Digital Audio White Paper

8LI:MHIS)RKMRIIVkW+YMHI
XS(MKMXEP%YHMS
.SLR;EXOMRWSR
John Watkinson is an independent consultant and author in the

broadcast industry with over 20 years experience in digital and
analog technology. With a BSc (Hons) in Electronic Engineering
and an MSc in Sound and Vibration, he has held teaching posts at
a senior level with the Digital Equipment Corporation, Sony
Broadcast and Ampex before forming his own consultancy.
He regularly delivers papers at conferences including AES, NAB,

SMPTE, IEE, ITS and IBC and has written a number of definitive
books including "The Art of Digital Audio".
An NVision2 Guide
1
The Video Engineers Guide to Digital Audio - John Watkinson
Publishers Forward
Our thanks to John Watkinson for agreeing to write this

book for us. John’s work has helped video and audio
engineers worldwide to better understand the techniques
required to make a safe transition to the digital environment.
At NVISION, we have endeavored to design and
manufacture product that ensures that the inclusion of
digital audio within a television facility can be achieved,
painlessly and cost effectively.
The most important task in designing a digital audio system

within a video facility, is understanding the importance of
issues that were irrelevant in the analog domain. It is our
belief that this new book will be an invaluable addition to
every engineers toolbox and will help us all to avoid the
pitfalls that exist on the road to digital.
Nigel Spratling
Vice President - Marketing
NVision2
ISBN 0-9640361-3-4
Some of the diagrams contained within this book

are reprinted with the permission of
John Watkinson from his book
The Art of Digital Audio (second edition).
2
'328)287
-RXVSHYGXMSR
'LETXIV;LEXMWHMKMXEPEYHMS#
'LETXIV'SRZIVWMSR
'LETXIV0IZIPWERHQIXIVMRK
'LETXIV8VERWQMWWMSR
'LETXIV7]RGLVSRM^MRK
'LETXIV6SYXMRK
'LETXIV,S[HS-KSHMKMXEP#
3
-2863(9'8-32
This guide is intended for the television engineer faced with the
problem of integrating digital audio systems in a television
environment. Digital audio is a complex subject which is often
described by specialists in a way which leaves many of us
confused. This guide is not like that. It is for normal people who
have a job to do and need to make decisions without getting lost
in pages of math. So there are no equations here, only clear
explanations. And there’s a Glossary where all the buzzwords are
defined.
Television production equipment is changing over to digital

technology at a rate which is accelerating. The advantages of
digital apply equally well to audio as they do to video. In fact,
retaining an analog audio system can prevent the full freedom of
digital video being realized.
There are many quality related reasons why digital audio is

advantageous, and it was the search for quality which led to the
development of the technology in the first place. However, the
main pressure today is economic. Essentially a sensibly planned
digital audio installation costs less to run than an analog plant
doing the same job. It will have less down time and need less
maintenance. Alongside the economic advantage, the excellent
sound quality is a useful bonus.
Television sound quality is becoming increasingly important.

With the expectations of the viewer set by his Compact Disc
system, TV sound has to match up. More and more people are
equipping their TV sets with quality speakers, often in surround
sound systems. In some countries, television broadcasts are
already accompanied by NICAM digital stereo sound. Before long,
Digital Video Broadcasting (DVB) will become the preferred way
to deliver digital audio and pictures to the home.
4
What is Digital Audio
Against this background, going digital in some way will be essential

in order to remain competitive, the only remaining question being
one of degree. Whilst a brand new facility can start out with a
totally digital system, in most installations there is an investment in
existing equipment and engineers who can handle it. A total
changeover to digital makes no sense and one of the jobs of this
guide is to illustrate how the replacement of a few key items could
offer a more cost effective solution in which digital and analog
co-exist for a transition period. In the real world, there is seldom
one universal right answer, and where possible we present more
than one solution to a given problem.
5
'LETXIV
;LEXMW(MKMXEP%YHMS#
1.1 Although there are other approaches, there is one system,

known as Pulse Code Modulation (PCM) which is in virtually
universal use for professional audio production. Fig.1.1 shows how
PCM works. The time axis is represented in a discrete, or stepwise
manner and the waveform is carried by measurement at regular
intervals. This process is called sampling and the frequency with
which samples are taken is called the sampling rate or sampling
frequency Fs. The sampling rate is fixed and the sampling clock is
jitter-free so that every sample will be made at an exactly even time
step. If there is any subsequent timebase error, the instants at
which samples arrive will be changed but the effect can be
eliminated by storing the samples temporarily in a memory and
reading them out using a stable, locally generated clock. This
process is called timebase correction and all properly engineered
digital audio systems must use it. Clearly timebase error is not
reduced; it is totally eliminated. A further consequence of this
approach is that the audio channels of a DVTR don’t have any wow
or flutter.
The sampling process seems to be taking no notice of what

happened between the samples. This is not true because all analog
signal sources from microphones, tape decks, pickup cartridges
and so on have a frequency response limit, as indeed do our ears.
When a signal has finite bandwidth, the rate at which it can change
is limited, and the way in which it changes becomes predictable.
When a waveform can only change between samples in one way,
nothing has been lost by sampling.
In a sense sampling has made audio more like video, which is

sampled into discrete frames. One consequence of using digital
audio with video is that we now have to deal with two sampled
signals. It is clear that for practical purposes some sensible
synchronization has to be achieved between the two rates. A
whole section of this guide is devoted to synchronization.
6

9
R
O
W
D
J
H
7 7 7 7 7 7 7 7 7LPH
'DWDWREHVWRUHG
RUWUDQVPLWWHGDUH
ZKROHQXPEHUV
Figure 1.1 In pulse code modulation (PCM) the analog waveform is

measured periodically at the sampling rate. The voltage (represented here by
the height) of each sample is then described as a whole number. The whole
numbers are stored or transmitted rather than the waveform itself.
Fig.1.1 also shows that each sample is also discrete, or represented

in a stepwise manner. The length of the sample, which will be
proportional to the voltage of the audio waveform, is represented
by a whole number. This process is known as quantizing and
results in an approximation, but the size of the error can be
controlled until it is negligible. The advantage of using whole
numbers is that they are not prone to drift. If a whole number can
be carried from one place to another without numerical error, it
has not changed at all. By describing audio waveforms numerically,
the original information has been expressed in a way which is
better able to resist unwanted changes in transmission.
Essentially, digital audio carries the original waveform numerically.

The number of the sample is an analog of time, and the magnitude
of the sample is an analog of the pressure (or the velocity) at the
microphone. In simple terms, the audio waveform is conveyed in a
digital system as if the voltage had been measured at regular
intervals with a digital meter and the readings transmitted. The rate
7
at which the measurements were taken and the accuracy of the

meter are the only factors which determine the quality. The quality
7KLVELWUHSUHVHQWV 7KLVELWUHSUHVHQWV

0RVWVLJQLILFDQWELW /HDVWVLJQLILFDQWELW
06%
1 1 1 1 /6%
7KLVELWUHSUHVHQWV 7KLVELWUHSUHVHQWV

([DPSOHV

Figure 1.2 In a binary number, the digits represent increasing powers of

two from the LSB. Also defined here are MSB and wordlength. When
wordlength is 8 bits, the word is a byte. Binary numbers are used as memory
+,*+
92/7$*( (/(&75,&$/
/2:

11 66 11 66 0$*1(7,&
237,&$/
0(&+$1,&$/
Figure 1.3 Several ways of handling binary data are shown here. In
electronic circuits, two different voltages are needed, commonly achieved
with a simple on/off switch. In magnetic recording, two flux directions are
used. Optical recording may use alternative opaque or dark areas and
mechanical storage may use raised or depressed areas on a surface.
8
is determined only by the accuracy of conversion and is

independent of the quality of the signal path.
1.2 Digital transmission has the advantage over analog in that it can
reject noise and jitter. If the whole numbers representing the video
waveform voltage are converted into binary, using the conversion
table shown in Fig.1.2, the binary digits, or bits, have only two
states, 1 and 0. These two states can then be represented by any
electrical, magnetic, optical or mechanical system which can exist
in two states. Fig.1.3 shows binary represented electrically by two
6/,&(
/(9(/
&/($1
6,*1$/
12,6<6,*1$/
6/,&(6
+(5(
&/($16,*1$/
6/,&(6+(5(
E
-,77(5'8(7212,6(
Figure 1.4 A Binary signal has the advantage that a simple slicer shown at
(a) can be used to recreate a square signal. The result is that noise only
causes jitter as shown in (b). This can be removed with a well designed
different voltages, magnetically by the reversal of the direction of

magnetization, optically by alternate opaque and transparent areas
of a recording, and mechanically by the presence of pits in the
surface of a laser disk.
9
With only two states, more noise can be rejected than in any other
system as Fig.1.4a) illustrates using the example of an electrical
interface. Although the signal transmitted is a clean, two-level
waveform, by the time it reaches the receiver it will be suffering
from noise and jitter. The receiver compares the voltage of the
signal with a reference which is mid-way between the transmitted
levels in a process called slicing. Any voltage above the slicing level
is considered a 1 and any voltage below is considered a 0. This
slicing process will reject considerable amounts of noise and
restore the signal to clean binary once more.
Slicing resists noise, but it is powerless against jitter which requires

a further process. Fig.1.4b) shows that jitter has the effect of
shifting voltage changes, or transitions, along the time axis in a
random manner. However, the average timing of a large number of
transitions is unaffected. A phase-locked loop (PLL) is an electronic
circuit which can average the timing of many transitions to
recreate a stable clock from a jittery one. It acts like the flywheel
on a piston engine which averages the impulses from the pistons
to produce smooth rotation.
The combination of a slicer and a phase-locked loop is called a

reclocker. The waveform leaving a reclocker is identical to the
waveform at the transmitter. Noise and jitter have been rejected so
there has been no loss of data due to the transmission.
Consequently, reclockers can be cascaded or tandem connected
as often as necessary yet the same data will come out of the end.
The reclocker is found everywhere in digital systems. Routers use

it, Digital VTRs use it in the video and audio signals. Hard disk
drives use it, CD players and DAT machines use it. The universality
of reclocking is because it eliminates generation loss. A device
which contains a reclocker launches a clean signal which is just as
robust as the signal launched from the previous device.
1.3 There are two main advantages which follow from the
characteristics of PCM audio. Firstly, in the absence of a
compression system, the quality of reproduction of a well
10
engineered digital audio system is independent of the transmission

or recording medium and depends only on the quality of the
conversion processes. Secondly, the freedom to interchange audio
data without generation loss between a variety of media allows
tremendous opportunities which were denied to analog signals.
Someone who is only interested in sound quality will judge the

former the most relevant. If good quality converters can be
obtained, all of the shortcomings of analog audio can be eliminated
to great advantage. Wow, flutter, particulate noise, print-through,
dropouts, modulation noise, HF squashing, azimuth error,
interchannel phase errors and time-code breakthrough all
disappear from recorders. Routing is performed without crosstalk
from other audio signals or from video signals.
Digital circuitry costs less to manufacture. Switching circuitry
which handles binary can be integrated more densely than analog
circuitry. More functionality can be put in the same chip. Analog
circuits are built from a host of different component types which
have a variety of shapes and sizes and are costly to assemble and
adjust.
Digital circuitry uses standardized component outlines and is easier

to assemble on automated equipment. Little if any adjustment is
needed and periodic maintenance is virtually eliminated.
Once audio is in the digital domain, it becomes data, and as such is

indistinguishable from any other type of data. Systems and
techniques developed in other industries for other purposes can
be used for audio. Computer equipment is available at low cost
because the volume of production is far greater than that of
professional audio equipment. Disk drives and memories
developed for computers can be put to use in audio products. A
word processor adapted to handle audio samples becomes a
workstation. There seems to be little point in waiting for a tape to
wind when a disk head can access data in milliseconds. The
difficulty of locating the edit point and the irrevocable nature of
tape-cut editing are hardly worth considering when the edit point
can be located by viewing the audio waveform on a screen or by
11
listening at any speed to audio from a memory. The edit can be

simulated and trimmed before it is made permanent.
Fig.1.5 shows a simple digital system consisting of an analog to

digital converter (ADC), a digital router and a digital to analog
converter. The router and the DAC reclock the data so that they
are identical to those leaving the ADC. Consequently the presence
of the router has no effect whatsoever on the audio quality. This is
true provided that there is no data corruption due to noise or jitter
which is in excess of what the reclockers can handle. Thus in
digital systems there are two important areas where attention must
be paid:
',*,7$/
5287,1*
$1$/2*
$'& 0$75,;
,1
$1$/2*
'$&
287
5(&/2&.(5
Figure 1.5 When reclocking is used, the routing system is transparent.

Quality is determined by the ADC and DAC, therefore well designed
products are essential.
1/ ADCs and DACs determine sound quality and the best available
should be used. Cheap converters are false economy as they drag
down the performance of the rest of the system.
2/ The quality is determined by the converters only if there are no

data errors. Therefore a major goal in digital systems is data
integrity. Data integrity cannot be measured with traditional analog
test gear but has to be built in by good installation practice.
12
Conversion
'LETXIV
'SRZIVWMSR
The quality of digital audio is determined by the accuracy of

conversion between the analog and digital domains. This section
considers the theory and practice of this critical aspect of digital
audio.
D
H
J
D
WO
R
Y

Q V 4
R X X
OH V R D
Q
S L WL
P D [ X ]H
D Q
LW
6 H
LP Q
W
R
&
'LVFUHWHWLPH
H
J
D
W
O H
R J
Y D
WO

V
X R
R Y
X H
W
Q
LW H
U
Q F
R LV
& '
&RQWLQXRXVWLPH 'LVFUHWHWLPH
E
4
X
YR DQ
OW WL H OH
D ]H J S
J
H D P
D RQ WO D
[L 6
V R
Y

WH
H
U
F
V
L
'
&RQWLQXRXVWLPH
Figure 2.1 Since sampling and quantizing are orthogonal, the order in
which they are performed is not important. In (a) sampling is performed first
and the samples are quantized. This is common with audio converters. In
(b) the analog input is quantized into an asynchronous binary code.
Sampling takes place when this code is latched on the sampling clock
edges. This approach is universal in video converters.
2.1 The ear can detect tiny amounts of distortion, and will accept
an enormous dynamic range. These characteristics require audio
13
signals to have a precision far in excess of that required by video.

