing lists of print, or other forms, of
advertising.
Personally relevant advertising is a key
concept in the advanced advertising tool kit, Trick-File – Is the placement of adverting
and revolves around matching a campaign’s “on top” of standard trick play (fast forward)
message to consumer interests. Additionally, video.
the collection of metrics to measure campaign
effectiveness is sought. Trick-file advertising eTV Additional Behavior
has the ability to achieve both of these Bandwidth Modification
objectives without massive upgrades to the Telescoping Yes Yes Yes
cable plant infrastructure. Addressable No Yes No
RFI Yes No Yes
The simplest types of trick-file advertising Trick-file No No No
can be implemented within the VOD
infrastructure. It will not consume additional
bandwidth like telescoping or linear In addition, the capabilities of trick file
addressable to the set top box, but will require advertising can incorporate a number of
additional VOD functionality and resources. advanced features in a two-way operating
Unlike eTV, the simpler forms of trick file environment. For example, trick modes may
advertising do not require a two-way plant. be captured as state changes on the VOD
Advertising content can be pre-produced and platform and then mapped to specific
checked for quality then staged on the VOD functionality. This type of state machine will
platform. For some VOD systems, a few

allow for the creation of a contest that is remote can be used for creating multiple
embedded within the content. One type of opportunities by associating different
contest could be an Easter egg hunt, where messages with each of the modes, such as
video widgets may be placed across the span Fast-forward, 2X Fast-forward, Rewind, 2X
of content and accumulated into a basket, Rewind are just a few. Added value may also
much like a shopping cart on the web. Each of be realized due to the nature of the
these interactions would be captured and advertising, the additional opportunities and
persisted. Once the consumer has interacted how the content that is constructed creates the
with all of the specified widgets they would option for different kinds of impressions.
be notified of contest results.
CONCLUSION
Other contests could span across multiple
pieces of content. For example, a Monopoly- Trick-file advertising offers the industry an
like game could be created where property innovative approach to harness ad skipping.
widgets are hidden and embedded in multiple It is activated at the request of the consumer,
pieces of content for collection by consumers. and sits squarely between ad-skipping and
disabling fast forwarding. Since much of the
A wide variety of contests are possible. required technology is already deployed,
Some may be complex and last months, or capital expenditure for implementation should
they may involve simple techniques, like be minimal. When deployed trick file
pressing fast forward twice followed by advertising offers a dynamic way to reach
reverse during an advertisement in order to specific demographic groups, and can collect
win a prize. The goal is to provide advertising metrics while upholding stringent
capabilities to both content providers and privacy guidelines.
advertisers for innovative campaigns to
connect with consumers. A flexible
advertising architecture is sure to make The next step for this exciting technology
advertising much more compelling and of is to get creative. Enlisting the talent of ad
interest to a wide segment of cable customers. agencies will help align the new technology to
consumer expectations and provide the best
Multiple Avails possible user experience. Organizations like
Comcast Spotlight and Canoe may be called
Trick-file advertising will allow for more upon to determine how to sell the new
avails per break. It is in effect a second placement opportunities and additional avails,
opportunity to reach the consumer. As as well as develop strategies for maximizing
previously stated, different buttons on a potential revenue.
v
“King of Beers Defies TiVo Timeshifting,”
Television Broadcast magazine, January 2008
vi
i
“Report: More Than Half of $100K Households Time “Fox Goes Sci-Fi to Measure ‘Fringe’ Ad
Shift,” By Anthony Crupi, Mediaweek, January 8, Effectiveness,” By Jon Lafayette, TVWeek,
2009 September 7, 2008
ii vii
Ibid Ibid
iii
“Study: DVR Ratings Impact Rises,” Steve
McClellan, Adweek, Nov. 6, 2008
iv
“MAGNA Global updates On-Demand forecasts,”
Radio Business Report, September 10, 2008

NEW MEGABITS, SAME MEGAHERTZ: PLANT EVOLUTION DIVIDENDS
Robert L. Howald, Ph.D., Amarildo Vieira, Ph.D., Michael Aviles

Motorola Home & Networks Mobility
Abstract proper architecture limitations and system

thresholds to ensure high-performance
The use of 256-QAM in the downstream delivery of 1024-QAM. These guidelines can
path has nearly completely replaced 64-QAM be used to enable MSOs to reliably extract
as the modulation of choice for MSO’s for two another 25% more bandwidth from their
simple reasons: digital multiplex.
(1) 256-QAM is more bandwidth

efficient, providing a 33% increase in
spectrum efficiency compared to 64- INTRODUCTION
QAM
(2) 256-QAM has been proven to work Appetite for more bandwidth has continued
reliably. to increase, and there is every expectation that
it will continue. In particular, as MSOs
A natural question to ask, given the migrate from broadcast-like models with a
relatively smooth transition from 64-QAM to VOD component, to delivery models trending
256-QAM, is whether there is a convenient towards all-switched unicast services, the
next step in terms of improved bandwidth Mbps required per service group climbs
efficiency. The industry has yet to make a dramatically. Existing tools (analog
significant move towards 1024-QAM. Prior reclamation, MPEG-4, SDV, etc.) can and will
papers, including by this author, have pointed continue to make the bandwidth explosion
out some of the potential hurdles. However, manageable. Nonetheless, continued
as the HFC network has evolved in support of bandwidth growth demands expansion of the
new service demands, and the downstream HFC toolkit. The requirements are driven by
multiplex has done the same, variables that a combination of high-consumption trends –
affect the ability to reliably implement 1024- personalized streams, the content itself
QAM are beginning to work in favor of this evolving to HD and beyond, the growth in
more bandwidth efficient approach. multiple and non-traditional consumption
venues and devices, and the desire to continue
It is now time to understand and quantify to increase data tiers for high-speed Internet
the practical performance and the potential service.
limitations in order that it may be deployed
properly. This paper will examine the The cable plant has kept up with the
required specification of impairments for bandwidth consumption by adding RF
successful transmission of 1024-QAM. The bandwidth and using efficient digital
discussion will summarize the effect of HFC- modulations to mine the capacity effectively,
specific impairments on 1024-QAM and and with robustness. What started as 64-
compare them to 256-QAM and 64-QAM. QAM digital signals became yet more
Finally, we will present some conclusions, bandwidth efficient with the deployment of
along with supporting measured data, on the 256-QAM downstream, which is the dominant

QAM approach today. The ability to efficiencies and implementing wider channels,
successfully deploy such schemes is due to the this could possibly be increased to four HDs
fact that the downstream channel in a cable per QAM [2].
plant has very high SNR, and a very low
distortion. This is because it was designed to Of course, there is no free lunch when it
ensure proper conditions for supporting much comes to modulation efficiency. To support
less robust analog video, which had 1024-QAM, a more stringent set of
historically dominated the downstream specifications must be met. The goal here is
payload. In addition to high linearity and low to identify how pristine the plant must be, or
noise, the downstream channel has a flat must become, in order to migrate to this
frequency response on a per-channel basis, modulation profile effectively in terms of the
minimizing both amplitude and phase likely candidates for disruptions to robust
distortion, although it can be prone to transmission: SNR, Beat distortion
reflection energy. interference, and to a lesser extent, phase
noise. It is worthwhile to point out that an
What has powered digital modulations intermediate step using 512-QAM, and its
historically is the ability to deliver robust link additional 12.5% efficiency, has merit in
performance and services over varying moving up the modulation complexity chain.
degrees of link quality by boiling the required All of the tools developed here apply to 512-
receiver function down to choosing between QAM as well, with of course a different set of
1’s and 0’s, instead of replicating and numbers drawn from them. For this analysis,
infinitely-valued analog waveform with high however, we will focus our attention on 1024-
fidelity. And, when channel conditions are of QAM, as these modems exist for the cable
high quality – such as the downstream cable space, and the downstream channel capacity
plant – the situation is ripe for exploiting favors the likelihood of adding increasingly
bandwidth using very efficient digital more bandwidth efficiency to the plant.
techniques.
PRIOR ANALYSIS, ESTIMATES, AND
As a simple example of the possibilities, TESTING
the theoretical capacity of a 6 MHz channel
with a 40 dB SNR is approximately 80 Mbps. An early analysis of 1024-QAM was
Yet, for 256-QAM, the transmission rate is presented at the 2002 SCTE Cable-Tec Expo
only about 40 Mbps. When accounting for [1]. While early modem ICs existed that
overhead, there is even less throughput. The could support this modulation mode [5], little
next higher order, square-constellation, attention had been given to it in the cable
modulation is 1024-QAM. This technique world by operators, and subsequently little
achieves an efficiency of 10 bits/symbol, or was understood about how well it would
another 25% efficiency over 256-QAM, and perform in a typical plant. The paper used
an impressive 67% improvement relative to communication theory and what was learned
64-QAM. Practically speaking, three MPEG- during the implementation phases of 64-QAM
2 HD’s per QAM fit comfortably, as and 256-QAM to draw conclusions about the
compared to the threat of visual artifacts of expectations for 1024-QAM, which was
jamming a third HD channel into a 256-QAM thought to be coming around the corner.
payload. Alternatively, it would represent at However, because of the potential
least two more SD streams per QAM. By also complexities of taking on this advanced
considering statistical multiplexing scheme and because of other more pressing

priorities, this modulation mode has not been Going from 64-QAM to 256-QAM means
exercised, nor has there been any significant 6 dB of lost margin in the example above,
trial activity. which sounds perilous – 36 dB of SNR
against an uncorrected 1E-8 performance
The conclusions of the 2002 paper covered threshold of 34 dB. However, the relative
a range of items, but ultimately focused on the digital power in this case is typically -6 dB, so
major items likely to be challenging to 1024- four of those lost dBs are recovered, at the
QAM under typical HFC performance expense of a larger impact on the total RF
delivered: SNR, analog beat distortions, and load of about 2.4 dB. Because that could
phase noise. In each case, the conclusion was mean 5 dB of third order distortion
that current RF characteristics represented degradation (such as CTB), development of
performance that would lead to minimal hybrids around digital loads had already
margin. And, for “below average” performing followed suit in the expectation of larger
HFC plants, the result could be operating in a loads. In fact, the expansion of analog
bit-error prone region that would add to the bandwidth in the plant has continued to spur
FEC’s work of delivering error-free data. The the development of new actives, and these are
concerns regarding these three parameters are often designed with the ability to have the
quickly summarized below. extended bandwidth still filled with analog
channels. So, as the needs of more digital
SNR SNR for 256-QAM payloads started to matter
from the RF load perspective, actives had
The difference in SNR requirements been keeping pace so as not to increase
between 64-QAM and 256-QAM is 6 dB, and distortion and degrade analog quality.
the difference between 256-QAM and 1024-
QAM is an additional 6 dB. This is nearly Now consider 1024-QAM. Figures 1 and 2
exact in the non-FEC case, and is close also in (see end of paper) show constellation
the case of added coding (of the same type), diagrams of 256-QAM @ 34 dB SNR and
although post-FEC error rate curves are much 1024-QAM @ 40 dB SNR. These are
steeper. So, while 64-QAM, which delivers a equivalent uncorrected BER cases supporting
1E-8 BER with no error correction at a 28 dB a 1E-8 performance. The similar relative
SNR, zips along comfortably on a cable plant, relationship of QAM symbol cloud to hard
each modulation order increase gets more decision boundary is apparent. The congested
difficult. At 28 dB, this allows the QAM load look of the 1024-QAM diagram, emphasized
to be implemented with up to 10 dB of signal by the small symbol decision regions, signals
power back-off relative to the analog carriers the sensitivity this scheme has to disturbance,
and yet still support a large link margin. For and is illustrative of the battle ahead. It takes
example, 46 dB delivered for analog, means only small impairments to move an otherwise
36 dB delivered for digital, resulting in more good symbol across a boundary. It is this 40
than 8 dB to spare without even including dB SNR and the sensitivity to analog beat
FEC. This amount of back-off is important, distortions that led to the suggestion that
because it allows the digital load to become 1024-QAM represents the first digital
much less consequential to the total power modulation choice that needs to be treated
load, adding only about 1 dB to the total RF more like analog modulation than digital.
load on an 870 MHz system, assuming a 12-
14 dB tilt. Consider what 40 dB means in terms of use
on the plant. In light of the above example for

digital loading, 46 dB of plant analog CNR used for 256-QAM in this testing example, the
becomes 40 dB of digital SNR. By adjusting authors noted errors accumulating. Increasing
the modulation order upward, and without any the digital power by 6 dB removed most, but
other steps taken, we have instantly not all, of these errors. It is not clearer
evaporated virtually the entire link margin, whether the original digital level was relative
and are now into a region of measurable bit to analog. A 6 dB increase of the single test
errors, relying on FEC to finish the job under channel on the multiplex would have no
even the most benign circumstance of thermal significant effect on loading whether the level
noise only. Clearly, for plants or areas of began at -10 dB or -6 dB. However, the
plant that deliver poorer SNR, the situation authors did note that the resulting SNR
becomes that much more challenging. measured at the receiver was 36 dB, which is
consistent with what is expected based on the
Additionally, on the STB side, there is sample calculations above. The authors
similar margin-limited mathematics. For STB additionally note that the system was tested on
noise figures in the 10-14 dB range, and for a relatively short cascade (N+3), which we
QAM signals arriving at the STB at the low will later see to be advantageous.
end of the power range, some simple math
shows the following: We will subsequently discuss how
changing HFC variables are improving this
Residual Thermal Noise Floor: scenario and helping change it for the better.
-58 dBmV/5 MHz
Add STB Noise Figure (middle of range): Distortion
-46 dBmV/5 MHz
Analog Level into STB: 0 dBmV Prior analysis [3], [4] had investigated the
Digital Level into STB: - 6 dBmV effects of analog beat distortions on 256-
(increased for 1024-QAM) QAM. Furthermore, laboratory
STB SNR contribution: -6 -(-46) = 40 dB characterization had developed relationships
for the comparative performance of 64-QAM
The SNR delivered to the STB and the and 256-QAM cable links against narrowband
SNR created by the STB combine to further interference, which analog beat distortion
aggravate the situation of being on the edge, represents. It was noted that in the modem
in this case by 3 dB, delivering a 37 dB SNR. receiver technology at the time, there was a
This led to the conclusion that existing 10-12 dB difference in susceptibility to a
conditions and typical deployment scenarios single, static, in-band narrowband interferer at
place the ability to ensure a smooth 1024- the main CTB offset frequency of interest. It
QAM roll-out at risk. It also reveals the seemed logical to extend this relationship
necessity of the 4 dB or so of coding gain – when discussing the difference between 256-
some coding gain reduction occurs between QAM and 1024-QAM for this same situation.
256-QAM and 1024-QAM for the same
scheme – just to ensure link closure at a Furthermore, analysis supported by test
reasonable error rate. results indicated that the receiver would begin
to count errors in a very high SNR
Field tests on live plant [2] bear out the fact environment for 256-QAM when C/I reached
that running 1024-QAM means dealing with about 36 dB, then slowly degrade over the
imperfections of transmission and limited next 5-7 dB or so before becoming
margin. At the original digital transmit levels unacceptably error prone. This seemed to

agree with field performance, as it would take constant [3]. This then taxes the interleaver as
poor CTB to achieve peaks of distortion of well as the error correction mechanism,
this magnitude, even given the noise-like possibly requiring these receiver functions to
quality to its amplitude. Minimum FCC be adapted for the increased relevance of this
requirements of 53 dB along with a noise-like disturbance.
peak-average, digital back-off, and the
addition of a thermal noise component Again, we will discuss how changing
correlated with field results that generally variables in HFC evolution are supporting
showed good deployment results with more robustness in this area as well.
occasional troubling installations, with a
mathematical explanation as follows: Phase Noise
53 dB CTBmin – 6 dB (back-off) – 12 dB (pk- Untracked phase error leads to angular

avg) = 35 dB symbol spreading of the constellation diagram
as shown in Figure 5 for 1024-QAM with .25
It was postulated that, relative to static CW deg rms of Gaussian-distributed untracked
RF interference (RFI), the interference phase error imposed. It was observed in [1]
cancellation mechanism in the receive that this represents a reasonable limit to
equalizer would struggle considerably more ensure that implementation loss due to phase
with an interferer such as CTB because it has noise is about 1 dB, assuming low uncorrected
some finite bandwidth and a randomly varying BER conditions, and with no practical phase
amplitude. This expectation turned out to be noise-induced BER floor. A floor in the 1E-8
the case. The concern echoed at the time for or 1E-9 region will be induced at roughly 50%
1024-QAM was that, under the measured more jitter, or .375 deg rms. Measurements of
performance of that generation of receiver on phase noise showed that for high RF carrier
256-QAM, another 10-12 dB of interference frequencies, typically associated with higher
sensitivity may not be tolerable in many more total phase noise, wideband carrier tracking
cases, and in cases where the CTB may be still left about .33 deg rms of untracked error,
average or better. enough to cause a BER floor to emerge at
very high SNR.
Consider Figure 3 in the context of the
above discussion about interference effects for The use of degrees rms is sometime easier
1024-QAM. A 1024-QAM constellation communicated as a signal-to-phase noise term,
diagram with a 35 dB S/I is shown in the and there are some simple rules of thumb to
figure. This simulation shows clearly how follow and make this simple, starting with 1
such a tone, without mitigation, would cause deg rms is equivalent to 35 dBc signal-to-
errors with noise added in a 1024-QAM phase noise. Doubling or halving entail 6 dB
system. In fact, Figure 4 shows the 38 dB S/I relationships. Thus, we have the following
case with a 40 dB SNR, and it is clear that conversions:
hard decision errors are occurring that would
require FEC support to correct. Analog 4 deg rms = 23 dBc SNRφ
distortion events also tend to be slow in 2 deg rms = 29 dBc SNRφ
duration relative to symbol times – a function 1 deg rms = 35 dBc SNRφ
of the frequency tolerance of the tone .5 deg rms = 41 dBc SNRφ
contributors, which results in an effective .25 deg rms = 47 dBc SNRφ
noise power bandwidth and associated time

The values .33 deg rms and .375 deg rms HFC ARCHITECTURE VARIABLES
represent 44.6 dBc and 43.5 dBc, respectively.
This is instructive to compare to the SNR for Analog Reclamation
the AWGN case, as it illustrates the more
threatening nature of the phase noise Analog reclamation is a “two birds with
impairment on M-QAM modulations of high one stone” architecture variable, offering
M. benefits to both SNR, through potential power
loading adjustment, and distortion. In the case
Recent BER measurements show that error of distortion, the benefit under constant total
flooring does indeed occur as measured by power loading is in the quantity of analog beat
pre-FEC errors, suggesting that there have not components. Constant power loading allows
been significant tuner noise improvements or an operator to take advantage of the
carrier tracking system changes enough to opportunity to increase SNR in the digital
mitigate this effect. However, although phase band with newly available RF power load
noise is a slow random process that challenges headroom resulting from extracting analog
burst correcting FEC to handle, the carriers and replacing them with lower power
combination of the interleaver, the Reed- digital ones. Even with this constant
Solomon encoding, and the relatively low operating point, the effect of analog
floor, does indeed result in zero post-FEC reclamation is to reduce the total number of
errors. Note however, that the phase noise analog beats accumulating that can fall
alone is requiring the FEC to work to clean up beneath a digital channel and create
the output data, consuming some FEC interference. Third-order distortions are the
“budget” in the process. ones that accumulate and cascade most
aggressively in the digital band, while worst
We will not carry forth any further analysis case second-order distortions populate the low
on phase noise, except to note that wideband, end of the band. Thus, we will focus on the
low-noise frequency synthesis is an art that impacts of third-order analog beat distortion,
has been developed in many other or CTB.
applications, at the expense of some cost, of
course. It can be costly in particular because SNR
wideband and low noise are competing
elements in frequency synthesis – high-Q The use of analog reclamation to free up
oscillators characterized by low noise do not bandwidth for digital channels has obvious
tune very far. As a result the designs often and well-understood implications for adding
involve multiple oscillators, frequency service value to the channel line-up. The
multipliers, switches, and more complex exposed bandwidth allows for increasing HD
direct digital and PLL-synthesis techniques content, more digital channels, more niche
compared to what is done today. The bottom channels, and more bandwidth for data
line is if there was a will for improved phase services. Also, because digital channels
noise, there is a way. typically run at lower power by 6-10 dB
relative to analog, replacement of analog
channels with digital results in an increase in
headroom that can be exploited for SNR
purposes while maintaining the same total RF
load on the optics or the RF actives. Table 1
shows what this headroom means in terms of

the RF power load when compared to a It would be ideal if there were 6 dB
reference power load, under different RF tilts, available to increase the digital levels.
for the case of occupied forward path However, it is not critical that there is not, as
bandwidth to 870 MHz. Table 2 shows the the existing modulations do run with
same, but for 1 GHz of loaded bandwidth. significant link margin under typical cable
Table 1 - Power Loading Effects of Analog Reclamation - 870 MHz
Channel Uptilt @ 870 MHz
Flat 12 dB 14 dB
Delta Ref QAM Increase Delta Ref QAM Increase Delta Ref QAM Increase
79 Analog Ref Load --- Ref Load --- Ref Load ---
59 Analog -0.7 2.5 -1.0 1.5 -0.9 1.5
39 Analog -1.6 3.5 -1.7 2.5 -1.6 2.0
30 Analog -2.1 4.0 -2.0 2.5 -1.9 2.5
All Digital -4.5 4.5 -2.8 3.0 -2.5 2.5
Table 2 - Power Loading Effects of Analog Reclamation - 1000 MHz
Channel Uptilt @ 870 MHz
Flat 12 dB 14 dB
Delta Ref QAM Increase Delta Ref QAM Increase Delta Ref QAM Increase
79 Analog Ref Load --- Ref Load --- Ref Load ---
59 Analog -0.7 2.0 -0.7 1.0 -0.6 1.0
39 Analog -1.5 3.0 -1.2 1.5 -1.1 1.5
30 Analog -2.0 3.5 -1.4 1.5 -1.2 1.5
All Digital -4.1 4.0 -1.9 2.0 -1.5 1.7
Note: For comparison of Tables 1 and 2, the plant conditions. For example, 45 dB of
delta of [1 GHz Ref Load – 870 MHz Ref analog CNR at end of line at a relative back-
Load] is as follows: off of 6 dB it delivers a 39 dB digital SNR.
Flat: 0.25 dB That’s 11 dB of margin for 64-QAM and 5 dB
12 dB Uptilt: 1.27 dB of margin to 256-QAM, not including coding
14 dB Uptilt: 1.56 dB gain. When considering coding gain, even in
the 256-QAM case there is substantial SNR
In the table, the left hand column for each margin to work with. As previously
case – flat load, 12 dB tilt, 14 dB tilt – described, we have the origins of the CATV
represents the decrease in total RF load network as an analog video network to thank
compared to the 79-analog channel reference. for this good fortune, as well as the fact that
In every case, the digital carriers run at -6 dB its origins were as an RF-only plant. Prior to
relative level. The right column for each case fiber optic carriage, very long amplifier
represents how much more power could be cascades were required, necessitating decent
allocated to each digital carrier in order to noise properties from the broadband
maintain very close (within 0.2 dB) to the amplifiers.
same total RF power load. This is the added
headroom available for SNR that was In the tables above, the flat case represents
previously mentioned. These seemingly small the effect on the optical loading of the analog
available dB become important as we consider reclamation process. It is apparent than in a
increased modulation order, as the move from mixed multiplex from a single transmitter, the
256-QAM to 1024-QAM comes with a 6 dB relative digital level allowable will be driven
SNR penalty. We will also see how fractions by limitations of the RF plant, where less
of dBs matter in cases of SNR thresholds vs QAM power can be allocated because of the
HFC conditions evaluated later. applied tilt. However, there is nonetheless
still headroom that can be exploited by

increasing the total power of the analog plus the digital channels are set at -3 dBc, while in
digital multiplex, gaining SNR for all the right-hand columns, the upper half of the
channels. Alternatively, Table 1 and Table 2 digital channels are set at -3 dBc. It is clear
show that 1-3 dB of increased QAM power from this balance of QAM power that even in
can be applied only to the digital part of the the worst case there is a very minor impact on
band of the tilted RF outputs, providing some total RF load of < 2 dB using split digital band
mitigation against the 6 dB increased SNR loading, and essentially no net power increase
requirement and subsequent lost margin. It to the RF load when the lower half of the
will become clear later as we discuss the digital band is used for 1024-QAM.
effect of cascade depth that these seemingly However, it is this part of the band that sees
incremental dBs can have an impact on the the highest level of CTB beat accumulation,
ability of the architecture to ensure the desired so there is an inherent trade-off between the
minimum 1024-QAM SNR objective is met. two.
Consider simply increasing QAM levels to Note that SNR discussion above refers to
-3 dB on the 79-channel cascade. This nets a the effect of a power increase on a fixed
1.5-2.0 dB total power load increase, which is thermal noise floor. However, there is a
enough to noticeably impact distortions if distortion component referred to as composite
nothing else is changed (no cascade intermodulation noise (CIN) that appears like
shortening). Now consider Table 3. Here, it a thermal noise floor, but is in fact a result of
is postulated that 1024-QAM is run at -3 dBc, distortion products with a digital carrier
with another set of digital channels (half of component. As digital carriers increase, there
them) remaining at -6 dBc. is more digital contribution to create CIN. In
system analysis, the CIN parameter combines
We saw in Tables 1 and 2 that the analog with the thermal noise floor to create the
reclamation effort can yield possible QAM parameter known as Composite Carrier-to-
level increases for the same power load, but Noise, or CCN. CIN looks like thermal noise,
with this typically being less than 3 dB. and has effects like thermal noise, and is
Nonetheless, to win back margin lost to SNR, mathematically treated like thermal noise in
in Table 3 we assume that 1024-QAM is run the calculation of SNR. However, it
at the -3 dBc level, and that the digital load aggregates as a distortion would aggregate
that is split into -6 dBc and -3 dBc segments. through a cascade, dominated by third order
This approach could limit the 1.5-2.0 dB effects. It can be easily isolated in system
additional power loading effects of having all cascade tests, and the CIN and AWGN
digital channels increase to -3 dB, but also components of CCN identified. However, the
provide that extra SNR boost to a set of 1024- CIN effect has not been included into the
QAM channels. In one case, the lower half of model at this point.
Table 3 - Split Loading for 1024-QAM - 1000 MHz In our examples,
14 dB Channel Uptilt @ 870 MHz performance analysis for
Lower Digital @ -3 dBc Upper Digital @ -3 dBc nominal and increased output
Delta Ref QAM Increase Delta Ref QAM Increase
levels lead to the conclusion
79 Analog 0.7 --- 1.7 --- that CIN contribution is
59 Analog 0.1 --- 1.4 ---
always smaller than the
39 Analog -0.4 0.5 1.0 ---
30 Analog -0.6 0.5 1.2 --- thermal noise contribution,
All Digital -1.0 1.0 1.1 --- and in some cases negligibly

so. Reduction occurs as expected as the into the analog band, and there is about a 1 dB
cascade shortens. During analog reclamation, difference in the worst case part of the digital
adding digital carriers adds more potential band – near the analog crossover. This
CIN contributors as described above. difference increases slightly as the total
However, the analog carriers drive the highest number of analog channels is reduced.
CIN3, generated as (2A+D), and the removal
of the highest level analog carriers more than The CTB beat count impact on the optical
offsets this increase. Nonetheless, to consider link is significant when the number of analog
the impact of a CIN effect, we evaluate the drops to its minimum. However, typically the
SNR threshold developed for the cascade over optical link of an HFC cascade will dominate
a range of link noise performance. From these the SNR aspect, but not be the primary driver
curves, the impact of allowing for increased of distortion. That would be the RF cascade,
CIN degradation of SNR can be observed by including the RF drive from the node. The
considering the plotted SNR and adding 2-3 effect on third-order digital distortion, CIN3,
dB maximum degradation. which adds to the optical link’s noise
characteristics, is less significant. CIN3 is the
Distortion dominant digital distortion that accumulates in
mixed cable multiplexes.
In addition to its positive effects on digital
SNR, analog reclamation offers benefits in the Finally, note that the CIN3 improvement of
distortion domain as well. Some modeled analog reclamation can be nearly washed out
relationships are shown and discussed below. with the addition of the digital band to 1 GHz
on the optical link (flat loading). Obviously,
Consider a load of 79 analog, with a digital this is just more digital spectrum to add to the
load to 870 MHz as a reference for this nonlinear mix, at the highest amplitude levels,
example. Observing an 870 MHz upper band so this is not unexpected.
will allow us to demonstrate the primary
impact of constant loading when bandwidth Now consider the tilted (in this case 12 dB
extension is also considered. to 870 MHz) examples. The impact of
removing analog is magnified, because the
Now observe Table 4, and the flatly loaded analog channels being removed are the highest
(optic) cases first. As can be seen, as a 79 level channels, and thus the strongest
channel analog load is reduced, the relative contributors to CTB. This is quantified as up
effect on worst case CTB is for it to drop. It to 15 dB improved worst case CTB for
also moves lower in frequency. However, the minimal analog channel count, a significant
digital channel band extends lower as well, gain. We can observe in Table 5 that nearly
and thus this movement of worst case CTB this full gain is achieved at the node output –
does not provide much assistance to the QAM the optical link and one RF amplification in
channel. The actual CTB peak extends just the node (N+0 case).
Table 4 - 3rd Order Distortion vs Analog Channel Count
CTB CIN3
Analog Channels Flat (Optics) Tilted (RF) Flat (Optics) Tilted (RF)
79 0 dB (Ref) 0 dB (Ref) 0 dB (Ref) 0 dB (Ref)
59 -2 dB -5 dB -1 dB -2 dB
30 -8 dB -15 dB -2 dB -9 dB
Extend Digital to 1 GHz Add 1.5 dB Add 3.5 dB

On the CIN3 side, we see more of the that these dBs can also be subject to variation
effect of removing the high analog carriers for across product types and models. The intent
the tilted multiplex, which drive the highest was to deliver some practical conclusions
digital distortion levels as (2A + D). While rather than accumulate worst case
the reclamation adds digital channels, it adds assumptions that deliver a skewed result.
them at lower power than the existing
channels on the uptilt, and, combined with the Table 5 shows modeled performance for a
reduction of highest analog levels, net gains in particular 1310 nm link, followed in the N+6
CIN3 are achieved. case by two amplifier types in a 2+4
configuration. The simulation uses
The extension of the band to 1 GHz has a mathematical models derived through
magnified effect in a tilt versus flat case, hardware verification of the particular laser
obviously a result of the digital channels now and receiver family, and of the individual
being relatively higher when installed at the amplifier characteristics. Of course,
high end of the band. parameters of different lasers, receiver,
amplifiers, etc can vary across product
Cascade Shortening families, vendors, implementation, etc., so a
“typical” arrangement using nominal levels
It is of curse not a secret that each and link lengths were chosen for a reference
additional amplifier placed beyond the node point. The data underscores the impact on
creates degradation. There are very well- noise and distortion of decreasing analog
understood rules in the RF world for cascaded channel loads, and shorter RF cascades.
degradations of equivalently performing
amplifiers: Again, CCN represents Composite Carrier-
to-Noise – a combination of the CNR or SNR
SNR → 10 Log N (optical link dominated) and digital
CSO → 15 Log N distortions.
CTB → 20 Log N
CIN → 20 Log N
Table 5 - Noise and Distortion @ 550 MHz vs Analog Channel Count
There are CCN CTB CSO
caveats to Analog Channels N+6 N+0 N+6 N+0 N+6 N+0
these rules 79 48 51 58 70 56 64
that have to 59 48 52 60 70 59 65
do with the 30 48 52 68 74 67 70
mix of
optical and RF distortions, noise and digital This table shows important trends in two
distortion, and the active technology used. It important directions that we will further
is with these rules and caveats, adjusted by quantify to develop threshold rules for 1024-
offsets associated with the variables QAM. Along the rows, the advantages of
introduced in the paper – primarily channel going from a 6-deep RF cascade to an N+0
loads and cascade depths, that we can estimate architecture, with the node as the last active,
distortion effects and the likelihood it will are on display for each parameter. This is also
impact 1024-QAM. While we will use sometimes called Fiber-to-the-Last-Active
particular numerical examples that represent (FTLA). Gains in both noise and distortion
typical characteristics, it is certainly the case are clear, both of which more ably support the

ability to handle higher order modulation such TEST RESULTS
as 1024-QAM. Moving down columns, the
benefits of doing analog reclamation becomes Based on field testing previously
clear. described, and lab performance
characterization testing, it has been clear for
From the perspective of noise, the some time that the output right at the end of
improvements available are of most the optical link (an N+0 link) was well-suited
significance in the shortening of the cascade, to 1024-QAM. The SNR and distortions at
and the impact of the lack of accumulation of this point are as good as they are going to get
amplifier noise. There is 3-4 dB additional n the cable, and the cascaded RF link sitting
optical SNR available relative to a typical between the node output and the STB can only
line-up and cascade depth of today. When serve to degrade this. It is this fact that
coupled with possible loading adjustments suggests that the N+0 architecture is an ideal
with the larger digital tier and added distortion HFC evolution supporting 1024-QAM, but
headroom available (from distortion also that any cascade shortening works in
improvements), there are ways to claw back favor of the higher order scheme. Testing was
close to the 6 dB of SNR – the amount of performed to verify this prediction.
increased sensitivity of 1024-QAM compared
to 256-QAM. Optical Link Testing
For both cascade shortening and Consider Table 6, which shows the results
minimizing the analog loading, Table 5 shows of typical 1310 nm optics through 20 km of
large available distortion gains. These analog fiber, with 79-channel analog loading and
beat distortions look like narrowband digital loading to 1 GHz. The QAM level is
interferers, but with random amplitude and ranged over -4 dBc to -8 dBc, and the 1024-
phase properties and a measurable bandwidth QAM channel inserted in several locations in
which, as previously described, makes them the loaded digital band, where it’s MER and
more difficult to cancel compared to static pre-RS FEC BER are measured.
CW interference. Cascade shortening reduces
the 20LogN accumulation of CTB distortion It is apparent, in particular given the -4 dBc
through the RF amplifiers, and these are the data and analog CNRs in the low 50 dB range,
HFC elements that tend to drive link that there is a measurement floor associated
distortion. Analog reclamation reduces the with the link, reflected in both BER and MER.
amount of beats entered
into the mixing process Table 6 – 1024-QAM Performance on Fully Loaded Optical Link
through which the 1024-QAM Carrier Frequency
CTBs and CSOs 603 MHz 747 MHz 855 MHz
accumulate. Both are QAM @ -4 dB to Analog MER 39.6 39.2 38.9
aided by changes in BER 6.1E-08 1.12E-07 3.76E-07
the directions shown in QAM @ -6 dB to Analog MER 39.0 38.9 38.6