The only concession is that much less bandwidth is required. The
input to an ADC is a continuous-time, continuous-voltage
waveform, and this is changed into a discrete-time, discrete-
number format by a combination of sampling and quantizing.
These two processes are totally independent and can be performed
in either order and discussed quite separately in some detail.
Fig.2.1a) shows an analog sampler preceding a quantizer, whereas
b) shows an asynchronous quantizer preceding a digital sampler.
Both approaches will be found in real equipment.
6LJQDO
F
2IIVHW
Figure 2.2 The sampling process requires a constant-amplitude pulse train

as shown in (a). This is amplitude modulated by the waveform to be
sampled. If the input waveform has excessive amplitude or incorrect level,
the pulse train clips as shown in (b). For an audio waveform, the greatest
signal levelis possible when an offset of half the pulse amplitude is used to
2.2 The sampling process originates with a pulse train which is

shown in Fig.2.2 to be of constant amplitude and period. The
audio waveform amplitude-modulates the pulse train in much the
same way as the carrier is modulated in an AM radio transmitter. In
the same way that AM radio produces sidebands or images above
14
Conversion
and below the carrier, sampling also produces sidebands although

the carrier is now a pulse train and has an infinite series of
harmonics as shown in Fig.2.3a). The sidebands repeat above and
below each harmonic of the sampling rate as shown in b).
The sampled signal can be returned to the continuous-time domain

simply by passing it into a low-pass filter known as an anti-image or
reconstruction filter.
D HWF
)V )
V )
V
%DVHEDQG
)LOWHUUHVSRQVH
E
HWF
)E )V )
V )
V
$OLDV]RQH
F HWF
)E )V )
V )
V
)V)E )V)E
1\TXLVWOLPLW
2XW ,Q
G HWF
D )V
E
D E
0LQLPXPSRVVLEOH
VDPSOLQJUDWH
H
)E ≡
)V [PLQLPXP
%DVHEDQG
Figure 2.3 (a) Spectrum of sampling pulses. (b) Spectrum of samples. (c)
Aliasing due to sideband overlap. (d) Beat-frequency production. (e) 4 x
oversampling.
This filter has a frequency response which prevents the images

from passing, and only the baseband signal emerges, completely
unchanged. It could be argued that the reconstruction filter is
unnecessary, since all the images are outside the range of human
hearing. However, the slightest non-linearity in subsequent analog
stages would result in gross intermodulation distortion and if such
a signal reached a TV transmitter the images would take the
transmitted spectrum way out of tolerance.
15
If an input is supplied having an excessive bandwidth for the

sampling rate in use, the sidebands will overlap, (Fig.2.3c)) and the
result is called aliasing, where certain output frequencies are not
the same as their input frequencies but instead become difference
frequencies (Fig.2.3d)). It will be seen from Fig.2.3 that aliasing
does not occur when the input frequency is equal to or less than
half the sampling rate, hence the rule which says that the sampling
rate must be at least twice the input bandwidth.
D E
Figure 2.4 In (a) the sampling is adequate to reconstruct the original

signal. In (b) the sampling rate is inadequate, and reconstruction produces
the wrong waveform (dashed). Aliasing has taken place.
Aliasing can be considered equally well in the time domain. In

Fig.2.4a) the sampling rate is obviously adequate to describe the
waveform, but at b) it is inadequate and aliasing has occurred.
Aliasing sounds quite unnatural and has to be prevented in
practical systems. A low-pass, or anti-aliasing filter is provided at
the analog input to prevent frequencies of more than half the
sampling rate from reaching the sampling stage.
The ideal filter with a vertical "brick-wall" cut-off slope is difficult

to implement. As the slope tends to vertical, the delay caused by
the filter goes to infinity: great quality but you don’t live to hear it.
In practice, a filter with a finite slope has to be accepted. The
cut-off slope begins at the edge of the required band, and
consequently the sampling rate has to be raised a little to drive
aliasing products to an acceptably low level.
16
Conversion
Every signal which has been through the digital domain has passed
through both an anti-aliasing filter and a reconstruction filter.
These filters must be carefully designed in order to prevent audible
artifacts, particularly those due to lack of phase linearity as they
may be audible. The nature of the filters used has a great bearing
on the subjective quality of the system. It is difficult to produce an
analog filter with low distortion, but fortunately today the problem
is sidestepped using oversampling. The audible superiority and
economy of oversampling converters has led them to be almost
universal. Strictly speaking, oversampling means no more than
using a higher sampling rate than theory demands. In the loose
sense an "oversampling converter" generally implies that some
combination of high sampling rate and various other techniques
has been applied. Oversampling is treated in depth in a later
paragraph of this section.
2.2 Sampling theory is only the beginning of the process which

must be followed to arrive at a suitable sampling rate. The finite
slope of realizable filters will compel designers to raise the
sampling rate. For consumer products, the lower the sampling rate
the better, since the cost of the medium is directly proportional to
the sampling rate: thus sampling rates near to twice 20kHz are
found. However, in professional digital audio recorders used in
record production, there is a need to operate at variable speed for
pitch correction. When the speed of a digital recorder is reduced,
the offtape sampling rate falls. With a minimal sampling rate the
first image frequency can become low enough to pass the
reconstruction filter. If the sampling frequency is raised without
changing the response of the filters, the speed can be reduced
without this problem. The sampling rate of 48kHz was adopted for
professional purposes and has been taken up by DVTR
manufacturers to the exclusion of other rates. Thus in the
television production environment this is the only rate to consider.
Variable sampling rates are seldom used in television as the
sampling rate is locked to video timing for practical reasons.
For broadcast purposes a 15kHz audio bandwidth is considered

sufficient and a sampling rate of 32kHz is more than adequate. This
17
H UUR U D
W
9
Figure 2.5 The effect of sampling timing jitter on noise, and calculation of
the rquired accuracy for a 16 bit system. (a) Ramp sampled with jitter has
error proportional to slope. (b) When jitter is removed by later circuits, error
appears as noise added to samples. For a 16 bit system there are 216Q, and
the maximum slope at 20KHz will be 20 000p x 216Q per second. If jitter is to
be neglected, the noise must be less than 1/2Q, thus timing accuracy t’
multiplied by maximum slope =1/2Q or 20 000p x 216Qt’ = 1/2Q
1 .
\ t’ = 2 x 20 000 x p x 216 = 121 ps
18
Conversion
Frequency is used in the NICAM 728 stereo TV sound system and

in DAB. The Compact Disc, like most consumer equipment has a
sampling rate of 44.1kHz. In the same way that 60Hz U.S. video is
incompatible with 50 Hz European video, and needs a standards
converter, audio with different sampling rates is incompatible and
a sampling rate converter is needed. Section 5 deals with rate
conversion.
2.3 The sampling clock has to be generated very precisely. Fig.2.5

shows the effect of ADC sampling clock jitter on a sloping
waveform. Samples are taken at the wrong times. When these
samples have passed through a system, the timebase correction
stage prior to the DAC will remove the jitter, and the result is
shown at b). A similar effect will occur if correctly sampled data is
converted by a DAC with a jittery clock. The magnitude of the
unwanted signal is proportional to the slope of the audio
waveform and so the amount of jitter which can be tolerated falls
at 6dB per octave. As the resolution of the system is increased by
the use of longer sample wordlength, tolerance to jitter is further

Q Q Q
V V V
S
H
D ELWQRLVHIORRU
N
W
R
S
% H
G D
N
M
LW
WH
LR
W U
D ELWQRLVHIORRU
U

H
LV
R
Q

R
W
O
D
Q ELWQRLVHIORRU
J
L
6

N N N
$XGLRVLJQDOIUHTXHQF\ +]
Figure 2.6 Effects of sample clock jitter on signal-to-noise ratio at different

frequencies, compared with theoretical noise floors of systems with different
resolutions.
19
reduced. The nature of the unwanted signal depends on the

spectrum of the jitter. If the jitter is random, the effect is noise-like
and relatively benign unless the amplitude is excessive.
Fig.2.6 shows the effect of differing amounts of random jitter with

respect to the noise floor of various wordlengths. Note that even
small amounts of jitter can degrade a 20-bit converter to the
performance of a good 16-bit unit. There is thus no point in
upgrading to higher resolution converters if the clock stability of
the system is insufficient to allow their performance to be realized.
Clock jitter is not necessarily random. One source of clock jitter is

crosstalk or interference on the clock signal. This is one reason the
AES/EBU digital audio interface uses a balanced line transmission.
If an external clock source is used, it cannot be used directly, but

must be fed through a well designed, well-damped phase-locked-
loop which will filter out the jitter. The phase locked loop must be
built to a higher accuracy standard than in most applications. Noise
reaching the frequency control element will cause the very jitter
the device is meant to eliminate. Some designs use a crystal
oscillator whose natural frequency can be shifted slightly by a
varicap diode. The high Q of the crystal produces a cleaner clock.
2.4 Quantizing is the process of expressing some infinitely variable

quantity by discrete or stepped values. In audio the values to be
quantized are infinitely variable voltages from an analog source.
Strict quantizing is a process which operates in the voltage domain
only; time can be assumed to stand still.
Fig.2.7a) shows that the process of quantizing divides the voltage

range up into quantizing intervals Q, also referred to as steps S. In
applications such as telephony these may usefully be of differing
size, but for digital audio the quantizing intervals are made as
identical as possible. When this is done, the binary numbers which
result are truly proportional to the original analog voltage, and the
digital equivalents of mixing and gain changing can be performed
by adding and multiplying sample values. When all quantizing
intervals are the same, the term uniform quantizing is used.
20
Conversion
9ROWDJH

1RPLQDO
YROWDJH
4XDQWL]LQJ
RI
LQWHUYDOV
TXDQWL]LQJ
4 LQWHUYDOV

E F
0LGWUHDG 0LGULVHU
4 4 4 4
4
4
4XDQWL]LQJ ,QSXW
G HUURU
Figure 2.7 Quantizing assigns discrete numbers to variable voltages. All

voltages within the same quantizing interval are assigned the same number
which causes a DAC to produce the voltage at the centre of the intervals
shown by the dashed lines in (a). This is the characteristic of the mid-tread
quantizer shown in (b). An alternative is the mid-riser system shown in (c).
Here 0 volts analog falls between two codes and there is no code for zero.
Such quantizing cannot be used prior to signal processing because the
21
Whatever the exact voltage of the input signal, the quantizer only
locates the quantizing interval in which it lies. The quantizing
interval is then allocated a code value which is typically some
form of binary number. The information sent is the number of the
quantizing interval in which the input voltage lay. The precise
location within the interval is not conveyed, putting a limit on the
accuracy of the quantizer. When the number of the quantizing
interval, i.e. the sample value, is converted back to analog, it will
result in a voltage at the center of the quantizing interval as this
minimizes the magnitude of the error between input and output.
The number range is limited by the wordlength of the binary
numbers used. For example, in a sixteen-bit system, 65,536
different quantizing intervals exist, although the ones at the
extreme ends of the range have no outer boundary.
2.5 Fig.2.7b) shows that the transfer function of an ideal quantizer

is somewhat like a staircase, and zero volts analog, corresponding
to all zeros digital or muting, is half way up a quantizing interval, or
on the center of a tread. This is the so-called mid-tread quantizer
which is universally used in audio. Quantizing causes a voltage
error in the audio sample which is given by the difference between
the actual staircase transfer function and the ideal straight line.
This is shown in Fig.2.7d) to be a sawtooth like function whose
amplitude cannot exceed +/- 1/2 Q peak-to-peak unless the input is
so large that clipping occurs.
The non-linearity of the transfer function results in distortion,

which produces harmonics. Unfortunately these harmonics are
generated after the anti-aliasing filter, and so any which exceed half
the sampling rate will alias. This results in anharmonic distortion
within the audio band. These anharmonics result in spurious tones
known as bird-singing. When the sampling rate is a multiple of the
input frequency the result is harmonic distortion. Where more
than one frequency is present in the input, intermodulation
distortion occurs, which is known as granulation. As a result
practical digital audio devices deliberately use non-ideal quantifiers
to achieve linearity.
22
Conversion
2.6 At high signal levels, quantizing error on program material is

effectively noise. As the audio level falls, the quantizing error of an
ideal quantizer becomes more strongly correlated with the signal
and the result is distortion. If the quantizing error can be
decorrelated from the input in some way, the system can remain
linear but noisy. Dither performs the job of decorrelation by
making the action of the quantizer unpredictable and gives the
system a noise floor like an analog system. The introduction of
dither prior to a conventional quantizer inevitably causes a slight
reduction in the signal to noise ratio attainable, but this reduction
is a small price to pay for the elimination of non-linearity’s. The
addition of dither means that successive samples effectively find
the quantizing intervals in different places on the voltage scale. The
quantizing error becomes a function of the dither, rather than a

ELWWUDQVIHUIXQFWLRQ

ELWWUDQVIHUIXQFWLRQ
ELW ELWWUDQVIHUIXQFWLRQ

WUXQFDWLRQ

ELW
WUXQFDWLRQ ELWTXDQWL]LQJLQWHUYDO
ELWTXDQWL]LQJLQWHUYDO
ELWTXDQWL]LQJLQWHUYDO

YROWV
DQDORJ
Figure 2.8 Shortening the wordlength of a sample reduces the number of

codes which can describe the voltage of the waveform. This makes the
quantizing steps bigger, hence the term requantizing. It can be seen that
simple truncation or ommission of the bits does not give analogous behavior.
Rounding is necessary to give the same result as if the larger steps had been
used in the original conversion.
predictable function of the input signal. The quantizing error is not

eliminated, but the subjectively unacceptable distortion is
converted into a broadband noise which is more benign to the ear.
2.7 Today's ADC technology allows 18- and 20-bit resolution to be
obtained. In order to input samples of such wordlength to systems
23
such as NICAM or DVB the words need to be shortened in some

way. If the word is simply truncated by discarding the unwanted
low-order bits or rounded to the nearest integer the linearising
effect of the original dither will be lost. Shortening the wordlength
of a sample reduces the number of quantizing intervals available
without changing the signal amplitude. As Fig.2.8 shows, the
quantizing intervals become larger and the original signal is
requantized with the new interval structure. This will introduce
requantizing distortion having the same characteristics as
quantizing distortion in an ADC.
In practice, the wordlength of samples must be shortened using

digital dithering prior to rounding. This is directly equivalent to the
analog dithering in an ADC. Digital dither is a pseudo-random
sequence of numbers.
2.8 Oversampling is an important topic in digital audio. It implies

using a sampling rate which is greater (generally substantially
greater) than that required by theory. Fig.2.9 shows the main
advantages of oversampling. At a) it will be seen that the use of a
sampling rate considerably above the conventional rate allows the
anti-aliasing and reconstruction filters to be realized with a much
more gentle cut-off slope. There is then less likelihood of phase
linearity and ripple problems in the audio passband.
Fig.2.9b) shows that information in an analog signal is two

dimensional and can be depicted as an area which is the product of
bandwidth and the linearly expressed signal-to-noise ratio. The
figure also shows that the same amount of information can be
conveyed down a channel with a SNR of half as much ( 6 dB less)
if the bandwidth used is doubled, with 12 dB less SNR if
bandwidth is quadrupled, and so on, provided that the modulation
scheme used is perfect.
The information in an analog signal can be conveyed using some

analog modulation scheme in any combination of bandwidth and
SNR which yields the appropriate channel capacity. Raising the
sampling rate potentially allows the wordlength of each sample to
be reduced without information loss.
24
Conversion
6 D P S OLQ J
* H QWOH D Q D OR J U D WH
ILOW HU VO RS H
$ XGLR
EDQG
D
) E
) V
! ) E
) V
)UH T X HQ F\
H
O
D
F G % G %
V

U
D
H
LQ
O

5
1
6
E
)UH T X HQ F\
1 RUP D O [ [
VD P S OLQ J R YH U R Y HU
U D WH VD P S OLQ J VD P S OL Q J
U
$ X G LR R
OR
I
EDQG H
LV
R
1
1 RL V H D IW H U
ILOW HU LQ J W R
D X G LR ED Q G
F
) V

Figure 2.9 Oversampling has a number of advantages. In (a) it allows the

slope of analog filters to be relaxed. In (b) it allows the resolution of
converters to be extended. In (c) a noise shaped converter allows a
Oversampling enables a converter element of shorter wordlength

to be used, suggesting a flash converter as used in digital video.
The flash converter is capable of working at very high frequency
and so large oversampling factors are easily realized. Certain types
of ADC architecture offer a noise floor which rises with frequency.
These are called noise-shaped converters. Such a converter can be
highly oversampled and the output can be fed to a digital low-pass
filter which has the same frequency response as an analog anti-
aliasing filter used for conventional sampling. The result is a
disproportionate reduction in noise because the majority of the
25
noise is outside the audio band. A high resolution converter can be

obtained using this technology without requiring unattainable
component tolerances.
Oversampling is confined to converter technology where it gives

specific advantages in implementation. The transmission system
will employ standard PCM, where the sampling rate is a little more
than twice the audio bandwidth. Fig.2.10 shows a digital
transmission system using oversampling converters. The ADC runs
at n times the Nyquist rate, but once in the digital domain the rate
needs to be reduced in a type of digital filter called a decimator.
The output of this is conventional Nyquist rate PCM which is
transmitted. On reception the sampling rate is raised once more in
a further type of digital filter called an interpolator. The system
now has the best of both worlds: using oversampling in the
converters overcomes the shortcomings of analog anti-aliasing and
reconstruction filters and the wordlength of the converter
elements is reduced making them easier to construct; the
transmission is made with Nyquist rate PCM which minimizes the
bandwidth required.
$1$/2*
$17,$/,$6
$1$/2* 2YHUVDPSOLQJ
'(&,0$7(
,1 $'&
1<48,67
Q[)V Q )V 5$7(
3&0
$1$/2* ,17(5)$&(
$17,,0$*(
$1$/2* 2YHUVDPSOLQJ P[)V

,17(532/$7(
287 '$&
Figure 2.10 Oversampling converters can be used for ADC or DAC

applications, but digital filters are used to allow Nyquist rate interfacing.
26
Levels and Metering
'LETXIV
0IZIPWERH1IXIVMRK
The subject of level in hybrid digital/analog systems is not

straightforward owing to the lack of standardization. In this section
we shed light on the principles involved and give practical advice.
3.1 In the analog domain, the 600 Ohm balanced line has been
widely used for audio. The use of balancing transformers has been
largely supplanted by balanced transformerless circuitry. Fig.3.1a)
shows a typical analog interconnect. The source amplifier device
has a negligible output impedance which is raised to 600 Ohms by
the incorporation of resistors. The load amplifier has a high input
impedance which is reduced by the parallel connection of a
resistor. Effectively the system acts as a potential divider wasting
half of the drive voltage so that the drive amplifier has to produce
+10dBu in order to obtain +4dBu on the line. Clearly if an attempt
is made to drive two 600 Ohm loads, the line will be shunted by
only 300 Ohms and a further 4dB drop will be suffered.
The 600 Ohm source impedance is a hangover from the days when
audio was transmitted over long land lines, between towns, for
example. In this case, the length of the lines approaches the
electrical wavelength of the audio signal and proper source and
load termination is needed to prevent reflections. This is why
video signals need proper termination because the frequencies are
so much higher and the cable lengths easily exceed the signal
wavelength. However, in a typical television studio, the distances
involved are simply not great enough for this to apply to analog
audio. With the exception of long land lines, 600 Ohm source
impedance is inappropriate as it has too many drawbacks.
In modern analog installations it is recommended that all source

devices should have an output impedance of only 65 - 75 Ohms
which is a better match to the cable actually used. Oscillation due
to capacitive loads is prevented, and accidental short circuits cause
no damage. Receivers should have an input impedance of a few
27
tens of Kilohms. It will be seen from Fig.3.1b) that the line level is
virtually independent of the number of loads connected and the
source does not have to produce an additional 6dB in level. In the
absence of a characteristic impedance, the concept of delivering
one milliwatt to get 0dBm is meaningless and all levels are
measured in dBu where 0dBu is 0.775 V rms irrespective of
impedance.
Ω
G% G%
Ω
6285&( /2$'
Ω
. . . .

Ω Ω Ω Ω
E 6285&( $1<180%(52)/2$'6
Figure 3.1 (a) Matched impedances waste power because 6dB is lost in the
resulting potential divider. (b) Low output impedance and low input
impedance allows multiple loads.
3.2 In the digital domain the range of signal levels has two limits.
At the bottom of the range is the noise floor due to the ADC. At the
top of the range is a hard clip where the positive and negative ends
of the number scale are reached. The dynamic range achieved
depends on the wordlength and quality of the ADC, but a good
rule of thumb is that each bit in the word gives nearly 6dB of
dynamic range. Thus a 16-bit system might achieve 93dB; a 20 bit
system 115dB. The largest signal which will fit into a digital
numbering system without clipping is defined as 0dBFs (dBs Full
28
Levels and Metering
Scale) or FSD (Full Scale Digital). Clearly on such a scale all

practical levels will be negative.
It is a characteristic of digital systems that over the whole dynamic

range the signal will be completely linear as Fig.3.2a) shows.
Accordingly the best transmission through a digital system will be
one where the signal amplitude is such that it just fails to clip.
)6'
&OLS/HYHO
/
(
/,1($5
9
'<1$0,&
(
5$1*(
/
',67257,21
02/
6DWXUDWLRQ
+($'5220
62/
/
(
9
/,1($5
(
5$1*(
/
',67257,21
Figure 3.2 The differing level characteristics of analog and digital.
Analog systems are quite different. At the bottom of the dynamic

range is the noise floor, but at the top of the range there is no
sudden onset of distortion. Instead the signal quality gradually
departs from ideal as Fig.3.2b) shows. The best signal quality will
be reached at a level where the noise is far enough down, but the
onset of distortion is avoided. This level is chosen as a Standard
Operating Level (SOL), and the levels above it are known as the
headroom. In analog systems, SOL is usually +4dBm. Transients
29
can be recorded in the headroom, as transient distortion is harder

for the ear to detect.
3.3 An all-digital system neither has nor needs headroom, but in

the real world it will be necessary to interface analog signals having
headroom to the digital domain. This requires the digital system to
have artificial headroom. This is obtained by adjusting the
sensitivity of the ADC such that FSD is not reached at +4dBm but
at some higher level. Naturally if the digital channel is to have a
0dB insertion loss the all DACs must have a corresponding gain.
0HWHUVUHDG
G%
GLIIHUHQW
176& $(6(%8 3$/

' ',*,7$/ '
$8',2
6RXUFHPDFKLQH 'HVWLQDWLRQ
DVVXPHV PDFKLQHDVVXPHV
)6' G%P )6' G%P
Figure 3.3 A transparent digital transfer results in different meter readings

because the two mchines assume different analog levels for FSD.
$QDORJ 'LJLWDOIRU 'LJLWDOIRU 'LJLWDOIRU
$XGLR 9LGHR (%89LGHR
)6' 1 R6SHF G%X RU G%XG%X
1 RP LQDO 9X G%X RU G%XG%X
+HDGURRP 1 R6SHF G% G% G%
Table 3.1 Common operating levels for audio production.
Unfortunately there is no international standard for the amount of

digital headroom to leave as Table 3.1 shows. Because of this, it is
possible for apparent level changes to occur when digital transfers
are attempted. This is difficult to grasp because a digital transfer
simply transports a series of numbers from one place to another; a
mechanism which cannot change the waveform in any way.
30
Levels and Metering
Fig.3.3 shows how it can happen. An NTSC D-2 DVTR having FSD
set at 20dB above +4dBm makes a transparent digital audio transfer
to a PAL D-2 DVTR which has FSD set at 18dB above +4dBm.
Although the numbers transferred are not changed, the level
meters on the source machine read 2dB higher than on the
destination machine. The 2dB gain error is due to the different
scaling of the two machines. Fortunately the digital transfer means
that no clipping can occur and the copy will be as good as the
original even if it is at the "wrong" level.
176& 3$/(%8
'975 '975
$QDORJ$XGLR2XWSXWV $QDORJ$XGLR,QSXWV
G% G%

29X G% 29X G% 29X G%
+HDGURRP G% +HDGURRP G% +HDGURRP G%
DVUHFRUGHG DVUHFRQVWUXFWHG DVUHFRUGHG
Figure 3.4 FSD level and unexpected gain.
Fig.3.4 shows an analog transfer between two machines with

different headroom. The source machine has 24dB of headroom
above +4dBm and the destination machine has 20dB of headroom
above +4dBm. Meters on the destination machine will read 4dB
higher, but of more concern is that signals in the top 4dB of the
source machine will clip in the destination machine. If the source
DAC is not adjustable, a 4dB pad will need to be inserted between
the machines to equalize the levels and to prevent premature
clipping.
31
3.4 In order to avoid level hassle, the following tips will prove
invaluable:
1. It is vital to decide upon a single FSD level for the plant and stick
to it. NVision ADCs and DACs make this easier because they can
easily be set to clip at 20, 24 or 28 dBm. If all ADCs, and DACs in
the system are set to the same FSD, level change cannot occur
unless program material is coming from elsewhere. If you regularly
take or supply significant numbers of digital tapes to another
installation, consider using the same FSD level as they do.
2. If possible, avoid transfers through the analog domain between

digital audio devices. Using digital transfer avoids generation loss
and level discrepancies are eliminated provided all devices have
the same FSD level. If different FSD devices cannot be avoided, for
example in standards conversion work, a digital level change will
be required. This is easily done using a simple mixer such as the
NV1055.
3. If analog transfers cannot be avoided, FSD levels of ADCs and