BER 1.5E-07 2.6E-07 2.5E-07
Table 5. We will QAM @ -8 dB to Analog MER 38.3 38.2 37.7
discuss what these BER 4.30E-07 2.02E-06 3.48E-06
results mean more
quantifiably in a later section. The digital SNR in the -4 dBc cases should all
be in the high 40 dB’s, and, even with
reasonable implementation losses, should be

error free or nearly so. Since they are not, amplifier to combine with this optical SNR.
there is evidence of an impairment affecting Furthermore, third-order analog distortion
the result, with possibilities in this setup being cascades as 20LogN to combine with the node
phase noise, beat distortions of 79 channels, output. It is these cascaded impacts that eat
I/Q imbalance (modem implementation loss), into margin for successful 1024-QAM.
STB-limiting noise figure, or a combination
thereof. However, in all cases in Table 4, the RF Testing
post-FEC BER was zero – it was error-free at
the output. To investigate thresholds of distortion
interference as a function of SNR, a laboratory
Not shown was another set of data run with test bed was put together, as shown in Figure
1550 nm optics. The essential result of that 6. While not a full cascade, power adjustment
link was that all pre-FEC BERs were in the variables in the setup allowed it to
1e-6 order of magnitude. Based on the above accommodate a range of distortion amplitudes
data, and similar noise and analog distortion at different SNRs.
numbers, it is likely that this result stems from
the effect of optical dispersion on the 1024- The RF testing performed include varying
QAM channel when operating at 1550 nm. In SNR under conditions of CW and live video
this case also, however, operation post-FEC CTB interference (NCTA practices define
was error free. CTB with CW carriers), across multiple
frequencies in the digital band (three chosen),
This 1310 nm optical link performance is and observing the pre-FEC and post-FEC
encouraging in that it was completely error performance. Table 7 shows the results of this
free post-FEC, but with minor flags on taking testing. Note that all dB values are relative to
pre-errors even at very high SNR (BER a digital QAM signal, not an analog reference.
flooring). It was also completely expected Also, note that the “Pre-FEC Error threshold”
that performance would be good, as the node used was 1E-7. Since the receiver bottoms
output – essentially the N+0 case, eliminates out around 1E-8 (slightly below), this was
any cascade effects. With noise performance selected as a point at which non-floor related
at the node output in the low 50 dB range, and errors are being taken. The “Post-FEC Error
distortion outputs in the high 60’s, conditions Threshold” is simply the point at which that
for 1024-QAM appear quite good. Of course, metric, representing data after it has been
quantifying what distortion levels are through its full clean-up, is still non-zero. It
acceptable is part of the objective of this may be the point at which an operator decides
paper. Nonetheless, given these excellent RF that CTB beyond this is, or some acceptable
characteristics at the node output, coupled dB higher, is consuming enough of the FEC
with previously described characteristics of budget, which then becomes unavailable for
the tuner performance, it points to phase noise other impairments. The post-FEC 1E-6
flooring and modem implementation loss threshold is logged as a reference point. Post-
(transmit fidelity) as the most likely sources of FEC curves tend to be very steep, and once
the pre-FEC BER floor. this level of error accumulation begins, the
system is not very far from operating with
While this N+0 scenario presents a clear unacceptable error counts or not at all. It
case as a sound environment for 1024-QAM, also represents a point of visual threshold-of-
unfortunately, as the RF cascade is visibility (TOV) in some circles. Finally, the
lengthened, noise degrades at 10LogN per

Table 7 - Error Threshold of Interference on Most Sensitive Channel
CW Interference CTB Interference
Pre-FEC Error Post-FEC Error Pre-FEC Error Post-FEC Error Post FEC Post FEC
SNR Threshold Threshold Threshold Threshold > 1E-6 Broken
50 dB 34 dB 33 dB 55 dB 55 dB 55 dB 45 dB
45 dB 35 dB 33 dB 55 dB 60 dB 55 dB 46 dB
40 dB 36 dB 34 dB 60 dB 60 dB 55 dB 49 dB
37 dB 38 dB 37 dB 60 dB 60 dB 55 dB 50 dB
point at which the system breaks under the according to the safe assumption that
weight of excessive CTB is identified. distortion is normally assumed to degrade.
Note CTB referred to in Table 7 is as it is Table 7 indeed proves a postulate made

measured in this setup, not per referenced here and originated in the 2002 paper – that
definition and typical practices. That is, rather mitigation of CTB is more difficult that CW
than CW beats under an analog virtual carrier, interference. Again, CTB’s random properties
it is live video beats underneath a QAM challenge the equalizer. In addition, CTB is
channel power. Beat distortion from live an averaging measurement of a noise-like
video beats are typically assumed 8-12 dB waveform. Thus, it has a high peak-to-
below “CW” CTB due to the downward average ratio, with 12-14 dB often used for
analog modulation, with most references practical noise. Theoretical analysis would
identifying 12 dB of “help” by turning on live predict higher, while measured values tend to
video. And, QAM power is, of course, this range or lower. Meanwhile, the peak-to-
derated by some amount to an analog average of a CW signal is 3 dB, and, it is of
reference – typically 6-10 dB, but as we have constant envelope amplitude.
been arguing here, this could perhaps be only
3-4 dB for 1024-QAM. Recall, the prediction that a threat could
exist for 1024-QAM was based on the CW
Also note that in this table, the interference case for 256-QAM, coupled with
performance noted is the worst performing of these factors as follows:
the three frequencies across the set of cascade
measurements taken. These are important CTB min (FCC) = 53 dBc
nuances, because the average measurement CTB pk ~ 40 dB
across the frequency set was typically 3-5 dB CTB pk, relative to QAM ~ 34 dBc
better than the worst measurement. And,
certain cascade combinations actually showed In 2002, 256-QAM showed BER
slightly better CTB at N+1, for example, sensitivity, pre and post-FEC, for S/I ratios in
compared to N+0. This phenomenon has been this mid-30’s range. It is encouraging to note
observed before in sample test of amplifiers that in testing, now seven years hence, that the
employing E-GaAs technology, which has receiver evolution has led to narrowband
different distortion-generating mechanisms ingress cancellation of 1024-QAM that is
than, for example, Silicon. The effect has comparable to what once was recorded to be
been attributed to this reality of the device the limits for 256-QAM. In [5], an ingress
physics, and random phasing relationships. cancellation example is shown that displays
While this has encouraging benefits, it is not roughly 30 dB of interference mitigation of a
considered a system “rule” that these static interferer in a noise-free environment
amplifiers are immune to degradation in a (>50 dB SNR).
general sense, and system design goes

From this simple example, it is clear by 5. The effect of typical gain and noise
comparing this 34 dBc value to the left-hand differences do not substantially changing the
side of Table 7, that the CTB impairment’s qualitative conclusion to be drawn. The net
peak is achieving levels in the range that CW noise figure impact of the gain variation of,
carrier amplitudes were seen to be of concern, for example, splitting the gains between trunk
but now for the more sensitive 1024-QAM amplification and line extenders, is a dB or so
case. It is this relationship that led to the in this N+6 case. This is small, but, as we will
conclusion that plants that were ok (but not point out below, adjusting the RF noise
great) could be a concern for 256-QAM at the contribution by a dB or two could make a
time, and thus even average systems could difference in some regions of the curves.
create issues for 1024-QAM. Now, however, Thus, it is desirable not to be operating in such
the 1024-QAM CW thresholds are much regions.
improved, comparable to prior 256-QAM
thresholds. This is much better situation than Each curve in Figure 6 and 7 represents a
had the S/I sensitivity been as previous for different value of SNR as set by the STB
256-QAM, knowing the added sensitivity alone, associated with the noise figure and
created by 1024-QAM. digital level (derated from analog) at its input.
Note from the figures that there is a wide
Table 7 reveals some possible threshold range of SNR combinations that essentially
pairing of noise and distortion for a quality offer no practical limit to RF cascade depth as
1024-QAM link. Again, note that these are it relates to noise degradation. Clearly,
the most sensitive readings recorded, with tolerating a 38 dB link requirement, which
averages across the board running 3-5 dB less would be relying very strongly on FEC for
sensitive. Nonetheless, let’s point out the key just thermal noise aspects, provides a
result: the 1024-QAM receiver saw error comfortable range of noise performance of the
degradation under some CTBs that are not link contributors to work with. However, the
inordinately high levels of distortion. range still includes conditions that could lead
to a sharp reduction in cascade acceptable.
HFC PERFORMANCE THRESHOLDS From a sensitivity analysis standpoint, such
conditions hinge on small dBs and even
SNR fractions thereof. This makes it more valuable
to be able to earn back SNR in the analog
Taking a look at SNR, we can draw some reclamation process by taking advantage of
conclusions about the depth of the RF cascade the load being lightened by the expanded
as a function of the expectation from the STB digital, or by taking advantage of headroom in
noise figure contribution, and the optical CCN distortion offered by the dwindling beat
contribution, as a function of a pre-defined counts. To ensure operation above the desired
overall SNR link objective. Consider Figure 7 threshold, it is not practical to rely on the
and 8. These figures illustrate the nature of outdoor plant to maintain key noise and
small SNR margin increments in their effect distortion parameters to within fractions of
on the tolerable cascade depth over a range of dBs over time and temperature, so operating
optical SNRs, for two chosen link in these sensitive regions should be avoided.
requirements: 38 dB and 40 dB. For Again, the 40 dB SNR value represents
simplicity, we have assumed equivalent RF approximately the 1E-8 point for uncorrected
amplifier noise and similar gain, and based 1024 QAM – a noble objective, setting up the
these estimates on the results shared in Table link budget to be acceptable without FEC

under the otherwise ideal conditions, and optical link to be the node receiver output,
reserving the FEC to deal exclusively with the where the multiplex is flat, and bundle the RF
set of channel issue. However, this may not stages of the node into the RF cascade for the
be practical without lower noise contributors analysis. Mathematically, then, our N+x case
on the CPE side (STB). will be modeled as N+x+1. We will assume
actives of the same characteristics per the
CTB Distortion original assumption across the link, including
the node RF, established at some nominal
We will make the similar simplifying reference output level. Again, this is a
assumption on the RF cascade as in the noise simplifying assumption, and has single digit
case – similar CTB performance and gains IP3 effects for typical ranges of parameters.
across the actives at nominal RF output levels. However, it allows creation of a model that
In this case, splitting the cascade as above provides a feel for some basic relationships.
among some typical performing amplifier Of course, the model can be extended to
types results in about a 2 dB increase (lower consider any number of individual
distortion – thus a conservative evaluation) in contributors. Note that there is not a simple
the third order compression point – the metric formulaic relationship for the RF cascade, as
that sets CTB. Again, the point here is to the tilt variation and channel maps both
recognize that around the data presented contributing to CTB variation.
below, consider that the exact CTB may not
be easily counted on to the dB level, and Figure 9 shows a sample of an analog beat
therefore understanding how performance map for 79 analog channels on a 12 dB tilt to
behaves over a range around the anticipated 870 MHz. Tools such as this are used to
CTB is important to ensure a robust actual calculate the impact of varying channel line-
system. The CPE end of the contribution is ups on relative distortion level (tone another
extracted from the analysis, because the and versus frequency) with results as depicted
threshold results will be based on measured in the Table 4 examples. Note that the peak
data that includes that receiver contribution distortion is third-order (CTB), and that it
when BER is measured, but the recorded CTB peaks just below the start of the digital band.
value is that delivered to the receiver input. As previously described, the worst digital
This leaves us with two contributors – the band CTB is about 1 dB lower than this peak.
optical link and the RF cascade – and the Figure 10 shows the case of flat loading, such
variables that modify the contributions to as would be carried on the optical link. In this
make the assessment. case, the maximum CTB beat location shifts
lower, and the difference relative to the worst
For the optical link, the beat map effect of digital band is closer to 2 dB. These results
flat loading and a fixed transmitter RF load are used in lieu of an available formulaic
can be relatively easily quantified representation to evaluate the reclamation
mathematically and applied as such during process on the titled multiplex distortions.
analog reclamation. While the link distortion
is dominated by the RF cascade, the optics We now evaluate the ability of the HFC
will be included because, as the cascade cascade to support a CTB threshold derived
shortens, it becomes more relevant, in from the data provided in Table 7. We will
particular with the magnified gains of the evaluate over a range of given RF amplifier
tilted channel load on CTB, compared to the CTBs specified at some typical RF output
flat loading gains. Thus, we will consider the levels to which these CTB values apply, and

evaluate for the varying analog channel counts analog channels from a 79-channel system and
used above. Two performance thresholds are going through two stages of analog
used, each of which is derived by using the reclamation.
measured case of -60 dBc interference as the
objective. The -60 dBc measurement In Figure 11, it is clear that relying on a 12
threshold was selected because dB CTB drop with live video and boosting the
QAM power by 3 dB, there is virtually no
• We care most about post-FEC practical cascade limitations for the cases of
performance analog reclamation employed. In the 79-
• Post-FEC above 1E-6 is too high, channel case, only poor or malfunctioning RF
beyond a video TOV, and crashing amplifiers appear to be a threat to the 1024-
occurs shortly thereafter on a Post- QAM system. However, note that there is no
FEC curve loading impact assumed of the added QAM
• It correlates with the range of the level. We know from prior discussion that
SNRs we would expect to see this amount of increase in relative level, when
delivered to the home applied to the entire digital load, can nudge
the total RF level up measurably, increasing
From the -60 dBc measured value of live distortion. For the cases employing analog
video underneath a QAM channel, we make reclamation, it has been shown in prior
the following adjustments to reference that sections that increased QAM level can be
value to a standard CTB measurement. A obtained with no effect on total RF load, but
standard measurement would be referenced to there is not enough headroom for a full 3 dB
an analog carrier level, and it would be using increase. There are also techniques such as
CW carriers as load signals. The net results of the split loading of 256-QAM and 1024-QAM
the former is a 6 dB offset (typical), and we that could limit the loading impact for the
will also evaluate a 3 dB offset to consider the existing 79-channel case.
case where more power is given to the QAM
load to support the higher SNR requirements For Figure 11, then, the assumption of
of 1024-QAM. When considering live video maintaining the total RF load means either the
carriers are the source of the distortion beats nominal power load is referenced to the
and not CW carriers, we have another offset to multiplex as described here (it includes
account for (in the opposite direction – live margin headroom for this small increase). Or,
video is more benign) in the range of 8-12 dB. only a subset of the channels employs 1024-
We will evaluate the endpoints of these cases QAM (such as the HD tier for more per-QAM
starting with a -60 dBc measurement efficiency, or the HSD tier). Or, analog levels
threshold, of which we will no longer refer to are compensated downward. In any case, we
as “CTB” given the assumptions that the term will remove the constraint regarding loading
entails. The endpoints are: in Figure 14 and recalculate this case.
CTBth, min = 60 + 6 – 8 = 58 dBc Now consider Figure 12, where the

CTBth, max = 60 + 3 – 12 = 51 dBc CTBth,min = 58 is used as the guideline. It is
clear the power of complete analog
Figures 11-14 display the results for these reclamation to support virtually any practical
cases for the depth of RF cascade, across a RF cascade depth in this case also. However,
range of RF amplifier CTBs, or net effective it also becomes clear how for 79-channel
CTB performance, as a function of the number systems and 59-channel systems, some

limitations are beginning to appear. This is SUMMARY
consistent with the expectation that the march
to shorter and short cascades supports the The use of 256-QAM is commonplace for
move to 1024-QAM, albeit not as powerfully DTV and HSD, and operators have learned
as what analog reclamation creates in margin. what it takes to implement it successfully on
Figure 13 removes the 30-channel analog mixed multiplexes. Its performance relative
case, where analog reclamation is complete or to 64-QAM is well-understood, and the
nearly so, thus adding granularity to the figure additional 33% bandwidth efficiency can be
such that reasonable cascade depths can be achieved now with minimal cost and pain.
better quantified as the requirement for them There is 25% more bandwidth efficiency
becomes shorter and shorter. Note that by the available if 1024-QAM is deployed. Modem
nature of the requirement chosen compared to ICs exist, and this paper sets out to get a feel
the RF CTB range, the curves are unable to for the pain part of the equation,
cross the x-axis, which represent N+0 (i.e. the understanding that this scheme is yet another
combined CTB of the optics and the CTB level more sensitive to impairments than 256-
range of RF used is better than the worst case QAM, and rivaling in many ways the
requirement of 58 dBc). sensitivity of analog video itself.
Finally, re-consider the CTBth,max of 51 The good news is that interference analysis
dBc under the conditions that the increase in conducted in 2002 based on findings
QAM power to a -3 dB derate applies to the extrapolated from 256-QAM performance of
whole load, and that the whole load is 1 GHz that era now seems to have been addressed in
of bandwidth. The total RF load then is the receivers available today. Narrowband
increased about 2 dB, increasing distortions, cancellation performance seems robust, and a
and in particular third order distortions by 4 measurable improvement over 2002. The
dB. This case is shown in Figure 14. Note performance against narrowband interference
the shift of a tolerable N+5 cascade to now today for 1024-QAM is roughly what it once
what might be an N+3 to maintain was for 256-QAM – the level of interference
performance against this threshold if not that used to disturb 256-QAM performs to
paying attention to the total power loading. roughly the same error rate now, but on the
Note also that this assumes a starting derate of more advanced modulation. This is very
QAM at -6 dBc and increasing to -3 dBc. If important for analog beat distortion
the starting derate is lower, than the impact on mitigation, which appears as noise modulated
total power is smaller. narrowband interference. The random
element of it makes the receiver have to work
With the curves above, operators are able harder, and as a result it is less capable of
to evaluate the likelihood of 1024-QAM can attenuating the interference. The improved
run successfully on the plant around some performance makes what once looked to be a
basic assumptions. Furthermore, a model has potentially troublesome CTB problem now a
been put in place that allows evaluation on a manageable one with some constraints on the
case-by-case basis for given combinations of HFC link, with performance characteristics
equipment, channel plans, and alignment. plotted against CTB and channel line-up
variations.
In terms of noise performance, not much

appears to have changed, and the system

display a pre-FEC floor likely associated with BIBLIOGRAPHY
modem and RF imperfections, such as phase
noise. The system relies on FEC to obtain [1] Howald, Robert, QAM Bulks Up Once
error-free performance under typical plant Again: Modulation to the Power of Ten,
conditions because of the 40 dB pre-FEC SNR SCTE Cable-Tec Expo, June 5-7, 2002, San
threshold for 1E-8 performance. For 64-QAM Antonio, Tx.
and 256-QAM, significant link margin existed
under normal conditions to set reasonable [2] Howard, Daniel et al, Methods to Increase
thresholds to qualify link acceptability. Bandwidth Utilization in DOCSIS 2.0
Because of the increased SNR requirement of Systems, NCTA Show, June 8-11, 2003,
1024-QAM, it is up to the operator to make a Chicago, Il.
different type of determination with respect to
link margin and acceptable threshold under [3] Katznelson, R., Delivering on the 256-
normal plant conditions. Acceptable margin QAM Promise: Adopting Aggregate
will have to be looked at differently for 1024- Degradation Measures for QAM Transmitters,
QAM. The figures in the paper are meant to SCTE Cable-Tec Expo, June 5-7, 2002, San
help operators see the trade space they are Antonio, Tx.
working with for SNR in that respect.
[4] Stoneback, D., Howald, R., Brophy, T. and
Given the results shown here, an awareness Sniezko, O., Distortion Beat Characterization
of the key drivers to link performance, and a and the Impact on QAM BER Performance,
modeling approach that can be used to assess NCTA Show, Chicago, Illinois, June 13-16,
HFC readiness to accept 1024-QAM signals, 1999.
operators can start netting themselves that
extra 25% of bandwidth efficiency that is [5] Tan, Loke Kun et al, A 70 Mb/s Variable-
currently going unused in already occupied Rate 1024-QAM Cable Receiver IC with
spectrum. That is, all the tools and knowledge Integrated 10-b ADC and FEC Decoder,
are in place to start getting new Megabits from IEEE Journal of Solid-State Circuits, Dec. 12,
the same Megahertz. 1998
ACKNOWLEDGEMENTS Robert L. Howald, Ph.D.

Customer Systems Architect
The authors wish to thank Rudy Menna and Technology Office
Chuck Moore for their valuable contributions Motorola Home & Networks Mobility
to this paper. 101 Tournament Drive
Horsham, Pa. 19044
407-242-9977
[email protected]

256-QAM @ 34 dB SNR
16 16
14
12
10
2
Im( Si)
0
10
12
14
− 16
16
16 14 12 10 8 6 4 2 0 2 4 6 8 10 12 14 16
− 16 Re( Si) 16
Figure 1 – 256-QAM @ 34 dB SNR
1024-QAM @ 40 dB SNR
32 32
30
28
26
24
22
20
18
16
14
12
10
8
6
4
Im( Si)
2
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
− 32 30
32
32 30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
− 32 Re( Si) 32
Figure 2 – 1024-QAM @ 40 dB SNR

1024-QAM @ 35 dB S/I
32 32
30
28
26
24
22
20
18
16
14
12
10
8
6
4
Im( Si)
2
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
− 32 30
32
32 30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
− 32 Re( Si) 32
Figure 3 – 1024-QAM with 35 dB Signal-to-Interference (S/I)

1024-QAM @ 39 dB S/I
32 32
30
28
26
24
22
20
18
16
14
12
10
8
6
4
Im( Si)
2
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
− 32 30
32
32 30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
− 32 Re( Si) 32
Figure 4 – 1024-QAM with 40 dB SNR and 38 dB S/I

1024-QAM @ .25 deg rms phase error
32 32
30
28
26
24
22
20
18
16
14
12
10
8
6
4
Im( Si)
2
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
− 32 30
32
32 30 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
− 32 Re( Si) 32
Figure 5 – 1024-QAM with .25 deg rms Gaussian Phase Noise

Analog
Channel Optical TX Optical Node Amplifier
Pad
Load (LM1000) (SG4000) (MB100)
Source
Fiber Pad
1024-QAM Amplifier
Modulator (BLE)
Noise Pad
Source
Tunable
Var.
CW Gain Band Pass
Atten.
Generator Filter
Var.
Atten.
Spec. Analyzer
1024-QAM PC
(CTB/SNR/Power
Demodulator (BERT)
Monitor)
Figure 6 – Distortion Interference Test Bed

Cascade Depth vs. Optical and STB Noise
QAM Link Req't = 40 dB
50
45 STB SNR, dB
40
45
RF Cascade Depth
35
44.5
30
44
25
43.5
20
43
15
42.5
10
5
0
45 44.5 44 43.5 43 42.5 42 41.5 41 40.5
QAM Optical CCN
Figure 7 – Sensitivity of Cascade Depth to STB and Optical Noise Contributions for a 40 dB Link
Requirement
Cascade Depth vs. Optical and STB SNR

QAM Link Req't = 38 dB
50
45
STB SNR, dB
40
41.5
RF Cascade Depth
35
41
30
40.5
25
40
20
39.5
15
39
10
5
0
45 44.5 44 43.5 43 42.5 42 41.5 41 40.5 40
QAM Optical CCN
Figure 8 – Sensitivity of Cascade Depth to STB and Optical Noise Contributions for a 38 dB Link
Requirement

Analog Composite
12
9
6
3
0
-3
-6
-9
-12
-15
-18 Total CSO
-21 Total CTB
-24 Carriers
-27
-30
-33
-36
-39
-42
-45
-48
-51
-54
0 100 200 300 400 500 600 700 800 900 1000
Figure 9 – Beat Distortion Map of 79 Analog Channels on 12 dB Uptilt

Analog Composite
45
42
39
36
33
30
27
24
21 Total CSO
Total CTB
18 Carriers
15
12
-3
-6
0 100 200 300 400 500 600 700 800 900 1000
Figure 10 – Beat Distortion Map of 79 Analog Channels Flatly Loaded

Cascade Depth vs RF Performance & Analog Channel Count
CTBreq = 51 dB
50.00
45.00
40.00
RF Cascade Depth
35.00
30.00 79
25.00 59
20.00 30
15.00
10.00
5.00
0.00
80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65
RF Amplifier CTB
Figure 11 – Cascade Depth Thresholds vs. CTB Performance and Analog Channel Loading,
CTBmax = 51 dB

CTBreq = 58 dB
50.00
45.00
40.00
RF Cascade Depth
35.00
30.00 79
25.00 59
20.00 30
15.00
10.00
5.00
0.00
80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65
RF Amplifier CTB
CTBmax = 58 dB

CTBreq = 58 dB
20.00
18.00
16.00
RF Cascade Depth
14.00
12.00
79
10.00
59
8.00
6.00
4.00
2.00
0.00
80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65
RF Amplifier CTB
CTBmax = 58 dB, Expanded for 59 & 79-channels only

CTBreq = 51 dB
50.00
45.00
40.00
RF Cascade Depth
35.00
30.00 79
25.00 59
20.00 30
15.00
10.00
5.00
0.00
80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65
RF Amplifier CTB
CTBmax = 51 dB, QAM load @ -3 dBc

OPTIMIZING UPSTREAM THROUGHPUT USING
EQUALIZATION COEFFICIENT ANALYSIS
Rob Thompson
Chuck Moore
Jack Moran
Rob Howald
Motorola Home & Networks Mobility
Abstract helped DOCSIS upstream signals work in

environments where some test equipment will
The current economic climate and not even work.
increasingly competitive marketplace is
driving cable operators to optimize utilization While repairing damaged waveforms for
and efficiency of their upstream plant. detection is the primary Pre-Eq function, in
However, a problem exists when more doing so the algorithm learns important
bandwidth is utilized, as more efficient, but information about impairments in the plant.
also more sensitive, modulation schemes are Simple query and analysis of the Pre-Eq
leveraged. Impairments that had previously coefficients can reveal the dominant
gone unnoticed now present a significant impairment for which the CM-CMTS link is
barrier. An understanding of these compensating. Conclusions may be drawn
problematic impairments is critical in order regarding the impairment contributions,
to properly identify, estimate, isolate, and including Micro-Reflections, Group Delay
minimize their impact to support this critical Variation (GDV), and Amplitude Distortion
transition cable operator’s now face. A goal (AD). A plurality of DOCSIS terminal
of this paper will be to expose these key devices, such as CMs, may be engaged in an
impairments that must be managed, and effort to isolate and minimize a dominant
present a strategy for successful optimization impairment.
of bandwidth utilization and efficiency.
INTRODUCTION
One powerful tool that cable operators
have at their disposal is the transmit pre- This paper will discuss the identification
equalization feature (Pre-Eq) of DOCSIS. of return path impairments that can be derived
Pre-Eq is mandated by CableLabs for by a comprehensive understanding and careful
interoperability among vendors, and is analysis of the data obtained as part of the
required of all certified cable modems (CMs) Pre-Eq function. Conclusions can be drawn
and cable modem termination systems based upon coefficient analysis of the
(CMTSs). Its primary function is to mitigate DOCSIS 2.0 Pre-Eq settings that the CM-
impairments which severely degrade CMTS link collaborate to create. Limits at
upstream DOCSIS signals. It is not which Pre-Eq will fail will be provided for
uncommon for the Pre-Eq feature to achieve conditions of Micro-Reflection impairment
performance improvements such that Bit- only, and a combination of Micro-Reflection,
Error-Rate (BER) and Modulation-Error- GDV, and AD impairments. These failure
Ratio (MER) support near error-free thresholds identify at which point the channel
operation in an upstream environment that is too impaired to function properly, or with
would otherwise be unable to support the reasonable margin. Definition of
transmission at all. Additionally, Pre-Eq has “reasonable” margin varies among cable

operators, but identifying limits will empower can be mitigated by equalization include
cable operators with the tools necessary to Micro-Reflections, AD, GDV, and
ensure their return plant is within their Narrowband Interference. Amplitude
criteria. Distortion is also referred to as Attenuation
Distortion (AD). GDV, a phase-related
This paper will demonstrate a distortion, is also commonly referred to as
mathematical simulation, supported by Group Delay Distortion (GDD) or Envelope
laboratory verification, as well as live CATV Delay Distortion (EDD) in the telephony
plant tests, all of which were used to establish world.
and validate the proposed limits. A
systematic approach for isolating discussed It is important to recognize that impulse
impairments within a CATV plant will be and thermal noise will reduce the equalization
proposed. This extension of the Pre-Eq algorithm’s ability to mitigate these
information into the realm of HFC aforementioned impairments. These
maintenance enables cable operators to commonplace impairments introduce random
identify suspect CATV plant components that amplitude and possibly decision errors that
may be contributing to an impairment the equalization algorithm operates on.
problem, and thus take corrective action pro- Successful application of coefficient analysis
actively. assumes that the CATV plant meets sufficient
impulse and noise requirements necessary to
EQUALIZATION support a desired modulation level use.
However, the fact that some level of these
This section introduces the fundamental noise contributors exists will introduce some
digital communications receiver function of error into the calculated Pre-Eq solution.
equalization, and how the operation of the These issues can be overcome with
equalizer generates coefficient information modifications to the Pre-Eq update approach,
that can be used for diagnostic purposes. but as a closed loop system, it requires careful
While equalization is part of virtually all analysis of the plant conditions to ensure
modern telecommunications platforms, for stability of operation.
cable it is instrumental in proper return
operation for advanced DOCSIS systems. In Digital Signal Characteristics
order to offer higher data rates to its
subscribers in the competitive world of high- The digital signal characteristics used to
speed Internet access, operators must take generate all of the data presented in this paper
advantage of the throughput benefits gained are as shown in Table 1. The 6.4 MHz
from leveraging more complex digital DOCSIS channel bandwidth was chosen
modulation schemes, such as 32-QAM and because its widened bandwidth makes it more
64-QAM. Unfortunately, these schemes are sensitive to impairments. Of the two formats,
also considerably more sensitive to digital clearly 64-QAM is the more sensitive of the
communication channel impairments than the two. This actually makes the more robust
16-QAM channels (maximum) they are wideband 16-QAM particularly valuable as a
replacing in the return band. diagnostic tool in this mode.
HFC Impairments Table 1 - Signal Characteristics

16-QAM 64-QAM
Equalizers are very powerful tools within 5.12 Msym/sec
the digital receiver, and they can hide a lot of 6.4 MHz
sins. Relevant return path impairments that 20.48 Mbps 30.72 Mbps

RRC Matching (α = 0.25) diagnostic data. The more DOCSIS
Square Constellation (max/min = ±1v) compliant devices located throughout the
HFC plant, the better equipped the cable
Equalization Fundamentals operator is to use this information for
proactive maintenance, potentially eliminating
Equalization is the process by which a a source of subscriber calls.
digital communications receiver (in
collaboration with transmitter in the Pre-Eq DOCSIS 2.0 Pre-Eq uses twenty-four,
case) estimates the inverse of the digital symbol-spaced coefficients, also called taps.
communication channel response, Hc(f), and For example, a CMTS estimates the values of
applies it to the incoming signal. The process these coefficients and forwards this
is as illustrated by the following transfer information to a CM via station maintenance
function equation [1]. messages. These coefficients are used for
amplitude and phase correction over a twenty-
Eq 1 - Equalization Transfer Function four symbol period time window. There are
1 1 several reasons for having equalization
He ( f ) = = e − jϑc ( f )
Hc ( f ) Hc ( f ) functionality in the transmitter in addition to
the receiver that are out of scope of this paper.
By using the inverse function, But, in general, the more complex and diverse
equalization attempts restore a received digital a channel transfer function may be, the more
signal to an ideal response - canceling the well-suited it is to deploying equalization
impairment encountered in the digital resources at both ends. Cable’s return path
communication channel. In doing so, has a large range of possible frequency
equalization minimizes inter-symbol- response signatures even within a single RF
interference (ISI). ISI is the mechanism leg of an HFC plant.
whereby the frequency response distortions
previously noted (AD, GDV) cause in the Coefficient Interpretations
time domain, adjacent digital symbols to leak
into one another and cause interference most DOCSIS 2.0 Pre-Eq coefficients are a list
clearly observed on an eye diagram. of 24 complex values and may be interpreted
in multiple ways, as demonstrated in Figure 1
DOCSIS Equalization through Figure 5. The figures presented here
represent a digital communications channel
Pre-Eq is required for DOCSIS with negligible levels of impairment. These
compliance. The endpoints of the DOCSIS can be used as a baseline to aid in the
link (CM and CMTS) collaborate to converge impairment identification process.
upon an estimate of the communication
channel response and bias transmission in Figure 1 is the magnitude of the
order to, ideally, cancel any impairment that equalizer’s impulse response, |he(t)|. A single
may be present. While the Pre-Eq function line at Tap = 0, known as the main tap, on the
significantly hardens the link to impairments, x-axis is the ideal response. In that case,
as an added bonus there is much more whatever stimuli excite the channel is
information that can be mined to help the perfectly replicated. Calculation of impulse
operator more broadly. Understanding what response magnitude is based upon complex
communication channel impairments the tap values of DOCSIS 2.0 Pre-Eq and is
DOCSIS link is attempting to cancel offers shown in Eq 2.
the operator with extremely valuable

Eq 2 - Impulse Response Magnitude coefficients and is shown in Eq 4. Note that
he (t ) = 20 log10 he (t ) the equalizer’s amplitude response is ideally
constant throughout the channel’s bandwidth,
The main tap represents the desired which is the Fourier result of the ideal
symbol energy, while the remaining taps impulse in Figure 1.
represent negligible correction magnitudes of
Eq 4 – Amplitude Response
< -35 dBc. The small random magnitudes of FFT
the non-zero taps are primarily due to H e ( f ) = 20 log10 he (t ) ↔ H e ( f )
simulated system noise.
For coefficient analysis of multiple CMs, Figure 4 is the equalizer’s phase response,
it is helpful to break-down impulse response θe(f). Calculation for phase response is based
measurements into regions in which dominant upon a 1024-point, Fast Fourier Transform of
impairments will have the greatest impact. the DOCSIS 2.0 Pre-Eq coefficients and is
Numerically sorting on these impaired regions shown in Eq 5. Note that the equalizer’s
facilitates efficient organization of similarly phase response is ideally linear throughout the
impaired CM groups, and this can help in channel’s bandwidth.
diagnosing issues.
Eq 5 – Phase Response
θ e ( f ) = arg ⎡⎢he (t ) ↔ H e ( f )⎤⎥

FFT
Two important regions of the impulse
response to focus on are the post-tap region ⎣ ⎦
and the main tap region. Dominant micro-
reflections typically impact the post-tap Figure 5 is the equalizer’s Group Delay
region, which consists of tap 1 through tap 16. (GD) response, GDe(f). Calculation of GD is
Dominant AD and GDV typically impact the based upon the phase response and is shown
main-tap region, which consists of taps in Eq 6. Note that the equalizer’s GDV is
adjacent to the main tap, numbers -3, -2, -1, 1, approximately 6 nsec across the channel’s
2, and 3. bandwidth. Group delay is anther way of
describing the phase characteristics, but in a
Figure 2 is the phase of the equalizer’s way more intuitively descriptive. Group
impulse response, θe(t). Calculation for phase delay represents the absolute time delay each
is simply the argument of the complex tap frequency component across the band will
values of DOCSIS 2.0 Pre-Eq and is shown in endure. As such, it is the variation of this
Eq 3. The impulse response phase appears delay (non-flat delay) that matters most, as
randomized between –π and π, except for the components of frequency arriving at different
main tap whose phase correction is 0 radians. times at the opposite end of the channel result
While this plot looks “noisy,” recognize from in distortion and ISI.
the Magnitude response that the amplitude
contribution of the randomly phased Eq 6 – Group Delay Response
imperfections is negligible. ⎛ δθ ( f ) ⎞
GDe ( f ) = −⎜⎜ e ⎟⎟
Eq 3 – Impulse Response Phase ⎝ δf ⎠
θ e (t ) = arg(he (t ))
The subsequent sections will show how
communication channel impairments will
Figure 3 is the equalizer’s amplitude
uniquely impact the DOCSIS Pre-Eq
response, |He(f)|. Calculation for amplitude
coefficient interpretations discussed above.
response is based upon a 1024-point, Fast
Fourier Transform of the DOCSIS 2.0 Pre-Eq

MICRO-REFLECTION Micro-Reflection impairment impact may
be measured on a spectrum analyzer as
This section introduces the micro- amplitude ripple. The peak-to-peak amplitude
reflection impairment and how it impacts the and frequency of the ripple, shown in Figure
DOCSIS Pre-Eq coefficients. As seen by a 7, are directly related to the micro-reflection
receiver, a micro-reflection is a copy of the impairment’s amplitude and delay. One look
transmitted signal, arriving late and with at the frequency response signature quantifies
reduced amplitude. The result of the the micro-reflection’s parameters. In this
additional copy is the familiar image ghosting case, the signal is impaired by a micro-
in analog video reception, but for digital reflection whose relative amplitude is -20 dBc
communications this is ISI. and whose delay is 4 symbol periods.
Micro-Reflection Fundamentals In a QAM constellation, a micro-

reflection causes the symbols to spread in a
Micro-Reflection sources are composed of miniaturized pattern similar to the full QAM
pairs of HFC components separated by a constellation itself. Additionally, phase
distance of cable. What’s important to distortion may cause the spread symbols to
understand about HFC components is that appear rotated. Consider first Figure 8 and
they facilitate the propagation of the signal Figure 9, which illustrate ideal constellation
copies in a variety of ways including return diagrams for 16-QAM and 64-QAM,
loss, isolation, mixing, and combining. respectively.
Figure 6 illustrates one of many possible Now consider Figure 10. Figure 10 shows
micro-reflection source configurations. Two the effect of a micro-reflection on a16-QAM
devices with poor return loss, acting as signal signal’s constellation diagram. The micro-
reflectors, are separated by a length of cable. reflection’s characteristics are those
The CM is acting as the second reflector, but previously depicted in Figure 7. Note the
any HFC component has the potential to spread throughout the symbol region on each
achieve a similar result. The reflector return 16-QAM point, and subsequently how now
loss and the loss between the reflectors less additive noise has more likelihood of
determine the amplitude of the micro- causing a symbol to jump a boundary and
reflection. The delay encountered as a signal create a hard decision error than the Figure 8
copy traverses the red path of Figure 6 will case.
determine which equalizer tap is responsible
for correction. Figure 11 repeats the same micro-
reflection scenario for 64-QAM. 16-QAM is
Note that the CM has as a design limit has less sensitive micro-reflections than 64-QAM
a high return maximum loss value of 6 dB, because of reduced decision boundary size of
meaning it may reflect up to 25% of its 64-QAM for the same average transmit
incident power. In the plant, design limits are power. In other words, 16-QAM symbol size
typically significantly better, but over time can spread more than 64-QAM. In comparing
will degrade as the plant ages and elements Figure 10 and Figure 11, the same level of
that contribute to good RF matching – micro-reflection impairment has spread the
connectors, cable, splitters, interfaces on symbols of the 64-QAM constellation
PCBs – degrade. appreciably closer to the symbol boundaries
than in 16-QAM constellation. The 64-QAM
QAM Signaling Impact situation is clearly a catastrophic situation
with some added noise unless some

equalization processing is applied to undo the Pre-Eq coefficients. AD is undesirable
micro-reflection. variation in the communication channel’s
amplitude response resulting in distortion of
Pre-Eq Coefficient Analysis the digital signal amplitude. Some common
forms of AD include tilt, ripple, and roll-off.
A single dominant micro-reflection
uniquely impacts the DOCSIS Pre-Eq AD Fundamentals
coefficients. The post-tap region of the
impulse response magnitude, illustrated in One common cause of AD in an HFC
Figure 12, reveals the characteristics of the plant is upper return band-edge operation of
Figure 7 micro-reflection: amplitude -20 dBc digital signals, combined with long reaches of
relative to the main tap, and delay 4 symbol coaxial plant. Long reaches of coaxial plant
periods later than the main tap. accumulate multiple diplex filters from
amplifiers and in-line equalizers. While
The impulse response phase reveals individually contributing small attenuation
negligible phase distortion of both the desired versus frequency, when part of a deep RF
symbol and the micro-reflection impairment. amplifier cascade, the combination may
contribute appreciable variation in a
The equalizer’s amplitude response of communication channel’s magnitude
Figure 13 shows the equalizer’s amplitude frequency response. An example of
response. This response derived should be amplitude roll-off has been provided in Figure
compared to Figure 7 in the context of Eq 1. 16.
The phase response shows some QAM Signaling Impact

nonlinearity across the channel’s bandwidth,
especially when compared with Figure 4 In a QAM constellation, amplitude roll-off
phase response. The GD response of Figure causes the symbols to spread in a pattern
14 clarifies the additional phase distortion similar in appearance to Additive White
with appreciably higher GDV than was Gaussian Noise (AWGN). For reference,
previously illustrated in Figure 5. Note Figure 17 and Figure 18 are constellations for
equalizer’s GDV is approximately 43 nsec 16-QAM and 64-QAM respectively, which
across the channel’s bandwidth, while the have been impaired by equivalent levels of
symbol duration itself is less than 200 nsec. AWGN.
DOCSIS Micro-Reflection Assumptions Increasing AWGN contribution by 6 dB

would show that 16-QAM is now just as close
CableLabs has identified via the DOCSIS to the decision boundaries as 64-QAM was
standards [2-5] certain assumptions regarding previously in Figure 18. Conversely,
the nature of a single dominant echo or micro- reducing AWGN contribution by 6 dB would
reflection present in an HFC environment. show 64-QAM is now just as far away from
DOCSIS compliant devices must interoperate the decision boundaries as 16-QAM was
at or below the values illustrated in the Figure previously in Figure 17. Use of 6 dB
15. demonstrates a well-known relationship
between AWGN and modulation complexity
AMPLITUDE DISTORTION (AD) on square constellations such as 16-QAM and
64-QAM. Each modulation order involves a
This section introduces the AD 6 dB increased sensitivity to thermal noise
impairment and how it impacts the DOCSIS

from 16-QAM on up (for QPSK to 16-QAM
it is closer to 7 dB). GROUP DELAY VARIATION (GDV)
The 6 dB relationship is isolated to the This section introduces the GDV

thermal noise impairment. Similar impairment and how it impacts the DOCSIS
assumptions must not be made regarding the Pre-Eq coefficients. GDV is undesirable
impairments discussed in this paper. Many variation in the communication channel’s
such additional impairment relationships have phase response resulting in distortion of the
been derived and discussed in the literature digital signal phase, or, as described, a
over time. For our case here, simulation and variation in the propagation of frequency
test is crucial for characterizing the true nature components of the signal across the channel.
of the relationship which exists between these
more complex impairments and modulation GDV Fundamentals
complexity, and in particular for multiple
simultaneous impairments. As is the case for AD, one major cause of
GDV in an HFC plant is upper-band-edge
Figures 19 and 20 represent the 16-QAM operation of digital signals, combined with
and 64-QAM constellations that result from long reaches of coaxial plant. The reasoning
the amplitude roll-off illustrated in Figure 16. is the same as in the AD case. Note that
Note the appearance of these as compared to filtering functions typically induce nonlinear
the AWGN cases in Figures 17 and 18. phase responses as the band edges are
approached, so the combination of AD and
As with prior impairments, 16-QAM is GDV in the same band region, understanding
less sensitive to amplitude roll-off than 64- that diplex filtering is the cause, is perfectly
QAM because of reduced decision boundary expected. Different filter functions induce
size of 64-QAM. In comparing Figure 19 and different GDV responses, just as different
Figure 20, the same level of amplitude roll-off filter functions induce different stop-band
impairment has spread the symbols of the 64- characteristics. It is common that the sharper
QAM constellation appreciably closer to the the roll-off, such as would be the case for long
symbol boundaries than in 16-QAM cascades, the worse the GDV will be.
constellation. And, again, the 64-QAM case
is bordering on a catastrophic link result QAM Signaling Impact
without some intervention.
In a QAM constellation, GDV causes the
Pre-Eq Coefficient Analysis symbols to spread in a pattern similar to what
has already been illustrated for AWGN and
Amplitude roll-off uniquely impacts the AD.
DOCSIS Pre-Eq coefficients. The near main-
tap region of the impulse response magnitude, As expected, 16-QAM is less sensitive to
illustrated in Figure 21, reveals main-tap GDV than 64-QAM because of reduced
spreading in the region of main tap +/- 3 taps. decision boundary size of 64-QAM.
The amplitude response of Figure 22 reveals
appreciable amplitude correction. Pre-Eq Coefficient Analysis
Lastly, there is linear phase and negligible GDV uniquely impacts the DOCSIS Pre-
GDV across the channel’s bandwidth. Note Eq coefficients. The main-tap region of the
equalizer’s GDV is approximately 12 nsec impulse response magnitude, shown in Figure
across the channel’s bandwidth. 23, reveals main-tap spreading as was

illustrated for the amplitude roll-off AD and GDV impairments contributed by the
impairment. However, the amplitude HFC network.
response, shown in Figure 24, reveals
appreciably less amplitude correction. Since An Ethernet link was established between
the induced impairment is phase related, this the subscriber side of the CM and the Wide
makes sense. Area Network (WAN) side of the CMTS.
The Ethernet connectivity was continuously
Of course, there is an appreciable amount monitored as increases in micro-reflection
of phase variation in the impulse response impairment were introduced into the path
phase and the phase response, while Figure 25 between the CM and CMTS. Loss of Ethernet
reveals appreciable GD correction over the connectivity was assumed to be the point at
GD correction present for the amplitude roll- which a High Speed Data (HSD) subscriber
off impairment scenario previously discussed. would log a service call with a cable operator.
Note equalizer’s GDV was approximately 30
nsec across the channel’s bandwidth. Simulation and measurement for both 16-
QAM and 64-QAM, illustrated in Figure 28
DOCSIS Group Delay Assumptions and Figure 29, reveal DOCSIS 2.0 mitigation
of impairments is appreciably higher than
CableLabs via the DOCSIS standards [2- what is assumed by DOCSIS to be present in
5] has also made assumptions regarding the the HFC environment. Additionally, there is a
nature of GDV present in an HFC reduction of correction capability caused by a
environment. DOCSIS compliant devices reduced decision boundary size as 16-QAM
must interoperate under the estimated signals are migrated to 64-QAM. Note that
conditions illustrated in the Figure 26. The this reduction is approximately 2 dB on
estimates shown in Figure 26 are based upon average and not the 6 dB expected from QAM
preliminary simulation and measurements of and AWGN impairment relationship
GDV and DOCSIS Pre-Eq interaction. previously discussed.
MAINTENANCE REQUIREMENTS Simulation and test of increasing micro-

reflection impairment were repeated with
Understanding the point at which Pre-Eq additional impairments, AD and GDV. AD
will fail is the first step toward leveraging the and GDV contributions were increased by
diagnostic benefits of equalization coefficient simply locating the test signal near the upper
analysis. Simulation and tests were band edge of a 5 – 40 MHz return path
performed of increasing single dominant spectrum, 36.8 MHz center frequency. Figure
micro-reflection impairment. The results of 30 and Figure 31 are measurements of the AD
these tests reveal the highest micro-reflection and GDV present at the upper band region of
impairment level that could be corrected by 5 – 40 MHz return path spectrum of the test
DOCSIS 2.0 Pre-Eq. 16-QAM and 64-QAM topology illustrated in Figure 27. These
were both evaluated. results, illustrated in Figure 32 and Figure 33
for 16-QAM and 64-QAM respectively,
The test topology is illustrated in Figure reveal a negligible change in correction
27. Seven amplifiers were cascaded with an capability of the DOCSIS 2.0 Pre-Eq even
optical link. The 6.4 MHz test signal was with the additional impairments.
centrally located within a 5 – 40 MHz return
path spectrum at 16 MHz center frequency, in Overall, the results shown in Figures 28
order to minimize contributions from both the through 33 are significant for the following
reasons:

1. Discussed impairment levels can exceed operators define maintenance requirements
DOCSIS HFC environmental assumptions and transition toward optimal upstream
and still be corrected by DOCSIS 2.0 Pre- throughput by minimizing cost and service
Eq calls.
2. Simulation results closely agree with
laboratory measurement ISOLATION PROCESS
3. Micro-Reflection impairment impact on
modulation complexity is different from There may be many ways to take
AWGN impairment impact advantage of the wealth of information
provided by Pre-Eq to isolate DOCSIS Pre-Eq
The HFC environment is dynamic in related impairments. The following process
nature, with causes including changing has been proposed in order to isolate
loading conditions, component decay, impairments using equalization coefficient
weather, and routine maintenance practices. analysis.
Allowing sufficient margin for this variation
will allow the HFC environment to breath. Step 1
However, exploiting the limits of acceptable
performance and maintaining margin will Ensure that majority of DOCSIS links are
optimize maintenance costs while also supporting at least DOCSIS 2.0 with Pre-Eq
minimizing service calls. enabled. The resolution of the 24-tap
equalizer of DOCSIS 2.0 is better suited to
In order to define the acceptable identify impairments, compared to the 8-tap
performance limits, simulation and equalizer of DOCSIS 1.1.
measurement are necessary. However,
simulation may bare the burden of exploring Step 2
impairment permutations while minimizing
the cost of testing resources. For example, the Query the DOCSIS 2.0 CM population
impact of multiple micro-reflection using an SNMP query tool similar to Modem
impairments can be studied and defined Pre-Eq Response Tool, illustrated in Figure
through simulation to establish acceptable 34. The Modem Pre-Eq Response Tool,
performance limits which can then be verified which is software developed by Marc
in the laboratory and HFC environment. Morrissette of Motorola, has many useful
features, the most important being the ability
Study of the single dominant micro- to query multiple DOCSIS terminal devices
reflection and a combination of micro- based on an IP address list. Periodic polls of
reflection, AD, and GDV impairments has coefficient values and other relevant physical
defined acceptable performance limits and layer (PHY) metrics are displayed and/or
behavior that is clearly different than stored into a log file for post processing. This
assumptions. tool also provides users with a graphical view
of either the impulse response or amplitude
Continued investigation of impairments response for each CM poll.
and combinations thereof can complete the
acceptable performance limit requirements of Using tools like the Modem Pre-Eq Tool
DOCSIS 2.0 Pre-Eq. Simulation can be can help cable operators establish a baseline
leveraged to help manage the cost of defining of performance, and identify problem
these limits. And finally, an understanding of DOCSIS links, based on CM IP addresses,
the impairment limits and relationships with within the HFC plant.
modulation complexity will help cable

Step 3 Identify suspected HFC components using
results from Step 4 and Step 5. As previously
Identify impairment problems by sorting, discussed, micro-reflection sources consist of
on increasing levels that sum the DOCSIS pairs of HFC components separated by a
Pre-Eq regions previously defined. For distance of cable. Again referring to Figure
example, determine which CMs experience 35, none of the CMs upstream from amplifier
the greatest amount of micro-reflection 1 are reporting a micro-reflection problem.
impairment by sorting on the levels which Additionally, diplex filters make amplifiers
result from summing the taps located in the located between suspected HFC components
post-tap region. unlikely. Therefore, all of the HFC
components fed from amplifier 1 are suspect.
Step 4
Step 7
Understand the relevant Pre-Eq
impairment problems, their impact, and how Inspect and repair as necessary all
they originate in HFC plant. For example, suspected HFC plant components resulting
one micro-reflection source has been from Step 6. Again, referring back to the
discussed in the micro-reflection example provided in Figure 35. Inspection of
fundamentals section, but many possible the suspect HFC plant components show the
permutations of micro-reflection sources must micro-reflection source to be a combination of
be understood for successful isolation. tap-to-output port isolation loss and an
unterminated cable splice at the end of the
Use results such as those in the amplifier 1 feeder run. This combination is
maintenance requirements section to define what is contributing to the micro-reflections
what impairment levels will likely result in experienced by CMs 1A, 1B, and 1C.
service calls, and consider these for potential Properly terminating the splice reduces the
areas to address proactively micro-reflections to negligible amplitudes.
Through repeated application of an Step 8

isolation process, understand how much
margin below service outages would optimize Repeat CM population query multiple
the cost/benefit ratio of proactive times and compare to baseline captured in
maintenance. Step 1 to ensure that the impairment problem
has been eliminated and the improvements are
Step 5 sustainable.
Leverage the CM population to CONCLUSION

differentiate between CMs experiencing an
impairment problem and those that are not.
For example, a query of the CM population of The DOCSIS Pre-Eq function has enabled
the HFC coaxial feeder segment illustrated in operators to deliver yet higher speeds of
Figure 35 reveals that CMs located off of upstream service to the subscribers. The use
amplifier 1 are reporting a micro-reflection of higher order modulation, with a strong
problem, while CMs located off of amplifiers push from DOCSIS tools such as equalization
2, 3, and T1 are not reporting a problem. has made this a reality. However, the
modulation order and bandwidth come at the
Step 6 expense of increased sensitivity to common
HFC impairments. Not only does the Pre-Eq

function ensure these higher speed links are [2] Data-Over-Cable Service Interface
robust, it also provides a wealth of insight into Specifications, Radio Frequency Interface
important plant characteristics. The type of Specification v1.0, ANSI/SCTE 22-1, 2002,
characteristics is many of those that are of Society of Cable Telecommunications
increased relevance as the upstream Engineers, www.scte.org
modulation complexity increases. [3] Data-Over-Cable Service Interface
Specifications, Radio Frequency Interface
By fully understanding the Pre-Eq function Specification v1.1, SP-RFIv1.1-C01-050907,
and deploying some simple tools to perform September 7, 2005, Cable Television
equalization coefficient analysis on the data Laboratories, Inc., www.cablemodem.com
gathered by this function, it is possible to [4] Data-Over-Cable Service Interface
identify the dominant impairments for which Specifications, Radio Frequency Interface
the DOCSIS 2.0 Pre-Eq is compensating. Specification v2.0, CM-SP-RFIv2.0-I11-
060602, June 2, 2006, Cable Television
Single dominant micro-reflection will Laboratories, Inc., www.cablemodem.com
mostly impact the post-tap region of the [5] Data-Over-Cable Service Interface
impulse response magnitude, revealing the Specifications, Radio Frequency Interface
amplitude and delay characteristics of the Specification v3.0, CM-SP-PHYv3.0-I05-
micro-reflection source. 070803, August 3, 2007, Cable Television
Laboratories, Inc., www.cablemodem.com
Dominant AD will mostly impact the
main-tap region of the impulse response
magnitude as well as the amplitude response.
Dominant GDV will mostly impact the

main-tap region of the impulse response
magnitude as well as the phase response.
Simulation and measurement are required

to determine all of the points at which
DOCSIS 2.0 Pre-Eq will work under various
impairment combinations and levels.
Understanding the limits of DOCSIS 2.0

Pre-Eq will help cable operators establish
when proactively maintaining the HFC plant
will be most beneficial, leading to a refined
process that helps cable operators leverage the
benefits of DOCSIS Pre-Eq coefficient
analysis.
REFERENCES
[1] Digital Communications;

Fundamentals and Applications, 2nd Edition,
ISBN 0-13-084788-7, 2001, Bernard Sklar,
www.phptr.com

DOCSIS 2.0 TX PRE-EQ Coefficient Magnitude
0
-5
-10
Magnitude (dB)
-15
-20
-25
-30
-35
-40
-45
-50
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Tap (symbol spaced - 5.12 Msym/sec)
Figure 1 – Negligible Impairment – Impulse Response Magnitude
DOCSIS 2.0 TX PRE-EQ Coefficient Phase
3.14
2.36
1.57
Phase (radians)
0.79
0.00
-0.79
-1.57
-2.36
-3.14
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Figure 2 - Negligible Impairment – Impulse Response Phase

DOCSIS 2.0 TX PRE-EQ Magnitude Correction
5.12 Msym/sec, 6.4 MHz Channel
8.00E-03
6.00E-03
Magnitude (dB)
4.00E-03
2.00E-03
0.00E+00
-2.00E-03
-4.00E-03
-6.00E-03
-8.00E-03
-2.56
-1.92
-1.28
-0.64
0.64
1.28
1.92
2.56
Relative Frequency (MHz)
Figure 3 - Negligible Impairment – Amplitude Response
DOCSIS 2.0 TX PRE-EQ Phase Correction

3.14
2.36
1.57
Phase (radians)
0.79
0.00
-0.79
-1.57
-2.36
-3.14
-2.56
-1.92
-1.28
-0.64
0.64
1.28
1.92
2.56
Figure 4 - Negligible Impairment – Phase Response

DOCSIS 2.0 TX PRE-EQ Group Delay Correction
1340
1335
Group Delay (nsec)
1330
1325
1320
1315
1310
1305
1300
-2.56
-1.92
-1.28
-0.64
0.64
1.28
1.92
2.56
Figure 5 – Negligible Impairment – Group Delay Response
Reflector2
HFC Coaxial
CMTS Reflector1 Cable
Network Network
Modem
Figure 6 – Micro-Reflection Source
Channel Amplitude Response

Averaging = 128 Samples, RBW = 1.875 kHz
2
Amplitude (dBmV)
-1
-2
-3
-4
-2.56 -1.92 -1.28 -0.64 0.00 0.64 1.28 1.92 2.56
Figure 7 - Micro-Reflection Impairment – Channel Amplitude Response

Ideal 16-QAM Constellation Diagram
1 1/3
2/3
Quadrature
-0
- 2/3
-1 1/3
-1 1/3
- 2/3
2/3
1 1/3
-0
In-Phase
Figure 8 - Ideal 16-QAM Constellation

Ideal 64-QAM Constellation Diagram
1 1/7
6/7
4/7
2/7
Quadrature
-0
- 2/7
- 4/7
- 6/7
-1 1/7
-1 1/7
- 6/7
- 4/7
- 2/7
2/7
4/7
6/7
1 1/7
-0
In-Phase
Figure 9 - Ideal 64-QAM Constellation

Pre-EQ Constellation Diagram
1 1/3
2/3
Quadrature
-0
- 2/3
-1 1/3
-1 1/3
- 2/3
2/3
1 1/3
-0
In-Phase
Figure 10 - Micro-Reflection Impaired 16-QAM Constellation

1 1/7
6/7
4/7
2/7
Quadrature
-0
- 2/7
- 4/7
- 6/7
-1 1/7
-1 1/7
- 6/7
- 4/7
- 2/7
2/7
4/7
6/7
1 1/7
-0
In-Phase
Figure 11 - Micro-Reflection Impaired 64-QAM Constellation

0
-5
-10
Magnitude (dB)
-15
-20
-25
-30
-35
-40
-45
-50
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Figure 12 - Micro-Reflection Impairment –Impulse Response Magnitude

1.000
0.800
0.600
Magnitude (dB)
0.400
0.200
0.000
-0.200
-0.400
-0.600
-0.800
-1.000
-2.56
-1.92
-1.28
-0.64
0.64
1.28
1.92
2.56
Figure 13 - Micro-Reflection Impairment – Amplitude Response

1480
Group Delay (nsec)
1430
1380
1330
1280
1230
-2.56
-1.92
-1.28
-0.64
0.64
1.28
1.92
2.56
Figure 14 - Micro-Reflection Impairment – GDV Frequency Response
DOCSIS Micro-Reflection Assumptions
0
-5
-10
Magnitude (dB)
-15
-20 5.12 Msym/sec
-25 2.56 Msym/sec
-30 1.28 Msym/sec
-35
-40
-45
-50
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Tap (symbol spaced)
Figure 15 - DOCSIS Micro-Reflection Assumptions

Channel Amplitude Response
Averaging = 128 Samples, RBW = 1.875 kHz
7
6
5
4
Amplitude (dBmV)
3
2
1
0
-1
-2
-3
-4
-2.56 -1.92 -1.28 -0.64 0.00 0.64 1.28 1.92 2.56
Figure 16 - Amplitude Roll-Off Impairment – Channel Amplitude Response

1 1/3
2/3
Quadrature
-0
- 2/3
-1 1/3
-1 1/3
- 2/3
2/3
1 1/3
-0
In-Phase
Figure 17 - AWGN Impaired 16-QAM Constellation

1 1/7
6/7
4/7
2/7
Quadrature
-0
- 2/7
- 4/7
- 6/7
-1 1/7
-1 1/7
- 6/7
- 4/7
- 2/7
2/7
4/7
6/7
1 1/7
-0
In-Phase
Figure 18 - AWGN Impaired 64-QAM Constellation

1 1/3
2/3
Quadrature
-0
- 2/3
-1 1/3
-1 1/3
- 2/3
2/3
1 1/3
-0
In-Phase
Figure 19 - Amplitude Distortion Impaired 16-QAM Constellation

1 1/7
6/7
4/7
2/7
Quadrature
-0
- 2/7
- 4/7
- 6/7
-1 1/7
-1 1/7
- 6/7
- 4/7
- 2/7
2/7
4/7
6/7
1 1/7
-0
In-Phase
Figure 20 - Amplitude Distortion Impaired 64-QAM Constellation

0
-5
-10
Magnitude (dB)
-15
-20
-25
-30
-35
-40
-45
-50
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Figure 21 - Amplitude Distortion Impairment –Impulse Response Magnitude

3.000
2.500
Magnitude (dB)
2.000
1.500
1.000
0.500
0.000
-0.500
-2.56
-1.92
-1.28
-0.64
0.64
1.28
1.92
2.56
Figure 22 - Amplitude Distortion Impairment – Amplitude Response

0
-5
-10
Magnitude (dB)
-15
-20
-25
-30
-35
-40
-45
-50
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Figure 23 - Group Delay Variation Impairment - Impulse Response

Magnitude

0.400
0.300
Magnitude (dB)
0.200
0.100
0.000
-0.100
-0.200
-0.300
-2.56
-1.92
-1.28
-0.64
0.64
1.28
1.92
2.56
Figure 24 - Group Delay Variation Impairment - Amplitude Response

1450
1430
Group Delay (nsec)
1410
1390
1370
1350
1330
1310
1290
1270
-2.56
-1.92
-1.28
-0.64
0.64
1.28
1.92
2.56
Figure 25 - Group Delay Variation Impairment - Group Delay Response
DOCSIS Group Delay Variation Assumptions
0
-5
-10
Magnitude (dB)
-15
-20
-25 200 ns/MHz
-30
-35
-40
-45
-50
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Tap (symbol spaced)
Figure 26 - DOCSIS Group Delay Variation Assumptions

Ethernet
CMTS
Fiber Protocol
BSR64000
Analysis
Coax Coax Coax Coax
23 Coax Coax Coax
µ-Reflection
Simulator
CM
SB5101
Ethernet
Protocol
Analysis
Figure 27 - Micro-Reflection Impaired Communication Channel Test
Topology

DOCSIS 2.0 Transmit Pre-Equalization
Maximum Correctable Micro-Reflection
16 MHz CF, 16-QAM, 6.4 MHz, 5.12 Msym/sec
0
Micro-Reflection Impairment Level (dB)
-5
-10
-15
-20
-25
-30
-35
-40
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Tap Location (T = Symbol)
Simulation Measurement Assumption

Figure 28 - Highest Correctable Micro-Reflection Using 16-QAM

16 MHz CF, 64-QAM, 6.4 MHz, 5.12 Msym/sec
0
Micro-Reflection Impairment
-5
-10
-15
Level (dB)
-20
-25
-30
-35
-40
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Tap (Symbol-Spaced)

Figure 29 - Highest Correctable Micro-Reflection Using 64-QAM
Channel Amplitude Response Estimate
4.5
4
Amplitude (dB)
3.5
2.5
2
-2.56 -1.92 -1.28 -0.64 0.00 0.64 1.28 1.92 2.56
Figure 30 – Cascaded Amplitude Distortion Estimate, BW = 6.4 MHz, CF =

36.8 MHz

Channel Group Delay Estimate
400
300
200
Group Delay (nsec)
100
-100
-200
-300
-2.56 -1.92 -1.28 -0.64 0.00 0.64 1.28 1.92 2.56
Figure 31 – Cascaded Group Delay Variation Estimate, BW = 6.4 MHz, CF =

36.8 MHz

36.8 MHz CF, 16-QAM, 6.4 MHz, 5.12 Msym/sec
0
Micro-Reflection Impairment Level (dB)
-5
-10
-15
-20
-25
-30
-35
-40
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Tap Location (T = Symbol)

Figure 32 - Highest Correctable Micro-Reflection with Cascaded Amplitude
Distortion and Group Delay Variation Using 16-QAM

36.8 MHz CF, 64-QAM, 6.4 MHz, 5.12 Msym/sec
0
Micro-Reflection Impairment
-5
-10
-15
Level (dB)
-20
-25
-30
-35
-40
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Tap (Symbol Spaced)

Figure 33 - Highest Correctable Micro-Reflection with Cascaded Amplitude
Distortion and Group Delay Variation Using 64-QAM
Figure 34 - Transmit Pre Equalization Query Tool

CM
T1C
T1 26 20 14 26
3
CM CM
T1B T1A
CM
23
3A
23 20 17 14
CM 20
1B
17 20 20 26
1
17
CM CM
1A 1C
2
CM
26
2C
CM CM
2A 2B
14 17 20 23
Figure 35 - Micro-Reflection Impairment Isolation Example

Separation of Video Processing and Multiplexing:
The Key to Network Scaling of HD-VOD and other Bandwidth
Intensive PersonalizedTV® Services
Ron Gutman, CTO

Lorenzo Bombelli, VP – Product Strategy and Management
Imagine Communications, Inc.
Abstract services, and delivers the following value to

system operators:
Cable operators need to rapidly
expand bandwidth capacity to accommodate • Clear and sustainable competitive
the emerging personalized and time-shifted advantage in video quality
viewing paradigm and the accelerating • Up to 50% gain in HFC bandwidth
availability and demand for HD content. • Reduced overall CAPEX and OPEX
Linear programming, traditionally broadcast • Uniform and controlled video quality
to subscribers using efficient high-end closed- across the entire network
loop Variable Bit Rate (VBR) technology, will • Improved system availability
increasingly be delivered over on-demand • Increased service velocity
networks which, using current Constant Bit
Rate (CBR) technology, consume
approximately 50% more bandwidth without ENABLING ON-DEMAND VBR
matching the video quality of high-end
broadcast VBR delivery. Limitations of Traditional VBR Technology
Although there have been a number of While VBR technology can provide
proponents of on-demand VBR, architectures significant bandwidth gains and video quality
and solutions based on broadcast-oriented improvements, a number of important factors
rate-shaping technology inherently rely on have impeded its adoption in on-demand
performing both video processing and applications.
statistical multiplexing on each on-demand
stream. Such approaches result in lower video During initial VOD rollouts, the
quality than high-end broadcast VBR and a challenge and complexity of developing the
cost structure that is not viable for wide scale underlying infrastructure, combined with the
deployment in on-demand applications. relatively low initial demand for on-demand
services, made optimizing video quality and
In this paper we propose a new bandwidth lower priority.
architecture that enables on-demand VBR by
separating the video processing of each video After initial rollouts, adoption of
asset from the downstream multiplexing of traditional VBR solutions was further deferred
each stream to achieve the same video quality because existing VBR solutions based on
as broadcast VBR and eliminate the broadcast-oriented rate-shaping technology
unnecessary and significant cost of re- could not generate enough incremental value,
processing the same asset multiple times. both in terms of video quality and bit rate
This “never cover the same ground twice” reduction, to justify the incremental
approach makes VBR viable for on-demand investment.

considerable value in the rapid expansion of
A key architectural limitation of these on-demand services, a new approach is
broadcast-oriented solutions is that they apply required to make VBR viable in on-demand
both video processing and multiplexing on networks.
each on-demand stream, which leads to two
key shortcomings. A New Approach for On-Demand VBR
The first is that broadcast-oriented We propose the following approach to

rate-shaping solutions attempt to compensate on-demand VBR to make it viable and
for the inherent cost associated with having to accelerate its adoption in cable:
apply video processing to each stream by
using bit rate as a less complex yet less 1. Decouple video processing from
accurate proxy for video quality. As a result, statistical multiplexing to enable
these solutions have no mechanisms to optimal technical and economic
establish and set the optimal bit rate to placement of each function in the
achieve a desired video quality benchmark, network.
and tend to sacrifice video quality in order to 2. Process each asset one time upon
meet bit rate targets. ingest using high-end VBR technology
rather than once for each stream using
The second and related shortcoming is rate shaping technology.
economic. Despite attempts to reduce cost per 3. Utilize the highest quality video
processed stream in broadcast-oriented rate- processing, giving priority to video
shaping solutions, it is impossible to quality rather than bit rate to ensure
overcome the inherent architectural limitation that quality is highly controlled and
of applying video processing on a per-stream not compromised.
basis by reducing component cost. As a result, 4. Optionally, create a simple linkage
broadcast-oriented VBR solutions have been between the video processor and the
unable to meet the system cost requirements SRM that enables the SRM to
for scalable deployment on all streams. incorporate quality in bandwidth
allocation decisions.
One proposed approach to mitigate this
limitation is to add an intelligent session The Case for Asset-based Video Processing
management layer which would dynamically
assign video processing resources only to Decoupling video processing from
over-provisioned QAM channels, relying on statistical multiplexing and applying video
predictable and static user behavior patterns to processing on a per-asset instead of a per-
allocate resources. While this approach does stream basis enables significant improvements
partially alleviate the cost and density burden in both the video quality performance and the
inherent in a broadcast-oriented rate-shaping economics of on-demand VBR. Take the
architecture, it does not provide the scalability following example. A cable system with a
required in a largely user-demand driven half million subscribers and a deployed base
delivery network and is therefore a short term of one million digital set-top boxes has a
workaround to a sub-optimal architecture, VOD system designed for 10% peak
rather than a transformative long term utilization, meaning that during peak usage,
solution. no more than 100,000 VOD or unicast
standard definition streams can be active
As a result of these shortcomings, simultaneously on the network. The operator
although VBR has the potential to deliver has taken into account user patterns across the

system and has deployed an intelligent session
management layer to enable over- Networking and System Benefits
provisioning of rate-shapers, and sufficient
broadcast-oriented rate-shapers to process In addition to the quality benefits
50% of deployed streams. In effect, the described above, shifting to asset-based video
operator has deployed 50,000 unique and processing has other important system
distinct video processors into the network that benefits.
can be assigned to user sessions as required.
The ratio of processing power required
Now consider a “super” video for video processing vs. multiplexing is
processor, with 100 times the processing approximately 98:2. Therefore, separating
power of a rate-shaper, applied on a per-asset video processing from multiplexing not only
basis at the ingest point of the on-demand unlocks the potential for higher video quality
library. Also suppose that at the ingest point on the video processing side, but also enables
the system requires a content refresh rate of significant reductions in cost, density, power
10,000 hours per month. Assuming real-time consumption and system delay on the
processing, this would require 10,000/30/24 = statistical multiplexing side.
14 processors.
Moreover, since assets are pre-
Even assuming 20 processors to allow processed, they can be encrypted upstream
overhead for peaks in refresh rates, per-asset from the multiplexer (after the video pre-
pre-processing with “super” video processors processing), and then distributed, multiplexed
would only require 2,000 video processing and transmitted to the subscriber without
units, saving an equivalent of 48,000 video compromising the content protection or digital
processing units. These gains are further rights management.
amplified if asset-based processing is applied
in regional clusters of 2-3 million subscribers Also, by consolidating video
or national clusters of 5-20 million or more processing operations and distributing the
subscribers. multiplexing function, system operators can
enjoy a new level of quality and operational
Video Quality Benefits control, ensuring that all subscribers have
same high quality viewing experience
The allocation of video processing on regardless of the particular and immediate
a per-asset, rather than per-stream, basis usage patterns in their neighborhood.
represents a fundamental shift in the approach
to on-demand VBR. With 100 times more
processing power available for each asset, IMPLEMENTATION
asset-based processing can take advantage of
advanced video processing techniques that are Statistical Multiplexing Efficiency
not economically viable when applied on a
per-stream basis. Look-ahead configurations The benefits of statistical multiplexing
can be maximized, two-pass and even multi- of VBR signals are well known and
pass encoding can be applied, the search for understood. The bandwidth efficiency
more precise motion vectors can be enhanced, benefits amounts to up to 30% gain in channel
scene changes can be more accurately capacity. When paired with the unlocked
identified, and the diligence applied to noise potential of a separated video processing
filtering can be both infinitely granular and engine, the efficiency of the VBR statistical
wholly comprehensive.

multiplexer can reach as high as 50%, but not
without some exceptions.
Our research suggests that when

unconstrained video processing is combined
with constant quality VBR multiplexing, the
aggregate rates of the individual programs
will peak above an allocated channel rate
approximately 15% of the time. The two Figure 2: HBO: bit rate histogram
examples given below plot instantaneous bit
rates for two different 30 minute standard In Example B, the input CBR rate is
definition programs and provide a histogram also 3.75 Mbps. The content is a nature
for the number of seconds at each bit rate. In documentary. The average bit rate is 2 Mbps,
both examples, the content is coded at a the standard deviation is 0.4 Mbps, the
constant “ICE-Q®” quality score of 96, maximum bit rate is 3.67 Mbps and the EBR
indicating a quality level of NMD (No is 2.46 Mbps.
Material Degradation), providing a constant
quality VBR signal that matches the quality of Example B
the source.
The complexity difference between the

two programs is reflected in their average and
maximum bit rates as well as their Effective
Bit Rate (EBR). The EBR is a new metric
that indicates how much bandwidth is
required to maintain constant quality for any
given asset. In Example A, the input CBR
rate is 3.75 Mbps. The content is an HBO Figure 3: Africa: bit rate over time
Sex in the City segment with an average VBR
bit rate of 2.3 Mbps, standard deviation of
0.45 Mbps, a maximum bit rate of 3.32 Mbps
and an EBR of 2.78 Mbps.
Example A
Figure 4: Africa: bit rate frequency histogram
In Example A, the percentage of

seconds that are above the EBR is 12%. In
Example B, the percentage is 15%. Similar
Figure 1: HBO: bit rate over time analysis has been performed on a larger
sample of 40 distinct HD and SD clips of
varying content type and complexity. VBR
bit rate distribution in this sample shows a
clear statistical match to normal distribution.
Overall, the SD clips have an average VBR
bit rate of 2.13 Mbps, standard deviation of

0.389 and an average EBR of 2.41 Mbps. If
the available QAM bit rate is 38 Mbps Improved Quality and Interoperability
(allowing some reserve) the average number
of streams per QAM is 15.8 according to the It is important to note that the asset-
EBR average. Assuming optimal time-shifting based approach for content processing can be
and buffer management at the edge applied to improve video quality as well as
multiplexer, the probability of the aggregated bandwidth efficiency. This is accomplished
average bit rate exceeding the QAM channel by starting with a higher input bit rate and
bit rate at 16:1 density is 0.7%. Similarly, the higher quality encoded source as the input to
probabilities of streams exceeding the QAM the asset-based processing stage, for example
bit rate for 3 HD plus 3 SD streams, 2 HD 7 Mbps CBR for SD. The resulting EBR for
plus 8 SD streams, and 1 HD plus 12 SD the processed asset will be higher to support
streams are 2.4%, 2.5% and 1.3% the higher quality level, resulting in a channel
respectively. Clearly, more streams per QAM capacity that is equivalent to today’s 10
provide larger statistical gain. streams per QAM, with each CBR at 3.75
Mbps.
The sensitivity of video and audio
communications is such that a ~2% error These processed streams will appear to
results in a completely unusable product. the network as typical 3.75 Mbps CBR. The
Therefore, some additional compression is video bit rate inside the 3.75 CBR transport
required on 2% of the content. rate, however, will be variable. In the valleys
of the VBR rate, additional information can
Second-Pass Asset Processing be stored that allows statistical multiplexers to
recreate the higher quality program derived
An additional benefit of the massive from the 7 Mbps source file. The file
amount of video processing enabled in the containing the extra layers of information
asset processing stage is that all of that must conform to standard VOD specifications
processing can be brought to bear in a second such that servers, pumps, QAMs and asset
pass of processing whereby the variable bit management systems will process the asset
rate signal can be more aggressively without requiring any additional processing or
compressed a second time to create an format conversion. The same formula can be
alternate VBR version of the signal to the used for high definition content, substituting
predominant constant quality “V1s.” This 25 Mbps input sources for the 15 Mbps CBR
alternate (“V2”) can then be transmitted to the content prevalent today. Using this method,
edge multiplexer and inserted during the 2% standard definition VOD assets can be “DVD
of the time in which it is required. quality,” while HD-VOD assets can approach
“Blu-ray quality.”
In this architecture, the statistical
multiplexer performs the critical process of
rate management without needing to perform
rate-shaping or re-compression of the signal.
And because the V2 version of the signal has
been compressed using the full capability of
the high-end video processor, it is of higher
quality than can be provided by a broadcast-
oriented rate-shaping platform, which is Figure 5: Improving VOD quality with ICE layers
severely constrained by cost and density Video Quality Aware SRM
requirements.