DACs must be set correctly. Be prepared to insert pads to obtain a
fixed level drop. If level has to be increased, some equipment may
require physical modification such as resistor changing in order to
get levels to match. Remember if you have a 600 Ohm output
impedance device, you can "win" 6dB of gain by feeding it to a Hi-Z
input.
32
Transmission
'LETXIV
8VERWQMWWMSR
The AES/EBU digital audio interface dates from 1985, but was
revised in 1992. It was designed to ensure interconnection of
professional digital audio equipment irrespective of origin; a goal
which has been met.
4.1 As the bit rate of digital audio is fairly low by electronic

standards, serial transmission is relatively easy and has the
advantage that multi-core cables are not needed. For many
purposes existing twisted screened analog cable will pass AES/EBU
digital audio, but for new installations cable designed for digital
audio should be used. Cable having an impedance of 110 Ohms +/-
10 percent and a capacitance of around 12 - 13 pF per foot will
work well.
Transmitting audio data serially is not as simple as connecting the

serial output of a shift register to the cable. In digital audio, a
common sample value is all zeros, as this corresponds to silence or
muting. If a shift register is loaded with all zeros and shifted out
serially, the output stays at a constant low level. At the receiver
there is nothing to indicate how many zeros were present, or even
what the bit rate was. Clearly serialized raw data cannot be
transmitted directly, it has to be modulated in to a waveform
which contains an embedded clock irrespective of the values of
the bits in the samples. At the receiver a circuit called a data
separator can lock to the embedded clock and use it to separate
strings of identical bits.
4.2 The process of modulating serial data to make it self-clocking is

called channel coding. Channel coding also shapes the spectrum of
the serialized waveform to make it easier to equalize against cable
losses. The AES/EBU interface uses the FM channel code which is
DC-free, strongly self-clocking and capable of working with a
changing sampling rate. In FM there is always a transition at the
33
bit-cell boundary which acts as a clock. For a data one, there is an

additional transition at the bit-cell center. Fig.4.1a) shows that for
encoding purposes each data bit can be represented by two
channel bits. A channel bit determines whether or not the voltage
on the cable changes. For a data zero, the channel bits will be 10,
and for a data one they will be 11. Since the first channel bit is
always one, each data bit begins with a clock edge. Since there can
be two transitions for each data bit, peak-to-peak jitter of half a data
bit can be rejected. The lowest frequency in FM is due to a stream
of zeros and is equal to half the bit rate. The highest frequency is
due to a stream of ones, and is equal to the bit rate. Thus the
fundamental frequencies of FM are within a band of one octave.
Effective equalization is generally possible over such a band. FM is
not polarity conscious and can be inverted without changing the
data. As a result an inadvertent exchange of Pins 2 and 3 which
would cause an inversion in analog doesn’t have the slightest effect
on a digital signal.
,1 3 8 7
' $ 7$
& + $ 1 1 (/ %, 76
)0 2 8 738 7
D : $ 9 () 2 5 0
6 H U LD OF K D Q Q H OE LWV
WU D Q VLWLR Q FI3 (

/R RNXS 6 K LIW
WD E OH U H J LV WH U
' 4
'
) 0 R X W

E
& OR FN
Figure 4.1 In FM coding, each data bit is converted into two channel bits
(a), a channel bit 1 causes a transition in the transmitted signal. The FM
coder is shown in (b).
34
Transmission
Fig.4.1b) shows how an FM coder works. Data words are loaded

into the input shift register which is clocked at the data bit rate.
This will be 64 times the sampling rate in use. Each data bit is
converted to two channel bits in the code book or look up table.
These channel bits are loaded into the output register. The output
register is clocked twice as fast as the input register because there
are twice as many channel bits as data bits. Ones in the serial
channel bit output represent transitions whereas zeros represent
no change. The channel bits are fed to the waveform generator
which is a one bit delay, clocked at the channel bit rate, and an
exclusive or gate. This changes state when a channel bit one is
input. The result is a coded FM waveform where there is always a
transition at the beginning of the data bit period, and a second
optional transition whose presence indicates a one.
Ω
7ZLVWHG3DLU

7UDQVPLW Ω
5HFHLYH

Ω

Figure 4.2 Recommended AES3 interconnect circuitry
4.3 The standard driver and receiver chips for RS-422A data
communication are employed in AES/EBU, but using transformer
coupling for better common mode rejection and avoidance of
ground loops. Equalization may be employed on longer cable runs.
Fig.4.2 shows the standard configuration. The AES spec does not
specify how to handle the shield (Pin 1) at the receiver. Good
practice requires Pin 1 to be connected to electrical chassis, not
signal ground, so that shield currents cannot produce interference
by flowing in common impedance’s. Not all equipment does this,
and the practice of leaving the shield floating at the receiver was
35
adopted as a palliative. Unfortunately this results in an open house

for RF with the potential for interference. One solution is to fit a
screen decoupling ( or shield bypass) capacitor between shield
and chassis which shunts RF energy to ground.
The output impedance of the drivers is 110 ohms, and the

impedance of the cable used should be similar at the frequencies of
interest. The driver was specified in AES-3-1985 to produce
between 3 and 10V pp into 110 Ohms but this was changed to
between 2 and 7 volts in AES-3-1992 to better reflect the
characteristics of actual RS-422 driver chips.
Ω
0DWFKLQJ5HVLVWRU
,QVLGH;/53OXJ
Ω 5HFHLYH

$(65HFHLYHUDQG;/55HFHSWDFOH
Figure 4.3 Impedance correction of AES3 1985 receivers

The receiver impedance according to AES-3-1985 was set at a high
250 ohms, with the intention that up to four receivers could be
driven from one source. This has been found to be inadvisable
because of reflections caused by impedance mismatches and AES-
3-1992 specifies a point-to-point interface with source, cable and
load impedance all set at 110 Ohms. It is important for highest data
integrity to convert all old equipment to have the lower input
impedance. It is impossible to measure input impedance with an
ohm meter because of the transformer, and so it will be necessary
to refer to the equipment spec or the schematics to establish the
impedance. Fig.4.3 shows how 250 Ohm equipment can be
converted without dismantling. A 196 Ohm 1 percent resistor is
fitted inside the XLR connector between pins 2 and 3.
This technique cannot be used if the system has the previously

legal configuration of several loads connected to one source. In
this case a digital audio distribution amplifier such as the NV1021
36
Transmission
must be used. However, if the DA is mounted close to the original

loads, they need not be modified.
Although not in the AES standard, loop-through can be used in

certain circumstances to reduce the number of DAs required. If a
large number of devices in the same rack require to be fed with the
same reference, they can be looped as shown in Fig.4.4 provided
that distances are short and that all devices except the last have a
high input impedance rather than the usual 110 Ohms. NVision
equipment has a Hi-Z setting to allow this to be done. Looping
should not be attempted with equipment which does not offer this
facility.
$(65HIHUHQFH
2KP
7HUPLQDWLRQ
19 19 19 19 19 19
Figure 4.4 Accurately synchronized CODEC’s with looping

In some cases it may be uneconomic to pull new cable runs to
replace an existing analog run. At the frequencies of digital audio,
conventional analog cable has an impedance of around 45 Ohms.
Best results will be obtained if the receiver is matched to this value.
The "resistor in plug" trick can be used to lower the input
impedance. 55 Ohms across a 250 Ohm input matches 45 Ohms,
as does 77 Ohms across a 110 Ohm input. The equalizing facilities
of the NV1021 will allow reception after considerable signal
rounding.
4.4 In Fig.4.5, the specification of the receiver is shown in terms of

the minimum eye pattern which can be detected without error. It
will be noted that the voltage of 200 mV specifies the height of the
eye opening at a width of half a channel bit period. The actual
signal amplitude will need to be larger than this, and even larger if
the signal contains noise. Fig.4.6 shows the recommended
37
7PLQ
7QRP +DOIRIDELSKDVHV\PEROSHULRG
7PLQ 7QRP
P9
7QRP
Figure 4.5 The minimum eye pattern acceptable for correct decoding of
standard two-channel data.

%
G

Q
L
D
J

H
Y
L
W
D
O
H
5

)UHTXHQF\ 0+]
Figure 4.6 EQ characteristic recommended by the AES to improve

reception in the case of long lines
equalization characteristic which can be applied to signals

received over long lines. The NV1021 contains such an equalizer
which will maximize data integrity on long runs.
4.5 There is a further standard, known as AES-ID, using coaxial

cable to achieve distances of around 1000m. This is simply the
usual AES/EBU protocol but with a 75 Ohm coaxial cable carrying
38
Transmission
a one volt peak to peak signal ( +/- 20 percent); the same level as is
used in analog video systems. Fig.4.7 shows what the electrical
interface looks like. It is not compulsory to use the transformers,
but as most implementations are adaptations of existing circuits
the transformers generally remain. Standard video BNC connectors
and analog video grade cable are used. Note the shield bypass
capacitor at the receiver. This is not in the standard, but allows
increased rejection of high frequency interference.
Ω
Ω
&RD[LDO&DEOH

7UDQVPLW Ω Ω 5HFHLYH

Ω

Figure 4.7 Recommended AES3 - ID interconnectivity circuitry.
Many television studios which have gone over to digital video are
left with surplus analog video DAs, routers and cabling. It might be
thought that this equipment would be perfect for digital audio
distribution, but this is not necessarily the case. Such installations
should be treated with caution. Whereas a proper digital audio DA
or router contains reclocking and slicing which launches a clean
signal at the output, video DAs and routers are purely analog
devices which neither slice nor reclock. They will introduce
generation loss which reduces data integrity. It is possible for two
analog video devices to work separately with digital audio, yet
when put in series the combination does not work. In such cases it
is hard to identify the fault.
4.6 In Fig.4.8 the basic structure of the AES/EBU transmission can

be seen. One subframe consists of 32 bit-cells, of which four will
be used by a synchronizing pattern. Subframes from the two audio
channels, A and B, alternate on a time division basis. Up to 24-bit
sample wordlength can be used, which should cater for all
conceivable future developments, but normally 20-bit maximum
length samples will be available with four auxiliary data bits, which
can be used for a voice-grade channel in a professional application.
39

/ / 0
6 6 $ X GLR 6 DP S OH : R U G 6 9 8 & 3
% % %
3 U H D P E OH 9 D OLG LW\ ) OD J
8 VH U ' D WD
& K D QQ H O6 WD WX V
3 D U LW\ % LW
Figure 4.8 AES3 subframe format.
The least significant bit is sent first. One advantage of this

approach is that simple serial arithmetic is then possible on the
samples because the carries produced by the operation on a given
bit can be delayed by one bit period and then included in the
operation on the next higher-order bit.
The format specifies that audio data must be in twos complement

coding. If different wordlengths are used, the MSBs must always be
in bit 27 irrespective of wordlength. Shorter words are leading
zero filled up to the 20-bit capacity. The channel status data
included from AES-3-1992 signaling of the actual audio wordlength
used so that receiving devices could adjust the digital dithering
level needed to shorten a received word.
Four status bits accompany each subframe. The validity flag will be
reset if the associated sample is reliable. AES-3-1992 described the
V-bit as indicating that the information in the associated subframe
is "suitable for conversion to an analog signal". Thus it might be
reset if the interface was being used for non-audio data. The parity
bit produces even parity over the subframe, such that the total
number of ones in the subframe is even. This allows for simple
detection of an odd number of bits in error, but it also makes
successive sync patterns have the same polarity, which can be
used to improve the probability of detection of sync. The user and
channel-status bits are discussed later.
40
Transmission
Two subframes make one frame and frames repeat at the sampling
rate in use. The first subframe will contain the sample from
channel A, or from the left channel in stereo working. The second
subframe will contain the sample from channel B, or the right
channel in stereo. At 48 kHz, the bit rate will be 3.072 MHz. In
order to separate the audio channels on receipt the synchronizing
patterns for the two subframes are different as Fig.4.9 shows.
These sync patterns begin with a pulse length of 1.5 bits which
violates the FM channel coding rules and so cannot occur due to
any data combination. The type of sync pattern is denoted by the
position of the second pulse which can be 0.5, 1.0 or 1.5 bits away
from the first. The third transition is positioned to make the sync
patterns DC free.
6\QF3DWWHUQIRU&KDQQHO$
6\QF3DWWHUQIRU&KDQQHO%
6\QF3DWWHUQIRU&KDQQHO6WDWXV
&
$ &K % &K &K % &K $ &K % &K $
6
6XEIUDPH 6XEIUDPH
)UDPH )UDPH )UDPH
6WDUWRI&KDQQHO6WDWXV%ORFN
Figure 4.9 AES3 frame format.
The channel status and user bits in each subframe form serial data
streams with one bit of each per audio channel per frame. The
channel status bits are given a block structure and synchronized
every 192 frames, which at 48 kHz corresponds to a period of four
milliseconds. The channel A sync pattern is replaced for one frame
only by a third sync pattern denoting the start of the channel status
block.
41
4.7 When 24-bit resolution is not required, which is most of the

time, the four auxiliary bits can be used to provide voice co-
ordination between studios as well as program exchange on the
same cables. Twelve bit samples of the talkback signal are taken at
one third the main sampling rate. Each twelve bit sample is then
split into three four-bit words which can be sent in the auxiliary
data slot of three successive samples in the same audio channel. As
there are 192 slots per channel status block period, there will be
exactly 64 talkback samples in that period. The reassembly of the
four-bit words can be synchronized by the channel status sync
pattern. Channel status byte 2 has a bit to indicate the use of
auxiliary data in this way.
%\WH
%DVLFFRQWUROGDWD VHH)LJ

0RGHDQGXVHUELWPDQDJHPHQW VHH)LJ
$XGLRZRUGOHQJWK VHH)LJ
9HFWRUHGWDUJHWIURPE\WH UHVHUYHGIRUPXOWLFKDQQHODSSOLFDWLRQV
$(6V\QFUHILGHQWLILFDWLRQ ELWV RWKHUZLVHUHVHUYHG
5HVHUYHG

6RXUFHLGHQWLILFDWLRQ E\WHVRIELW$6&,,QRSDULW\

'HVWLQDWLRQLGHQWLILFDWLRQ E\WHVRIELW$6&,,QRSDULW\

/RFDOVDPSOHDGGUHVVFRGH EELWELQDU\

7LPHRIGD\VDPSOHDGGUHVVFRGH ELWELQDU\

&KDQQHOVWDWXVUHOLDELOLW\IODJV VHH)LJ
&5&&
Figure 4.10
42
Transmission

R
R U W R
W
W R X W

W F X
W
O O D
OH D
R
X D O
Q X D H O
D
I G V D G
H Q Q X X
W H D Q OH R OH
D ' LR
W D E
W Q
E
U
P
X D
F D D
J U 0 ] D
G R H
O
LV G
0
LV
LQ H
W H ] G + O G
O
D
] V +

W .