Optionally, with minor modifications, based on quality with a guarantee that total
the Session Resource Manager (SRM) can channel capacity will not be exceeded.
play a vital part in further optimizing
bandwidth while maintaining the highest The EBR is equal to the combination
video quality in an on-demand VBR system. of the average VBR bit rate combined with
the maximum V2 bit rate, thus guaranteeing
From a video coding perspective, the quality of the multiplex at any given time.
content varies in complexity. Sports content
typically includes high temporal complexity EBR = Max(Average(V1),Max(V2))
(motion) while other high-complexity content
is spatially detailed or includes spatial The EBR value is calculated on a per-
subtlety (ramp-type edges) that requires asset basis and could be delivered to the SRM
advanced processing. Other content such as using the ADI interface making the standard
static images, dark backgrounds or non- SRM quality aware and enabling it to load-
moving head-and-shoulders video is less balance according the quality rather than bit
complex and can be processed much more rate. Note that this approach does not require
easily. adding more elements to the session
management flow.
The fundamental problem with
providing the same bit rate for all streams Furthermore, while the EBR approach
without taking into account content described above is very simple and effective,
complexity is that it results in sub-optimal it is possible to further optimize the solution.
picture quality and bandwidth efficiency. Just For example, the asset video processor could
as 3.75 Mbps may be inappropriate (overkill) provide the SRM with the averages and
to adequately encode a head-and-shoulders standard deviations of V1 and V2 per asset:
news clip, the 3.75 Mbps rate can be equally
inappropriate (underkill) for encoding a µ – Asset average bit rate
complex scene. Even with virtually unlimited σ – Asset standard deviation
encoding, it can be impossible to maintain
quality during dramatic scene changes, fast In this case the SRM would open a
motion and high spatial complexity if strictly session if both of the following conditions are
constrained to 3.75 Mbps. true:
For example, the processing-rich V1 – Normal_Dist(38, M1, S1, 1) ≥ 99%

encoding in Example A resulted in an
Effective Bit Rate (EBR) of 2.78 Mbps. If one V2 - Normal_Dist (38, M2, S2, 1) ≥ 99.999%
tries to multiplex 16 or even 15 such streams,
severe degradation in video quality is virtually Where:
guaranteed, especially if done by a M=
processing-limited rate-shaping device. The S=
problem is amplified in HD because the
aggregated standard deviation is much higher Therefore, enabled by the EBR
due to the lower number of streams per QAM. (Effective Bit Rate), and using very simple
methods that can be enhanced and optimized
A simple solution to this problem is to by independent SRM vendors, the Session
provide the EBR to the SRM thereby enabling Resource Manager (SRM) can actively load-
the SRM to make QAM allocation decisions balance the QAM sessions according to video

quality rather than stream counts, thereby video processing platforms. By relying on
guaranteeing constant video quality to all edge video processing, we unnecessarily
sessions while further optimizing bandwidth impair service quality by sub-optimally
utilization. forcing video quality decisions to lower
common denominator devices.
In other words, the SRM is
transformed into a video-quality aware device In contrast, by separating the video
that can ensure and control video quality processing and performing it at the video
while delivering more streams at any given server ingest point on a per-asset basis, we
system capacity. ensure not only a consistently higher video
quality during times of non-contention, but
also a dramatically improved picture during
CONCLUSION the most critical periods of aggregate peak
rate channel utilization.
The growing demand for bandwidth-
intensive PersonalizedTV services such as In addition to enabling superior video
HD-VOD, switched unicast, and Internet quality and bandwidth efficiency, a separated
video (whether streaming or downloads) will asset-based processing architecture is
drive substantial HFC bandwidth pressure. uniquely equipped to scale with advancements
in video and audio coding, while maintaining
While on-demand VBR can deliver centralized control over delivered video
significant video quality improvements and quality, optimizing the cost and density of the
bandwidth gains, current solutions based on edge multiplexing solutions, and minimizing
broadcast-oriented rate-shaping technology delay in the end user experience.
have significant shortcomings that challenge
their viability for scale deployment. In summary, an on-demand VBR
approach that separates advanced video pre-
We propose that while edge processing from statistical multiplexing offers
multiplexers must perform rate management a compelling solution for scale deployment
to accommodate the small percentage of the that delivers the full value of on-demand VBR
time when aggregated bit rate peaks exceed and provides system operators a new and
the available channel rate, it does not follow powerful tool in the ongoing battle for
that those same edge devices must also be bandwidth and the consumer.
burdened with the computational intensity of

TelePresence over DOCSIS
John T. Chapman, Cisco, [email protected]
Harsh Parandekar, Cisco, [email protected]
Jeffrey Finkelstein, Cox Communications, [email protected]
Abstract SERVICE PROVIDER

DEPLOYMENT MODELS
Telepresence is a new technology
category that delivers a unique, in person Large enterprises were the early
experience for virtual meetings. A adopters of telepresence, which sought to
telepresence solution integrates advanced increase productivity and reduce expenses
visual, audio, and interactive technologies by conducting virtual meetings with an in-
with broadband networking to bring person experience. These enterprises
people together from across the campus typically deployed telepresence over a
and around the world. private corporate network, or secured the
appropriate service level agreements
To date, telepresence has been (SLAs) from service providers to support
deployed mostly by large enterprises to their enterprise telepresence systems.
enable employees to conduct virtual
meetings with fellow employees and Enterprises are now looking to extend
customers at remote offices that are linked their telepresence systems to more
via a private corporate network. locations, including additional corporate
offices, key customer and partner
As telepresence becomes more facilities, and executives’ home offices.
prevalent, businesses will want the ability As enterprises continue to grow their
to extend the system to locations beyond telepresence systems, cable operators will
the corporate network. However, as a increasingly have the opportunity, and the
real-time, two-way service with high- challenge, to support telepresence services
definition video and spatial audio, on their DOCSIS access networks.
telepresence places stringent requirements
on the underlying network to provide a The stringent network requirements to
high-quality user experience. support the real-time, bi-directional, high-
definition video and spatial audio of
This paper describes methods for telepresence will likely require cable
extending telepresence to locations served operators to offer new levels of service to
by a DOCSIS access network, such as subscribers who want to access corporate
executives’ homes. The challenges of telepresence systems via the DOCSIS
delivering telepresence over DOCSIS network.
networks are investigated, and potential
solutions for addressing these challenges Although the enterprise will typically
are proposed. bear the cost of the network services
required to provide remote access to the
enterprise telepresence system, the cable

operator will likely bill the subscriber While this paper is focused primarily
directly as part of a residential services on enabling telepresence over DOCSIS
package. (TPoD) as an extension of an enterprise
telepresence system, the challenges and
As telepresence becomes a more proposed solutions described herein are
prevalent medium for business-to-business also applicable to the other deployment
communications, enterprises will want to models described above.
enable their systems to interoperate with
their ecosystem partners. Smaller
businesses will also want to deploy THE TELEPRESENCE
telepresence, but may not have the USER EXPERIENCE
resources or know-how to manage their
own telepresence system. Telepresence is a technology that
allows people who are in physically
Cable operators and other service separate locations to communicate with
providers may see an opportunity to offer each other as if they were in the same
a managed telepresence service to these room. Telepresence combines professional
customers as part of a commercial services video, professional audio, and networking
package. This could entail hosting a to create a real-time in-person experience.
telepresence call management server
(CMS) and managing access control and So what does that really mean in more
billing in addition to providing the technical jargon? Video is the main
network services. component. Enterprise telepresence
systems, as shown in Figure 1 typically use
As telepresence technology evolves, one to three 65 inch plasma monitors
one can imagine consumer electronics operating at 1080p and 30 frames per
devices incorporating telepresence second. The monitors are set on one side
capabilities for personal use. As this of a desk in such a way that the desk at
evolution takes place, cable operators will both ends of the call combines together to
increasingly be able to offer the ultimate create one virtual desk with everyone
visual networking experience to their sitting around it.
residential subscribers.
The audio uses professional
Figure 1 – Telepresence Endpoint with Three Screens

microphones and speakers. In a larger adjustments have already been done. It
system, multiple microphones are used, just works.
and audio streams are coordinated so that
audio from an individual on the left side of The quality of the experience and the
the room is played back on the left speaker careful placement of monitors, cameras,
at the far end. This is known as spatially and microphones create the illusion of an
aware audio. in-person experience.
Even the lighting of the room, one of

the most important considerations, SYSTEM OVERVIEW
borrows from the professional world.
Reflective or diffused lighting is Behind the Scenes
recommended as direct lighting causes
shadows. Light sources are located in There are other components in a full
front of the people to light up their faces, telepresence system.
not behind them as often happens in the
webcam experience. In the room, there is a telepresence
endpoint that contains the CODEC
The lighting is chosen to have the best (Coder-Decoder) and the call management
color temperature (4100K) to make skin software. On one side, the chassis
tones look good. The camera is a fixed connects to the monitor, camera, speaker,
focal length 1080p camera and is and microphone. On the other side, the
calibrated to the room. There are even chassis connects to an Ethernet network. If
design guidelines for the exact size of the more than one set of monitors is needed,
room, and the color and composition of then this system is duplicated and then
the walls, floor, and ceiling. interconnected via the Ethernet network.
After all, it’s all IP at this point.
There is a method for conveying slides
from a PC. This may be done through The telepresence endpoint may contain
picture-in-picture or a separate monitor or an auxiliary channel that sends and
projector. receives audio and video from a personal
computer. This is for sharing slides and
And finally, there is the user interface. other audio-visual material.
Classical video conferencing systems had
a complicated user interface that required At a remote location, there is a call
lots of configuration and usually didn’t manager that manages the telepresence
work. A properly designed telepresence calls and the IP phone. For telepresence
system uses a simple user interface such systems that use SIP (Session Initiated
as an IP phone. For example, a call can be Protocol) for signaling, the call manager is
booked in Outlook. Automated often implemented with a SIP proxy or an
provisioning software causes the meeting IP PABX.
to show up on the IP phone screen. The
call is established by pushing a single In order to make multi-party calls, the
button. remote location may also have a
telepresence conference bridge.
In a properly designed telepresence
system, there is nothing to adjust. All

To permit interoperability between TV at the end of his six foot desk and sits
vendors, the remote location may also at the other end for calls. The camera is
offer gateways that perform signaling mounted to the top of the TV. The
conversion. Once example of a gateway controller was put into a cabinet with
would be to convert between older H.323 ventilation. The phone and microphone sit
systems and newer SIP based systems. on the desk.
Adapting Telepresence to the Home The lights in the room were changed
Office to get the right color, and a reflective light
source was added above the TV to help
The home office is not as easily the camera. The windows in the room had
configured as the enterprise environment. to get blackout shades so that they would
The easiest solution for a home office is to not create back lighting.
use an all-in-one floor standing system
and put it on the far side of the desk. To connect back to the enterprise
environment, a good quality router with
The author of the paper configured a build in VPN (IPsec based) is needed. As
home office telepresence system with will be seen later in the paper, the
discrete components. He put a 40” LCD existence of this router and the VPN
Figure 2 – TPoD End‐to‐End System

create one of the two major problems that telepresence system and the DOCSIS CM
have to be overcome in order to set up should be a wired path. Wireless
QoS on the DOCSIS access network. connections are not recommended as they
have a higher packet loss rate that can
End-to-End System impact real-time video.
Figure 2 shows an entire TPoD system In a worst case scenario, there could
that spans both the enterprise and the be three routers in the home network.
service provider environment.
1. A home router which aggregates all
The enterprise contains one or more the traffic from the home network
telepresence systems. It also contains its
own IP PBX system (call manager system) 2. A telepresence router that places
that is used to connect IP phones and to telepresence traffic onto a VPN. A
connect the telepresence systems. There is third-party service provider may
a local telepresence conference bridge as remotely manage this router.
well.
3. An enterprise telecommuter router to
The home office as described earlier that is managed by the enterprise for
contains a telepresence system that sits data and VoIP connectivity. (Ideally,
behind a VPN router. This connects to the this is the same router as the
cable operator through a local cable telepresence router)
modem (CM) and cable modem
termination system (CMTS). The main difference between these
three routers is that a different individual
For TPoD systems that are deployed or organization manages each router. In
for intra-enterprise purposes, all calls are the simplest scenario, there is one router
typically routed to the CMS managed by that manages both the home network and
the enterprise. If the enterprise CMS is not the telecommuter and telepresence
compatible with the home office networks.
telepresence system, a third-party
telepresence service may be used. Third- One connectivity option is to have the
party service may also be used to enable home router and telepresence router
inter-enterprise calls. The third-party connected separately to the CM. For this
service provider will require secure access to work, the cable operator has to be able
to each enterprise. This is significant as it to configure the CM to classify and
can impact call flow considerations. provide QoS treatment for telepresence
traffic. This would require separate IP
addresses from the cable operator for each
HOME NETWORK router and NAT in the CM to be disabled.
CONSIDERATIONS
If the telepresence router is connected
The home network consists of all the through the home router, the home router
networked components in the home and should be configured to provide priority
the interconnectivity, including the access to the CM for the telepresence
telepresence endpoint and the DOCSIS service.
CM. The connection between the

Figure 3 – VPN and NAT Scenario
For best results, the telepresence router

should be directly connected to a port on • VPN
the home router or the CM. This will • NAT (multiple)
eliminate any contentions within the home
network for traffic. A VPN tunnel is built between home
and enterprise routers. The VPN tunnel
puts an encrypted IP packet inside another
ACCESS NETWORK IP packet. Without the ability to read the
CONSIDERATIONS inner encrypted packet, a SP (and the
CMTS) can only use the outer IP header to
The core problem that this paper is route between home and the enterprise.
addressing is how to deliver QoS on the
DOCSIS network to support a The home and enterprise routers
telepresence call. To do that, the DOCSIS terminate the VPN tunnel, remove the
system needs to identify the flow of outer header and route only the inner
telepresence packets, separate them from packet onwards. The VPN tunnel is
the rest of the packets, and apply specific transparent to the telepresence system and
QoS algorithms. the enterprise call manager. Since the
outer IP header is only added/removed by
To understand this issue better, lets the routers terminating the VPN tunnel,
look at a packet as it leaves the the telepresence system and the enterprise
telepresence system and travels upstream call manager only communicate using the
to the CMTS. inner IP header and have no knowledge of
the outer IP address. So who is going to
There are two major issues with the tell the CMTS how to build a classifier?
way that telepresence is connected that
prevent easy identification of telepresence The situation is complicated further by
packets. NAT. Often the port on the router that is

offering a VPN service also runs a local
NAT. That means that the telepresence
packet had its IP addresses changed prior MEDIA FLOW SPECIFICATIONS
to being encrypted and encapsulated. This
is the first level NAT. Specifications vary between the
various manufacturers of telepresence
If the VPN router is connected to the systems. The following set of tables is
home gateway router, the home gateway intended to be a reference point for a
router often runs a NAT. Now the outside typical telepresence system. Actual
IP address of the VPN tunnel is changed. performance may vary.
This is the second level NAT.
Video
Most CMs today come with built in
NAT. As the packet travels through the Specification Description
CM, the previously NATed addresses are Image Size 1920 x 1080 or
NATed again. This is the third level NAT. 1280 x 720
Frame Rate
The packet is now ready to be - Main Video 30 fps progressive
classified by the DOCSIS side of the CM - Aux Video 5 fps progressive
and then be sent to the CMTS. Encoding H.264, VBR
Encapsulation RTP, IPsec
When the packet leaves the DOCSIS Video Bit Rate
MAC domain, it may be subject to yet - 1080p 3 – 4 Mbps average
another NAT. It is a network NAT that - 720p 1 – 2.5 Mbps average
may get deployed after IPv4 addresses Max Frame 65 KB in 33 ms.
become scarce and before IPv6 can be Burst (12 - 16 Mbps peak)
deployed. Although it does not impact the Aux bit rate 500 kbps avg
upstream packet, by changing the packet 2 Mbps peak
address yet again, it will impact the packet Max Latency 150 ms one way
classifier needed on the corresponding including all delays.
downstream of the CMTS at the far end of Tolerated Jitter
the connection. - packet jitter 10 ms
- video jitter 50 ms
In summary, the upstream telepresence Tolerated Packet
packet, once it left the safety of the Loss 0.05%
telepresence endpoint is grabbed by the
network, NATed, encrypted, encapsulated Table 1 – TPoD Video Specifications
in a tunnel, and then NATed up to three
more times. This is worse treatment than In Table 1, the options for high
Houdini used to get. definition video are 720p or 1080p
resolution at 30 frames per second (fps).
And now the CMTS is supposed to 1080i does not make sense since there is
build a classifier for this packet? How is no television legacy in this system.
that going to work? Before we discuss our Smaller screens may be fine with 720p.
proposed solution, lets look at the second Aux video from a PC can be sent at a
part of the QoS equation that is the traffic lower frame rate.
characteristics of the telepresence flows.

Ironically, 720p does not necessarily The peak rate in Table 1 is calculated
help out the cable environment. While the over a 33 ms frame interval, even though
home monitor may be okay with 720p, the the burst less if it is spread over a longer
enterprise monitor is likely to be larger time interval if there is room in the system
and do better with 1080p. Further, the latency budget. In the latest telepresence
executive using the TPoD system at home systems, these bursts can be as infrequent
is usually more concerned about being as every 5 minutes or more.
presented well to other participants than
about how well the others look to him. RTP encapsulation is used so the video
streams may be multiplexed into common
That can lead to the odd case where a transports. RTP also provides better
1080p signal is sent upstream from the management of the data path.
home camera to the corporate 60” screen,
while a 720p signal is send downstream The maximum one-way latency goal
from the corporate camera to the 40” of 150 ms is the same latency goal that is
home screen. This runs contrary to the used for voice calls. The value includes all
cable system bandwidth that typically has network and equipment delays.
more bandwidth in the downstream than
the upstream. The jitter goal is also very similar to
what is allowable for voice calls since
The common codec choice for jitter tolerance is achieved through jitter
encoding of high definition (HD) video buffers that in turn add latency. Video
these days is H.264. Aggressive H.264 jitter does get more complicated as there
codecs can achieve a target video bit rate are differences between what is allowable
of 4 Mbps or better. There is a catch. for actual packet jitter caused by packet
People familiar with MPEG-2 encoded queuing versus video frame jitter which
video on demand (VOD) and switched can be caused by serialization delays on
digital video (SDV) systems are used to slow links.
3.75 Mbps constant bitrate (CBR) for
standard definition (SD) video streams. Audio
These streams are encoded such that I-
frames do not cause traffic peaks that Specification Description
would exceed the target 3.75 Mbps encode Encoding AAC-LD
rate. Raw Bit Rate 64 kbps
Packet Interval 20 ms
Maintaining a constant bit rate while Encapsulation RTP
encoding high-definition video in a low- Max Rx Channels 4
latency, interactive system such as Max Tx Channels 2
telepresence requires substantial video
processing capabilities in the encoder. Table 2 – TPoD Audio Specifications
Thus, telepresence systems typically use
variable bit rate encoding, and thus Table 2 uses an audio codec of the
streams will see a burst when an I-frame is quality of Advanced Audio Encoding Low
sent. That burst can typically be 2x to 4x Delay (AAC-LD). More compressed
the average bit rate. audio codecs do not make sense since the
video already takes up so much
bandwidth.

• Assumed Min Reserved Rate Packet
A home office telepresence system Size.
may receive up to four channels of audio
from the enterprise system - three Downstream Only:
channels from the three microphones and
one aux channel. Assuming a single • Downstream Peak Traffic Rate.
screen home system, it could send up to
two channels of audio to the enterprise DOCSIS data rates are enforced by the
system – one from the microphone and CMTS through rate shaping of the service
one from the aux channel. in the downstream, and by metering of
grants in the upstream. DOCSIS uses a
Aggregate Traffic Profile token bucket based rate limit for service
flows. The number of bytes forwarded is
Specification Description limited during any time interval T by
1080p 3 – 6 Mbps Max(T), as described by the expression:
720p 2 – 4 Mbps
Table 3 – TPoD Aggregate Specifications Max(T) = T * (R / 8) + B
Table 3 sums up the audio and video Where:

from previous tables. Table values are for
a home office telepresence system. The
T = time interval under consideration
presumption is up to a three-screen system
at the enterprise may be communicating R = maximum sustained traffic rate
with a one-screen system at the home (so (bits/sec) [7 - C.2.2.5.2]
three audio feeds and one video feed). B = maximum traffic burst (bytes) [7
- C.2.2.5.3]
Auxiliary audio and video may or may
not be present. Approximately 20% The DOCSIS specification includes an
overhead is added to the raw bit rates to optional parameter for the downstream
allow for encapsulation and for signaling. called Downstream Peak Traffic Rate:
DOCSIS Configuration
Peak(T) <= T * (P / 8) + 1522
The DOCSIS transmission path easily
handles bursty traffic as all traffic is rate Note that DOCSIS does not limit the
shaped with a formula that allows for instantaneous rate of a service flow.
traffic bursts. The common traffic Individual packets will always travel at the
parameters used are: native rate of the media (~40 Mbps for an
Annex B downstream, and ~10 Mbps for a
Downstream and Upstream 3.2 MHz, 16 QAM upstream).
• Traffic Priority
The minimum value for B is 1522 (one
• Max Sustained Traffic Rate (R)
minimum size Ethernet frame) and the
• Max Traffic Burst (B)
default is 3044 bytes although it can be as
• Min Reserved Traffic Rate high as 20 million byes for the DS and 3

million bytes for the upstream. This is Problem Definition
known as “Power Boost” in the cable
industry. To implement QoS with a DOCSIS
Service Flow, the CMTS must have two
As a general rule, the calculation is important sets of information.
done based upon the Ethernet frame size
1. Packet Classification
and does not include DOCSIS framing
(except for upstream concatenated 2. Traffic Descriptor
frames).
The packet classifier is used at the
CMTS in the downstream direction and
To accommodate telepresence traffic the CM in the upstream direction to direct
as described in these tables, the value B a particular flow of packets to a particular
should be greater than 65 KB and the QoS queue. A DOCSIS classifier is
value R should be at least 6 Mbps. The analogous to an RSVP filterspec. The
downstream peak traffic rate should be 19 typical items of interest are:
Mbps (16 Mbps plus 20% packet
overhead). • Destination IP address
• Source IP address (optional)
Different codec configurations would
• Destination Port address
result in different requirements. A system
in which telepresence calls are • Packet Type
dynamically signaled and setup could
optimize the values of the service flow The service flow encodings need to
parameters on a per call basis. know about the size and characteristic of
the media flow. The telepresence traffic
Since actual peak traffic is infrequent, has three general flows, each with
admission control can be done on the multiple sub-flows. These are:
average data rate numbers.
• One or more video streams
• One of more audio streams
SOLUTION OVERVIEW • Signaling to one or more end
points
This section provides an overview of
three proposed solutions for providing DOCSIS has the toolset to classify
QoS for TPoD, and discusses their pros each of these flows individually, and even
and cons. The next sections will then to provide different upstream scheduling
respectively focus on each solution and and different QoS treatment to each type
dive into the technical details. of flow. But does it make sense to do so?
The video flows completely dominate the
But first, lets define exactly what other flows in a TPoD system. Video
information is required in any signaling requires the most bandwidth and the least
messages that are trying to convey QoS latency and jitter. The audio and signaling
information. are along for the ride.
Also, the job of packet classification is

difficult due to the existence of the VPN

and NATs. Still, the IP destination address messages so that they will not be tunneled.
of the IP signaling packets will be The RSVP message has extensions that
different than the IP destination address of permit a bi-directional reservation.
the audio and video packets.
The advantage of this approach is that
Each solution will have to address by only involving the local telepresence
these issues. system and the local CMTS, all DOCSIS
QoS reservations can be made. This
Solution #1: Pre-provisioned DOCSIS benefits the deployments in which the
enterprise telepresence system is behind a
This solution tries to get the most out firewall and unable to help out.
of the existing DOCSIS specification. The
solution is focused on setting the correct It also is useful for scenarios where the
parameters in the CM configuration file cable operator does not own the call
and on the CMTS. manager or where the call manager cannot
be easily modified for PCMM. Also,
The advantage of this solution is rapid because this solution operates on-path, it
time to market. It allows early TPoD can operate through NATs transparently
systems to be deployed to see what the and through a VPN gateway that performs
market interest is and how well two-way the right signaling processing.
real-time video works over a DOCSIS
network. Because only a small number of
network elements need to change, this
The disadvantage of this solution is the technique can be deployed with a
lack of visibility into when a telepresence reasonable time to market advantage.
call is set up or torn down. This
complicates or eliminates the ability to do The main disadvantage of this system
admission control. There may also be a lot is network security. The telepresence
of manual configuration of addresses system is an untrusted entity. Bandwidth
rather than the network just figuring things reservations may come from that IP
out. This prevents this solution from address that are not actual telepresence
scaling well. calls.
Thus, this solution is targeted at field Solution #3: Off-Path Reservation

trials with limited deployment.
This solution looks like a classical
Solution #2: On-Path Reservation PacketCable solution. The call manager
communicates through the policy engine
This solution uses RSVP over UDP to to the CMTS to reserve bandwidth for the
communicate between the home call.
telepresence endpoint and the CMTS. In
this solution, the call manager is not The advantage of this solution is that
involved in setting up the bandwidth the call control is more under the control
reservation within the network. of the call management system.
This solution requires the VPN router The disadvantage is that a

to filter and forward the RSVP over UDP sophisticated call management system has

to be in place. Today, enterprise class call using the downstream frequency
managers do not support PacketCable parameter in the modem configuration
Multimedia (PCMM) interfaces, and file.
service provider class call managers do
not support video calls. As PCMM If it’s too cumbersome to add the
becomes more widely deployed, this downstream frequency in the modem
solution provides a flexible path to configuration files for existing
providing bandwidth and QoS on demand subscribers, a simple feature can be
services implemented in the CMTS to designate
specific upstreams and downstreams as
“telepresence-only”. This could be done
SOLUTION #1 with a specific CM Service Type TLV.
PRE-PROVISIONED DOCSIS
During registration, the CMTS would
This solution uses provisions and look for this TLV in the cable modem
hooks in the existing DOCSIS protocol in configuration file and, depending on
order to ensure that the telepresence traffic whether the TLV is present or absent, it
gets the required QoS treatment over the would move the cable modem to the
access network. appropriate downstream.
We also need to make sure that the Since the entire upstream and
introduction of telepresence units in the downstream channel is dedicated to a
field does not disrupt service to other telepresence setup, the QoS profile for the
modems deployed in the field. This can be telepresence modem would simply
achieved in multiple ways dedicate the full channel bandwidth as
CIR bandwidth for that modem.
Option 1: Separate Upstream and
Downstream for Telepresence Pros
This option would implement a 1. Very simple. TPoD should always

separate upstream and downstream work since resources are dedicated.
channel on the CMTS for telepresence
service. This would guarantee that service 2. TPoD does not interfere with the
to deployed cable modems remains production network. Good for early
unchanged and the telepresence traffic is field trials.
isolated from other traffic in the field. If
the home telepresence subscriber is also Cons
an existing cable modem user, they would
get a second modem that is dedicated for 1. Inefficient use of HFC plant
telepresence service. bandwidth and CMTS ports.
While this option has severe limitation Option 2: Shared Upstream and
when it comes to scaling the telepresence Downstream
service, it’s a safe option to start field
trials and to roll out the service initially. In this option, the telepresence cable
Nailing certain cable modems to certain modem shares the upstream and
downstream channels can be done easily downstream channels with other cable

modems connecting to the same fiber
node. The modem configuration file for A hybrid option is also possible when
the telepresence modem has to be we use multi-channel DOCSIS 3.0
carefully configured in order to minimize modems. A subset of channels can be set
impact on the other cable modems. aside for telepresence service and the
remaining channels could be used other
For this option, if the telepresence user traffic to and from the non-telepresence
already has a cable modem for residential CPEs in the home.
internet service, we could consider
connecting the telepresence system to the The CMTS would be configured in
same modem. The simplest way to set up such a manner that the secondary
the home network is to connect the upstream and downstream service flow
telepresence endpoint to the home VPN created on the CMTS for the telepresence
router, and connect the VPN router traffic would pick up the appropriate
directly to the cable modem (with NAT channel set. With this approach, a single
disabled) instead of connecting through a modem would be deployed in the
home gateway or NAT device. The VPN telepresence user’s home but the
router would be statically provisioned telepresence traffic would still be kept
with a routable IP address. isolated from other residential traffic in
the cable upstream and downstream.
This address can be used to define the
appropriate upstream and downstream Once again, CIR reservations could be
classifier in the modem configuration file used to guarantee bandwidth. That would
to ensure that other non-VPN CPEs in the prove inefficient. A more reasonable
telepresence user’s home that are approach is to use a priority service flow
connected to the shared modem do not combined with a scheduling algorithm
disrupt the QoS for the telepresence such as real time polling service (RTPS).
system. The CMTS would provide QoS
for all the traffic generated by the VPN There is no dynamic admission control
router. in this approach. Thus, the number of
TPoD systems provisioned per channel
The telepresence endpoint can also be should be managed carefully.
configured to mark outbound (upstream)
traffic with a DSCP code point. The VPN Pros
router could be configured to preserve
these markings when tunneling traffic. 1. The pre-provisioned bandwidth
The VPN router would do this by copying functionality based on classifiers and
the DSCP values from the inner packet service flows defined in the cable
(telepresence packet) to the outer packet modem configuration file is
(VPN encapsulation). compliant with standard DOCSIS
behavior and hence is supported by a
The DSCP could then be used in the current CMTS.
upstream classifier at the CM to identify
telepresence call traffic. Note that once the 2. Easy to provision and manage the
VPN traffic hits the CMTS, the CMTS telepresence service through the
typically is configured to rewrite all DSCP modem configuration file
values on outbound packets.

3. If priority service flows are used for tunnels. These are described in detail in
TPoD sessions, the bandwidth can be the next few sections.
over-subscribed.
NAT Traversal:
Cons
Bandwidth reservation is complicated
1. Telepresence service has strict QoS when RSVP signaling has to traverse one
requirements. If static CIR or more NATs. First, NAT devices may
provisioning is used to permanently not handle raw RSVP packets properly.
reserve bandwidth for TPoD sessions, Hence the telepresence client will use the
that bandwidth is not available to RSVP over UDP packet format that
other CIR services such as VoIP or encapsulates the RSVP message inside a
other TPoD calls. The bandwidth is, UDP packet destined to a well-known
however, available to best effort UDP ports, 1698 and 1699 for RSVP
flows when not in use by the TPoD encapsulation.
session.
The telepresence client would use the
2. No admission control. Thus a newly destination address of the remote
added TPoD session could interfere telepresence client for these RSVP over
with an existing TPoD session if there UDP messages. The CMTS would have to
is not enough bandwidth on the intercept these RSVP over UDP packets
channel. on the upstream by filtering the well-know
RSVP encapsulation UDP ports.
SOLUTION #2 Another issue with NAT traversal is

ON-PATH RESERVATION that the client requesting the bandwidth
would use the “inside” NAT address and
Resource Reservation Protocol port by default, while the actual traffic that
(RSVP) is the standard “on-path” traverses across the network would have
bandwidth reservation protocol to ensure the external IP address and port.
that the right QoS treatment is setup along
the path of a media flow through the In order to avoid this problem, the
network. CMTS would have to use the source
address of the actual packet instead of the
The RSVP message contains a “filter local endpoint address specified in the
spec” and a “session”. Together, these filter spec of the RSVP message while
specs allows a node in the network to creating the upstream and downstream
identify the flow – source IP, destination classifiers.
IP, protocol as well as source and
destination port if the traffic is UDP/TCP. There are several types of NAT
The RSVP message also contains a “Flow algorithms. The basic NAT algorithm
spec” that define the traffic parameters known as one-to-one NAT replaces a
(e.g. bandwidth) of the reservation. packet’s source IP address and source port
number with a new external IP address
There are several challenges with and external source port. There is only one
dynamic bandwidth reservation such as internal to external mapping of the source
NAT traversal and IPsec based VPN

IP address and port for all IP destination that it can accept RSVP requests from
addresses and ports. clients connected to that modem.
Another NAT algorithm is known a Teleworker VPN:

symmetrical NAT. Symmetrical NAT
changes the external mapping of the As shown in Figure 4, in most cases the
source IP address and source port for each telepresence endpoint would be connected
unique instance of the destination IP to a home VPN router. In the VPN case,
address and port. the CMTS will not be able to snoop
packets sent from the telepresence client.
With symmetric NAT, the local port This is because the VPN tunnel would
for the media flow cannot be predicted or extend from the home VPN router to the
determined easily. In order for the solution enterprise VPN router and the CMTS
to work even with symmetric NATs, the would be in the middle. By default, RSVP
telepresence endpoint can set a wildcard over UDP packets sent inside the IPsec
for the source port in the RSVP filter spec. VPN tunnel would also get encrypted
along with the actual telepresence flow.
Hence, the CMTS will program the
DOCSIS classifier without the local This enterprise VPN challenge can be
endpoint port. The upstream will match solved by adding a new feature to the
based only on the source and destination home VPN router. The VPN router will
IP address, protocol type and destination have to filter for RSVP over UDP packets
port. The downstream would match based and will forward the RSVP messages
only on the source and destination IP unencrypted outside the VPN tunnel in the
address, protocol type and source port. upstream direction.
Authorization of the Telepresence Client: While this change in the VPN router
would ensure that the CMTS can snoop
As the CMTS is going to process the RSVP over UDP message, there is a
resource reservation requests from the further challenge in the VPN scenario.
home, it has to be ensured that the solution The source and destination address and
is not susceptible to Denial of Service port information in the RSVP message
attacks from malicious cable modem would not be useful for the CMTS, since
users. The standard mechanism to the packets of the telepresence flow would
authorize a user requesting bandwidth still be encrypted within the IPsec VPN
using RSVP is by using Pull-COPS /AAA tunnel. (and therefore encapsulated with
(Authentication, Authorization and packets whose IP addresses are those of
Accounting) where the CMTS would the VPN gateways).
offload the authorization decision to a
COPS /AAA server. To solve this problem the CMTS will
instead have to store the source and
Pull-COPS /AAA authorization may destination IP address in the IP header of
not be deployed in a specific cable the packet carrying the RSVP over UDP
network. Hence another option is to add a message, and use those to create the
vendor-specific TLV in the modem classifier. When the VPN router forwards
configuration file to indicate to the CMTS the RSVP over UDP message to the
CMTS, it uses the same source and

Figure 4 ‐ RSVP over UDP Packet Flow
destination address that it will use to send and a source address of @2, which
the corresponding media packets within represent the near end of the VPN tunnel.
the IPsec VPN tunnel.
Figure 4 also shows an additional NAT
Hence, the CMTS can create a operation that is performed by the cable
DOCSIS classifier that matches all the modem which changes the source address
traffic within the VPN tunnel. The VPN of the packet to @3 and keeps the
router would have to insert a flag in the destination address unchanged as @4. The
RSVP message that the CMTS would CMTS filters the RSVP over UDP packet
interpret as “don’t use the address and port and uses the destination address @4 and
information within the RSVP message. source address @3 to create the upstream
Instead, create the DOCSIS classifier and downstream classifiers to provide
using the source and destination address of appropriate QoS to the bi-directional
the RSVP over UDP packet”. telepresence media.
In Figure 4, the telepresence endpoint Bandwidth Reservation in the

sends an RSVP over UDP packet with a Downstream Direction:
destination address of @5 (the far end
telepresence endpoint) and a source Typically RSVP messages describe
address of @1. The VPN router intercepts unidirectional flows. If bandwidth
this packet, sets the above mentioned flag reservation is required in both directions,
in the RSVP message, and transmits the it is expected that each endpoint sends
packet unencrypted using a destination RSVP requests for its own transmission in
address of @4 (the far end VPN router) the forward direction.