D OH ]
W
S + F X E + F
F
L R . Q
P . W H
O D . H
O
G X D LV
D H H
6 LQ D V 0 G V

8
%
J J
(

LQ LQ 6 O
O O G
S S ( R
H
$ U
P G P N W
D H D F H Q
V N V R R
F O K
H R H Q W
F
F O F X H
U U Q W
X H
W X H
W
L D
R D R D Q U

6 U 6 U
R J
L Q

W LO

D

P S

U
R P

I D

V

Q
L

V
L
G
V Q V Q
U LV H
D
D X

H K LR
W
V
G G W
Q L V
LY F D U D L

H
S
H LR K OH U W V

F P O W S E H V D

G H H H LV F D Y O K

H U V
V HO P
LV
R H
W R O D H H
G
Q S
H
D V Q D K V V V
W F L X G L
G G Q P
\ L V
W S O

H - V H D H
G D O Q D H H
E

D P X O LG D
7 K O Y
U K
K D U F K
V
LQ S X H Q E U 7
, S
E
H W
X I P D D V L
W
R P H

U R D LV
H
Y & P LV
H
D H K Z
W
1 H G R 1 P G R & H G 5 W
V \

O I OL

H R
U
Q D
Q

H
W
D
\ P
L
K
& E
W
U

V S

U V

L O
I

D

H H

K G
7 G
LR

U
D
)LJXUH
G

X R
L G
D
G Q
O X D

D W
D V
P H H
U G Q G
R R R R
1 P 1 P

N N
F I F
I R
O R R
O

R
E
H E

V V V
H X
V X X
X W O W
D D D
U W Q W
H V
V

O R
L O
P H V H
X Q V Q
V Q H Q
Q I
D R D
R K U K
& F 3 F

43
4.8 The professional channel status structure is shown in Fig.4.10.

Byte 0 determines the use of emphasis and the sampling rate, with
details in Fig.4.11. Byte 1 determines the channel usage mode, i.e.
whether the data transmitted are a stereo pair, two unrelated mono
signals or a single mono signal, and details the user bit handling.
Fig.4.12 gives details. Byte 2 determines wordlength as in Fig.4.13.
This was made more comprehensive in AES-3-1992. Byte 3 is
applicable only to multichannel applications. Byte 4 indicates the
suitability of the signal as a sampling rate reference.
%\WH

$XGLRFKDQQHOXVDJH 8VHUELWVPDQDJHPHQW
0RGHQRWLQGLFDWHG5HFHLYHU
GHIDXOWWRFKDQQHOPRGH 'HIDXOW1RXVHUGDWDLQGLFDWHG
0DQXDOVHOHFWHQDEOHG
ELWEORFNVWUXFWXUH
7ZRFKDQQHOPRGH DVIRU&ELWV
6LQJOHFKDQQHO PRQR PRGH

+'/&SDFNHWVFKHPH $(6
3ULPDU\VHFRQGDU\PRGH

8VHUGHILQHG
&KDQLVSULPDU\

6WHUHRPRGH &KDQLV/HIW
WR 5HVHUYHG

5HVHUYHG
WR

9HFWRUWRE\WH

UHVHUYHGIRUPXOWLFKDQQHO
Figure 4.12 Format of byte 1 of professional channel status.

There are two slots of four bytes each which are used for
alphanumeric source and destination codes. These can be used for
routing. The bytes contain seven bit ASCII characters (printable
characters only) sent LSB first with the eighth bit set to zero
according to AES-3-1992. The destination code can be used to
operate an automatic router, and the source code will allow the
origin of the audio and other remarks to be displayed at the
destination.
44
Transmission
%\WH

$X[LOODU\ELWVXVDJH 8VHGWRLQGLFDWH
5HVHUYHG
ELWPD[DXGLRVDPSOH DFWLYHDXGLR
$X[ELWVXQGHILQHG VDPSOHZRUGOHQJWK
VHHWDEOH
ELWVPD[DXGLRVDPSOH$X[
ELWVXVHGIRUPDLQVDPSOHGDWD
ELWPD[DXGLRVDPSOH$X[

ELWVXVHGIRUFRRUGLQDWLRQFKDQQHO
5HVHUYHGIRUXVHUGHILQHG

DSSOLFDWLRQV

WR 5HVHUYHG

%LWVVWDWHV $XGLRZRUGOHQJWK $XGLRZRUGOHQJWK
ELWPRGH ELWPRGH
1RWLQGLFDWHG 1RWLQGLFDWHG

ELWV ELWV
ELWV ELWV
ELWV ELWV
ELWV ELWV
ELWV ELWV
Figure 4.13 Format of byte 2 of professional channel status.
Bytes 14-17 convey a 32-bit sample address which increments

every channel status frame. It effectively numbers the samples in a
relative manner from an arbitrary starting point. Bytes 18-21
convey a similar number, but this is a time-of-day count, which
starts from zero at midnight. As many digital audio devices do not
have real time clocks built in, this cannot be relied upon. AES-3-92
specified that the time-of-day bytes should convey the real time at
which a recording was made, making it rather like timecode. There
are enough combinations in 32 bits to allow a sample count over
24 hours at 48 kHz. The sample count has the advantage that it is
universal and independent of local supply frequency. In theory if
the sampling rate is known, conventional hours, minutes, seconds,
frames timecode can be calculated from the sample count, but in
practice it is a lengthy computation and users have proposed
alternative formats in which the data from EBU or SMPTE timecode
45
The next to last byte contains four flags which indicate that certain
sections of the channel-status information are unreliable (see
Fig.4.14). This allows the transmission of an incomplete channel-
status block where the entire structure is not needed or where the
information is not available. For example, setting bit 5 to a logical
one would mean that no origin or destination data would be
interpreted by the receiver, and so it need not be sent.
&KDQQHOVWDWXV%\WH

7LPHRIGD\ XQUHOLDEOH

5HVHUYHG
/RFDOVDPSOHDGGUHVV
XQUHOLDEOH
6RXUFHDQGGHVWLQDWLRQGDWD
XQUHOLDEOH
Figure 4.14 Byte 22 of channel status indicates if some of the information

in the block is unreliable.
The final byte in the message is a CRCC which converts the entire
channel-status block into a codeword. The channel status message
takes 4 mSec at 48 kHz and in this time a router could have
switched to another signal source. This would damage the
transmission, but will also result in a CRCC failure so the corrupt
block is not used. Error correction is not necessary, as the channel
status data are either stationary, i.e. they stay the same, or change
at a predictable rate, e.g. timecode. Stationary data will only
change at the receiver if a good CRCC is obtained.
4.9 It is possible to include digital audio in the serial digital video

interface signal (SDI) using the auxiliary data capacity which exists
during vertical and horizontal sync pulses. This technique is
known as embedded audio. The audio data are identified by a
special code which video receivers will ignore.
46
Transmission
The higher clock rate of component SDI means that there is

capacity for up to sixteen audio channels sent in four Groups.
Composite SDI has to convey the digitized analog sync edges and
bursts and only sync tip is available for ancillary data. As a result of
this and the lower clock rate composite SDI only has capacity for
four audio channels in one Group. SDI has various levels of
support for the wide range of sampling rates and wordlengths
encountered in audio. A default level of the standard has been
designed specifically for the television production environment
and it is unlikely that any more complex system will be required in
practice. This is known as Level A which operates only with a
video-synchronous 48 kHz sampling rate and transmits V,U,C and
the main 20-bit sample only. As Level A transmission is a default
receivers will assume it unless told otherwise by signaling and so
only Audio Data Packets need be sent.
Fig.4.15 shows the structure of the audio data packing. In order to

prevent accidental generation of reserved synchronizing patterns,
bit 9 is the inverse of bit 8 so the effective system wordlength is 9
bits.
$GGUHVV [ [ [
%LW
BB BB BB
% % % %
% $ $ 3
% $ $ &
% $ $ 8
% $ $ 9
% $ $ $06%
% $/6% $ $
% &+ 06% $ $
% &+ /6% $ $
% = $ $
Figure 4.15 AES/EBU data for one audio sample is sent as three 9 bit
symbols. A = audio sample. Bit Z = AES/EBU channel-status block start bit.
47
Three nine-bit symbols are used to convey all of the AES/EBU

subframe data except for the four auxiliary bits. Since four audio
channels can be conveyed, there are two "Ch" or channel number
bits which specify the audio channel number to which the
subframe belongs. A further bit, Z, specifies the beginning of the
192 sample channel status message. V, U and C have the same
significance as in the normal AES/EBU standard, but the P bit
reflects parity on the three nine bit symbols rather than the
AES/EBU definition. The three-word sets representing an audio
sample will then be repeated for the remaining three channels in
the packet but with different combinations of the CH bits.
One audio sample in each of the four channels of a Group requires

twelve video sample periods and so packets will contain multiples
of twelve samples. At the end of the packet a checksum is
calculated on the entire packet contents.
As the ancillary data transfer is in bursts, it is necessary to provide

a little RAM buffering at both ends of the link to allow real time
audio samples to be time compressed up to the video bit rate at the
input and expanded back again at the receiver. In such a system all
that matters is that the average audio data rate is correct.
Instantaneously there can be timing errors within the range of the
buffers. Audio data cannot be embedded at the video switch point
or in the areas reserved for EDH packets, but provided that data are
evenly spread throughout the frame, 20-bit audio can be embedded
and retrieved with about 48 audio samples of buffering. The
buffering stages cause the audio to be delayed with respect to the
video by a few milliseconds at each insertion. Whilst this is not
serious, Level I allows a delay tracking mode which allows the
embedding logic to transmit the encoding delay so a subsequent
receiver can compute the overall delay. If the range of the
buffering is exceeded for any reason, such as a non-synchronous
audio sampling rate fed to a Level A encoder, audio samples are
periodically skipped or repeated in order to bring the delay under
control.
The extractor recognizes the ancillary data TRS or flag and then
decodes the ID to determine the content of the packet. The Group
48
Transmission
and channel addresses are then used to direct extracted symbols to

the appropriate audio channel. A FIFO memory is used to timebase
expand the symbols to the correct audio sampling rate. Care is
needed with embedded audio when it passes through SDI routers.
Special techniques are necessary to prevent clicks on hot cuts.
Section 6 deals with this issue.
49
'LETXIV
7]RGLVSRM^MRK
5.1 When multiple digital audio signals are to be assembled from a

variety of sources, either for mixing down or for transmission
through a TDM (Time Division Multiplexing) system, the audio
samples from each source must be synchronized to one another.
Such a technique has been used since the earliest days of television
in order to allow vision mixing, but now that audio is conveyed in
discrete samples, these too must be genlocked or backtimed to a
reference for most production purposes. Each source of digital
audio must be supplied with a reference sampling rate signal from
some central generator, and must return samples at that rate.
Failure to properly genlock audio sources results in intermittent
pops and crackles as sample values are corrupted. Simply having
the same nominal sampling rate is not enough. If digital audio is
being used in the television environment further synchronizing
will be required. In VTRs the digital audio and video are recorded
on the same tape tracks. The scanner or drum speed is locked to
video. It follows that the audio block rate and consequently the
audio sampling rate must be locked to video. Fig.5.1 shows how to
lock external converters to a DVTR. A spare digital audio output is
used as a reference.
7\SLFDO'975
6 \Q F
19
&RGHF
/
5
$'

/
5
'$
$( 62 X WSX WV
19
&RGHF
/
5
$'
$( 6,QS XWV
/
5
'$
Figure 5.1 Small island synchronization
50
Synchronizing
Whilst all professional digital audio devices have an AES/EBU or

SDIF-2 reference input, some recent devices may also have a video
reference input which multiplies the video frequency by an
appropriate factor to produce a synchronous audio sampling clock.
It is, however, unknown for video devices to genlock to audio.
There are two basic ways of locking digital audio in a television

environment. In the first, a master generator such as the NV5500
produces locked video and audio references, and these are
distributed separately. In the second, only video syncs are
distributed, and all audio devices are fitted with video-locking sync
generators such as the NV1080. In both cases the NV1022 DA can
be used to obtain the required fan-out. In cases where older audio
equipment requires SDIF-2 wordclock, the NV1082 can be used as
shown in Fig.5.2.
6',):RUG&ORFN
0XOWL7UDFN
5'$7
$(6
6',)
19$'
19$' $(6
3URGXFWLRQ
19
0L[HU
19$'
$(6
19$'
$(66\QF/RRS7KUX
19
$(6
6\QF
*HQHUDWRU '975
+RXVH9LGHR6\QF
Figure 5.2 The NV1802 reference generator and local island

At first sight it would appear that the two approaches achieve
identical results, but this is not quite the case. Fig.5.3 shows how a
video-locked audio sampling rate reference is obtained. First, video
H-rate is multiplied in a phase locked loop to a frequency of 13.5
MHz which is a common multiple of PAL/NTSC H-rate.
51
9LGHR+RUL]RQWDO
176&N+]
3$/ N+ ]
&RPSDULVRQ
)UHTXHQF\
#N+]
÷

0+ ]
3L[HO&ORFN
÷
&RPSDULVRQ
)UHTXHQF\
#N+]
÷
0+]
$(60DVWHU&ORFN
$(6$XGLR
÷ 6DPSOH5DWH
Figure 5.3 Locking digital audio to video horizontal frequency.
The highest common subdivision frequency between 13.5 MHz

and 48 kHz is 12 kHz. 13.5 MHz is divided down to 12 kHz and 48
kHz is obtained by multiplication in a further phase locked loop.
Whilst the frequencies are in exact ratios, the starting conditions of
the four dividers are unknown and so the relative phase of the two
signals is arbitrary. This is not a problem in the digital domain, but
if such a clock is used as a converter reference, analog phase errors
can occur. If two ADCs are driven by two different video-locked
references, the relative phase of the two analog signals is lost and
if these signals are a stereo pair the imaging will be impaired. The
simple solution is that if video locking is used, all converters which
are expected to operate with phase coherence should run from the
same 48 kHz reference.
5.2 Fig.5.4 shows that the standard 48 kHz sampling rate has fixed
relationships to all standard video and film frame rates and to other
audio sampling rates. This makes it possible to synchronize digital
audio transfers between different picture formats. In most dual-
format installations, PAL and NTSC equipment are not locked to
52
Synchronizing
N+]