Telepresence traffic is also bi- downstream bandwidth requirements.
directional, but it’s easier to provision the The source port is wildcarded in the
network such that the CMTS is expected filter spec to avoid issues with
to process RSVP messages only from the symmetric NATs.
local telepresence endpoint. There are
several issues with trying to support RSVP 2. If a VPN router is present, then the
messages for downstream traffic from the RSVP over UDP packet is intercepted
remote telepresence endpoint by the VPN router, which changes the
source and destination IP address of
1. In case of the teleworker VPN case, the packet to be the same as the source
traffic from the remote telepresence and destination of the IPsec tunnel.
endpoint, including RSVP messages, The VPN router also sets a flag in the
would be sent encrypted inside the RSVP message requesting the CMTS
VPN tunnel and cannot be seen by the to “ignore the filter spec” within the
CMTS RSVP over UDP message
2. Even for the non-VPN case, RSVP 3. When the CMTS intercepts the RSVP
messages coming from the remote over UDP message, it ensures that the
telepresence endpoint may be filtered message is being received from a
out somewhere in the network – client behind an authorized modem. If
especially if that endpoint belongs to a authorization fails, the CMTS would
different service provider. drop the packet. If it succeeds it would
continue to process the RSVP request.
3. The CMTS would be open to Denial
of Service attacks from the Internet. 4. If the “ignore the filter spec” flag is
It’s easier to authorize the local not set, the CMTS creates an upstream
telepresence endpoint requesting QoS classifier based on source, destination
than it would be to authorize a remote IP address, protocol, and destination
telepresence endpoint that the port. It also creates a downstream
operator has no control over. classifier based on source and
destination IP address, protocol and
In order to avoid these issues, source port.
bidirectional RSVP can be used as defined
in the PacketCable specification. This To avoid problems with symmetric
would enable the local telepresence NATs, in both classifiers the local
endpoint to also request bandwidth from telepresence endpoint UDP ports are
the CMTS for the reverse flow in the not used for classification. If the
downstream direction. “ignore the filter spec” flag is set,
then the CMTS creates upstream and
Step-Wise Solution downstream classifiers based only on
the source and destination IP address
1. As part of the call setup, the video of the received RSVP over UDP
resolution is negotiated. The local packet.
telepresence endpoint calculates the
bandwidth to match this resolution and 5. The CMTS tries to admit the upstream
sends an RSVP over UDP message and downstream service flows based
that describes the upstream and on the TSPEC. If it fails, the CMTS

sends back an RSVP call admission
control (CAC) reject notification. 1. This is an on-path bandwidth
When the telepresence endpoint reservation solution that can be
receives this notification, it can try to deployed independent of various
negotiate a lower video resolution and flavors of off-path solutions
motion with the remote endpoint and (PacketCable, PCMM etc) that the
send a new RSVP request, thereby operators may be using for existing
restarting operation at step 2 above. voice services.
6. Once the local telepresence endpoint 2. Complex NAT and VPN scenarios can
succeeds in reserving the required be supported with this solution.
bandwidth, it completes the SIP call
setup and telepresence media begins to 3. An on-path RSVP over UDP solution
flow between the two telepresence for TPoD can simultaneously exist
endpoints. with an off-path PCMM system for
VoIP.
7. When the telepresence call is done, the
local telepresence endpoint sends an Cons
RSVP teardown message which is
again intercepted by the CMTS. When 1. Even though the reference PCMM
the CMTS receives this message, it architecture does include support for
deletes the upstream and downstream RSVP clients that request QoS
service flows created for the resource directly from the CMTS, this
telepresence call. approach has not been fully defined in
the standards specification. Hence
8. RSVP uses soft state to manage the initial implementations of this solution
QoS reservation in the network. This would be vendor specific or need to be
soft state has to be periodically further standardized.
refreshed by the local telepresence
endpoint by sending RSVP message
periodically. If the CMTS does not SOLUTION #3
receive these periodic RSVP OFF-PATH RESERVATION
messages, it will eventually timeout
the bandwidth reservation and will tear Nailing down bandwidth as
down the service flows created for the recommended in solutions #1 and #2 will
RSVP request. be a challenge for many cable operators.
An alternative approach is to use the
This behavior ensures that bandwidth DQoS infrastructure as defined in
is reclaimed even in the case where the PacketCable MultiMedia to dynamically
TP call is not gracefully torn down. An create and tear down service flows as
example where this is need would be if needed.
the local telepresence endpoint is
powered down before it can end the Rather than statically provisioning the
telepresence call and send the RSVP QoS within the access network, the first
teardown message. initial SIP Invite from the Telepresence
client to the call manager would trigger a
Pros secure sequence of events that would

allow the CMTS to dynamically provision
the QoS while still meeting the goals of 4. Application Manager translates SIP
solutions #1 and #2. It is envisioned that request to PCMM (COPS) and sends
there would be no need for dedicated to Policy Manager
bandwidth and that using a shared model
would provide sufficient capacity. 5. Policy Manager validates request,
determines resources needed and sends
The call manager could be located PCMM gate-set to CMTS
within the Service Provider domain or be
provided by a third party telepresence 6. CMTS determines resource
service. The general call flow remains the availability, creates gate and
same with the addition of authentication. communicates with cable modem
The use of secure protocols with using DSx (Dynamic Service Flow)
authenticated is a requirement as the QoS messaging
requests will be triggered by parties
external to the cable operators network. 7. Cable Modem sets up service flow
The basic flow for requesting QOS 8. Telepresence media flows flow bi-
from a hosted call manager external to the directionally with the proper QoS.
cable network is as follows.
Multiple checkpoints are used to
1. Local TelePresence (CPE) issues SIP maintain security:
Invite to call manager.
• the 3rd party call manager must have
2. Call Manager sends SIP based QOS their specific certification signed by
request to the Service Edge Proxy the site they are contacting,
using a secure transport protocol such
as SIPS, TLS or HTTPS • the edge proxy must have details on
the contacting site,
3. Edge Proxy validates request and
sends the request to application
manager.

Figure 5 ‐ TPoD PCMM System
being static flows they are created

• the application manager must have dynamically using PCMM [4][5].
knowledge of the edge proxy and
allow requests from that device, The addressing challenges from the
home NAT and VPN still exist with this
• the policy manager must have policies call setup scenario. The policy messages
that support the requests being made, that arrive at the CMTS have to contain
and the actual IP addresses that the CMTS will
see, as opposed to the IP addresses that the
• the CMTS must allow requests from telepresence system uses. This is a non-
the policy manager. trivial problem and will be the subject of
additional research.
The actual flows created are
equivalent to the ones created in the on-
path reservation system, but rather than

Pros bandwidth reservation and quality of
service. Three solutions were proposed
1. Uses PCMM for on demand and assessed for viability in TPoD
bandwidth management. PCMM is a deployments.
well-defined specification and CMTS
products on the market support Pre-provisioning the DOCSIS network
PCMM. for telepresence services can be a suitable
approach for field trials and limited
2. Does not require dedicated bandwidth deployment, but not for large-scale
deployment due to the inefficient use of
3. Will not impact the quality of voice DOCSIS network resources.
calls
On-path reservation enables the home
4. Secure protocols with authentication office telepresence system to request
DOCSIS network resources directly from
Cons the CMTS, thereby eliminating the need
for integration with an external call
1. Requires the operator to deploy a management system. However, additional
PCMM infrastructure. authentication mechanisms may be
necessary since the TPoD end point is
2. Multiple points to manage increases considered to be untrusted.
complexity.
Off-path reservation enables the
3. Introduces potential single points of telepresence system to manage the
failure. reservation of end-to-end network
resources, but requires further integration
of DOCSIS network resource management
SUMMARY functions with the telepresence call
management system.
Telepresence is a new technology for
conducting virtual meetings with a real- With further investigation and
time, in-person experience. As the development, the on-path and off-path
adoption of this new technology reservation approaches proposed can be
accelerates, users will increasingly want to viable solutions for large-scale TPoD
communicate via telepresence in locations systems.
served by DOCSIS networks, such as
executives’ home offices. Cable operators
will have an opportunity to offer advanced ACRONYMS AND ABBREVIATIONS
network services to support customer-
owned telepresence systems, and could AAA Authentication,
also offer a managed telepresence service Authorization and
to customers without a complete in-house Accounting
telepresence system.
AAC-LD Advanced Audio Encoding
This paper described the challenges in Low Delay
supporting telepresence over DOCSIS,
particularly in the areas of dynamic CAC Call Admission Control

RSVP Resource Reservation
CIR Committed Information Protocol
Rate
RTP Real Time Protocol
CM Cable Modem
SD Standard Definition
CMS Call Management Server
SDV Switched Digital Video
CMTS Cable Modem Termination
System SIP Session Initiated Protocol
CODEC Coder – Decoder SIPS Secure SIP
COPS Common Open Policy SLA Service Level Agreement

Service
TLS Transparent LAN Services
CPE Customer Premise
Equipment TLV Type Length Variable
DOCSIS Data Over Cable System TPoD Telepresence over DOCSIS

Interface Specification
TSPEC Transmission Specification
DSCP Differentiated Services
Code Point UDP User Datagram Protocol
DSx Dynamic Service Flow VOD Video on Demand
HD High Definition VPN Virtual Private Network
HTTPS Secure HTTP

REFERENCES
IP Internet Protocol
1. Dierks, T. and E. Rescorla, RFC5246,
IPsec IP Security The Transport Layer Security Protocol
Version 1.2,
MAC Media Access Control http://tools.ietf.org/html/rfc5246
NAT Network Address 2. Freesoft.org, SSL V3.0 Specification,

Translation http://www.freesoft.org/CIE/Topics/ssl
-draft/3-SPEC.HTM
PABX Private Automated Branch
Exchange 3. Wagner, David, and Bruce Schneier,
Analysis of the SSL 3.0 Protocol, The
PCMM PacketCable Multimedia Second USENIX Workshop on
Electronic Commerce Proceedings,
QoS Quality of Service USENIX Press, November 1996

4. Cablelabs, PacketCable™ Dynamic _genericcontent0900aecd80554cb2.ht
Quality-of-Service Specification, I03, ml
2007
http://www.cablelabs.com/specificatio 7. Cablelabs, MAC and Upper Layer
ns/PKT-SP-DQOS1.5-I03-070412.pdf Protocols Interface Specifications,
DOCSIS, I09, 2009,
5. Cablelabs, PacketCable™ Network- http://www.cablelabs.com/specificatio
Based Call Signaling Protocol ns/CM-SP-MULPIv3_0-I09-
Specification, I03, 2007, 090121.pdf
http://www.cablelabs.com/specificatio
ns/PKT-SP-NCS1.5-I03-070412.pdf 8. Cisco, Telepresence Network Systems
2.0 Design Guide, August 1, 2008,
6. Cisco, Telepresence Room http://www.cisco.com/en/US/docs/solu
Requirements, 2009, tions/Enterprise/Video/TelePresence_
http://www.cisco.com/en/US/solutions Network_Systems_2.0_DG.pdf
/ns669/networking_solutions_products

THE CHALLENGES OF STOPPING ILLEGAL PEER-TO-PEER FILE SHARING
Kevin Bauer, Dirk Grunwald, and Douglas Sicker
Department of Computer Science, University of Colorado
Abstract used to distribute copyright-protected media

illegally.
Illegal file sharing within peer-to-peer P2P file sharing involving copyright
networks has become a significant threat to protected content presents significant
the film and recording industries. problems for network management and
Furthermore, such peer-to-peer protocols copyright enforcement. P2P networks utilize a
often use a significantly large amount of large amount of bandwidth, particularly
bandwidth relative to other protocols, upstream bandwidth, complicating network
complicating network management. In the management for broadband Internet service
past, copyright enforcement agencies have providers (ISPs), particularly during times of
been hired to investigate users who appear to peak network utilization. In addition, the
be sharing files illegally. In addition, illegal dissemination of copyright-protected
broadband Internet service providers (ISPs) media is an obvious problem for the
have actively throttled peer-to-peer traffic in respective copyright holders that may result in
an effort to reduce load on their networks. We a loss of revenue. As a consequence, there is
observe that an “arms race” has begun ample incentive for both broadband ISPs and
between file traders and copyright copyright holders to work to stop the
holders/ISPs in which the file traders have proliferation of file sharing within P2P
started to develop techniques for hiding their networks.
involvement in the transfer of copyright- Our primary goal in this paper is to assume
protected media files. In response, the a proactive position toward understanding the
copyright holders’/ISPs’ investigative tactics current techniques for distributing and hiding
are evolving to match the changing strategies. copyright-protected content within P2P
In this paper, we provide a survey of the networks. We focus our discussion primarily
current tactics used by file traders to hide on BitTorrent, since it is currently the most
their involvement in illegal file transfers and popular P2P protocol for file sharing. We
speculate about future strategies that may observe that an arms race has already begun
emerge on both sides of the arms race. between file traders and copyright holders in
which the file traders have started to develop
techniques for hiding their involvement in the
1. INTRODUCTION transfer of a copyright-protected media file. In
response, the investigative tactics used by
Peer-to-peer (P2P) networks have recently copyright holders are evolving to match these
grown in popularity for a variety of changing strategies. We provide a survey of
applications such as content distribution, the current tactics used by file traders to hide
streaming multimedia, and voice-over-IP. P2P their involvement in illegal file transfers and
networks are often built around a speculate about future strategies that may
decentralized architecture to distribute data in emerge on both sides of the arms race.
a manner that offers high availability of The remainder of this paper is organized as
content, inherent fault-tolerance, and follows: In Section 2, we provide an
efficiency. While P2P networks offer several introduction to BitTorrent, the most common
important advantages over traditional P2P network in use today. In Section 3, we
client/server architectures, experience has describe the most common techniques that
shown that these networks are sometimes copyright holders have used to track the

distribution of their copyright-protected
content. These strategies often include
locating individual users, issuing DMCA
takedown notices, or even pursuing more
serious legal actions against suspected file
sharers. We also discuss the past tactics used
by broadband ISPs to throttle BitTorrent
traffic. In response to the copyright holders’
desire to protect their content, there is now
significant incentive for P2P users to shed
their network identities and enjoy a certain
degree of anonymity. In addition, to avoid
traffic shaping, P2P users have incentive to
try to hide the nature of their traffic using Figure 1: A file transfer with BitTorrent.
encryption. We discuss the current tactics
used to evade ISP traffic shaping practices out-of-band mechanism. These metadata files
and copyright enforcement in Section 4. In are often hosted by sites such as isoHunt [4]
Section 5, we describe the most common and The Pirate Bay [8]. Once a peer obtains
methods for achieving anonymity online and the metadata file for a desired file, the peer
present evidence from a prior study that P2P contacts the tracker server to obtain a list of
users are beginning to use BitTorrent other peers who are sharing the file. In the
anonymously. We also briefly outline prior process, the peer also registers itself with the
proposals to incorporate anonymity tracker. Other peer discovery mechanisms are
mechanisms into P2P networks themselves. available, including distributed trackers built
We also speculate about the future tactics that upon distributed hash tables (DHTs) and
may be employed to distribute copyright- gossip protocols; however, the centralized
protected content. Finally, we provide tracker server method is simple, and thus, the
concluding remarks in Section 6. most commonly used. The peer finally issues
requests for blocks, or sub-pieces (typically
2. BACKGROUND ON BITTORRENT 16KB), from other peers. Peers who possess
the complete file are called seeders and peers
BitTorrent has become one of the most who do not are referred to as leechers. A file
popular peer-to-peer protocols for file sharing. transfer using BitTorrent is illustrated in
A key feature of file transfers with BitTorrent Figure 1.
is that files are not transferred sequentially, as The protocol's precise sequence of
in protocols such as HTTP or FTP. Instead, messages to initiate a data transfer is provided
files are broken into fixed-size pieces and are in Figure 2. First, a leecher establishes
transferred in parallel. This enables BitTorrent communication with another peer by
to transfer data very quickly and efficiently exchanging handshake messages. The
among a large number of peers. As a result, handshake consists of a plain-text protocol
the protocol can be particularly greedy with identifier string, a SHA1 hash that identifies
regard to bandwidth. the file(s) that are being shared, and a pseudo-
To share a file with BitTorrent, a metadata random peer identification string. After both
file containing the piece length, a SHA1 hash peers have exchanged handshake messages,
of each piece to ensure integrity, and a URL the leecher sends a bitfield message, which
to a tracker server is published through an contains a bit-array data structure that
concisely describes the pieces of the file that
the peer has already obtained. After

exchanging bitfields, the leecher knows which However, given the unfortunate reality that
pieces the other peer can supply, and it BitTorrent is being used to distribute
proceeds by requesting specific blocks. Once copyright-protected movies and music,
a leecher has obtained a piece, it notifies other BitTorrent has caught the attention of
peers by sending a have message. More organizations such as the Motion Picture
details about BitTorrent can be found in its Association of America (MPA) and the
protocol specification document [1]. Recording Industry Association of America
(RIAA). Furthermore, as a consequence of
BitTorrent's aggressive network behavior, it
consumes excessive amounts of bandwidth
relative to other protocols, thus complicating
network management for broadband ISPs.
Since copyright holders and network
operators both have incentive to curtail
BitTorrent usage (though for different
reasons), both have initiated campaigns aimed
at slowing the proliferation of BitTorrent
usage. In this section, we present an overview
of the strategies employed by broadband
network operators and entities representing
copyright holders.
Figure 2: BitTorrent’s message exchange to 3.1 Broadband ISP Tactics
initiate a piece transfer.
Due to the challenges that BitTorrent presents
Due to its aggressive behavior with regard for network management, some broadband
to bandwidth usage (often described as ISPs have recently adopted policies aimed at
swarming behavior) BitTorrent presents disrupting or even blocking BitTorrent traffic
significant network management challenges within their networks [3]. In particular,
for ISPs. BitTorrent is often configured to Comcast received extensive publicity for their
open many TCP connections simultaneously, use of Sandvine to specifically target
sometimes using all bandwidth available to BitTorrent flows with forged TCP RST (reset)
the user. For ISPs, this behavior may packets, causing a targeted TCP connection to
complicate network management, especially be prematurely and abruptly closed. This
during times of peak utilization. As a result, policy has been criticized by network
many ISPs have actively attempted to regulate neutrality proponents and consumer advocates
BitTorrent’s bandwidth usage. In the next in part because there was little transparency
section, we discuss the common tactics used and disclosure regarding these practices. In
in the past to investigate copyright violation response, researchers have produced a variety
and to control bandwidth consumption. of techniques and tools [12, 21, 26] to detect
this type of traffic manipulation by ISPs.
3. INVESTIGATIVE TACTICS
3.2 Copyright Holder Tactics
BitTorrent is not used solely for copyright
violation. There are many legitimate uses Since the tracker servers that enable illegal
including obtaining software updates, file transfers are often hosted in foreign
downloading Linux ISO images, and sharing countries where legal recourse against such
non-copyright protected movies and music. activity is limited [11], the representatives

such as the MPA and RIAA acting on behalf transfers and attempt to launch a variety of
of copyright holders have initiated a large “attacks” on leechers [19]. In particular, this
scale investigative effort to identify and study identified two distinct attack strategies:
pursue individual users participating in illegal fake-block and unresponsive peer attacks.
file transfers. Such companies as Media The fake-block attack occurs when peers
Defender [5] and Safenet [10] have been hired operated by copyright enforcers deliberately
to passively monitor the tracker servers for reply to piece requests with invalid blocks of
copyright infringing file transfers to obtain the data. When an entire piece is obtained, the
list of IP addresses of the users who are leecher verifies the piece's integrity with a
participating in the file transfers. Recall that SHA1 hash. However, the hash fails due to
BitTorrent's primary peer discovery the invalid block(s). This requires the leecher
mechanism requires that the IP addresses of to download the entire piece again (which
other peers participating in the file transfer be wastes time and bandwidth), since it does not
publicly advertised. know precisely which block is corrupt.
A recent study [25] found that these The unresponsive peer attack occurs when
investigators obtain the list of IP addresses a peer completes a valid BitTorrent handshake
from the trackers and send an ICMP echo and bitfield exchange (which is the prelude to
(ping) message to each end-host to ensure that the data transfers), but the peer refuses to send
it is alive. These investigators often target any data. This attack also causes leechers to
suspected file sharers with DMCA takedown waste time and bandwidth exchanging control
notices and even have initiated more formal messages with peers that have no intention to
legal proceedings in some cases. provide pieces of the file.
However, as the authors of [25] observe, The aforementioned study found that both
this type of investigative strategy is of these attacks are relatively common. In
problematic, since it is easily prone errors, addition, while these attacks may cause a
especially false positive identification. False download to take up to 50% longer, they are
positives occur when users are wrongly ineffective at stopping BitTorrent file
accused of actively participating in the file transfers altogether.
sharing. False positives may occur as a result
of normal network activity, for example, if a 4. RESPONSE TO ANTI-P2P CAMPAIGNS
user obtains a DHCP lease on an IP address
that had previously participated in the file Given the techniques used to mitigate
transfer. However, false positives may also BitTorrent usage by network operators and
occur by actively polluting a particular copyright holders, file sharing tactics have
tracker's peer list with arbitrary IP addresses. begun to evolve to incorporate mechanisms to
It is possible to explicitly register arbitrary IP prevent blocking by ISPs and to avoid legal
addresses to a tracker, thus implicating any sanctions by entities representing the
end-host in the file sharing. The authors of copyright holders. In this section, we provide
[25] poignantly demonstrated the an overview of the next phase in the arms race
shortcomings of the current investigative between the file sharers and ISPs/copyright
tactics by registering devices such as holders.
networked printers and wireless access points
to tracker lists, and subsequently receiving 4.1 Concealing BitTorrent from ISPs
DMCA takedown notices for these devices'
alleged involvement in illegal file transfers. In an attempt to frustrate traffic shaping or
Another study has found that blocking by ISPs, an obfuscation technique
representatives of the copyright holders called Message Stream Encryption has been
actively participate in illegal BitTorrent file proposed as an optional extension to the

BitTorrent protocol [6]. Message Stream to frustrate these investigations. One common
Encryption requires that pairs of strategy is to intentionally introduce randomly
communicating peers perform a Diffie- selected IP addresses into the tracker lists
Hellman key exchange to agree on a shared (called pollution). For instance, the Pirate
secret and then encrypt the BitTorrent header Bay, a popular tracker-hosting site, has
(and optionally the payload) using the RC4 implemented this policy [9]. This tactic
stream cipher. This feature is available in increases the potential for false positives to a
Vuze [14], μTorrent [13], and other level that may not be tolerable for the
BitTorrent clients. In order to use the investigators. For instance, wrongly accusing
encryption feature, a peer can only innocent users of sharing files illegally could
communicate with other peers that support the have serious consequences for the copyright
encryption feature. holders including negative public opinion or
However, protocol header encryption and even sanctions from government regulators.
payload encryption are relatively ineffective In addition, services such as PeerGuardian
at obfuscating the traffic type, since the [7] have emerged to provide IP address
packet size characteristics remain intact. blocking capabilities for P2P applications. For
BitTorrent traffic has a distinctive signature instance, this service could be used to block
consisting of large bidirectional data transfers, all IP addresses that are suspected of active
thus it would still be relatively easy to detect pollution or monitoring.
despite encryption. Furthermore, sophisticated More extreme techniques to evade the
techniques based on statistical or machine copyright enforcement authorities are even
learning methods could be applied to starting to become common. For instance,
detecting if an encrypted stream is BitTorrent BTGuard [2] offers a pay proxy service in
traffic [24, 27, 28]. Encrypting BitTorrent which subscribing users can encrypt and
does, however, require the ISP to develop and tunnel their BitTorrent traffic through a proxy
deploy these types of sophisticated traffic server hosted in a foreign country. Using such
classification techniques, which may be a service, when a BitTorrent client registers
expensive and time consuming. Furthermore, itself with a tracker server, the tracker server
these traffic classification techniques are not knows only the proxy's IP address, and
perfect and may have non-negligible consequently, the copyright enforcers also can
classification errors. This could result in a observe only the proxy's IP address. Provided
scenario in which other protocols are that the proxy service does not keep records
misclassified as BitTorrent. of its clients' activity, it is difficult to
In addition to encryption, it is possible that determine the identity of the real client. The
BitTorrent may adopt a UDP transport encrypted tunnel may also frustrate ISPs'
mechanism, which is rumored to be included BitTorrent traffic throttling, but as described
to a future version of µTorrent [13]. The UDP in Section 4.1, traffic analysis techniques exist
transport would render the traffic shaping that may reveal the underlying type of traffic
practices using forged TCP RST packets within the encrypted flow.
ineffective.
5. EMERGING STRATEGIES TO HIDE P2P
4.2 Evading Copyright Authorities
The changing tactics employed by file sharers
Since the large scale investigations carried out and copyright holders/ISPs can best be
by entities representing copyright holders described as an arms race of evolving
have resulted in DMCA takedown notices and strategies and counter-strategies. In this
the potential for more serious legal sanctions, section, we discuss the current cutting-edge
counter-strategies have emerged in an attempt and possible future strategies that P2P users

may apply to obfuscate their activities from circuit and encrypts the user's data with each
their ISPs to avoid traffic throttling and to key in a layered fashion. Upon receiving a
hide from copyright enforcement authorities. packet, the Tor router removes its layer of
Technologies that enable end-users to shed encryption and forwards the packet to the next
their network identities and enjoy anonymity router in the path. Once the final layer of
while online are one line of emerging encryption has been removed, the last Tor
strategies. We present evidence to suggest that router forwards the payload to the destination
P2P users may be beginning to use server.
anonymous networks to avoid traffic It is important to note that only the first Tor
throttling by their ISPs and avoid router on the path (called the entry guard)
identification and subsequent legal action by knows the true identity of the client, and only
copyright enforcement authorities. the last Tor router on the path (called the exit
router) knows the identity of the destination
5.1 P2P and Anonymous Networks server. Tor provides a strong degree of
anonymity, subject to the assumption that it is
The Internet and its fundamental protocols difficult for a single entity to control both the
(i.e., TCP/IP) were designed with no regard first and last Tor routers on a user's virtual
for anonymous network access. However, circuit [15]. However, an ISP or group of
recent research in anonymous colluding ISPs could feasibly monitor the
communications has provided the designs and links entering and exiting the Tor network and
implementations of anonymous overlay perform traffic analysis to link the clients and
networks based on onion routing [22] and mix destinations.
networks [17]. Anonymous networks are
currently being used throughout the world for
a variety of applications, often enabling
freedom of speech and press within repressive
countries.
Tor has become the most popular overlay
network for anonymizing TCP-based
applications [20]. Tor is able to provide a Figure 3: Tor's system architecture.
stronger form of anonymity than the proxy
server approach (described in Section 4.2) In prior work, we characterized how Tor is
because it is built around a decentralized used in practice [23]. In particular, we
design; therefore, no single entity knows both analyzed the application-layer protocols that
the source and the destination of an are commonly used with Tor. We discovered
anonymous flow. Tor's system architecture that individual users are starting to use Tor to
(illustrated in Figure 3) consists of three conceal BitTorrent activity. While operating a
components: Tor routers, Tor proxies, and Tor router for four days, we observed over
directory servers. Tor routers forward TCP 430,000 BitTorrent connections leaving the
traffic on behalf of participating users by Tor network, accounting for approximately
employing a layered encryption scheme 285GB of traffic. While the number of
similar to onion routing. A user running Tor BitTorrent connections was relatively low in
proxy software creates a virtual circuit of comparison to other protocols such as HTTP
precisely three Tor routers. First, the Tor and SSL, the amount of traffic transported
proxy obtains a list of all available Tor routers over these connections was surprisingly high.
from the set of trusted directory servers. Next, However, there are plug-ins for popular
the Tor proxy establishes shared secret keys BitTorrent clients (such as Vuze) that make it
with each of the three Tor routers on the easy to connect the BitTorrent client to the

Tor proxy software. In addition, given the past provide a speculative discussion of the tactics
practices of monitoring and profiling users that may be used by file sharers in the future.
suspected for participating in illegal One potential technique for hiding content
BitTorrent file transfers, it is reasonable to is to use a distributed and anonymous data
suspect that the number of users who turn to store. Freenet [18] is a P2P network in which
strong anonymity mechanisms like Tor may peers can store and retrieve files that are
increase in the future. named by location-independent keys. To
In addition to anonymizing overlay retrieve a file, a user computes a hash of the
networks like Tor, it is possible that P2P users content's description - which is used as the
may look to other sources for anonymity. For look-up key - and forwards a retrieval request
example, the design of an anonymity layer to another peer in the network. The request is
specifically tailored for BitTorrent has been forwarded through potentially many peers
published [16]. The protocol, called until the content is found, upon which, the
BitBlender, works by introducing special content is sent back to the original requester
peers called relay peers into the BitTorrent through each peer that forwarded the initial
system architecture. These peers do not request. In doing so, the replying peer does
actively share any file(s), but merely proxy not know who actually initiated the request,
piece requests and responses on behalf of and the requesting peer does not know where
other users actively sharing the file(s). the data is stored. Furthermore, peers hosting
BitBlender's primary goal is to introduce a files only know the hash of the file's
certain degree of plausible deniability for description, so they remain agnostic regarding
peers listed by the trackers. With BitBlender, the content they host.
a copyright enforcement authority cannot This content hiding strategy offers
simply examine the tracker's peer list to obtain significant advantages over BitTorrent. The
an accurate view of the peers who are Freenet-style of content hosting and retrieval
involved in the sharing. The copyright offers relatively strong deniability for both the
enforcer must actively participate in the file hosts and the retrievers. Furthermore, this
sharing and conduct sophisticated traffic strategy significantly complicates
analysis in order to have any chance of investigations launched by anti-piracy
isolating the real active peers. However, since agencies.
the relay peers exhibit many of the same In addition to Freenet-style P2P networks,
protocol-level behaviors as the real peers, it Tor offers the ability to host hidden services
may still be difficult to isolate the real peers. within the Tor network. A hidden service can
While BitBlender is only a proof-of-concept be established in such a manner that the
design (i.e., there is currently no available service's owner does not reveal their identity.
implementation), it is possible that this relay It is difficult to shut down such a service,
strategy may be incorporated into popular since its location is hidden. More details on
BitTorrent clients in the future. hidden services in Tor can be found in Tor's
design document [20].
5.2 End Game Tor’s hidden services provide strong
anonymity for both the service's host and
Until this point, we have discussed the current those who download content, and represent
and emerging strategies used for hiding illegal perhaps the most radical counter-measure to
file sharing within P2P networks. Next, we anti-piracy efforts. While there is a significant
examine how the shifting strategies used to performance penalty associated with using
stop this type of file sharing may cause a hidden services (i.e., additional download
radical shift in content hiding strategies and time), users may be willing to cope with this
limitation if there is sufficient incentive,

perhaps such as avoiding prosecution. If [5] Media Defender - P2P anti-piracy and P2P
widespread usage of Tor's hidden services for marketing solutions.
illegal file sharing becomes a popular counter- http://www.mediadefender.com.
strategy, there may be little recourse for anti-
piracy authorities to stop it. [6] Message stream encryption.
http://www.azureuswiki.com/index.php/Mess
6. CONCLUSION age_Stream_Encryption.
In this paper, we presented an overview of the [7] Peer Guardian 2.

current strategies for identifying illegal file http://phoenixlabs.org/pg2.
sharers and a survey of the counter-measures
that file sharers have employed in response. [8] The Pirate Bay. http://thepiratebay.org.
We observe that a strategic “arms race” has
started as the tactics for pursing illegal file [9] Pirate Bay tricks anti-pirates with fake
sharers and hiding evolve. In addition, if this peers. http://torrentfreak.com/the-pirate-bay-
arms race continues, we speculate about the tricks-anti-pirates-with-fake-peers-081020.
future tactics that may be used to hide illegal
file sharing and conclude that strong [10] Safenet Inc: The foundation for
anonymity mechanisms and location-hidden information security. http://www.safenet-
services may be the final resort of the illegal inc.com.
file sharing movement. Since this implies a
somewhat bleak outlook for the anti-piracy [11] Secrets of the pirate bay.
authorities, we conclude that alternative http://www.wired.com/science/discoveries/ne
strategies – including tiered bandwidth pricing ws/2006/08/71543.
models to discourage high bandwidth usage
on broadband networks and lower-cost media [12] Switzerland network testing tool.
distribution methods – should be investigated http://www.eff.org/testyourisp/switzerland.
to provide individuals with more economic
incentives to obtain content from legitimate [13] µTorrent - the lightweight and efficient
sources. BitTorrent client. http://www.utorrent.com.
REFERENCES [14] Vuze HD network.

http://www.vuze.com/app.
[1] BitTorrent protocol specification.
http://wiki.theory.org/BitTorrentSpecification. [15] K. Bauer, D. McCoy, D. Grunwald, T.
Kohno, and D. Sicker. Low-resource routing
[2] BTGuard - BitTorrent Anonymously. attacks against Tor. In Proceedings of the
http://btguard.com. Workshop on Privacy in the Electronic
Society. October 2007.
[3] Comcast is using Sandvine to manage P2P
connections. [16] K. Bauer, D. McCoy, D. Grunwald, and
http://www.dslreports.com/forum/r18323368- D. Sicker. BitBlender: Light-weight
Comcast-is-using-Sandvine-to-manage-P2P- anonymity for BitTorrent. In Proceedings of
Connections. the Workshop on Applications of Private and
Anonymous Communications. September
[4] isoHunt - the BitTorrent and P2P search 2008.
engine. http://isohunt.com.

[17] D. Chaum. Untraceable electronic mail, [23] D. McCoy, K. Bauer, D. Grunwald, T.
return addresses, and digital pseudonyms. Kohno, and D. Sicker. Shining light in dark
Communications of the ACM, February 1981. places: Understanding the Tor network. In
Proceedings of the 8th Privacy Enhancing
[18] I. Clarke, O. Sandberg, B. Wiley, and T. Technologies Symposium. July 2008.
W. Hong. Freenet: A distributed anonymous
information storage and retrieval system. In [24] A. W. Moore and D. Zuev. Internet
Proceedings of Workshop on Design Issues in traffic classification using Bayesian analysis
Anonymity and Unobservability. July 2000. techniques. In SIGMETRICS. 2005.
[19] P. Dhungel, D. Wu, B. Schonhorst, and [25] M. Piatek, T. Kohno, and A.

K. W. Ross. A measurement study of attacks Krishnamurthy. Challenges and directions for
on BitTorrent leechers. In Proceedings of the monitoring P2P file sharing networks - or -
International Workshop on Peer-to-Peer Why my printer received a DMCA takedown
Systems (IPTPS). February 2008. notice. In 3rd USENIX Workshop on Hot
Topics in Security, July 2008.
[20] R. Dingledine, N. Mathewson, and P.
Syverson. Tor: The second-generation onion [26] N. Weaver, R. Sommer, and V. Paxson.
router. In Proceedings of the 13th USENIX Detecting forged TCP reset packets. In
Security Symposium. August 2004. Proceedings of NDSS. February 2009.
[21] M. Dischinger, A. Mislove, A. [27] C. Wright, F. Monrose, and G. Masson.

Haeberlen, and K. P. Gummadi. Detecting On inferring application protocol behaviors in
BitTorrent blocking. In IMC '08: Proceedings encrypted network traffic. Journal of Machine
of the 8th ACM SIGCOMM conference on Learning Research. 2006.
Internet measurement. October 2008.
[28] S. Zander, T. Nguyen, and G. Armitage.
[22] D. M. Goldschlag, M. G. Reed, and P. F. Automated traffic classification and
Syverson. Hiding routing information. In application identification using machine
Proceedings of Information Hiding: First learning. In LCN. 2005.
International Workshop. May 1996.

The Headend Revisited: A Multi-Service Video Data Center
for the Modern MSO
S.V. Vasudevan, R. Wayne Ogozaly
Cisco
Abstract across headend, transport, and last mile

networks. These new services must coexist
Cable MSOs are under competitive with and even complement existing services
pressure to deliver an increasing number of such as linear broadcast, switched digital
new services to effectively compete with rival video, and on-demand video services. Unlike
service providers, as well as emerging over- the passive linear broadcast delivery model,
the-top players. This paper proposes a these advanced services implement much
strategy which evolves the traditional cable more dynamic traffic flows between
headend into a multi-service Video Data subscriber and headend, resulting in
Center which is well positioned for the increasing operational considerations.
challenges of the modern era.
But video delivery systems are not the only
The Video Data Center combines the best
area that has recently experienced great
of the traditional video headend with proven
infrastructure and traffic growth.
data center techniques. This paper discusses
Consumption of internet web content,
the intersection of these digital delivery
information which is compositionally
systems and describes a next-generation data
populated with a mixture of database content
center model. Internet data center
and increasingly rich media components, has
architectures are examined and compared
fueled the evolution of data centers, super-
with the data delivery requirements of a
evolved versions of the “computer room” of
modern digital cable system. Specific
the 1970s. Traditional internet content and
insertion strategies will describe how cable
service providers have deployed data center
delivery networks can evolve to a multi-
computing systems to support the 24x7
service Video Data Center, spanning
operation of web, application, and database
broadcast, on-demand, and a next generation
servers, supporting the efficient delivery of
IPTV services, while integrating the best of
interactive web content to large user
today’s regional headend with advances in
populations. In order to meet the increasing
data center technologies.
performance and scale demands of a global
web audience, enterprise computing, network
INTRODUCTION and storage architectures have evolved into
Cable MSOs face considerable operational organized and tiered architectures to support
challenges as they seek to smoothly integrate high-transaction information throughput.
innovative new video services into their
existing delivery platforms. As we witness the Most modern cable headends and hubs
deployment of more interactive video already use IP as a common transport for data,
services, IPTV overlays, internet streaming to voice, and video traffic. Accordingly, video
set top receivers, and other advanced delivery to consumers is rising in scale,
technologies, of equal importance is the growing in interactivity, and being distributed
smooth integration of these new functions to an increasingly diverse set of video-capable
with existing video, voice, and data services devices. The increasing adoption of

interactive services such as VOD, SDV and we propose some appropriate data center
IPTV will present a similar increasing techniques and best practices that can help
transactional traffic load on service delivery cable headends more effectively scale the
platforms. As these changes occur and content throughput of their delivery systems.
networks expand their reach to a national and
global scale, and application server compute
1.0 THE DATA CENTER EVOLUTION
requirements grow to respond to the
increasing transactional load, the headend of Introduction
old begins to resemble a data center that is
serving a variety of data types, led by video,
Much like the cable video headend, data
to a large subscriber population.
centers in the enterprise are under increasing
performance pressures. The enterprise data
The service provider industry has already
center is being influenced by shifting business
benefitted from the leverage of technology
pressures and increasing operational
that was originally developed for orthogonal
limitations. The new demands of the
markets. The very use of IP switching and
enterprise require enhanced video services,
routing for video delivery serves as an
greater collaboration, near-instantaneous
excellent example. The switches and routers
access to applications and information, and
that now carry triple-play traffic were
compliance with ever-tightening regulatory
originally developed for the enterprise
compliance. These strains have manifested as
networking market. All of the staff-years and
operational issues relating to power and
capital investment in ASICs, software and
cooling, efficient asset utilization, element
hardware development in the first 10 years of
and system management and monitoring,
switch and router development went towards
escalating security and provisioning needs,
products that were intended for the wiring
and increasing operational expense.
closet, and essentially funded by the global
enterprise IT market. By adapting these
A fundamental metric of data center
technologies to the performance, scale, and
efficiency relates to cost, power, space,
reliability requirements of the service provider
computational power, and ingress and egress
market, and through clever media processing
throughput. Any data center that is designed,
techniques that enable the safe carriage of
trades off between these metrics, depending
media streams over IP networks, IP switching
on the underlying requirements of the offered
and routing of entertainment grade video is
service. Beyond this, there is increasing
now seen as the norm. The service provider
attention paid to the “-ilities”: [1]
industry enjoyed a drastic reduction in per-bit
transport costs as a benefit of this technology
transfer. • Scalability
• Reliability
The remainder of this paper examines the • Availability
evolution of the enterprise/internet data • Serviceability
center, and provides an overview of • Manageability
networking and computing requirements that • Security
have driven the evolution of data centers in
their efforts to scale and server their The data center architecture has evolved to
requirements. We compare and contrast accommodate these demands in an efficient
between internet data center information way. With each architectural shift, the
processing requirements and cable headend network has become an increasingly
information processing requirements. Finally, important ingredient to enable the latest round

of transitions. The transformation to the
newest evolution of Data Center 3.0
technologies focuses on a service-oriented
design, consolidation of server farms,
virtualization across disparate networked
elements, and enhanced automation to
manage the heavy load of requests for
content, applications, and services.
Figure 2 – The Data Center architecture implements

a layered approach designed to diffuse massive
loading across thousands of server elements [2].
It should be first noted that in enterprise data

center terminology, the use of the descriptors
core, aggregation, and access are used in an
almost opposite manner to the way these
Figure 1 - The Enterprise Data Center is undergoing
an extreme makeover, which is focused on server terms would be used an interpreted by a cable
consolidation, service virtualization, and controlled headend engineer.
automation.
The core layer provides the high-speed
packet switching backplane for all flows
The net result of this transformation is a
entering and leaving the data center. The core
massively scalable architecture; some data
layer provides connectivity to multiple
centers are built to contain upwards of
aggregation modules and provides a resilient
100,000 servers, stitched together as a fabric
Layer 3 fabric for this redundant system. This
of virtualized elements.
layer runs an interior routing protocol, like
Many design elements of the enterprise OSPF or EIGRP, and load balances traffic
data center are applicable to the next between the core network and the aggregation
generation cable Video Data Center. To better layer.
appreciate the similarities and differences, it is
The aggregation layer includes multi-
useful to present a brief overview of
service switches which manage the interaction
enterprise data center architecture.
between servers through a Layer 2 domain.
Server-to-server traffic flows through the
Attributes of the Modern Data Center
aggregation layer and use features such as
Most modern data center architectures are firewall and server load balancing, to optimize
based on a proven layered approach, which and secure applications.
has been deployed in many of the largest data
The access layer is where servers
centers in the world. This layered approach
physically attach to the network. Unlike the
provides the foundation of the contemporary
cable world, “access” in the data center refers
data center and includes 3 major components:
to core of server design, where the network
a core layer, aggregation layer, and access
provides access to thousands of servers.
layer [2].
Server components consist of 1RU servers,

blade servers with integral switches, blade application or database resources as the
servers with pass-through cabling, clustered service requirements dictate.
servers, and mainframes. The access layer
The multi-tier model may use software that
network typically consists of modular
runs as separate processes on the same
switches and integral blade server switches.
machine using inter-process communication
These switches manage Layer 2 and Layer 3
(IPC), or on different machines with
topologies which distribute flows across
communications over a switched network.
various server pools in different broadcast
domains.
Distributing Requests Across Server Pools

Many applications run inside an enterprise
data center. Each application may have one or
more IP addresses associated with it,
providing an identity to which Internet users
send their requests. As Internet requests are
received through the core and aggregation
layers, specialized load balancers distribute
these requests across a pool of targeted
servers, using a combination Layer 2 VLANs
and internal virtual IP addresses. A series of
complex algorithms sift through requests,
associate a virtual IP address and VLAN
domain, and distribute the requests to the
private addresses of physical servers [3].
These complex load balancers may confine
requests for specific applications to a pre- Figure 3 – For most HTTP-based applications, Web,
allocated pool of servers, which are Application, and Database servers work across clearly
segregated into manageable clusters. In that defined server pools using a tiered model, to process
way, server performance, redundancy, and HTTP-based service requests [2].
security can be managed and scaled in a
predictable way.
2.0 THE VIDEO DATA CENTER
The multi-tier model in today’s data center
is dominated by HTTP-based applications. Comparing and Contrasting Enterprise
Scalable web-based services are built with and Video Data Centers
multi-tier processing layers which include
web, application, and database tiers of servers. An important distinction between
Web servers field http:// requests and pass enterprise and video data centers can be
transactional information. Application servers understood by analyzing the traffic pattern of
fulfill these transactional requests by each service.
consulting one or more databases or sub-
services. This information is presented back to For a 3-tired web site, the amount of
the web server, which composes the properly information offered is relatively small (an http
formatted HTML page containing all GET method), the computational demands are
requested information. This model enables the moderate, and the amount of information
independent scaling of presentation, returned is also relatively small (the average
web page size is about 300kB). For a search

engine web site, while the information offered packet loss requirements. In the process of
is also small, the computational demands are being delivered, the stream may be spliced,
very large (a search request may be encrypted, or otherwise transformed by
dispatched to dozens of index servers working another cascading device. Additionally,
in concert) [4], and the information returned is sophisticated processes and systems are
also very small. For a video service provider needed to manage the increasing rotation of
with a service such as VOD, the information licensed on-demand content, as well as ad
offered is relatively small, the computational content.
demands are moderate, but the information This evolving mix of new services,
returned is very large (a 2 hr movie in HD expansive capacity, and more interactive
moves 14GB of data). Thus we can see that services, has triggered a evolution of the
there is an immediate distinction in both video headend. To accommodate this new
computational as well as egress throughput service mix, the Video Data Center aims to
requirements between the internet data canter combine the best of data center and traditional
and the video data center. Accordingly, the headend models into a more scalable and
architecture can be optimized for the specific manageable system.
service requirements. But it cannot be
mistaken that both types of data center have Video Data Center Design Goals
shared goals in the areas of scalability,
A number of key components have
reliability, availability, serviceability, and
emerged to address the needs of the Video
manageability, and security.
Data Center in the modern era:
In this spirit, a video headend architecture
• Multi-Service Support:
implementing selected data center best
practices is proposed. The next generation Cable Video Data
Center not only needs to support core
cable video services including broadcast,
Combining the Best of Data Center and Switched Digital Video, and VOD, but it
Video Headend Technologies must also modularly accommodate new
As MSOs continue to expand the number services such as IPTV, internet streaming
and scale of video services, the resulting to the TV, and video streaming to the PC
permutation of content, formats, interactive or handheld device.
features, and end-devices has placed a burden • Service Independent Scaling:
on the traditional cable headend. This burden
may be felt within the headend in a number of Each video service (broadcast, SDV,
ways. First, services such as VOD, SDV, and VOD, IPTV, etc) should be able to
IPTV require a resilient 2-way independently grow in capacity, without
communications network, including outside disrupting existing services. The IP
plant and home connections. Upstream traffic network and multicast control plane will
will organically grow with the number of play a key role to both identify and
subscribers and applications. independently manage each video service.
Correspondingly, the performance of the • Resource Modularity and Demarcation:
application servers that process these
Each video resource within the Video
upstream requests will need to improve. In
Data Center (acquisition, grooming,
many cases the end result of the subscriber
encryption, etc.) is organized as a modular
request will be the switching or streaming of
component with clear demarcation points.
video content. This information must be
A pair of multicast addresses are used to
delivered via a resilient IP network with strict
identify the input and output points of a

video stream as it traverses the delivery shared across multiple video services. The
network. The IP network and multicast “IP Early Acquisition” module ensures
control plane are used to identify, grow, that this content is available to all services
and independently manage each video in an IP format, at the earliest point in the
resource. This methodology will prepare content acquisition process.
the way for service virtualization and
• Improved Operations and Management:
improved scaling. It also enables the
development of IP appliances to provide Stream visibility and quality monitoring
future stream processing functionality, or across all services at strategic
super-appliances that consolidate more measurement points is always important.
than one stream processing function. Since video streaming/processing devices
and networking devices comprise the
• Service Virtualization: video delivery chain, it is important to
Service virtualization has many meanings, have monitoring tools that can present a
both in enterprise and service provider synthesized and unified view of the
information processing. From the service delivery system. The Video Data Center
provider perspective, it describes the integrates video service monitoring across
implementation of a service that is the delivery network.
logically viewed as a single client-server
entity, but may physically be realized by
3.0 BROADCAST SERVICES
more than one application instance. In
USING VIDEO DATA CENTER
content networking, virtualization refers to
TECHNOLOGIES
the abstraction of a video object from its
physical attributes or location. For Implementing Switched Digital Video
example, a movie may be stored in within the Video Data Center
multiple locations, and be transcoded into
multiple formats to suit multiple receiving While there are some fundamental
devices, but in a virtualized differences between the volume and type of
implementation it need only be known as information exchanged from a headend versus
a single entity. a database-driven website, there are also many
similarities between each data center’s design
• Fault Containment and Resiliency:
goals. Many data center architecture and
The combination of embedded quality networking practices can be applied in the
monitoring and enhanced redundancy cable headend. As an illustrative example, we
techniques allow video faults and outages
to be rapidly identified, circumvented, and
contained to the Video Data Center. The
Video Data Center provides a fully
redundant system in which most failures
are contained and not propagated
throughout the video network.
• IP Early Acquisition:
The next generation Video Data Center
will acquire large volumes of content, Figure 4 – The Switched Digital Video prototype
both real-time and asset-based, sourced service is implemented using Video Data Center
from diverse locations in a wide range of techniques and includes 100 HD and 200 SD
channels, mapped to 30 Service Groups, in a 15,000
formats. Much of this content will be household hub site.

system resource utilization. When the
present a hypothetical system design for a opportunity is presented to augment the
Switched Digital Video (SDV) service amount of switched programming, the Video
implementation. The SDV service described Data Center should be able to accommodate
in this paper was designed and tested using capacity expansions without any major
the design parameters presented in Figure 4. disruption to the system in place.
The advent of Switched Digital Video
Layered Network Approach
(SDV) technology provides a fundamental
change in the way the industry delivers digital The Video Data Center implements a
video entertainment. With SDV, service layered network approach to manage resource
providers have the ability to offer a wider access, security, and scaling across a diverse
variety of programming while managing HFC set of video elements. Figure 5 highlights the
network bandwidth in a sustainable way. The Core, Aggregation, and Video Resource
SDV architecture switches only selected Access layers within the Video Data Center.
content onto the HFC based on channel Similar to the Enterprise Data Center, these
change requests from users within a service layers provide a Layer 3 core connection to
group. Thus, content that is not requested by the Regional Network, aggregation and load
any user in a particular service group does not balancing of video flows across various cable
occupy HFC bandwidth. services, and access to a range of modular
video resources. Individual video resources
MSOs considering the deployment of
are linked together using the multicast control
SDV technology face three operational
plane to create a SDV cable service.
challenges. The first includes the integration
of a large number of operational components
required by the SDV service. Unlike the linear
broadcast model, SDV implements a much
more dynamic service in which SDV Servers
dynamically map requested content to various
QAM channels across hundreds of service
groups. This increases operational
complexity.
The second challenge for Cable MSOs
considering SDV is how to smoothly integrate
an SDV service with existing video services
across the headend. The SDV service must
co-exist and even complement existing
services such as linear broadcast, VOD, IPTV,
and other streaming techniques. The SDV
system must also be managed and scaled
concurrently with these adjacent video Figure 5 – Similar to the Enterprise Data Center, the
services. Video Data Center implements a layered network
A final challenge is in capacity planning approach, including Core, Aggregation and Access
layers.
for future growth. Since the SDV system has
the ability to admit a great number of linear
programs with only a modest increase in The Aggregation Layer includes redundant
required HFC stream resources, operators multi-service switches which provide GE
need to rely on tools to provide visibility of access to individual resource elements. Since

broadcast services within the Video Data servers, regional networks, and national
Center operate on a much smaller scale than backbones. Resource managers that are
the traditional data center, a fewer number of present for interactive services will play a
multi-service switches are required. In this principal role in the Video Data Center.
SDV example, two high density switches These managers will optimize the
provide dual-homed GE inter-connects to all allocation of resources to satisfy a number
video resource elements. of parameters, including cost, latency,
even potential revenue opportunities.
Enabling Service Virtualization
• Masking Complexity within Multi-
Service virtualization is a critical principle Service Environments:
upon which many of today’s advanced data
centers are built. The goal of service Improved resource management systems,
virtualization is to provide a more efficient advanced load balancing, and multi-tiered
delivery platform which is better aligned with access networks hide the underlying
the massive scale and complexity of today’s technology from the end user. Content is
networked environment. Although service available in multiple formats to a myriad
virtualization in the data center continues to of consumer devices. Exactly how this
evolve, different components are directly content is acquired, formatted, and
applicable to the Video Data Center: delivered is hidden from the end user.
• Network Enabled Modularity Similar techniques are implemented within

the Video Data Center. As described in
Resources are defined as modular
Figure 7, video resources operate within a
components with clear network
demarcation points, providing a more modular architecture, defined by clear
scalable and flexible design. Capacity can network demarcation points.
typically be added with minimal impact to
existing services. (Example: SAN-based
network storage which is independently
managed and scaled from the server
farms)
Figure 7 – Similar to the Enterprise Data Center,

Figure 6 – Data Center storage is a modular resource, video resources are split into modular components
which is networked via the SAN, and can be and available as distributed network elements
independently managed and scaled [2].
These video resources are shared by
• Distributed Service Models multiple video services including linear
Providers have the flexibility to distribute broadcast, SDV, time-shift TV, PCTV/IPTV
various processing stages across multiple and other streaming services. This

combination of resource modularity and In this example, the same multicast group
network demarcation enables specific address is used for each SPTS stream across
functions like Ad Insertion or VOD Streaming the four processing stages. There is one group
to be contained within a centralized Headend address “GP“ mapped to the 300 primary
or distributed throughout a regional network. video streams in the SDV tier, and a second
group address “GS“ mapped to the 300
SDV Service Creation secondary or backup streams. This SSM
The multicast control plane plays a vital feature greatly reduces the number of
role in the Video Data Center design. Figure 8 multicast groups used throughout the Video
describes how the multicast control plane Data Center since a single group address (with
strings together modular resources to create varying source addresses along each stage of
video content for the SDV service. In our the delivery chain) can follow a stream
example, 300 SDV SPTS streams are mapped throughout the Video Data Center. This
to unique IP multicast group addresses. technique also provides a clear network
Devices along each processing stage issue demarcation point between each stage and
IGMPv3 JOIN messages to draw specific offers MPEG monitoring tools complete
multicast streams to each device over the visibility into all SDV multicast streams.
layer 3 GE network. Redundant multi-service Within this Video Data Center architecture,
switches provide the routed control plane to the SDV video path contains four processing
manage the handoff of these streams between stages. The IP Early Acquisition stage
each processing stage. provides redundant access to all SDV video
Each of the major processing stages content. During this stage, SDV video
utilizes a unique multicast source address to content is acquired from redundant sources
identify primary and backup video sources, including satellite, off-air, and terrestrial
and obtain the correct content. Source- links. SDV content is transcoded to an IP
Specific Multicast (SSM), a feature of MPEG format at this early stage. Where
IGMPv3, also allows each stream to maintain necessary, video streams are converted from
a single multicast group address (but varying multi-program transport streams (MPTS) to
source addresses), spanning all of the video single program transport streams (SPTS), as
processing stages. used by SDV.
This second stage provides Stream
Conditioning of all SDV channels by
redundant groomers. Since redundant copies
of the SDV video streams are available, this
stage performs a stream selection process in
which each groomer independently selects the
best available copy based on ETR-290 MPEG
quality measurements. After stream selection,
groomers rate cap the resultant SPTS streams
from Variable Bit Rate (VBR) to Constant Bit
Rate (CBR), as used by SDV.
The third stage of processing provides Ad
Figure 8 – The 300 channel SDV service is created by Insertion, delivered by ad servers and
stringing together processing elements, using the standards-based SCTE 30/35/130 MPEG
SSM multicast control plane. insertion techniques.

In the final stage of SDV processing, two processors require their own sophisticated
independent encryption devices bulk-encrypt session and resource management systems.
all SDV SPTS channels. The redundant These management systems string together a
encryption devices independently create two number of video resources to generate each
instantiations of the SDV channel lineup. The properly encoded, conditioned, ad spliced,
complete SDV channel lineup is now and encrypted video stream. Fortunately, for
available for use by remote hub sites. In this broadcast video services, the scale of today’s
fully redundant system, 300 primary streams Video Data Center is typically much smaller
are identified by multicast group address than its enterprise counterpart with respect to
“GP“, in addition to 300 secondary video number of elements. Given these differences,
streams branded by group address “GS“. Video Data Center designs continue to
Throughout this data center design, employ resource managers and the multicast
processing elements compare stream quality control plane to provide flow control and load
between the primary and backup SDV copies. balancing for most broadcast services.
In many cases, redundancy mechanisms For the SDV design, the SSM multicast
identify MPEG level faults, and perform a address scheme provides a stable and
cutover to the backup stream to contain the predictable load balancing technique, which is
fault to the data center. These techniques suitable to the size and scale of the SDV
prevent the propagation of video faults service. Specific SDV streams are mapped to
throughout the video network, pre-empting individual GE ports via IGMPv3 JOINs.
QAM-level cutovers to backup streams in Figure 9 illustrates how a subset of multicast
many cases. group addresses are mapped to specific GE
ports within the Stream Conditioning stage of
Source Specific Multicast (SSM) Load the Video Data Center. In this example, 300
Balancing SDV SPTS streams are evenly distributed
across 3 GE ports. The streams are divided
A key element of the enterprise data center into 3 groups of 100 streams (e.g. G1-100, G101-
is the ability to load balance millions of flows 200, G201-300). SSM allows the same group
across hundreds and potentially thousands of address to be reused across different

servers. These complex load balancing processing stages, simplifying stream
techniques typically use a creative management. This processing stage employs
combination of L2 VLANs, virtual address different multicast source addresses to acquire
schemes, and server pools to evenly distribute content (from S3) and handoff the groomed
requests across thousands of servers. This streams (from S5) to the next stage.
design is driven in part by the
commoditization of server hardware in which
enterprise-class servers are being replaced
with thousands of low-cost blade servers.
With the assistance of distributed computing
and distributed systems management, load
balancing across thousands of server elements
is both efficient and reliable in the traditional
data center design.
By contrast, most elements within the
video headend remain specialized and far Figure 9 – SSM multicast provides a well-defined load
from commoditization. These specialized balancing technique to distribute 300 SDV video
streams across 3 GE ports.

This technique achieves a well-defined
load balancing of SDV video flows. At each
processing stage, multicast groups associated
with video programs are joined to predefined
GE ports. Each processing element is assigned
a work load of multicast groups by the
resource management system.
Gaining Control over Massive Scaling

The SDV system should be able to scale to
a larger number of channels without impact to
the current SDV deployment. Figure 9
provides the estimated bandwidth if we Figure 11 - Modular video resources and extra
doubled the number of SDV channels. Based multicast groups are simply added to each processing
on these calculations, the total bandwidth stage to accommodate 600 SDV video channels.
required to deliver 600 channels of SD and
HD content in the SDV tier is 4.5 Gbps.
4.0 IPTV INSERTION INTO THE
VIDEO DATA CENTER
IPTV Insertion Strategy

The power and modularity of the Video
Data Center are highlighted with the insertion
Figure 10 - Bandwidth required if the SDV Service of a future IPTV service. As described in
was doubled in capacity to include 200 HD and 400 Figure 12, IPTV video streams may require
SD programs. unique stream conditioning, ad insertion, and
bulk encryption. To accomplish this unique
To accommodate this additional capacity, processing, a separate set of multicast group
extra video resources are required by the addresses (GIPTV) are used to draw the new
Video Data Center. Figure 10 highlights this IPTV streams through additional video
growth in capacity, without interfering with resource elements. Similar to other cable
the currently deployed SDV service. At each services, a single group address for the IPTV
stage, additional multicast group addresses are streams will follow those streams throughout
assigned to the extra SDV channels. These the video network. The combination of
new SDV multicast groups are drawn through resource modularity, network based
the different switch ports, groomers, ad demarcations, and SSM based flow control
insertion, and encryption systems, ultimately allow this insertion technique with minimal
providing an efficient and scalable technique disruption to existing video services.
to double the SDV service. Since each video
processing stage is modular and utilizes clear
network demarcation points, video resources
within each stage can scale independently.
The multicast control plane and resource
management system simply directs new SDV
streams to the new processing elements.

RF, SDI, and ASI video into IP MPEG SPTS
streams. Similarly, terrestrial video sources
from remotes sources will be converted to an
IP MPEG SPTS format.
In this example, common MPEG video
streams, acquired from diverse sources are
available for linear broadcast, SDV, VOD,
and IPTV services.
Figure 12 – The Video Data Center provides an IPTV

insertion strategy with minimal disruption to already
deployed services.
The IP early acquisition stage provides a
common source of content for all cable
services, including IPTV. As improvements
to resource and element managers continue,
next generation session-resource management
systems can support both traditional cable
services and newer IPTV systems, enabling Figure 13 – IP early content acquisition collects and
further consolidation. This IPTV insertion transcodes video content to an IP MPEG format for
use by Broadcast, SDV, VOD, and IPTV services.
strategy delivers a manageable insertion
process while minimizing the impact to
existing cable video services.
5.0 ON-DEMAND SERVICES
IP Early Acquisition USING VIDEO DATA CENTER
TECHNOLOGIES
The next generation Video Data Center will
acquire large amounts of video content, Second generation video-on-demand
sourced from diverse locations in a wide systems take advantage of Data Center
range of formats. Much of this content will content caching techniques and distributed
be shared across multiple video services. The video streaming to more efficiently absorb the
“IP Early Acquisition” model insures that this massive scale and unpredictable loading of
diverse content is made available to all on-demand services. Figure 14 provides an
services in an IP encapsulated format, at the example of an advanced on-demand network.
earliest point in the acquisition process.
The intelligent IP infrastructure allows
The goal of the IP Early Acquisition stage VOD content to be stored in large vault arrays
is to convert all streams acquired from diverse at strategic locations across the network. On-
sources to an IP MPEG SPTS format. demand vaults can provide a centralized and
Figure 13 provides a typical Headend video consolidated repository for large quantities of
acquisition process in which content is content, which are acquired from a wide range
acquired from satellite, off-air, and terrestrial of sources to support both live and on-demand
video sources. RF video content from applications [5]. The ability to grow content
satellite and off-air sources is processed by a storage independently from content streaming
series of encoders, and groomers to convert

is a critical enhancement, and was inspired
from the Web caching architectures. When combined with a national network,
an on-demand caching scheme allows for a
significant consolidation of content ingest
points. Advanced caching protocols enable
real-time ingest and the delivery of content
throughout a national footprint within a 250
ms period. HD bit rates are also supported at
a massive scale. Based on these data center
design improvements, terrestrial content
distribution will likely supplant the traditional
on-demand “pitcher/catcher” distribution
techniques of today. This networked
infrastructure also provides a flexible platform
to implement robust resiliency schemes,
Figure 14 – The VOD content caching model is
where multiple copies of content are stored
similar to Web-based cached systems in which the and accessed from distributed locations. All
most popular video content is efficiently cached near of these elements work together through an
the network edge, and delivered to the user via multi- intelligent network to more efficiently deliver
format streamers. the next massive wave of on-demand content.
In this distributed caching model, on-
demand titles are distributed upon request
using a tiered caching scheme. Cached video
makes its way through the intelligent transport
to video streamers located near the edge of the
access network. As content streamers are
moved closer to the edge, providers benefit
from improved scaling, video quality, and
reduced transport costs. This distributed
caching model has been shown to reduce
transport network bandwidth by as much as
95% since the most popular content is
Figure 15 – Advanced content caching protocols
delivered once, cached, and reused across enable real-time ingest and real-time distribution of
many requests for the same title. video content across a national footprint.
Multi-format video streamers in the hub

sites handle traditional MPEG video formats, SUMMARY
but also support internet streaming formats
such as Flash or Windows Media. These Enterprise computing is the midst of
multi-format streaming improvements another evolutionary transformation, as data
coupled with in-home gateway devices can center architectures give way to “cloud
enable a single video infrastructure to now computing” systems such as Amazon Dynamo
support multiple streaming services to various [6]. Advances in this area are driving
in-home devices. Backup video streamers investments in advanced and highly-scalable
located within the regional headend can be networking and distributed computing
positioned to handle unexpected peak periods, architectures. The expected increase in
when streamers in a specific hub site are interactive video traffic requirements is
saturated with requests. placing higher requirements on cable

headends to deploy similarly scalable video
delivery systems that are easily adaptable to
growth and performance. From this
standpoint, the evaluation and application of
data center technologies and practices can
greatly benefit service providers’ efforts to
scale and grow their video delivery systems.
REFERENCES
[1] “Datacenter Reference Guide”, Sun
Microsystems, July 2007.
[2] “Cisco Data Center Infrastructure 2.5

Design Guide”, Cisco Systems, 2007.
[3] “Towards a Next Generation Data Center

Architecture: Scalability and Commoditization”,
A. Greenberg, P. Lahiri, D. Maltz, P. Patel, and S.
Sengupta. Microsoft Research, 2008.
[4] “The Anatomy of a Large-Scale Hypertextual

Search Engine”, L. Page and S. Brin, 1998.
[5] “Climbing Mount Everest”, S. Kotay (Comcast)

and J. Pickens (Cisco). NCTA Technical Papers,
2008.
[6] “Dynamo – Amazon’s Highly Available Key-Value

Store”, G. DeCandia et. al., ACM Symposium on
Operating Systems Principles, 2007.

VIDEO QUALITY ASSURANCE ACROSS IP NETWORKS:
CHALLENGES, REQUIREMENTS & WINNING STRATEGIES
New Technologies Permit Operators to Proactively Manage, Identify,

And Solve Problems to Control Operational Costs and Boost Efficiency
Asha Kalyur, Marketing Manager, Cisco
Abstract viewing content (e.g., on-demand) and

additional devices for displaying video
As customers demand increasing flexibility programs (e.g., PCs, game systems and
to receive video services anywhere and handheld mobile devices). The industry is
anytime with the best possible quality of relatively young, but viewers still expect a
service, cable operators of all types have flawless experience. Testimony arises from
deployed IP next-generation networks, which anyone who has endured a widespread video
provide new levels of scalability, flexibility outage during a major Pay-Per-View event,
and fault resilience, while acting as a Final Four tournament game or major football
platform for the rapid deployment of new game. But poor quality of experience triggers
video services. more than inconvenience.
Operators will be challenged with For instance, a major boxing event

monitoring and troubleshooting the growing transmitted by a medium-size Multisystem
number of video flows across these dynamic Operator (MSO) region can generate more
IP networks. A single lost IP packet can than $1 million in revenue. Unhappy
create visible artifacts on a large number of customers will cancel subscriptions if
television screens, promoting the need for a confronted by a major outage. Further, cable
new approach to video quality monitoring. television represents a lifeline service during
local or national emergencies. Cable networks
A new set of technologies can be used to disseminate urgent news and critical public
ensure high-quality transmission of video information, warning the public of severe
across an IP network, using network-centric weather or other impending dangers.
and service-centric views. This new,
integrated solution – bridging the gap Service providers require a comprehensive
between “IP engineers” and “video service assurance solution in order to achieve
engineers” – enables cable operators to the expectations of continuous high-quality
precisely isolate and rapidly resolve video viewing experiences. Such a solution enables
issues, while providing a common reference service providers to proactively identify
for diverse operational groups that will quality issues before they impact customers
reduce the duration of impairments and and leads to more efficient troubleshooting
consequently reduce subscriber support calls. and resolution.
INTRODUCTION
CHALLENGES FOR PROVIDING
The world of entertainment video is CONSISTENTLY GOOD
changing to support alternative methods for QUALITY OF EXPERIENCE
The cable community understands the

complexity of troubleshooting IP networks.