N+] N+] ISV ILHOG ILHOG ILHOG
)LOP 3$/ +'79 176&
Figure 5.4 Multi-format synchronization relationships.
each other because video standards converters are designed to run

with asynchronous signals. With analog audio transfers there is no
problem. With digital audio, PAL and NTSC equipment must be
locked together. This is easily done as Fig.5.5 shows. The NV5500
Universal Sync generator provides simultaneous PAL and NTSC
sync outputs and a synchronous 48 kHz audio output. As NTSC and
PAL have no meaningful phase relationship, the NV5500 does not
attempt to provide one. Existing video sync generators are
genlocked to the NV5500 outputs. In this way all video references
within the plant are locked, and any device which multiplies video
syncs to produce an audio sampling rate will be automatically
genlocked to any other device doing the same thing even if the
two devices are on different video standards.
176&6\QF
176&6\QF*HQHUDWRU 19 3$/6\QF*HQHUDWRU
176&'$ 3$/'$
176&' 3$/'
N+] N+] N+]
$XGLR $XGLR $XGLR
,GHQWLFDO6DPSOH5DWHV
Figure 5.5 NV5500 master timing for system synchronization.
53
With locked video generators, the digital audio soundtracks can be

easily dubbed during a standards conversion operation. Note,
however that video standards converters necessarily cause a
significant delay in the video path. An NV 1060 digital audio delay
unit shown in Fig.5.6 allows the standard converter delay to be
matched in the audio, maintaining precise lip-sync.
176&'975 3$/'975
176&# 3$/#
+] 6WDQGDUGV +]

9LGHR2XW 9LGHR,Q
&RQYHUWHU
$(62XWSXWV $(62XWSXWV
$(6# $(6#
N+] N+]
19'HOD\
$(6,QSXWV

176& 3$/
9LGHR6\QF 9LGHR6\QF

19
$(6,QSXWV
6\QF*HQHUDWRU
Figure 5.6 Synchronous delay compensation
3$/'$ 3$/6\QF*HQHUDWRU
19
3$/'
176&'$
176&'
N+] N+] N+]
$XGLR $XGLR $XGLR
,GHQWLFDO6DPSOH5DWHV
Figure 5.7 NV5500 Slave timing for system synchronization.
Sometimes the video timing is determined elsewhere, for example

if two studios are genlocked for transfer over a distribution
network. In this case slave synchronization can be used. Fig.5.7
shows that the NV5500 is connected as a slave and generates NTSC
syncs from a PAL input or vice versa. When slave locking is used,
54
Synchronizing
frequency accuracy needs to be considered, as AES/EBU, NTSC and

PAL have different tolerances. Thus in theory signals genlocked to
an NTSC signal which was just within tolerance might be outside
their stricter tolerances. In practice problems are unlikely to arise.
5.3 AES11-1991 documented standards for digital audio

synchronization and requires professional equipment to be able to
genlock either to a separate reference input or to the sampling rate
of an AES/EBU input. The shift register in a receiver is generally
buffered with a parallel loading latch which allows some freedom
in the exact time at which the latch is read with respect to the
serial input timing. Accordingly the standard defines synchronism
as an identical sampling rate, but with a no requirement for a
precise phase relationship. Fig.5.8 shows the timing tolerances
allowed. The beginning of a frame (the frame edge) is defined as
the leading edge of the X preamble. A device which is genlocked
must correctly decode an input whose frame edges are within +/-
25 % of the sample period. This is quite a generous margin, and
corresponds to the timing shift due to putting about a kilometer of
cable in series with a signal. In order to prevent tolerance buildup
when passing through several devices in series, the output timing
must be held within +/- 5% of the sample period.
The reference signal may be an AES/EBU signal carrying program

material, or it may carry muted audio samples; the so called digital
audio silence or digital quiet signal. Digital quiet is useful as a
router input whereas tone is useful for continuity checking and
DAC alignment. The NV5500 and NV1000 family allow quiet or
tone to be selected. Tone can be 500 or 1000Hz and the level may
be set to FSD (0dBFs) or -20dBFs.
5.4 Although desirable, in practical situations, genlocking or

backtiming is not always possible. In a satellite transmission, it is
not really practicable to genlock a studio complex halfway round
the world to another. When genlock is not achieved, there will be
a slow slippage of sample phase between source and destination.
In some cases a totally different standard sampling rate, e.g. 44.1 or
32 kHz will need to be handled. Laser discs are mastered on DVTRs
55

Q

LR

Q V
H H
R F LV
F
LW
Q P Q
S D V D
H U Q U
F H
O D H
O
H R U R
5 W 7 W

H
K F
W

U
W
H D
Y
R Q

V
W Z
X R
S K
V

Q H
,
F

Q
D H
U
H H
F I
H
Q U

H H
U
H K
I W

H
U
R
W

D
J
R
W
LQ
W
F P
L
H W
S
V Q
H L
U
U
K H
W V
L
Z OR
% F

H V
O H
P D E

D
U Q W
V
LJ
I
E V X
X
6 P

8 V
% W
( X
G S
6 W
LR X
U ( R
H $
S
V
H Q D
O L
H
S U
G H
P H K
D U
L
V

Z

X
T G
H H
U
W
\ S
$
F H
H D F
U F
P X D

D
U
F H
I F E
E H
D

W
X W J V
6 D X
WH O Q
L
D
O P

P
L
W E

H
K W
D

7
Q
)LJXUH
Z
R
O\ K
U V

D
H O\ H
U J
D Q
H
D

U

D E F
56
Synchronizing
having 48 kHz audio, yet they require 44.1 kHz. The same problem
will occur if it is required to make a CD master from a videotape
soundtrack. These problems can be handled by sampling rate
conversion.
176&'975 3$/'975
176&# 3$/#
+] 6WDQGDUGV +]
9LGHR2XW 9LGHR,Q
&RQYHUWHU
$(62XWSXWV
$(6 $(62XWSXWV

$(6# 5HIHUHQFH

N+] 19 19
'HOD\ 65&
$(6,QSXWV $(6# $(6#

N+] N+]

$(6,QSXWV
Figure 5.9 A standard NTSC to PAL transfer.
Fig.5.9 shows a standards conversion between two DVTRs. Whilst

the ideal solution is to use an NV5500 as shown above, this may
not be appropriate where only occasional use is contemplated. All
video standards converters are designed to work without lock
between the two standards and so the video transfer takes place
with both DVTRs free running. This results in the audio sampling
rate between the two machines being slightly different. The
NV1050 sampling rate converter resolves the discrepancy by
expressing the input audio waveforms in a different sampling
structure. Note how the output sampling rate of the converter is
determined by a reference taken from an audio output of the
destination DVTR. The NV1050 can handle all four audio channels
simultaneously.
Standards converters need to interpolate along the time axis

between several video fields and this results in a significant
processing delay. The NV1060 delay unit compensates by delaying
the audio data a corresponding amount.
5.5 When film is converted to video in a 59.94 Hz system, the 24

Hz frame rate is handled by running the film 0.1% slow and
57
converting alternate frames to two and three fields. This field

repeating has the effect of portraying motion as two unevenly
spaced jumps in a five field sequence. Whilst this is acceptable for
59.94 Hz television, it produces dreadful artifacts after standards
conversion to 50 Hz. Even motion compensated standards
converters fail on such material because they attempt to reproduce
the bizarre motion. The solution is to use a special standards
converter which is identifies the third field in the 3/2 sequence
and deletes it. Pairs of fields are then de-interlaced to produce the
original film frames at 23.97 Hz. These frames are interpolated to
change to 625 lines and are recorded on a 50 Hz VTR which is
modified to run 4.004% slow so that it actually runs at 23.97 Hz.
The audio clocks change pro-rata resulting in an audio sampling
rate of 46.0339 KHz. Fig.5.10 shows that the NV 1050 sampling
rate converter can produce this frequency simply by genlocking to
the audio output of the destination DVTR which multiplies up
from 23.97 Hz. An NV 1060 audio delay compensates for the video
processing delay.
176&'975 3$/'975
176&# 3$/#
+] 6WDQGDUGV +]

9LGHR2XW 9LGHR,Q
&RQYHUWHU
$(62XWSXWV $(62XWSXWV
$(6# $(6#
N+] N+]
19 19
'HOD\ 65&
$(6#
$(6,QSXWV
N+]

$(6,QSXWV
Figure 5.10 Sample rate conversion for complex video transfers.
When the resulting tape is played at 50 Hz and 48 kHz, the video

has the same motion portrayal as if the original film had been
played on a telecine at 25 Hz and the audio stays in sync with it.
5.6 Consumer DAT recorders are extremely inexpensive and

portable, and make excellent field recorders. Although they have
no timecode or genlock facilities, they are crystal controlled and
stable and can be used unlocked with clapperboard
58
Synchronizing
synchronization. In a simple production, asynchronous field DAT

tapes may be combined with effects from a 44.1 kHz Compact
Disc player using the configuration shown in Fig.5.11. The NV
1055 Mixer needs all four inputs to be video-locked 48 kHz. The
flexibility of the NV 1050 is demonstrated particularly well here
because it can simultaneously accept two pairs of inputs at
different rates and convert them to a common rate.
5.7 Sampling rate conversion is analogous to standards conversion
&'3OD\HU
$(62XWSXWV
$(6#
N+]
$(6N+]5HIHUHQFH
$(6,QSXWV

19 19
65& 0L[0LQXV
$(6#
N+]
$(6#
N+]
5'$7
Figure 5.11 Multiple source rate conversion for audio production.
in video. It is an interpolation process using a low-pass filter to

compute the value of the input waveform at some a point between
the input samples. One of the most difficult aspects of the
conversion is to estimate the position on the time axis of the
output sample with respect to the input sample in order to
generate the filter coefficients. Any inaccuracy here is the
equivalent of having clock jitter in the original ADC. This
estimation is especially tough if the two rates are totally unlocked
as any sample phase relationship can occur. If, however the two
rates are related by a ratio of two integers, the number of filter
phases becomes finite and the performance can be enhanced
because the coefficients are more accurate. This is known as
synchronous rate conversion.
59
The NV4448 has this capability. It contains a phase locked gearbox

which has a ratio of 160:147. In addition to accepting AES/EBU
clocks, the NV4448 also accepts and generates SDIF-2 and SPDIF
signals so it can be used to interface a wide range of equipment to
the video environment. In addition to the synchronous conversion
facility, the NV 4448 operates in asynchronous mode should it
become unlocked.
$ (6 # $ (6 #
&' N + ] 1 9 N+ ]

5'$7
3OD \ H U 6 \Q F
Figure 5.12 Synchronous conversion with AES3 clocking
Fig.5.12 shows a 44.1 kHz CD player driving the NV4448 which

locks to the input data rate. The NV4448 then produces a
synchronous 48 kHz output and rate converts the input to it. An
RDAT machine locks to the 48 kHz output.
$(6# /
5'$7
N+] 19 5 'LVN
&'520
6\QF :& 0DVWHULQJ
0DVWHU
6',)
#N+]
Figure 5.13 Synchronous sample rate conversion for video.
Fig.5.13 shows an alternative situation where a 44.1 kHz RDAT

recording is synchronously converted to video-locked 48 kHz to
suit a DVTR. The RDAT machine is simply fed with a video
reference so that it produces video-locked 44.1 kHz. The DVTR
produces video locked 48 kHz. The ratio between these is exactly
160/147.
$ ( 6 # $ ( 6 #
N + ] 1 9 N+ ]
5' $ 7
6 \Q F
' 975
9 LG H R 6 \ Q F
Figure 5.14 CD ROM mastering

60
Synchronizing
Fig.5.14 shows an example of DVTR audio being rate converted to

make a laser disc. In this case the flexibility of the NV4448 is useful
because many laser disc cutters require SDIF-2 input.
5.8 In the digital domain, delay is easy and in most video

installations delay is everywhere. Line delays in decoders, field or
frame delays in noise reducers and DVEs, arbitrary delays in frame
synchronizers all conspire to give you the sound before the lips
moved. Less obvious is that the delays experienced by the video
signal also shift it with respect to the timecode.
$QDORJ975 '975
9LGHR2XW 9LGHR$' 1RLVH5HGXFHU 9LGHR,Q
7LPH&RGH 19 7LPH&RGH
$(6,QSXWV