Adding newer video services adds to the complexity for those charged with
complexity, straining network operational maintaining the quality of those services.
resources. Isolating video anomalies to a
specific problem domain and allocating VIDEO ASSURANCE MANAGEMENT
proper operational resources is critical to SOLUTION REQUIREMENTS
supporting high-quality video over an IP
network and controlling costs. To assure excellent quality, service
providers delivering broadcast television
Video relies on the network to provide services over IP networks need powerful end-
customers with a high quality of experience. to-end, fault-management systems for
An effective video fault management system continuous monitoring and correlation of
allows providers to detect and resolve quality service and network performances. These
issues before customers complain. Limiting systems must meet several requirements:
help desk calls is preferred. Service calls
reduce profit and increase customer churn. • Identify faults in real-time and
The average cost of a single customer case determine the impact to the specific
equals or exceeds the profitability of that video services.
consumer for an entire year, according to • Support visibility into several
some cable providers. operational domains (such as head-
end, core, aggregation or access) for
In the broadcast television industry, fault isolation.
success depends on consistently providing • Integrate with the network
excellent video quality. When broadcast video infrastructure and operations support
services are delivered over an IP network, systems (OSS).
quality of experience is affected directly by • Share common information pools and
packet loss and jitter. One IP video packet automation scripts for fault isolation,
contains approximately 1,400 bytes of identification and resolution.
information, and each IP packet contains • Promote communication among
multiple MPEG encapsulated video packets. personnel from different operational
The loss of even one IP packet can lead to domains (head-end, core, aggregation,
video impairments lasting half second or access).
more. A single dropped packet also may cause • Present information in language
“tiling.” Many dropped packets can lead to familiar to both network operations
video freezes or loss of signal at the and video operations personnel.
customer’s location.
• Provide an architecture that supports
devices from multiple vendors.
Lengthy content delivery chains further
• Provide operational dashboards with
complicate video networks. The delivery
intuitive drill-down navigation to
chain includes the broadcaster's head-end, the
detailed quality statistics per domain,
transport network consisting of multiple
per program and per stream.
network elements, video manipulations such
• Extensible architecture to include
as multiplexing, ad insertions and finally
components supporting future
through to the customer's home network and
set-top box. services.
A video assurance solution must account CRITICAL FACTORS TO CONSIDER

for the various technologies transporting the FOR VIDEO ASSURANCE
video services and reduce the operational

The critical factors a video assurance service should IP transport issues be
solution must consider include: operational detected.
costs, problem domain isolation and
determination of services impacted. Related Figure 1. Video Assurance Statistics
statistics are shown in Figure 1. 83%: Problem calls related to poor video
experience.
• Operational Costs: This inversely
affects cable operator revenue. Four 30 – 50%: Digital cable customers
out of every five digital-program experience latent issues after a
subscribers who placed service calls technician’s visit.
disconnected within the same year,
28%: New installs require service call
according to studies. Up to one-third
within one year.
of all service calls requires at least one
80%: New digital customers with a service
follow-up by a service technician.
call in a year disconnected within that
Increasing operational costs and
same year.
decreasing subscriber quality of
experience is an unsustainable model.
20- 25%: New installs requiring service
• Problem Domain: Successfully
call within one month.
reducing the duration of a video
5%: Digital customers disconnected
delivery issue depends on locating the
within one month of install.
problem. Rolling out a truck to a
customer site when the problem 25 - 33%: Service calls resulting in at least
originates in the core transport 1 repeat visit
network represents an unnecessary
cost to the service provider. Similarly, EXAMPLES OF CRITICAL PROBLEMS
dispatching network engineers to RELATED TO VIDEO ASSURANCE
isolate and troubleshoot the IP AND HOW THEY CAN BE SOLVED
network infrastructure would be
unnecessary if the problem originated
in the video head-end. An effective 1. What is the most important characteristic
video assurance management solution of an efficient video assurance
must quickly identify the problem management solution?
domain in the event of a service
disruption, so providers don’t incur Today, there are many fragmented
wasted costs. solutions for video assurance management.
• Service Impact: Delivery of video Many existing video monitoring solutions are
services over an IP network represents either point solutions or rely on vendors who
the combination of two distinct address only a small sub-section of overall
technologies. A video assurance delivery. Video assurance requires an
management solution must bridge this architecture that permits interoperability
technology gap. Since video issues are between monitoring and troubleshooting.
likely to be reported as program Video operations dashboards should condense
issues, there must be an ability to the amount of information presented to the
correlate those services to the IP operator and provide a unified service status.
transport streams responsible for the Many video impairments are transient, and
delivery. Similarly, an operator should service technicians often fail to properly
be able to assess the impact on a video detect and isolate the video-service disruption.
Without an end-to-end video assurance

management strategy, technicians lack a of the video service affected may mean the
logical trail to follow and instead work difference between troubleshooting a rerun vs.
instinctively, changing cables, connectors, the Super Bowl. Through problem domain
STBs and other equipment, trying to solve the reduction and service impact assessment,
issue. Technicians frequently are dispatched operators can assign the personnel best-suited
to a reporting customer site, even though the for a particular problem, reducing both
fault actually may have been miles away at resolution time and costs.
the head-end, since video has dependencies
within the IP network. That saddles the
service provider with increased operational 3. How can an NOC monitor a channel
costs. across multiple multicast streams and Ad
Zones?
An efficient video assurance solution
provides a complete end-to-end service Cable operators use IP transport addressing
assurance architecture for video services. The to route packets through their networks. For
architecture must account for multiple broadcast video services, multicast addresses
domains including the video head-end, the allow a single IP stream to reach large
providers transport and set-top boxes at home. numbers of subscribers. Advertising content is
inserted to video services as the multicast
2. Many operations centers fail to flows travel through the national network and
communicate with one another. regional networks to the local hubs. The
insertion of advertising at these different
Productivity increases when there is an points in the network is commonly referred to
ability to measure video flow as it transfers as Ad Zones.
between operational responsibilities. This
occurs when an anomaly is isolated to a The use of Ad Zones allow service
specific problem domain in the end-to-end providers to increase revenue by offering
path, and the appropriate operational groups more specific advertising as the video service
receive the assignment. moves closer to the end subscribers. However,
ad insertion presents a challenge to service
Most cable operators maintain separate providers. As the content of a video service is
video operations centers (VOC) and network changed with new advertising, so must the
operations centers (NOC). A video assurance multicast address of the IP transport. It is not
management solution that alerts network unexpected that a video service may travel
operations personnel which part of the through several different multicast flows
network requires attention (or what network before reaching the subscriber. When a
event caused an anomaly) and simultaneously problem is reported on a particular video
tells video operations personnel which service, operators face the challenge of
specific programs were affected is critical to determining which multicast flows are
timely resolution. The ability for a video responsible for the delivery of that complete
assurance solution to track the IP flows that service.
contain a particular program (e.g., pre-ad-
splice, post-ad-splice, pre-encrypt, post- A video assurance solution must be able to
encrypt) and all associated IP addresses used associate a video service with the multiple
in the distribution of these services would multicast flows controlling the transport of
allow service providers to determine both that service through the network. The video
problem impact and the priority of the video assurance solution also must be capable of
services affected. Understanding the priority determining through which points in the

network these multicast flows exist. Without A video assurance solution must scale to
this capability, operators would be unable to the number of subscribers in the last mile,
reduce the problem domain for an issue and providing error correction and quality
would be forced to examine the entire notifications as set top boxes come on and
transport network. offline. Services such as rapid channel
change, forward error correction and video
4. Monitoring the head-end network error repair are all components affecting video
in a video assurance management quality of experience. A monitoring solution
solution. that seeks to be end-to-end must effectively
address monitoring in the last mile.
Head-ends are the heart of a video network
and can impact the largest number of 6. The role of video probes in the larger
customers should a serious or catastrophic video assurance management solution.
outage occur.
Video probes are common tools for video
Video content and payload quality on the monitoring. Video probes traditionally are
ingress at the head-end is critical. Any flaws placed at key demarcation points along the
associated with the payload at this point will video path. Monitoring at the ingress and
simply appear as poor quality on subscribers’ egress points of different domains in the
screens. A great deal of processing takes place transport informs the operator if the flow
at the head-end that can potentially affect the successfully has been transported through this
video. This is where most of the video domain.
encoding, transcoding, rate shaping, program
insertion, multiplexing and encryption takes Most probes collect data about each video
place. Head-end monitoring not only service by examining the MPEG transport
guarantees the quality of the content being headers carried within a multicast flow. Other
delivered to the subscriber, but also enables probes, which are generally more expensive,
the service provider to monitor the quality of are capable of examining the MPEG data
the content being supplied. The ability to contents directly for errors.
identify video content errors at the acquisition
source alleviates wasted truck rolls or time The information gained from the probes,
spent troubleshooting the IP transport, all of when combined with other video metrics, help
which reduces operational costs. isolate problems quickly in the delivery chain.
Video probes are an important part of any
The ability to identify video content errors at video assurance solution. The key is tying
the video source assures service providers these devices into the larger video monitoring
with the highest quality of service for their architecture that accounts for the many other
subscribers. aspects of video assurance and video service
delivery.
5. Monitoring the last mile is critical.
7. How important is the user interface in a
Last mile and home network problems video assurance management solution?
account for a majority of all quality issues
reported to service providers and must be Being able to monitor multiple service
effectively addressed. While probes are silos means viewing multiple screen outputs
effective monitoring tools, deployment to to determine where a video service may have
scale in the last mile is cost prohibitive. become impaired. This is not a simple task.

The use of a unified dashboard will Consumers do not care where the problem
prevent operators from associating diverse in the network originates. They do not care
alerts from various screens to a common whether it is in the head-end or at the set-top
video service. box or anywhere in between. The only thing
that consumers care about is consistent
A “single pane of glass view” is critical for delivery of quality video and a quality
network operations personnel. Complexity on experience.
how the events are collected, and from what
sources, should be hidden at this level. The Telecom operators, broadcasters, satellite
video service status is the critical element that operators and cable operators are delivering
should be represented and monitored. An more video services to end consumers today.
operator should be able to quickly diagnose Video is a complex, performance-sensitive
that an anomaly exists, the domain where an service that requires a network to provide
anomaly is occurring and the services that are excellent quality of experience to its
impacted. customers. To assure quality, cable operators
who deliver video services need a powerful
For lower-level troubleshooting, specific end-to-end video assurance solution.
engineering operators relevant to the problem We have reviewed the key drivers that can
domain where the anomaly is occurring can help service providers re-evaluate, redesign
be dispatched for further additional and build reliable, end-to-end and fault
diagnostics. tolerant video assurance management
solutions. Reduced subscriber churn and
8. The challenge for cable operators to create increased market share will belong to service
unity among various vendor offerings. providers who seize the opportunity to ensure
customers receive the highest quality of
Equipment from many vendors comprises experience through strategic and timely
the broadcast video network, while the IP network management investments.
network is primarily composed of routers and
switches. Head-ends and hubs have Cable operators constantly look at new
specialized equipment that captures or service models and quality metrics that
encodes video signals for insertion into the directly relate to customers' video
network and transport to the customer experiences. There are myriad variables to
premises. An effective video assurance measure and control in order to deliver an
management system can view, map and uninterrupted stream of broadcasts with a rich,
communicate with every component of this clear signal and crisp sound. If service
network. It must offer multivendor support providers neglect the management aspect of
and be flexible to accommodate additions, this new opportunity, they may face higher
changes and upgrades to the network. For costs when dealing with future operational
example, this flexibility would recognize the issues. One of the determining factors in their
addition of last mile drop points to support continued success will be an overall end-to-
new subscribers; understand software end video assurance management solution.
upgrades in switches and routers; and
accommodate new bandwidth-management
schemes over time.
SUMMARY Asha Kalyur is a marketing manager for Cisco. She

can be contacted at 408-527-4065 or
[email protected].

Wireless and Home Networking:
A Foundation for Service Provider Applications
Tim Burke - Liberty Global

Michael Eagles - UPC Broadband
Abstract
EVALUATING THE DRIVERS FOR
The explosion in the variety of SERVICE PROVIDER SUPPORTED HOME
Consumer Electronic (CE) devices and NETWORKING
applications that deliver video and internet In this section of the paper we explore the
experiences has produced the need for simple technology drivers for adoption of home
and integrated home networking solutions. networking. We consider the following
The cable TV community has analyzed, tested drivers:
and debated the viability of offering home
networking solutions for many years but  Proliferation of wireless networked
competitive pressures and technology devices.
advancements has finally prompted near term  Rise of Wireless Home Networking
action. Standards and Ease of Use.
 Rise of Personal Web Applications,
Although the home networking environmental Place-shifting, and Social Networking
conditions are reaching an inflection point,  Rise of Personal Digital Media and
the standards, consortiums and technologies Low Cost Home Storage, and the
are very fragmented and the network operator decline of Digital Rights Management
economics justifying a service offering can be (DRM).
marginal.  Access Competition, Product
Relevance and Substitution.
This paper focuses on various wireless home  Wireless Home Networking is
networking technologies and solutions. The happening today !
drivers for home networking are considered
and a variety of wireless home networking Proliferation of Wireless Networked Devices
configurations are discussed.
Today‟s digital home includes many
Preliminary test results from both a networking capable devices, with the range of
performance and economic basis are multi-media networked devices continuing to
evaluated. Of particular importance will be to increase. End users require home networking
assess the user and operator experience from in order to support many of these new
a set-up and maintenance perspective. devices.
The conclusion section will contrast the Multi-media Devices: Manufacturers are
technical and economic characteristics and starting to include network connections or
benefits of the various wireless home Wi-Fi technology in common devices like
networking solutions. A recommendation will digital still cameras and MP3 players and
be proposed that identifies areas of printers. Examples include Apples Wi-Fi
opportunities for wireless home networking enabled iPod touch, or iPhone; Eye-Fi‟s Wi-Fi
solutions for Multiple System Operators SD memory card; Archos personal media
(MSO’s). player; and the Hitachi Wooo camcorder or

Kodak Zx1camcorder that can stream live CPE devices, application layer standards for
video using Wi-Fi to the HDTV set. device discovery, and service provider.
Media Extenders & Streamers: A range of IP (b) Ease of use

connected media extender and streamer
devices have entered the market in recent Manufacturers such as Linksys (with its
years. Examples include Apple TV, LELA – Linksys EasyLink Advisor -
Netgear‟s recently announced ITV2000 software) have significantly improved the
Internet TV player.1 In addition we see installation, configuration, and maintenance
emerging streamers such as the Netflix Roku software for wireless networking devices.
device, and networked gaming consoles such
as the X-Box 360 which recently included the However it is still quite complex for most
Netflix service. It is no surprise that today‟s mainstream consumers and causes additional
TV require many HDMI ports ! call volume to ISP customer care centers.
Multi-room Audio: Multi-room audio Service providers can drive home networking
solutions that leverage home networking further toward mass market by building a
capabilities are emerging. Examples include proper integrated support ecosystem.
Sonos, Linksys by Cisco Wireless Home Installers and service providers could
Audio system 2, and Apple‟s Airport-based incorporate available tools to allow easy
wireless audio streaming; in addition to audio maintenance and upgrades.
streamers such as the Logitech‟s Squeezebox,
and Internet radio devices such as Tangent‟s
Quattro Internet radio. Rise of Place-shifting, Media multi-tasking,
Personal Web Applications and Social
Networked TVs: TV sets with built in Networking
networking are beginning to emerge. At CES
2009 for example Sony, Samsung, LG and (a) Place-shifting and multi-tasking
Toshiba were all introducing TVs with
Ethernet and/or Wireless connections 3 which Place shifting devices such as the Slingbox
could be used to display Yahoo! widgets. have emerged to enable services to be used in
any room. Place shifting devices and services
require a home network ideally based on a
Rise of Wireless Home Networking Standards solution that supports location flexibility or
and Ease of Use portability.
(a) Standards Emerging home server solutions such as

Microsoft Windows Home Server, and
Standards are critical in evaluating the drivers Linksys by Cisco Media Hub4 are emerging
for service provider supported home that support multi-user and remote access to
networking. Support for forwards and user generated digital content, requiring a
backwards compatibility with today‟s home network.
networked devices determines the quality of
experience subscribers experience and helps With the proliferation of devices there is
resolve end user issues. The key to achieving anecdotal evidence that multi-tasking is taking
this support is the extent wireless home place in the living room, combining for
networking technology can be embedded in example, the TV viewing experience with

laptop browsing. Wireless networking is an With today‟s explosion of multi-media digital
enabler of multi-tasking. devices such as Camera‟s, Handycam‟s, and
Multi-media Handheld device, users are
Figure 1: Living-room Multi-tasking generating more personal digital content than
ever before.
As highlighted in Table 1 below, it is

estimated that the typical U.S. broadband
household will have almost 1 terabyte of
personal digital media in the home by 2012.
Table 1: U.S. Household Digital Media

Growth 5
2008 2010 2012
(GB/hh) (GB/hh) (GB/hh)
Music 11 17 24
Photos 14 47 151
Video 201 347 723
(b) Personal Web Applications & Social With an expanding library of personal digital
Networking content including music, photos and movies;
the home network starts to play a pivotal role
The personalization of the web is evident in in allowing a user‟s multiple devices to
the explosion of web-based email platforms synchronize or access personal content
such as Gmail, Yahoo Mail and Windows between rooms and across devices.
Live Mail.
Furthermore, applications are emerging to
With the emergence of social networking support multi-device synchronization of
applications such as MySpace and Facebook, personal digital media such as Microsoft Live
combined with micro-blogging platforms Mesh, driving the need for home networking.
Twitter and Friendfeed; web-based
applications reinforce the personalized nature (b) Low Cost Home Storage
of networked communication.
In 1988 1 GB of storage in the home would
The speed at which web-based personal cost about US$40,000, so storing an 11GB
applications have been developed for wireless music library would cost US$440,000! That
networked devices, such as the Apple iPhone cost has dramatically declined so that in 2008
and Google‟s Android platform emphasizes 1 GB of storage cost of approximately
the importance of operator-supported wireless US$0.20 meaning that 11GB music library
home networking. could be stored for US$2.20. This dramatic
shift in the economics of storage has enabled
Rise of Personal Digital Media and Low Cost households to store large libraries of multi-
Home Storage, and the decline of DRM media content for consumption around the
home.
(a) Personal Digital Media

Figure 2: Consumer Price Per GB Declines 6 architecture requires an in home solution to
US$ Per GB for Magnetic Disk Storage enable multiple devices to connect to the same
100,000
service. This can be compared to Cable‟s
10,000 service group which can support multiple
1,000
devices with a dedicated CPE if required.
US$ Per GB
Figure 3: Telco Access Architectures Depend

100
10
on Home Networking
1
0
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
Additionally, small form factor storage, such
as SD cards, and smaller Mini and Micro SD
cards are supporting multi-media from a range
of new home wireless networked devices such
as Apple‟s iPhone and Google‟s G1 Android
handset. Such devices are able to both access
and contribute to the personal digital media As a result Telco‟s have deployed advanced
library. residential gateways for several years by
necessity.
(c) The demise of Digital Rights Management
(DRM) (b) Product Relevance and Substitution
Driving Service Provider Wireless Home
Recently Apple announced it was abandoning Networking
DRM protection for iTunes song downloads
in favour of a non-DRM model7. The fall of The traditional RBOC or Incumbent Telco,
DRM means that a barrier to multi-room has typically placed an emphasis on the
audio has fallen and this facilitates a wider residential gateway CPE device with the latest
device ecology for the consumption of multi- features as a way to attract subscribers,
room around the home. The key question is reporting the residential gateway CPE as a
whether this trend will extend to the video subscriber acquisition expense. This has
world? In particular, the ability to move provided a head start for Telco‟s in residential
MPEG4/H.264 HD content to various devices gateway adoption and wireless home
within the home at its much lower bandwidth networking penetration.
requirements could be a key enabler for whole
home wireless networking of video, data and A subset of Telco residential gateway
voice. deployments can be seen below in Table 2:
Access Competition, Product Relevance and

Substitution
(a) Access Competition Driving Wireless

Home Networking
Telco access architectures have been a driver

of home networking. The Telco point to point

Table 2: Telco Residential Gateways 8 Wireless is playing a larger role in home
Telco Gateway Sales/Units networking. As seen in the table below,
Product Deployed outlining the shift in European online
British Hub 3.5m HUB consumer home networking, there is a trend
Telecom (integrated boxes, about toward wireless becoming the default method
Wi-Fi and 100k per of home networking.
DECT) month
France Livebox 7.5m sold, Table 3: European Online Consumer Use of
Telecom (Integrated selling 300k Home Networking 10
Wi-Fi, USB per month
DECT dongle)
Deutsche Speedport Selling 200k to Q4 2006 Q4 2008
Telekom (Integrated 250k per Yes – Wired 12% 8%
Wi-Fi and month Yes - Wireless 13% 20%
DECT) Yes – Mixed /not sure 7% 12%
No 68% 60%
Substitution challenges can also be a driver Summary of Wireless Home Networking

for service providers. Wireless home Drivers
networking could be a unique selling point to
counter wireless mobile broadband offerings Proliferation in networked devices, ever-
in the market place. expanding user generated content libraries,
the desire for „anywhere‟ place-shifted
Wirelesss home networking is happening content access, the rise of the dynamic multi-
today !! media web, the rise of real-time social
networking, and new viewing behaviors are
Wireless home networking is happening today all driving the end user adoption of wireless
and solutions are emerging to support not only home networking.
basic connectivity but also other devices such
as gaming consoles, multi-media extenders Competitive access products from Telco‟s
and entertainment devices as outlined in the already include wireless home networking,
networked home in Figure 4. and the threat of mobile broadband
substitution creates additional urgency for
Figure 4: The Networked Home 9 fixed line service providers.
The message for the service provider: wireless

home networking is happening today, service
providers can either choose to participate and
add service provider supported features and
value for the subscriber, or watch from the
sidelines at the risk of market share loss,
mobile broadband substitution, and loss of
relevance in the home multi-media
experience.

THE POSSIBILITIES OF WIRELESS a reliable in home transport medium and
HOME NETWORKING further raises the bar for wireless as a total
home networking solution. As a total home
If a service provider, like a Cable TV networking tool the bandwidth, performance
Company, could utilize in home wireless and quality issues of video over wireless have
technology to help solve its home networking limited in-home wireless technology and
needs then many complexities of whole house could hamper its evolution beyond basic data
distribution could be solved. Wireless is and voice applications.
inherently a simple, flexible, low cost and
convenient medium for both the subscriber In this paper we will look at transmitting HD
and network operator. Wireless is a video throughout the home, replacing HDMI
particularly attractive option for the author‟s video cables within a room via “wireless
family of international companies because, HDMI” and traditional data/voice home
unlike the United States, most homes in networking.
Western Europe, Eastern Europe, Asia and
South America are smaller size, multi- Regulatory, technological and standardization
dwelling units and not pre-wired with coaxial advances in the wireless sector over the past
cable. few years has placed wireless home
networking in a position where it could
Wireless technologies (Wi-Fi, cordless and possibly meet the home networking challenge
DECT phones) are currently the solution of just described.
choice for in-home data communications and
phone services as illustrated by the chart Regulatory
below.
For instance, on the regulatory front, higher
Figure 5: Global Households with Wi-Fi power transmission of signals has been
Networks10 allowed in the 5 GHz Wi-Fi spectrum in
Europe. The spectrum available in Europe
and the U.S. are becoming more aligned for 5
GHz Wi-Fi, UWB (6-10 GHz) and even 60
GHz. Finally, larger blocks of contiguous
spectrum are being made available (~ 600
MHz at 5 GHz, 1.7 GHz at UWB, 7 GHz at
60 GHz).
Technology
Overall, wireless technology advances have

made huge leaps. Orthogonal Frequency
Division Multiplexing (OFDM) modulation,
Multiple Input Multiple Output (MIMO)
antenna schemes, large channel bandwidths
The progress in viable technologies (e.g.- (e.g.- 40 MHz channels), chip integration that
DLNA, UpNP, MPEG4 video compression) puts baseband and RF functions on a single
and applications (e.g.- Internet TV, Apple TV, chip and high quality video compression
Hulu, multi-room DVR…) that move video techniques provide economical and bandwidth
around the home has accelerated the need for efficient solutions. The combination of these
technological advances gives operators the

appearance that wireless whole home  Transmit power levels versus
networking is feasible even when interference trade-off‟s
entertainment applications are assumed.  Quality of service capability for video
applications
Standards  Compression versus latency versus
bandwidth tradeoff‟s for video
Finally, the standards bodies, and in particular applications
the IEEE, have made tremendous strides at
continually updating and improving their The interplay and relationships between the
wireless specifications. The 802.11 working success criteria results in complex systems
groups have created new amendments and architectures.
standards in many areas (e.g. - high
throughput/802.11n, Quality of Service/ Consumer Electronics (CE) Device Volumes
WMM, Security/WPA2, Power Save/APSD)
that help improve service levels and A large ecosystem of wireless home
capabilities. Although these specifications networking chipsets and devices are required
take longer than desired to become standards to reach the proper economics in the C.E.
the progress made in the difficult political world. But success in obtaining the proper
environment of the standardization process is device scale in the unlicensed, unregulated
impressive. spectrum realm of the home network means
too many wireless devices operating in the
THE REALITY OF WIRELESS HOME same spectrum which causes interference,
NETWORKING quality and capacity issues. It has taken about
5 years for the 60 MHz of Wi-Fi 2.4 GHz
All the positive indicators just mentioned spectrum (three 20 MHz channels) to become
must be tempered with the reality of the too congested using just voice and data
wireless medium. The wireless channel is applications. In some ways, the great success
unstable and unpredictable in an outdoor of Wi-Fi has bred failure for wireless as a
environment but becomes extremely variable home networking solution.
indoors as floor plan layouts, furniture, walls
and living quarter sizes vary widely. The Spectrum Availability
characteristics of the Radio Frequency (RF)
channel can change rapidly over a time period As a positive consideration, the 5 GHz Wi-Fi
due to fading and interference. Consequently, frequency is relatively greenfield, contains
the capacity of the channel and signal strength ~500 MHz of contiguous spectrum (twenty-
(signal to noise ratio or SNR) seen by the four 20 MHz channels) and is available
receiver fluctuates constantly. almost uniformly across both the U.S. and
Europe. The future success of 802.11n
The evaluation of wireless whole home devices combined with the higher bandwidth
networking solutions must be looked at along requirements of video applications means a
five critical success criteria. similar 2.4 GHz congestion problem could
eventually occur. The capacity and quality
 High volume components & chips to limits at 5 GHz would occur first in the denser
meet consumer electronic (CE) device urban areas of apartments and multi dwelling
economics units and in areas where public Wi-Fi
 Sufficient and uniform spectrum at the networks are operating in the same
right frequencies frequencies. Anticipating this eventuality,
standards bodies and start-up companies are

venturing into the higher frequencies of UWB Quality of Service (QoS)
(6-10 GHz) and 60 GHz spectrum where
much larger blocks of spectrum are available. Whole home wireless solutions that
Unfortunately, higher frequencies have the accommodate delay sensitive applications
well known disadvantages of severely limiting such as video and voice must at a minimum
the range the signal can go and requiring be able to prioritize various traffic types. The
higher power and costs. For this reason UWB 802.11 standard that specifies QoS is called
and 60 GHz solutions have been currently WMM (Wireless Multi-Media) and has the
relegated to in room solutions only. potential to ensure quality video transmissions
within the home. Unfortunately, wireless
Transmit Power Levels home networks have not reached the point
where they will guarantee and reserve
In general, transmit power limits in wireless bandwidth by offering parameterized QoS as
devices have similar economic, cost and is done within the cable TV and DOCSIS
interference trade-off‟s. The EU has recently network of a cable operator.
increased the allowable transmit power limits
across most of the 5 GHz Wi-Fi frequencies Compression, Latency and Bandwidth
to 1 Watt.11 Although this certainly helps in
getting whole home networking solutions to In the wireless home networking arena a
work in houses and apartments that are larger debate rages over the effects of transporting
and made of stronger materials the RF energy uncompressed (e.g.- HDMI) versus a
from adjacent networks will raise the noise compressed video signal (e.g.- MPEG2 or
floor and interference levels. Higher power MPEG4/H.264). No wireless technology
levels in devices also translates into additional transports true uncompressed HDMI as a
costs. Transmit power levels above a certain baseband video signal. Other than the large
point (~100 mW) limits the ability to integrate spectrum range of the 60 GHz frequencies, all
a power amplifier (PA) into a chip as non- other in-home wireless technologies use some
linearity and peak to average power ratios sort of real time compression. Compression
cause problems in chip designs. Outboard (e.g.- H.264) will be needed to send video
PA‟s typically translates into higher costs. over wireless especially as video scales up
with higher frame rates, deeper colors and
Figure 6: Wi-Fi 802.11n 5 GHz Frequencies higher resolutions.13 Some vendors introduce
and Transmit Power12 the concept of lossless versus lossy
compression to further differentiate their
products. Traditionally the industry has
regarded any compression above 4X
compression ratios as lossy compression.
Compression at lower compression ratios
allow it to be categorized as lossless
compression.14
Linked in with this discussion is the latency

issues associated with transcoding from one
codec to another when dealing with
compressed video, the ability to move a
compressed signal across devices while still
complying with all DRM considerations and
the ability to add the graphic overlay for an

EPG (Electronic Program Guide) over the voice applications. The major alternatives
compressed signal. Obviously a compressed investigated in this paper include:
signal has substantial advantages in moving
video content around the home as the  DECT (Digital Enhanced Cordless
bandwidth requirements of MPEG2 HD are Telecommunications)
15 to 20 Mb/s (7 to 10 Mb/s for  High Throughput Wi-Fi (802.11n)
MPEG4/H.264) while uncompressed HD  Variations To High Throughput Wi-Fi
requires > 3 Gb/s of capacity. (802.11n) Specification
 Optimized Video At 5 Ghz Or
WHOLE HOME NETWORKING Proprietary Wireless HDMI.
REQUIREMENTS  Ultra Wide Band (UWB) and (IEEE
802.15.3c or 802.11ad) technology
The CableLabs OpenCable Home Networking
study group has spent a considerable amount Each of these four alternatives have a number
of time defining the various use cases for of start-up and established companies pushing
moving video, data and voice applications their particular technology and specification.
throughout the home. As a result of this effort Many have established consortiums of
the minimal bandwidth, performance companies and industry associations and
requirements and architectures of a home consortiums of companies with the intent of
network have been proposed. bringing their specification to a
standardization body for approval. The table
A typical deployment scenario of one Set Top below summarizes the major technology
Box (STB) with storage capability and two options. In all cases the standardization
STB‟s without storage depicts a requirement process is ongoing and in many cases at an
for four times the maximum bandwidth for a early stage. Only the Wi-Fi 802.11n standard
single HD stream of MPEG2 video content is very near completion and has a large
being transported between network elements ecosystem of chipsets and consumer
in the home. Assuming a 20 Mb/s per HD electronic devices currently being built.
stream requirement (MPEG2) a consistent and
reliable 80 Mb/s wireless network is needed Table 4: Wireless In-Home Alternatives16
within the home that offers full Quality of Asso
Spectr Standar
Major
Chip
Data
Rate Indoor
Service (QoS) and prioritization (or better yet Name ciatio
n
um d Body Supplie Clai Range
reserved capacity) of video media content.

rs ms
1.88–
64Kb
CableLabs assumes another 20 Mb/s for best DEC
1.9GHz
in DSP /s,
efforts based data traffic and prioritized voice DECT
T
Foru
Europe,
1.92–
ETSI
Group
and
300K
b/s
~50m
traffic. Therefore the MAC throughput data m 1.93GH

zin the
SiTel (CA
T-iq)
rate of at least 100 Mb/s is required to support US
Intel,
Wi-Fi 2.4
all types of in home networking video, data High
™All GHz IEEE
Broadco 300-
Throughput m 600 ~100m
and voice streams.15 Wi-Fi
iance
17
and 5
GHz
802.11n
Marvell Mb/s
…
Wirel
OVERVIEW OF TECHNOLOGY OPTIONS ess
Hom
FOR WIRELESS HOME NETWORKING Optimized
Video at
e
Digit IEEE
5GHz 1
al 5 GHz 802.11a Amimon ~50m
(Proprietary Gb/s
There is a range of wireless technology Wireless
Interf
ace
a, ac
HDMI)
options, associated frequencies, existing or (WH
DI
emerging standards and industry associations ™) 18
that could possibly meet the demanding home Ultra Wide WiM
6 to 10
Formall TZero
480
Band edia y IEEE Radiospi <10m
networking requirements of video, data and (UWB) ™ 19
GHz
802.15.3 r
Mb/s

a now
ECMA-
Sigma
Designs,
led to the tremendous growth in the industry.
368 PulseLin The chart below outlines some of the key
k…
IEEE milestones in the 802.11 committee‟s
Very High
Throughput
Wirel
ess
802.13.3
c and progress.
60 GHz HD 60 802.11a 3
SiBeam <10m
(Proprietary ™ GHz d Gb/s
Wireless (WiH formally Table 5: Key IEEE 802.11 Standards22
HDMI) D) 20 802.11v Standard Date Description
ht 802.11a 1999 First 5 GHz standard. Incorporates
OFDM technology with speeds up to
54 Mb/s
DECT Digital Enhanced Cordless Telephone 802.11b 1999 First standard to gain wide adoption.
Operates in the 2.4 GHz using DSS
CDMA technology and 11 Mb/s
DECT was developed by ETSI but has since 802.11g 2003
speeds
Operates at 2.4 GHz but employs
been adopted by many countries all over the OFDM technology at speeds of 54
Mb/s and is backwards compatible
world. The original DECT frequency band with 802.11b
(1880 MHz–1900 MHz) is used in all 802.11n 2008 for Draft
2.0
Next generation standard that uses 2.4
& 5GHz and leverages MIMO,
countries in Europe. Outside Europe, it is beamforming to produce 600 MB/s
speeds
used in most of Asia, Australia and South 802.11e 2005 Provides support for Multimedia
America, making it well suited to the author‟s Applications with Quality of Service,
called WMM
international systems. 802.11i 2007 Adds security features to improve on
WEP using Advanced Encryption
Standard (AES)
In the United States, the Federal The innovative technologies employed in the
Communications Commission in 2005 802.11n specification results in greater
changed channelization and licensing costs in throughput and reliability over previous Wi-Fi
a nearby band (1920 MHz–1930 MHz, or 1.9 solutions. Data rates as high as 600 Mb/s, 10
GHz), known as Unlicensed Personal times greater than 802.11 a/g previous
Communications Services (UPCS), allowing standard, are possible with 802.11n. Figure 7
DECT devices to be sold in the U.S. illustrates the dramatic boost in network
capacity and ultimately the speeds possible
Although the DECT data rate makes it for in-home wireless networking applications
unsuitable for high speed data or video, many using 802.11n technology. Contrasting this
voice service providers consider that “DECT development with today‟s 802.11a/g speeds of
is the go to platform for voice calling in the 54 Mb/s and typical wired Ethernet 100 Mb/s
home”21. Further, as a home wireless throughputs begins to put this advancement in
technology for voice, DECT has the perspective.
advantage of a well established ecology of
handset device manufacturers. Additionally Figure 7: Evolution of Wi-Fi Throughput 23
many subscribers are familiar with DECT, the
technology does not require additional
subscriber education.
High Throughput Wi-Fi (802.11n)
Wi-Fi is the wireless LAN technology brand

developed by the Wi-Fi Alliance to certify
IEEE 802.11 devices. In a little over a decade
Wi-Fi has evolved from an innovative idea
into an indispensible technology for
consumers. The original standard has been
continually updated and enhanced, which has

The 802.11n specification is built on the performance and throughput from its
cornerstones of OFDM and 802.11a and predecessors are:
802.11g standards. In this technique the  Multiple Input Multiple Output
usable bandwidth is divided into a large (MIMO) technology
number of smaller bandwidths or subcarriers.  Enhancements in modulation and
These subcarriers are mathematically coding schemes
orthogonal or unique and can be tightly  Packet overhead improvements called
packed next to each other to gain maximum packet or frame aggregation.
spectral efficiency. The high speed  Channel bonding (40 MHz channels)
information to be transported is then divided
into multiple lower speed signals and (a) MIMO Technology
transmitted simultaneously on different
frequencies (subcarriers) in parallel.24 Both MIMO uses multiple radios and antennas to
802.11a/g and .11n specifications utilize 52 allow different data to be simultaneously
subcarriers spread across a 20 MHz channel transmitted across multiple transmit antennas.
bandwidth. Figure 8 illustrates the various On the receive side of the wireless data link
subcarriers and how the data to be transported the separate unique data streams go across
is distributed across the subcarriers in both free space on different paths and are received
frequency and time. by the separate receive antennas. Because of
the spatial diversity of the multiple paths
Figure 8: OFDM Technology 25 through the air, MIMO systems are able to
transmit two unique data streams. This results
in data transmissions at twice the data rate of
single antenna and radio systems. This
concept is called MIMO Spatial Multiplexing
(SM) and is illustrated in Figure 9. In Figure 9
the original data stream or message (101011)
is sent simultaneously as two different
messages (101 and 011) across two different
radios, antennas and spatial stream paths to
the receiving device antennas. The 802.11n
standard can support four spatial streams
across four transmit and receive antennas.
Figure 9: MIMO Spatial Multiplexing

A major advantage of OFDM and 802.11n is
Technology26
its ability to tolerate multipath fading by
carrying small amounts of information on
individual subcarriers or frequencies. If one
or two frequencies/subcarriers are lost due to
a fade then only a small amount of
information is lost which can be compensated
for via error correction coding and
retransmissions.
The primary innovations that allowed the