/ 19
19$'
5 19
6\QF
Figure 5.15 Delay compensation for analog to digital conversion with

image enhancement.
The solution is to put compensating delays in the audio and

timecode signals which cancel the differential delay. Fig.5.15
shows a typical application. An analog tape is being dubbed to a
DVTR using video noise reduction. The noise reducer causes
significant delay. The timecode passing between the machines is
delayed by the NV1061 and the audio is delayed by the NV1060.
This requires ADCs in the shape of an NV1035. A video-locked 48
kHz clock comes from a NV1080 fed with house reference.
In Section 5.2 above the use of a compensating audio delay in

standards conversion applications was illustrated. In these cases
61
the timecode is not transferred because SMPTE and EBU timecodes

are incompatible. Instead the destination machine re-stripes using
its own generator from a given starting code and a time code
compensator is not required.
62
Routing
7IGXMSR
6SYXMRK
6.1 There are two basic types of routing to consider: synchronous

and asynchronous. Each has its own advantages and disadvantages.
The first question to ask is whether "hot switches" are required. A
hot switch is where a router output selects a different source on
air. This may be done in a muted pause between, for example,
programs and commercial breaks. Clearly during such a quality-
critical switch no pops or clicks are allowed. In presentation
suites, most transitions between sources are done with a mixer,
but in the case of a mixer failure it is possible to stay on air by using
the router to bypass the mixer. All transitions then have to be hot
switches.
If hot switch capability is required, then only synchronous routing

will do the job. However, for many purposes, such as in edit suites,
routers will only need to change configuration when no wanted
material is passing through. In this case pops and clicks at the
switch point are acceptable as they never appear on the finished
product. Asynchronous routing can then be used; an option
which, in the right application, can offer savings over
synchronous.
6.2 Imagine what would happen on a video tape if instead of

performing an assemble edit the machine was just forced into
record. An asynchronous AES/EBU switch does the same thing to
the audio data. The AES/EBU serial data stream is precisely created
so that the position of each bit after the sync pattern determines
the function of that bit. Consider a switch made between two
AES/EBU signals having the random phase relationship shown in
Fig.6.1a). In the time after the switch point, but before the next
sync pattern, the receiver is fooled into putting the bits from the
second source in the wrong place in the sample. The result is a
corrupted sample value which on conversion will result in an
analog voltage which is likely to have a very different level to those
either side - in other words a click. A further problem is that the
63
random jump in sync pattern phase can cause downstream devices

to lose lock momentarily causing further clicks or mutes.
$(6
$ 5 % / $ 5 % / $ 5 % / $ 5
$(6
/ $ 5 % / $ 5 % / $ 5 % / $
$(6
$ 5 % / $ 5 % / $ 5 % / $ 5
$(6
/ $ 5 % / $ 5 % / $ 5 % / $
$(6
/ $ 5 % / $ 5 % / $ 5 % / $
$(6
/ $ 5 % / $ 5 % / $ 5 % / $
E
VHFPD[
7DNHDQG$(6IUDPH
F
9LGHRYHUWLFDO
Figure 6.1 (a) Random inputs. (b) Phase alignment. (c) Video transition
6.3 Data corruption can only be avoided by having all inputs to the
matrix genlocked as in Fig.6.1b). It is now easy to switch during
the sync pattern or frame boundary so that the output changes
effectively between samples. In this case no corruption of the
sample value takes place. Having to genlock all inputs to the same
64
Routing
phase is highly inconvenient and in practice is not necessary

because a small buffer memory on each router input can be used to
phase align all inputs to a single reference. When this is done, it is
only necessary for the sampling RATES to be locked; the exact
input phase does not matter. This is the operating principle of a
re-phasing synchronous router such as the NV3512SA. With all the
matrix inputs phase locked, switching does not disturb the output
data phase. In the television environment a synchronous digital
audio router will need to switch in the vertical interval to match
any associated video material. Whilst the audio sampling rate is
video locked, the audio phase is arbitrary. Consequently a video-
slaved audio router will switch on the next sync pattern after video
vertical, causing a delay of up to 20 microseconds as shown in
Fig.6.1c). This is generally insignificant.
6.4 Asynchronous routers don’t have the re-phasing buffers and so

are less expensive. Although they can pop and click on transitions,
this may not matter for applications such as monitoring. They also
have some advantages. There is no need for the audio sampling
rate to be locked to video and in fact any sampling rate can be
used. This can be useful for handling signal sources such as
consumer DAT machines or CD players which have no genlock
input. However, it is possible that a downstream device may not be
able to lock to a free-running signal. Some devices have a very small
lock-in range obtained by tweaking a crystal oscillator and are only
really happy with a genlocked input.
6.5 When digital audio is embedded in serial digital video (SDI), the
act of switching a video router usually interrupts the AES/EBU
framing and results in a click. If hot switching is needed, the audio
must be stripped out of the video (disembedded) first. Audio
routing must be done in a synchronous audio router prior to
re-embedding. Fig.6.2 shows how this is done. Assuming a cut is
being made between two SDI signals, each signal must have a
disembedder so that clean digital audio data are available from both
SDI signals. The video router performs the video cut, and the audio
cut is performed in a synchronous audio router prior to re-
embedding. Installing a de-embedder on every input to an SDI
65
6', ',6 9,'(2

(0%(''(5 6',
(0%(''(5 5287(5
287
6', ',6
(0%(''(5
$8',2
5287(5
Figure 6.2 Disembedding prevents pops and crackles at SDI switches.
video router would be very expensive, and the alternative is have a

smaller number (at least two) of de-embedders on dedicated SDI
router outputs. In this way a de-embedder can be assigned in
advance to a video signal to which the SDI router is about to cut.
The de-embedded audio is then available to the parallel
synchronous audio router which cuts to it on the next audio frame
boundary after the video cut. After the cut, the outgoing de-
embedder is no longer in use and can be re-assigned to the next
source. Care is needed to ensure that the SDI router does not
switch any de-embedder input when the output is in use. It should
be recalled that the embedding and de-embedding processes cause
a slight delay to the audio which may build up if these techniques
are cascaded.
6HULDO,QSXWV 6HULDO2XWSXWV
6HULDO
$(6,QSXWV $(62XWSXWV 9LGHR
5RXWHU
(PEHGGHU
196$
6\QFKURQRXV 19 19
$(65RXWHU
19 19
Figure 6.3 A large audio break-away configuration.
66
Routing
Synchronous routing does not cause sample corruption, and so a

hot switch during a quiet section between programs will not be
audible. However, a hot switch during program material could still
cause a click if there is a disparity between the audio waveforms at
the switch point. The same click would occur switching analog
signals and the solution is the same for both: a short crossfade is
needed. Fig.6.3 shows how a small digital mixer such as the
NV1055 can be slaved to a synchronous digital audio router to
perform crossfades instead of cuts.
6', 6', 6',

,1 5287(5 287
(0%(''(5 ',6(0%(''(5 ',6(0%(''(5
&5266
)$'(5
HJ19
Figure 6.4 Hot cutter using cross-fades. Incoming audio is dis-embedded

and fed to fader. After the cross-fade the previous input dis-embedder is no
longer required and can be assigned to the next input.
Fig.6.4 shows the ultimate hot cutter for embedded audio. Two
assignable de-embedders provide clean audio prior to video cuts
and a crossfade is made between the audio signals prior to re-
embedding.
6.6 Conventional routers operate on the entire AES/EBU signal

which contains a pair of audio channels. What happens if you want
to route individual audio channels? The answer is to use a sub-
frame router. Fig.6.5 shows that in this useful device a pair of
AES/EBU inputs carrying four audio channels are phase aligned and
converted internally into four AES-like signals where the same
audio channel appears in both sub-frames. This technique is used
67
$(6
$ &K % &K $ &K % &K $ &K % &K
$(6
& &K ' &K & &K ' &K & &K ' &K &
&KDQQHO$
$ &K $ &K $ &K $ &K $ &K $ &K $
&KDQQHO%
% &K % &K % &K % &K % &K % &K %
&KDQQHO&
& &K & &K & &K & &K & &K & &K &
&KDQQHO'
' &K ' &K ' &K ' &K ' &K ' &K '
E
Figure 6.5 (a) Two AES3 inputs. (b) Phase aligned, four channel audio.
in the NV 1055 digital mixer which is shown in Fig.6.6. The NV

1055 is a four layer 4 x 1 mixer where each output is a linear mix
of the four inputs. Using zero and unity gain coefficients the NV
1055 can route any audio channel to either sub-frame of either
output. Because of the mixing facilities, the routing function can be
implemented with true cross fades, eliminating audio waveform
jumps on hot cuts.
6.7 Whilst major installations will require a large central

synchronous router, there are many applications, such as small
suites and project rooms which need smaller, simpler
68
Routing
,QSXW%XV /D\HU''
/D\HU&&

/D\HU%%
D E /D\HU$$
$ $
$(6&&''
$(6&'
D E
% %
D
)RUPDW
$(6$$%%
'H)RUPDW D E

& &
$(6$%
D E
' '
Figure 6.6 NV1055 conceptual architecture.
asynchronous routers. The NV1038 is an 8 x 8 router element

which is easily cascaded to make larger routers. Fig.6.7 shows how
a 16 x 16 router is built using four elements.
,QSXWV
[$(6 [$(6
5RXWHU 5RXWHU
,QSXWV
[$(6 [$(6
5RXWHU 5RXWHU
19&RQWURO
2XWSXWV 2XWSXWV
:LUHG2U&RQQHFWLRQ :LUHG2U&RQQHFWLRQ
Figure 6.7 A 16 x 16 Matrix with 8 x 8 building blocks.
The input wiring requires some care in order to comply with

transmission line practice. As section 4 showed, loop through can
be used provided all but the last device is set to high impedance.
Fig.6.8 shows the example of four elements in an 8 x 32 matrix.
69
19&RQWURO
,QSXWV
[$(6 [$(6 [$(6 [$(6
5RXWHU 5RXWHU 5RXWHU 5RXWHU
D
2XWSXWV 2XWSXWV2XWSXWV2XWSXWV
,QSXW G G G

+L= +L= +L= +L=
'ULYHU 'ULYHU 'ULYHU 'ULYHU
,QSXW ,QSXW ,QSXW ,QSXW

E
%ORFN %ORFN %ORFN %ORFN
Figure 6.8 A conceptual 8 x 32 matrix (a). A detail of one AES3 input bus
Note how three of the elements have their terminating switch left
open; the last one is closed to give the line 110 Ohm termination.
Distances d1, d2 etch must be short compared to the cable run
from the source.
Output cascading requires a little more thought because we can’t

have two outputs trying to drive the same cable. This problem is
solved using special drivers which can adopt a high impedance
state. Fig.6.9 shows that several drivers can then be wired to the
same output provided that the control system set all but one to a
high-Z state. Once more transmission line practice must be
considered. When operating, a driver has an output impedance of
110 Ohms, however, all but the last driver sees an unterminated
stub in parallel with the terminated output cable. This stub can
cause reflections which potentially reduce the system data integrity
unless distances d1, d2 etc. are kept to less than a couple of feet.
70
Routing
,QSXW ,QSXW ,QSXW ,QSXW
%ORFN %ORFN %ORFN %ORFN
+L= +L= +L= +L=
'ULYHU 'ULYHU 'ULYHU 'ULYHU
G G G 2XWSXW
Figure 6.9 A detail of one AES3 output bus.
The control system is simplified because each 8 x 8 element can be

set to a different address on the RS 422 control bus and is
configured based on the input and output addresses associated
with the whole matrix. In this way an external intelligent
controller can operate the cascade as if it were one large router.
6.8 Timecode is a vital signal yet it is often taken for granted.

Although timecode signals fit within the bandwidth of audio, it is
often forgotten that they are actually digital signals. Time code has
a data rate of 2.4 kBits/sec which is modulated with bi-phase
coding to produce a DC-free signal where most of the power is
concentrated in a bandwidth from 1.2 to 2.4 kHz. Good practice
requires that digital signals are sliced and re-clocked at each stage
of transmission to ensure data integrity. Reclockers output a clean
constant amplitude trapezoidal waveform as specified in SMPTE
12M. Analog distribution and routing doesn’t do that and the result
is signal rounding and jitter which impairs the signal. To make
matters worse, many recorders don’t reclock the replayed T/C,
they simply amplify the signal from the heads resulting in a signal
amplitude that varies with tape speed. In the absence of standards
for the electrical interface, timecode signals are found in balanced
XLR and unbalanced BNC form.
71
6.9 Fortunately there is an alternative to this flaky, crossed fingers

approach. All NVision T/C routers use correct slicing and
reclocking. Fig.6.10 shows that the input signal is sliced to a binary
signal, rejecting noise and waveform distortions. Routing is actually
done with this binary signal in simple logic gates.
Input A
Input B
Processed Internal Input
Figure 6.10 Time-code input signal processing.
Following the routing step, the binary signal is processed as shown

in Fig.6.11 to re-create the textbook-perfect SMPTE 12M T/C signal
having trapezoidal edges whose rise time is a function of
frequency. This performance is maintained over a tape speed range
from 1/30 normal to 100 x normal. The balanced/unbalanced
dilemma is resolved by inputs and outputs which are designed to
work with both. All inputs use a balanced amplifier which
accommodates balanced signals directly or unbalanced signals by
tying the inverting input to the shield. The output stages use an
electronic virtual transformer which normally produces a balanced
Internal Output
Frequency Content
Output
Rise time varies with rate.

Proportion is constant.
Figure 6.11 Time-code output signal processing.
72
Routing
output. However, if one output is tied to shield, the result is an

unbalanced signal with the same amplitude. Fig.6.12 shows these
simple unbalancing techniques.
%DODQFHG,QSXW %DODQFHG2XWSXW

6LQJOH

(QGHG
,QSXW &
&
6LQJOH
(QGHG
2XWSXW
Figure 6.12 Balanced and single ended time-code connections.
Whilst giving optimum data integrity to T/C signals and interfacing

balanced or unbalanced equipment, the NVision T/C routers
actually use less space and power than traditional analog routers.
For example the NV 3064TC provides a 64 x 64 matrix in only 6RU
and requires 45 watts. An analog router might need three times
that space and ten times the power. In practice the power saving
is greater because you don’t pay extra for the air conditioning to
take the heat away. NVision T/C routers easily slave to your
existing control system, eliminating changeover grief.
6.10 Unlike audio, video and timecode signals which only travel
from the source to the destination, control signals such as the RS
422 standard used in SMPTE 207M are bi-directional. Commands
can be issued on one signal pair and status can be returned on
another. Instead of source/destination we need terms such as
master/slave and controller/controlled as Fig.6.13 shows. Edit
controllers are easy to grasp; they always want to be the master. In
contrast, VTRs have split personalities. During the day a VTR might
be a slave to an edit controller; by night it’s the master for
duplication. SMPTE 207M calls for commands to be transmitted on
73
(GLWRU 0DFKLQH
&RQWUROOLQJ &RQWUROOHG

&RPPDQGV
7UDQVPLW 5HFHLYH

U U
R R
W W
F F
H H
Q Q
Q Q
R R
& &

Ð Ð
' '
Ó Ó

5HVSRQVH
5HFHLYH 7UDQVPLW

Figure 6.13 SMPTE 207M Port definition and pin-out.
975$ 975%
&RQWUROOLQJ &RQWUROOHG
7UDQVPLW 7UDQVPLW