802.11n standard to make such a large leap in

(b) Enhancements in Modulation and Coding (d) Channel Bonding
Schemes
802.11n networks gain an immediate capacity
802.11a/g Wi-Fi networks use a single stream increase by implementing the concept of
and eight possible modulation and coding channel bonding. Combining two 20 MHz
schemes resulting in eight possible data rates. channels into a 40 MHz channel is a very
802.11n MIMO technology can implement a straightforward way of increasing the
concept called rate adaptation and variable bandwidth and capacity of the system. This
modulation and coding schemes. If the RF technique is utilized more effectively in the 5
conditions are good and the signal strength of GHz range as there are twenty-four 20 MHz
the receiver is strong a higher modulation channels to work with versus the three 20
level (e.g. 64QAM) and weaker error MHz channels in the 2.4 GHz frequency.29
correction code can employed at that instant
for the data stream. This results in a higher The combination of a 40 MHz channel, two
data rate for that particular data stream. As an spatial streams and the packet aggregation
example of the power and complexity of the gains of frame aggregation will provide a 300
802.11n specification there are 77 different Mb/s maximum data rate. If four spatial
modulation and coding schemes (MCS) streams are used this peak data rate can
possible where 8 MCS schemes are increase to 600 Mb/s.
mandatory.27 The end result of the multiple
spatial streams, coding and modulation (e) Implementation Issues
options, are the possibility for much higher
data rates. To ensure a seamless transition to the newer
802.11n technology the specification was
(c) Packet and Frame Aggregation designed for backwards compatibility. Legacy
clients (802.11a/b/g) will operate without
Reducing the packet overhead required for problems in an 802.11n network. For
data transmission is critical for reliable high instance, 11a clients operate in the 5 GHz
speed delivery of data operating in the spectrum and will continue to do so in a 5
802.11n mode. In conventional wireless GHz 802.11n network.30
transmission systems the amount of overhead
is fixed regardless of the size of the data The downside to this configuration is a
packet. As the data rate increases the reduction in performance when older devices
overhead remains the same. The 802.11n operate in a 802.11n system. Legacy clients in
specification has fixed this problem by an 11n network will reduce the overall
aggregating multiple packets of overhead into throughput of the 11n system. The peak
a single transmission frame.28 performance of an 11g client is ¼ that of 11n.
So if an 11g client is operating at 10 Mb/s
Unfortunately, aggregating multiple overhead then the 11n capacity will only be at 40 Mb/s
packets can cause an increase in system (1/4 the 11n system).31
latency as the radio must hang onto packets at
the transmitter until it creates a transmission As previously mentioned transmit power is
frame of the desired size. For that reason important and can be a key implementation
some real time applications (e.g.-voice) do not issue for 802.11n systems. Additional power
utilize the benefits of packet aggregation. is required for each additional MIMO antenna
and to operate in the wider 40 MHz channel
mode. Each MIMO antenna will require a
separate PA and RF chain so power

consumption and resulting cost is increased.32 taken into account when Wi-Fi systems are
Likewise, to keep an equivalent range shown being implemented.
in a 40 MHz system for a 20 MHz
implementation will take much more transmit Variations To High Throughput Wi-Fi
power. (802.11n) Specification
Unfortunately the gain of a MIMO Spatial Because the 802.11n specification has
Multiplexing (SM) system assumes the provisions for many optional features there
existence of multipath and uncorrelated exists the ability of a chip designer and access
signals for each of the spatial streams being point developers to offer a variety of
transmitted. Radio signals need to reflect off standards compliant 802.11n
of walls and furniture to cause the receiver to implementations. The major suppliers of
see multiple representations of the same 802.11a,b,g chipsets and start-up companies
signal arriving at different times and have begun to optimize video transport within
amplitudes. In typical residential homes this the home over the 5 GHz spectrum associated
phenomena occurs but situations could arise with 802.11n.
where the intended multipath effect is not
present thereby decreasing the benefit of The ability to use the additional 20 and 40
MIMO and increases in system capacity. MHz channel bandwidths, aggregate baseband
and RF functions onto a single System on a
In 802.11n systems it is important for all Chip (SoC), utilize MPEG4/H.264
traffic to be classified as priority, best efforts compression and take advantage of MIMO
or background type traffic. Implementing the technologies have led to many advancements
Quality of Service traffic classifications called in video distribution within the home.
Wireless Multi-Media (WMM) will be a
difficult but important process. The correct Some new and innovative chip design shops
tagging and classification will be critical to have creatively utilized the various core
properly manage the traffic in a wireless home technologies listed below to great advantage
network. Not all consumers will be up to this while still complying with the standard.
task and could cause performance issues that
will reflect poorly on the technology.  Beamforming MIMO (simultaneous
transmission of the same data streams
A final implementation and design over multiple antennas)
consideration in the 802.11n specification are  Adaptive channel modulation and
the use of two interference control coding options
mechanisms called Dynamic Frequency  Enhanced QoS (prioritizing video,
Selection (DFS) and Transmit power Control data and voice services)
(TPC). DFS is a feature that checks for the  Radio resource techniques (utilizing
presence of military radar operating in the 5 feedback information from the client
GHz frequencies and if detected requires the devices while transmitting)
802.11n device to utilize other available  Linking compressed video (H.264)
frequencies. TPC requires 802.11n devices to chip and MIMO RF technologies
reduce their transmit power if they are  Integration of advanced Power
operating very close to each other.33 Amplifier (PA) and antenna designs
Although these interference controls are a
burden to the 802.11n system developer they As these many variations of the 802.11n
are not hugely disruptive issues but must be standard enter the market it will be very
important for the Wi-Fi Alliance and network

operators offering wireless home solutions to over integrated MIMO antennas to
properly certify interoperability across access accomplish the beam forming or optimized
points and client devices. The beamforming RF patterns.35
variants discussed here add the most gains and
consequentially the greatest complexity and (b) Chip Level MIMO Beamforming Designs
interoperability issues.
There are many creative companies
(a) 802.11n Standards Compliant developing chip level MIMO beamforming
Implementations: Beamforming MIMO designs. Because of the inherent bandwidth
restrictions on beamforming MIMO, most
MIMO technology has the two major design chip level MIMO beamforming designs utilize
choices of either Spatial Multiplexing (SM) or an H.264 codec chip along with the baseband
beamforming. By choosing the beamforming and RF chipset to ensure guaranteed video
approach the video transport can be optimized quality and wireless transmission of multiple
for high performance and distance. HD video content.
Beamforming MIMO allows for the control of
the RF signal and actually steers the signal Figure 11: 802.11n Chip Level MIMO
from the access point to where the client Beamforming Architecture36
device is located. These RF pattern
adjustments are made very quickly from
moment to moment if either the client device
moves, multiple clients are in the range of the
Access Point (AP) or the RF conditions vary
naturally. By choosing the beamforming
MIMO option over SM MIMO the 802.11n
chip or system designer typically foregoes the
bandwidth gains of SM MIMO. In many cases
multiple unique data streams are no longer
In one such design a 20 MHz channel carries
being transmitted from each antenna
three MPEG4/H.264 streams (1080p, 60fps)
simultaneously.
at between 8 to 10 Mb/s each. Vendor testing
claims of a consistent 30 Mb/s (UDP
Figure 10: 802.11n Beamforming MIMO
transmission) of capacity across 25 meters
Architecture34
and through multiple walls in a suburban
home are one of many published 802.11n
results.37
Another key beamforming MIMO chip design

option involves either having the
beamforming performed at one end (AP side)
or both ends (AP & client). In one particular
implementation the semiconductor vendor
In the 802.11n specification there are chip
focused on delivering optimized
level and system level implementations of MPEG4/H.264 HD video content where its
beamforming. Chip level designs typically use
chip is located at the transmit end of the video
mathematical and DSP based processing stream only. Therefore, it is placed in a
power to manipulate the phase of the central distribution point (access point, home
baseband RF signal. As shown in Figure 10, gateways, multi-room DVR‟s) of HD content
these phase shifted signals are transmitted out

and interworks with standard off the shelf streams) is sent simultaneously over the 2 RF
802.11n chips (and MPEG4 codecs) in the chains using 2 PA‟s where 4 antenna choices
client device such as a remote STB or HD TV are possible.39 Another chip level
display.38 The reliability and distance benefit beamforming vendor chose to utilize 4 RF
of beamforming MIMO combined with a low chains, four PA‟s and 4 antennas.40 This
total cost for a whole home implementation design probably has more power and more
(because of the ability to interoperate with beamforming patterns possible but probably at
low end 802.11n chipsets) is a very a higher cost.
advantageous design. A second major chip
vendor decided to implement their The beamforming software in a particular
beamforming MIMO design by using their vendors design is the “secret sauce” for their
chip is used at both ends of the link. The MIMO implementation. Software that
result is probably a higher performance design chooses the best antennas and in effect creates
but at a greater cost. an optimum transmit antenna pattern can
make all the difference in performance. The
The decision to use either 20 MHz or 40 MHz ability to have many beamforming pattern
(two bonded channels) channel bandwidths is options and continually (and quickly) adjust
likewise important. If the 802.11n the pattern due to the varying conditions of
implementation utilizes a single 20 MHz the RF environment within the home cannot
channel in the 5 GHz frequency to transport be underestimated. The 802.11n standard
MPEG4 HD content then the transmit power allows for information from the client side of
and antenna gains can be more focused and the link to be sent back to the transmitters so
contribute to a better range result. An that these software decisions can be made.
implementation using two bonded 20 MHz
channels (40 MHz) will need higher transmit Unfortunately, there are multiple options and
power to get the same distance and much complexity written into the standard for
throughputs (all things being equal). The this critical feedback information. For
second start-up vendors beamforming chip instance, the implicit feedback option means
design utilizes 4 transmit and 4 receive the 11n client provides very limited
antennas in their chip with a 40 MHz channel. information and the chip vendors
This chip design is an excellent way to get implementation calculates how it should
both high capacities (40 MHz channel) and change its beamforming from moment to
also reliable and long range performance (4x4 moment. The explicit feedback option in the
Beamforming MIMO). standard spells out the continuous feedback of
information the client must send. To date, no
Additionally, the transmit power and number chip manufacturers of client end only chips
of power amplifiers decision is crucial as it is have implemented this portion of the standard.
a major contributor of cost and ability to Start-up chip vendors that have their MIMO
integrate on a chip and circuit board level for beamforming solution on one end only are
vendors. Transmit power level of up to 100 utilizing implicit feedback while those with
mW are known to provide the best linear chips on both ends use some form of explicit
characteristics of the highly modulated signals feedback to adjust the beam.41
(e.g.- 64QAM) of the 802.11n specification.
Likewise, the number of RF chains and Because of the many options and complexity
quantity of MIMO antennas to choose from the interoperability testing and certification of
are critical performance and cost decisions. In the beamforming portion of the 802.11
one illustrative beamforming design the entire standard is very much in its infancy and will
video stream (comprised of 3 MPEG4 be an extremely difficult area to enforce.

Suffice it to say, when beamforming is used chipset from a traditional 802.11 chip vendor
between different vendors equipment the is combined with 3 PA‟s and RF chains to 6
performance will most likely fall off directional antennas. At any point in time an
considerably and firmware or even hardware optimized beamforming pattern is sent out
changes will be require for interoperability. over the best 3 antennas using knowledge
obtained from the RF channel and the far end
(c) System Level MIMO Beamforming wireless adapter. At any point in time
Designs hundreds of different antenna patterns are
possible that can add gain to the system
Companies that utilize off the shelf 802.11n performance or even reject interference.43
chips and add their own hardware and
software implementations of beamforming As in many of the other creative
MIMO technologies are known as system implementations of the 802.11n standard the
level MIMO beamforming designs. Their system level MIMO beamforming designs are
area of expertise is in the design of enhanced only able to get the improved range and
multi antenna systems and beamforming bandwidth performance claimed when their
software algorithms that continually create MIMO beamforming units are used at both
and direct new RF patterns. ends of the video or data stream. Optional
software programmable configurations are
Figure 12: 802.11n System Level MIMO possible to allow the main access point to
Beamforming Architecture42 communicate with industry standard, off the
shelf 802.11n adapters and extenders in
laptops and other devices, but the
performance will be reduced.
A nice advantage of the system based solution

is that it is not optimized for video at the chip
level and therefore offers a true bidirectional
home networking solution where both video
and data services can be transported over the
data stream. QoS (802.11 WMM) is
implemented so prioritized video and best
Key enablers for system level MIMO efforts data transmission share the 300 Mb/s
beamforming designs is the ability to create of capacity.
many different patterns using multiple
antenna choices, directional antennas, A key challenge for the network operators
sophisticated software that continuously incorporating higher end implementations of
adjusts and changes patterns, and efficient the 802.11n standard in their home
power amplifiers. networking solutions is the ultimate
performance these devices obtain when
One such vendor‟s implementation is able to deployed in a mixed operating mode. The in
get high capacity, performance and range by home network performance may match claims
utilizing directional antennas (6 in 1 of their 5 in a perfect end to end single vendor
GHz implementations), beamforming environment but degrade severely when a
technology and the higher transmit power variety of vendors and chipsets are used in a
possible with outboard power amplifiers home network. In addition, because the
(250mW). A standard 3x3 MIMO baseband beamforming implementation at their access
point is a unique implementation of MIMO

beamforming it may not be compatible with video source allows the Amimon reference
standard 802.11n client device design shown in Figure 13 to optimize the
implementations (such as explicit chip level wireless connection for video applications.45
feedback designs).
Figure 13: Optimized Video Proprietary
Architecture46
Optimized Video At 5 Ghz or Proprietary
Wireless HDMI technology.
Targeting the large, unused, contiguous

spectrum available to Wi-Fi technology in the
802.11n 5 GHz frequencies some vendors
have chosen to create a unique specification
targeted for video transmission and wireless
HDMI applications.
Amimon is one such start-up silicon vendor

that has designed a very effective chip design
As in all proprietary designs the key for this
optimized for video transmission at 250 to
technology will be to obtain standards
800 Mb/s. They have created a consortium of
approval for the reference design in the
vendors (called WHDI), that includes CE
802.11 study groups and then build an
manufacturers, to create a specification and
ecosystem of supportive CE vendors.
standard for wireless HDMI.44
Ultra Wide Band (UWB) Technology
Their technology also takes advantage of the
(ECMA-368)
multi antenna gains possible with MIMO
technology (4x5 MIMO) and 40 MHz wide
Ultra-Wide Band is a radio technology that
channel bandwidths.25 As a result, large
uses a 528 MHz channel bandwidth
capacity video streams are possible with their
(compared to the 20 or 40 MHz channels of
implementation.
802.11n) within various frequencies from 3.1
to 10.3 GHz. The technology is heavily
In addition, they claim their transmission does
promoted by the WiMedia™ Alliance and
not use compressed video. Amimon appears
utilizes the well known OFDM technology
to perform some form of compression but at
and is being used to transport compressed
lower compression ratios (< 4X) that allow it
video (1080p at 30 fps) within a single room
to be categorized as lossless compression, or
only.
for their definition, uncompressed.
WiMedia UWB can support data rates of up
The key to their uncompressed video is their
to 480 Mb/s and transmits at very low power
ability to combine real time video processing
levels. The FCC has placed transmit power
of the video source with channel coding and
levels (.1 mW) on these frequencies as a
modulation function present in traditional
majority of the bands compete with other
wireless transmission systems. The Amimon
users operating in these bands. As a result of
chip prioritizes the source video components
a .1 mW peak transmit power limit spread
according to their importance and only keeps
across 528 MHz the Power Spectral Density
the most significant bits. By tightly linking an
(PSD) is quite low for UWB technologies
understanding of the varying RF channel with
whch results in operation across very short
the actual encoding and processing of the

distances (< 10 meters). Because UWB offering UWB technology over coax as the
technology operates across such a wide WiMedia group seems to be retrenching
channel bandwidth it is very resistant to slightly from the pure wireless applications.
frequency fading. By utilizing 128 OFDM 4
MHz subcarriers across the full 528 MHz One of the biggest disappointments for the
wide channel it is able to guarantee quite high UWB technology and the WiMedia
quality HD video with its built in frequency Association has been their inability to obtain
diversity. IEEE 802.15.3a Personal Area Networks
(PAN‟s) standards approval as the
Additionally, there is over 7 GHz of spectrum specification languished for three years.
available in the various band groups from 3.1 Eventually the committee was disbanded and
GHz to 10.3 GHz. Only the U.S. has access the WiMedia Association was forced to gain
to the full allocation as Europe is allowed to approval from the ECMA international
use only 3.250 GHz of spectrum. Also in standards organization49.
many of the bands (e.g.- 3.5 GHz) UWB
operation is required to Detect And Avoid It appears that the biggest issue associated
(DAA) outside operation from WiMAX, radar with the UWB technology is that the very low
operators and others before transmitting in power limits imposed by the FCC have
those frequencies. limited its applications to in-room compressed
HD video. This limited application is being
Figure 14: Frequency Diagram47 outstripped by in-room uncompressed video at
60 GHz or whole home compressed HD video
possible with the many 802.11n technologies
and their variants at 5 GHz. At this writing
the UWB technology appears to have become
a shelved technology that has never lived up
to its original hype.
Very High Throughput 60 GHz Technology

(Proprietary Wireless HDMI)
The millimeter wave spectrum is very

advantageous for the high bandwidth
requirements of transporting HD video signals
because of the large amount (7 GHz of
spectrum between 57 and 64 GHz) of unused
spectrum available in the 60 GHz frequencies.
Although the UWB specification originally
This technology certainly comes closest to
started out originally with much promise but
pure wireless HDMI as full HD quality video
has had much difficulty gaining traction in the
at 1080p and 60 fps is possible over a 4 Gb/s
marketplace and standards bodies over the
stream including the A/V control signaling
years. Seven major vendors such as T-Zero,
associated with HDMI. The well known
WiQuest, Radiospire, Pulselink, Focus
benefits of transporting uncompressed video
Semiconductor, Artemi and Intel were
are therefore its major attraction. Additional
recently forced to either shut down or merge
benefits of 60 Ghz include the uniformity of
their UWB operations leaving just two or
this spectrum availability across the U.S.,
three vendors left in the WiMedia space.48
Europe, Japan, China and other major
Most remaining vendors appear to begin
markets. The combination of large spectrum

bands and wireless technology advances of TEST RESULTS OF VARIOUS WIRELESS
MIMO makes multi gigabit rates possible. HOME NETWORKING SOLUTIONS
Unfortunately, the well known downside of Preliminary Findings

60 GHz transmission is the short distance
limits due to signal strength losses through Preliminary test results of unmodified
free space. Although the antennas possible at 802.11n draft technology, based on a sample
these high frequencies can be made very small of standalone gateway devices that utilize
and compact the maximum transmit power 2.4GHz and Gigabit Ethernet LAN and WAN
requirements for even very short distances are ports provided an insight into the potential
quite large. SiBeam is the major chipset challenges associated with delivering triple
vendor in this area and is the leader in play services over today‟s wireless home
promoting their specification for network solutions.
standardization. They require a 7 Watt power
amplifier to transmit one video stream 10 The test configuration included multiple
meters in a Non Line Of Site (NLOS) mode services (Video, Voice and Data) over each
or 25 meters with Line Of Site (LOS).50 wireless home networking device in specific
Further integrated circuit advancements may directions as outline in Table 6.
reduce the power to 4 Watts and eventually 2
Watts but the large power requirements and Table 6: Triple Play Test Traffic Direction
resulting high total system costs continue to
make this technology very much a niche Abbreviation Description
offering. L2L LAN to LAN
W2L WAN to LAN
The ability to reduce power, costs and obtain L2W LAN to WAN
standardization will be very important for 60 W2R WAN to Radio
GHz technology adoption. At this writing the R2W Radio to WAN
SiBeam specifications are well positioned for L2R LAN to Radio
being pushed through the 802.11ad Very High R2L Radio to LAN
Throughput (formally 802.11vht) task R2R Radio to Radio
group.51 The WirelessHD™ (WiHD)
consortium has also been effective at Key findings indicate that when the home
promoting this specification within the wireless devices were loaded with service
standards bodies. As is sometimes the case, provider traffic that Gigabit Ethernet ports
this specification is also being promoted in the including LAN to LAN performance and
IEEE 802.15c working group which is part of LAN to WAN performance met performance
the Personal Area Networks (PAN) study thresholds, where as most test that required
groups. use of the Radio (i.e. Radio to Radio and
L:AN to Radio did no meet requirements.
Because this technology and the frequency it
operates in will most likely always be an in-
room solution only, its applications and uses
will most likely are limited and therefore be a
primarily a niche technology for home
networking solutions.

Figure 15: 802.11n Wireless Test Flows Testing in 5GHz bands indicates some
significant improvement over 2.4GHz and as
more devices become available in the 5GHz
band this will be the subject of additional
testing.
Further testing with a wider range of wireless

home networking devices during 2009 will be
the subject of a future paper.
Future evaluation and testing 802.11n

More detailed results are summarized in the Solutions for Home Networking
table below in Table 7, with results in bold
indicating they fell outside the threshold The key criteria in evaluating
requirement. implementations will be:
Table 7: 802.11n Triple Play Performance  Capacity of multi-play capabilities, for

example to determine whether the
802.11n, 2.4GHz
Service Data 1-Way 12 Mbps Voice
entire channel is dedicated for video
Video G.711 transmission or other services are
L2L: 390.1 L2L: 11.9 Not tested
possible on the bandwidth
Throughput
(Mbps) W2L: 117.6 W2L: 11.9
Required > 100 L2W: 41.3
W2R: 8.9
L2W: 11.9
W2R: 11.9
simultaneously.
R2W: 15.1 R2W: 11.8  Consistent and reliable performance
L2R: 11.7
such as voice/video quality (latency,
L2R: 8.6
R2L: 7.5 R2L: 11.8
% Bytes Lost
R2R: 3.2
Not tested
R2R: 10.9
L2L: 0.0 L2L: 0.0 jitter and packet loss) over range and
Required: < 0.1 W2L: 0.0
L2W: 0.0
W2L: 0.0
L2W: 0.0
network load.
W2R: 0.6 W2R: 0.2  Effective throughput and capacity over
R2W: 1.0
L2R: 1.8
R2W: 0.4
L2R: 0.1 path loss.
R2L: 1.4
R2R: 8.5
R2L: 0.2
R2R: 3.0
 Receiver sensitivity and spectral
Jitter (ms) Not tested Not tested L2L: 0.0 efficiency.
Required: < 2 W2L: 0.5
L2W: 2.3  Service specific quality matrix such as
W2R: 2.4
R2W: 4.0
video startup and zapping latency.
L2R: 2.8
THE SERVICE PROVIDER SUPPORTED

R2L: 3.0
R2R: 8.5
1-way
(ms)
delay Not tested Not tested L2L: 0.8
W2L: 1.3 WIRELESS HOME NETWORKING
Required: <120 L2W: 3.5
W2R: 4.7
BUSINESS MODEL
R2W: 8.0
L2R: 7.0
R2L: 6.5 What is the cost of adding wireless home
MOS Not tested Not tested
R2R: 28.2
L2L: 4.4
networking for the service provider? Are the
Required: > 3.5 W2L: 4.4 costs different between the various wireless
L2W: 4.3
W2R: 4.2 technologies? We explore the economics
R2W: 4.0
L2R: 4.0
across multiple wireless home networking
R2L: 4.1
R2R: 4.0
solutions.
Key Finding Radio speeds to Radio, esp. Radio, esp.
not support 100
Mbps data
Radio to Radio
does not support
Radio to Radio
does not support Developing the wireless home network “Pain
product speeds <0.1 bytes lost
for video.
< 2ms jitter, or <
0.1 bytes lost for
Threshold”
voice.

We assume that the average revenue from How does the pain threshold described above
wireless home networking is $10 per month compare to the cost of the wireless home
per subscriber. Of this $10 per month we networking technologies? To make this
assume that 80%, or $8 is required for sales comparison we assume a 24 month life span
and marketing, customer care, billing and for wireless CPE accommodating the shorter
G&A per subscriber per month. This leaves life cycle associated with emerging wireless
$2 per month per subscriber to cover all home networking technologies. The resulting
wireless home network related technology cost per sub per month can be compared to the
costs. $2 pain threshold in Figure 16 below.
Considering the technology costs we initially Figure 16: Wireless Home Networking “Pain
reviewed the incremental costs of wireless Threshold” and Technology Costs
home networking technology we considered Wireless Home Networking Cost Per Subscriber Per Month
bit the A-end or Access Point incremental cost $35

WHN. Technology Incremental Cost / Sub / Month
to add wireless and the B-end, in the case that $30 WHN. Pain Threshold / Month $29.17
a specialized B-end device is needed to $25

$20.83
complete the link. $20
$15
$12.50
Table 8: CAPEX Cost of Adding Wireless $10
Access Client Cost Total

$4.17
$5 $1.04
$0.17 $0.21
Point ($US) Service $0
Incremental Provider DECT 802.11n

Draft
802.11n
Custom RF
802.11n
Custom
WDMI UWB 60GHz
Cost ($US) Cost Antenna
DECT $4 Any DECT $4

Handset The analysis of EBIT margin, highlighted in
Figure 17 below, indicates that unmodified
Wi-Fi $5 Any $5
802.11n Wi-Fi
802.11n and 802.11n with minimal
(Baseline) customizations do not add substantially to the
Wi-Fi $25 Any $25 CPE cost or require a custom or proprietary
802.11n Wi-Fi chip at both ends of the link, support a
(Celeno) * business case for wireless home networking.
Wi-Fi $50 $50 $100
802.11n
(Ruckus) ** Figure 17: Wireless Home Networking EBIT
WDMI $150 $150 $300 Margin
(Amimon) Wireless Home Networking EBIT Margin
*** 50%
18% 18%
UWB $250 $250 $500 0%
10%
(Tzero) -22%
***
-50%
60GHz $350 $350 $700 -100%

-105%
(Sibeam) -150%
*** -200% -188%

* Note: Celeno are targeting a chip-based solution for embedding in
CPE/RG devices so no additional cost is needed for casing, -250%
packaging, cabling. -272%

** Note: Ruckus had no plans for a CPE/RG integrated device and
-300%
DECT 802.11n 802.11n 802.11n WDMI UWB 60GHz
would only consider on a business case basis so the cost is based on Draft Custom RF Custom
Antenna
an add-on access point MediaFlex 7000 at the estimated cost to the
Telco rather than the list price. This clearly demonstrates the economic
*** Note: WDMI, UWB, and 60GHz solutions are primarily
targeting cable replacement in 1st generation of products so the cost is advantage of leveraging established device
including casing, packaging, cabling, power for each end of the link. and certification ecosystems; in addition to
scale and forwards and backwards

compatibility in selecting a suitable wireless END NOTES
home networking solution.
1
We can summarize that 802.11n, or a variant CES 2009, Jan 7, 2009, “Netgear
ITV2000 Internet TV Player Lets You
of 802.11n, that supports a large established Watch BBC News Whenever You Feel Like
ecosystem of devices and also provides It”,
robustness for video, in conjunction with http://i.gizmodo.com/5125266/netgear-
DECT for telephony; can potentially provide itv2000-internet-tv-player-lets-you-
watch-bbc-news-whenever-you-feel-
a wireless home networking foundation that is like-it &
economically viable for service providers. http://www.netgear.com/Products/Enter
tainment/DigitalMediaPlayers/ITV2000.
CONCLUSION aspx
2
CES 2009, Jan 7, “Linksys by Cisco
Wireless Home Audio system unveiled”,
This paper highlights that there is not a “one- http://www.engadget.com/2009/01/07/li
size fits all” approach for wireless home nksys-by-cisco-wireless-home-audio-
networking in support of service provider system-unveiled/
applications.
3
CES 2009 Preview: Whole-home
1080p60 HD via Wi-Fi, Jan 6 2009,
Gadgetell,
Testing of 802.11n devices against service http://www.gadgetell.com/tech/comment
provider triple play services indicates that it is /ces-2009-preview-whole-home-hdmi-
technically challenging to utilize only a single 1080p60-via-wi-fi
4
20 MHz radio channel for all services and CES 2009, Jan 7, 2009 “Cisco
Simplifies Digital Media Access for
service provider product requirements. Consumer” &
http://www.linksysbycisco.com/US/en/p
The wireless home networking technologies romo/Media-Hub
that are most suited to integration into the
5
U.S. Broadband Households, “Home
Servers & Consumer Storage”, Parks
service provider CPE are those that (a) have a
Associates, 2008, p.27
significant established ecosystem, (b) have 6
Seagate magnetic storage price
volumes that support economies of scale, (c) trends using a blend of retail
are low cost, (d) support backwards pricing for magnetic Hard Disk
interoperability, (e) have the flexibility to Drives.
7
TCM Net, Jan 6, 2009,
support multi-regional variations, and (f) are http://blog.tmcnet.com/design-vs-
able to operate in well penetrated radio- functionality/2009/01/itunes_abandons
congested environments. _drm.html
8
Jos van der Loop, SiTel
Preliminary findings indicate that 802.11n, or Semiconductor BV, Jan 2009, DECT
World Forum.
a variant of 802.11n that is backwards 9
Forrester Research, Home Networking
compatible with the standards, for data and Dynamics & Opportunities, Feb 11-12
video and DECT for voice both have well 2009, Charles S. Golvin
established ecosystems and compelling
10
Forrester Research, European
Consumer Technology Adoption Study Q4
economics and are strong candidates for 2006 & Q4 2008 Online Devices, Access
service provider CPE integration. And Telco Survey
11
ETSI Standards, EN 301 893 V1 4.1
(2007-07), ETSI EN 302 502 V1.1.1
(2006-11), EN 301 489-17 V1.3.2
(2007-06), SIEMIC Inc., 2008
12
ETSI Standards, EN 301 893 V1 4.1
(2007-07), ETSI EN 302 502 V1.1.1

31
(2006-11), EN 301 489-17 V1.3.2 802.11n Wireless Technology
(2007-06), SIEMIC Inc., 2008 Overview, Cisco White Paper, 2009
13 32
SiBeam Presentation, “Wireless 802.11n Delivers Developing
Without Boundaries, 2009 Ecosystem, McCartney, July 2008,
14
Network Zone Web Article, “Wireless CommsDesign Article
33
Chipsets that Deliver Guaranteed 802.11n: Look before you Leap”,
Bandwith for High Speed Wireless Motorola White Paper, 2008
34
Networking”, November 3, 2008 IEEE Wireless LAN Medium Access
15
Open Cable Home Specifications Home Control (MAC) and Physical Layer
Networking 2.0, Reserved Services (PHY) Specifications: Enhancements
Domain Technology Specification (OC- for Higher Throughput, IEEE
SP-RSD-TECH-101-080630), 2008 Cable P802.11n/D2.0, 2008
35
Television Laboratories, Inc. Beamforming Webinar, Ruckus
16
“802.11n: The New Wave in WLAN Wireless, February, 2009
Technology”, Yarali & Ahsant, http://ruckus.na3.acrobat.com/p315245
September 2007, Mobility 2007 61
36
conference Technical Paper Celano product Brief, “CL1300 Wi-Fi
17
Wi-Fi™ Alliance is a trade group SoC”, 2009,
that owns the trademark to Wi-Fi. http://www.celeno.com/Resources/pdf/
18 37
WHDI™ is a trademark of AMIMON, Connections 2009 Conference,
Ltd. Blogspots, “Celano Responds to
19
WiMedia™ is a registered trademark Follow-up panel questions”, January
or service mark of the WiMedia 20, 2009
38
Alliance Celano Test Report: CL1000, Celano
20
WirelessHD™ is a trademark of the. Wireless Whitepaper, 2008
39
WirelessHD consortium Celano Test Report: CL1000, Celano
21
Kaido Kert, "Skype and DECT: Taking Wireless Whitepaper, 2008
40
Skype Beyond the PC", DECT World Network Zone, “Wireless Chipsets
Forum, Feb 3, 2009 that Deliver Guaranteed Bandwidth for
22
Comtrend, “Access and Home High Speed Wireless Networking”,
Networking”, Andrew Morton, 2008. November 3, 2008, EN-Genius
23
HP Whitepaper, 802.11n drives an Publishing Limited
41
Architectural Evolution, BitPipe, Celano Test Report: CL1000, Celano
2008 Wireless Whitepaper, 2008
24 42
The Beamforming Techniques in MediaFlex 7000 Series Product
Wireless MIMO Systems, IC Access Lab, Description and Specification,
graduate Intitute of Electrical 802.11n Multimedia Smart Wi-Fi Acess
Engineering, April 2008 Point and Receiver, 2008, Ruckus
25 “
The future of wireless networks: Wireless
43
Understanding 802.11n”, Paulin and MediaFlex 7000 Series Product
Rabjohn, Oct 2008, CommsDesign Description and Specification,
article 802.11n Multimedia Smart Wi-Fi Acess
26
WLAN MIMO Operation, Peter Cain, Point and Receiver, 2008, Ruckus
Agilent Technologies, October 2007. Wireless
27 44
Test Strategies for 802.11n MIMO Enabling wireless uncompressed HDTV
Devices, Minarsky, Octoscope, October connectivity with a unique video
2006 modem approach, Meir Feder, Amimon
28
802.11n Wireless Technology Ltd., 2008
45
Overview, Cisco White Paper, 2009 WHDI Transmitter and Receiver
29
“The Practical realities of 802.11n datasheet, Amimon Ltd., 2008
46
in home networking applications”, Enabling wireless uncompressed HDTV
Morris, May 2008, CommsDesign Article connectivity with a unique video
30
Reliable Wireless Multimedia modem approach, Meir Feder, Amimon
Streaming: A comparision of UWB and Ltd., 2008
47
802.11n, T-Zero White Paper, January WiMedia Alliance, “Worldwide UWB
2009 Regulatory Status”, 2009

48
In-Stat, “UWB Market sees major
setbacks, but some progress”, January
26,2009
49
WiMedia Alliance White Paper, “UWB
Standards”, Stephan Wood, January 1,
2008
50
Wireless HD Specification,SiBeam,
October, 2007
51
IEEE 802.11 report, Netgear Report,
Jones-Petrick & Associates, January
2009

ISBN 0‐940272‐01‐6; 0‐940272‐08‐3; 0‐940272‐10‐5; 0‐940272‐11‐3; 0‐940272‐12‐1; 0‐940272‐14‐8; 0‐
940272‐15‐6; 0‐940272‐16‐4; 0‐940272‐18‐0; 0‐940272‐19‐9; 0‐940272‐20‐2; 0‐940272‐21‐0; 0‐940272‐
22‐9; 0‐940272‐23‐7; 0‐940272‐24‐5; 0‐940272‐25‐3; 0‐940272‐26‐1; 0‐940272‐27‐X; 0‐940272‐28‐8; 0‐
940272‐29‐6; 0‐940272‐32‐6; 0‐940272‐33‐4; 0‐940272‐34‐2; 0‐940272‐35‐0; 0‐940272‐36‐9; 0‐940272‐
37‐7; 0‐940272‐38‐5; 0‐940272‐39‐3; 0‐940272‐40‐7; 0‐940272‐41‐5; 0‐940272‐42‐3; 0‐940272‐43‐1; 0‐
940272‐44‐X; 0‐940272‐45‐8; 0‐940272‐46‐6; 0‐940272‐47‐4; 0‐940272‐48‐2; 0‐940272‐49‐0; 0‐940272‐
50‐4; 0‐940272‐51‐2; 0‐940272‐52‐0; 0‐940272‐53‐9; 0‐940272‐54‐7
© 2015 National Cable and Telecommunications Association. All Rights Reserved.

2009 Ncta Technical Papers PDF

Uploaded by

Copyright:

Available Formats

2009 Ncta Technical Papers PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2009 Ncta Technical Papers PDF

Uploaded by

Copyright:

Available Formats

THE COMPLETE

TECHNICAL PAPER PROCEEDINGS

Abstract The TV is King.

2009 NCTA Technical Papers - page 71

Figure 1. User-Generated Video Views

Cable has an opportunity to tap into this

2009 NCTA Technical Papers - page 72

2009 NCTA Technical Papers - page 73

involve using web development technologies

2009 NCTA Technical Papers - page 74

2009 NCTA Technical Papers - page 75

Such services amount to a natural offering,

2009 NCTA Technical Papers - page 76

Abstract are often unable to read the mark after small

2009 NCTA Technical Papers - page 77

2009 NCTA Technical Papers - page 78

2009 NCTA Technical Papers - page 79

2009 NCTA Technical Papers - page 80

2009 NCTA Technical Papers - page 81

Abstract After the DTV transition, all of the full

solutions. This paper provides an overview of 698 MHz 806 MHz

TV: 294 MHz 108 MHz

In the United States, 402 MHz of

2009 NCTA Technical Papers - page 227

TVBDs may be used to provide the

2009 NCTA Technical Papers - page 228

2009 NCTA Technical Papers - page 229

2009 NCTA Technical Papers - page 230

Interference from TVBDs to the cable

2009 NCTA Technical Papers - page 231

have the potential to significantly increase the

instances of DPU becoming a problem.

Sufficiently strong DPU can completely

disrupt services on the affected channel,

2009 NCTA Technical Papers - page 232

100 mW (+20 dBm) EIRP

The FCC has established the RF DPU 40 mW (+16 dBm) EIRP

E Field V/m Far Field

standard for analog cable ready TV receivers

digital TV's but we can assume the same

the widespread use of Channels 3 and 4 as the 10 100

in-home interconnect frequency and has

The specified minimum tap level for 256

2009 NCTA Technical Papers - page 233

on the end of the cable was also found to have

a large impact on SE, and hence DPU Figure 7. The Effects of

Table 3 summarizes the mid-band 500 MHz

2009 NCTA Technical Papers - page 234

antennas and receivers. 10.000

A combination of remediation approaches

may be needed to resolve DPU interference in

subscriber homes. One of the most important -6 dBmV

techniques is to properly terminate open ports. 0.100

Cable drops, splitters and terminations should

unused cabling/splitters in the home should be Figure 8. DPU Susceptibility as a

2009 NCTA Technical Papers - page 235

2009 NCTA Technical Papers - page 236

[1] FCC Auction 73, 700 MHz Band.

2009 NCTA Technical Papers - page 237