5HFHLYH 5HFHLYH

Figure 6.14 Controlled/ControllingVTR circuitry.
pins 3 and 8 and cables which are wired pin-for-pin. If a device

wants to change from a slave to a master, it has to stop receiving on
pins 3 and 8 and drive them instead. Fig.6.14 shows how a VTR
changes its personality; it simply swaps over the pin pairs on the
connector so that command and response signals exchange.
74
Routing
Unless the device at the other end follows suit we have two drivers
fighting over the voltage on the wires. A conventional router can’t
change and often blown driver chips were the result. On later
VTRs two connectors were fitted, and hardwired in such a way
that one would have the right pinouts for master operation and
one would have the right pinouts for slave operation. This concept
is shown in Fig.6.15. The idea was that a two layer router would
handle the master and slave signals separately. Unfortunately, the
hardwiring causes command signals entering the slave socket to
appear on the response input of the master socket. Unless the
router output driving the response signals to the VTR can go to a
high impedance state, once more there is a contention. Having a
two layer router raises the cost and reduces flexibility.
&RQWUROOLQJ
'HYLFH
;07 &RPPDQGV

5&9 5HVSRQVH
2XW ,Q

5RXWLQJ
975
6ZLWFK
;07 ,Q 2XW ,Q 2XW

&RQWHQWLRQ
5&9

;07

5&9

Figure 6.15 RS-422 data contention with standard routers.
The approach taken in the NV3128D machine control router is

simply to provide a single layer bi-directional router which can
dynamically change its ports between master and slave operation.
Unless a control path is set up, the router remains in a high
impedance state to prevent contention damage. Only a single cable
is needed to each machine and because the master/slave switching
75
is automatic, there are no switches to fiddle with and no time

wasted. Fig.6.16 shows a dynamic port in detail.
7UDQVPLW

5HFHLYH
&URVVSRLQW
0DWUL[2XWSXW
&URVVSRLQW
0DWUL[,QSXW
3RUW&RQWUROOHU
)URP&RPPDQG
,QWHUSUHWHU

5HFHLYH
7UDQVPLW
Figure 6.16 RS-422 Dynamic port detail.
A further machine control problem occurs when routers switch a

different VTR to an editor. If the switch happens two quickly the
edit controller doesn’t realize that the VTR has changed and may
carry on using the wrong protocol. The NV 3128 overcomes this
problem by inserting a time-out pause between dropping the old
machine and selecting the new one. The editor senses the
communication loss and re-establishes protocol. Fig.6.17 shows a
machine control router with dynamic ports.
6.11 In addition to the balanced RS 422 standard, there is a lot of RS

232 equipment around. RS 232 uses a single unbalanced wire for
each direction. A master or controller device is called DTE (Data
Terminal Equipment) whereas a slave or controlled device is called
DCE (Data Communication Equipment). RS 232 ports do not
change dynamically: if a piece of equipment is built as a DTE, that’s
what it always is. Fig.6.18 shows an RS 232 link. Note the
76
Routing
'\QDPLF3RUWV '\QDPLF3RUWV

&RQWUROOLQJ
0DFKLQH

"
V
W
X
S
"
,Q

"
"

&RQWUROOHG
0DFKLQH

2XWSXWV
3RUW
&RQWUROOHU
&RPPDQG
,QWHUIDFH 195RXWLQJ6ZLWFK
Figure 6.17 Dynamic ports in an X - Y architecture.

additional signals which are used for handshaking. If the software
in the machine is designed for "smart modem" operation, these
extra handshake lines can be tied up as shown in Fig.6.19 so that
each device effectively shakes its own hands. The interface then
defaults to a transmit line, a receive line and a common line which
is needed because the signals are not differential.
When configuring a system with RS 232 devices, it is only

necessary to know whether a given device is DTE or DCE. The
77
7'
5'
576
&76
'7( '&(

'65
'DWD 'DWD
&RPPRQ
7HUPLQDO &RPPXQLFDWLRQ
(TXLSPHQW '&' (TXLSPHQW
'75
5,

565&

3LQVRUDUHQRWVXSSRUWHG
RQPRVWPLFURFRPSXWHUV
Figure 6.18 Industry standard 25 pin RS-232 connection.
'7( '&(
7' 7'
5' 5'
576 576
&76 &76
'65 '65
&RPPRQ &RPPRQ
'&' '&'
'75 '75
5, 5,
Figure 6.19 An RS-232 smart modem connection
router can then be programmed to act as the opposite type of

device. In fact this can be done by the same dynamic port principle
as is used for RS 422. The only difference is that the port never
changes. The port type is set up using the NVUTILS configuration
program and remains in non-volatile memory until the
configuration needs to be changed.
78
Routing
19'5RXWLQJ6ZLWFK
563RUW
7UDQVPLW

5HFHLYH

7R)URP
5RXWHU
0DWUL[
3RUW&RQWUROOHU

5HFHLYH

7UDQVPLW
563RUW
7UDQVPLW
5HFHLYH
7R)URP
5RXWHU
0DWUL[
3RUW&RQWUROOHU
5HFHLYH
7UDQVPLW
Figure 6.20 routing and conversion )URP&RPPDQG
of RS-232 and RS-422 ,QWUHSUHWHU
79
Fig.6.20 shows that it is easy to interface RS 422 to RS 232

equipment capable of "smart modem" operation.
6.12 Router control equipment can represent a significant

investment. Configuring a large system to work the way you want
it can take many man hours of effort. This effort should not be lost
when the router is replaced. A further consideration is that the
existing control system is familiar to the users. NVision philosophy
is that a new router should be able to work with the existing
control system.
All NVISION routers are provided with a Control Interface Module.

This device is capable of interpreting the commands of a large
number of router control systems. RS 422 and parallel interfaces
are supported allowing your new NVision router to function at any
level with any address space. Table 6.1 shows the currently
supported protocols. Please call if your system is not listed here.
6HULDO56
0)* )250$7
%76 5HPRWH%LQDU\3URWRFRO
(6%XV
3UR%HO 5656
8WDK $96$96
3/3/
9LVWHN 56
6HULDORU3DUDOOHO
+DUGZDUH
0)* )250$7
*9* +RUL]RQ
3(6$ 3LQ3DUDOOHO
7DEOH&XUUHQWO\VXSSRUWHGSURWRFROV
80
Routing
The router matrix status and configuration options are stored in

SRAM with capacitor backup. Table 6.2 shows the dynamics of the
SRAM. Should the control system fail the last map is retained
indefinitely.
0DWUL[ RXWSXWVLQOHVVWKDQ
5HIUHVK7LPH PVHFFRQWLQXRXV
0DWUL[6WRUDJH PLQXWHVJXDUDQWHHG
(OHPHQW5HWHQWLRQ KRXUVW\SLFDO
0DWUL[,QLWLDOL]H /DVWNQRZQVWDWHKDUG
RQ3RZHU8S FRGHGGHIDXOWVHOHFWDEOH
7DEOH 0DWUL[PHPRU\G\QDPLFV

Router transition timing is critical and depends upon the

references available. Table 6.3 shows the priority structure of the
reference system.
5HIHUHQFH 9LGHR9HUWLFDO176&RU3$/
3ULRULW\
5HIHUHQFH $(6([WHUQDO)UHTXHQF\
3ULRULW\ 5HIHUHQFH
5HIHUHQFH ,QWHUQDO)UHH5XQQLQJ
3ULRULW\ $(62VFLOODWRU
5HIHUHQFH 6RIWZDUH:DWFKGRJ([HFXWH
3ULRULW\ HDFK&RPPDQGDV5HFHLYHG
7DEOH7UDQVLWLRQWLPLQJSULRULW\VWUXFWXUH
6.13 To further speed the incorporation of an NVision router into

your system, a software package known as NVUTILS is provided.
This works on any IBM compatible PC running DOS 3.1 or above
and communicates with the diagnostics port of the router. This is
located near the front of the Control Interface Module. Possibly the
most useful program in NVUTILS is called NVMAP. NVMAP allows
the following functions:
81
1/ Configuring router levels. Allows setting up of two partitions

which is useful for creating AES/EBU 1/2 and 3/4 layers or for
routers which have asynchronous audio and timecode modules in
the same frame.
2/ Physical and logical addresses can be assigned. A new router will

probably not have the same physical addresses as the old. This is
no problem because address reassignment allows the old control
system to function as normal. All multi-level or multi-layer salvos
programmed into the old system will still work.
3/ Layers can be logically linked. Any time a take command is

issued, its linked source and output are automatically connected.
AES/EBU 1/2 can be linked to 3/4. This is useful for audio-follow-
video operations.
4/ For maximum flexibility, a router can be configured as 1/2 and

3/4 linked as in function 3, but retaining the ability to function as a
flat router. The linked router and the flat router are logical levels. A
priority structure is provided to determine which layer wins should
a conflict arise.
5/ Because they have dynamic ports, NVision machine control

routers can be configured for broadcasting. NV3128 routers allow
many destinations to listen to one host but with only one
destination responding. This is done with virtual layers as in
function 4.
If you are starting out rather than replacing a router, NVTAKE is

software which allows complete router control. Single takes and
salvos can be executed. Salvos can be stored and retrieved.
Execution is by a single keystroke.
6.14 The AES/EBU interface works well within its distance

limitations, but for transmission over longer distances, perhaps
between remote studios, an alternative is needed. The NV2000
system allows up to 10 channels of digital audio to be multiplexed
into one bit stream which can be transmitted on digital coaxial or
fiber optic lines, or on wideband analog video links.
82
Routing
Fiber optic technology uses two different techniques depending

on the length of the link. When light pulses are launched down an
optical fiber, they travel by internal reflection. Light traveling
nearly parallel to the axis reflects less often than light traveling
more obliquely and these light components travel different
distances. At the receiver this multi-mode reflection shows up as
edge smearing which gets worse with distance and limits the range
to a few kilometers. The more expensive single mode fiber only
allows propagation parallel to the axis of the fiber. This eliminates
the edge smearing and allows much greater distances to be
covered without a repeater; typically 50 kilometers.
Many Telcos have single mode fibers with enormous bandwidth.

At each end, these fibers have multiplexers which allow several 6
MHz analog video signals to be combined into one optical signal.
The NV 2000 can be used with such a single mode mux because it
can create a 75 Ohm analog pseudo-video signal carrying the audio
data.
Wave Division Multiplexing (WDM) is an exciting technology

which allows more than one signal to share the same optical fiber
without interference. Each signal uses a different wavelength,
typically 1550 nm and 1300 nm. The wave division multiplexer is
essentially an optical prism which separates light according to its
wavelength. Fig.6.21 shows WDM being used to provide full
duplex operation on a single optical fiber.
The NV 2000 produces a 18.432 Mb/sec data stream directly on
optical fiber, on 75 Ohm digital coaxial cable, or as a pseudo-
analog video signal. Each data stream can carry up to 10 audio
O 2QH)LEHU O
:'0 :'0
O O
' '
0 0
0 3 3 0 3 3
,QSXWV 8 2XWSXWV ,QSXWV 8 2XWSXWV
8 6 6 8 6 6
; ;
; ;
Figure 6.21 WDM’s increase channel capacity.
83
channels. These can be input as analog or 48 kHz AES/EBU digital.

The multiplexer locks to 48 kHz or to a video reference and all
input modules get their clocks from the multiplexer. The
demultiplexer automatically creates the same clocks from the
transmitted signal which drive all the output modules. In duplex
installations, the remote mux may be locked to the remote demux
so that returning signals are synchronous.
In some applications, such as accessing a remote DVTR for audio

layback production, it is necessary to transmit timecode and
control signals in addition to the audio. Fig.6.22 shows that two
audio channels in the multiplex can be replaced with an auxiliary
channel carrying RS-422 control data, time code and cue audio.
$XGLR(GLWRU '975
56&RQWURO 56&RQWURO
9LGHR
2XWSXWV 2XWSXWV 6\QF
7& 7&
,QSXWV ,QSXWV
7& 7&
19 19
' '
0 0
$(6 $(6 $8; $(6 $(6 $8; 0 3 3 $(6 $(6 $8; $(6 $(6 $8; 0 3 3
8 8
,Q ,Q ,Q 2XW 2XW 2XW 8 6 6 ,Q ,Q ,Q 2XW 2XW 2XW 8 6 6
; ;
; ;
6\QF
Figure 6.22 Remote audio layback.
84
How do I go digital
'LETXIV
,S[HS-KS(MKMXEP#
If you are starting afresh, there is really only one choice today.
Digital audio routing equipment offers significant savings over
analog technology. It takes up less space, uses less power, needs
less maintenance.
7R$QDORJ 7R'LJLWDO
2XWSXWV 2XWSXWV
([LVWLQJ 19
$QDORJ 'LJLWDO$(6
19
$XGLR $XGLR
&RGHF
5RXWHU 5RXWHU
,QSXWV ,QSXWV
)URP$QDORJ )URP'LJLWDO
Figure 7.1 Mixed analog and digital audio routing.
Not many people are in the position to scrap an old router and start
over. There is an investment in existing equipment and it is
senseless to throw it out. In any case this is quite unnecessary.
Since most new digital audio equipment will be equipped with
AES/EBU interfaces, the best way of getting into the act is to add a
digital audio router so that new devices can be incorporated into
the system. Eventually, the digital router will replace the analog
router, but initially, both work side by side. Fig.7.1 shows what
happens. Instead of equipping all existing analog devices with
ADCs and DACs at great cost, the analog router is retained and
joined to the digital router with a smaller number of ADCs and
85
DACs. The control system of the digital router is slaved to the old
router so that they work as one. When an old analog device is
replaced, the new digital device goes on the new router. Eventually
all devices are replaced and the old router goes.
Table 7.1 Mixed format and digital solutions
Step 1. Determine the router size needed for analog equipment,

digital equipment, and the analog/digital crossover. If
the crossover exceeds 50%, use an all digital router.
For calculation, use stereo audio crosspoints. AES3
digital accomodates stereo in one signal.
Step 2. Add the number of crossover inputs and outputs to the

dimensions of the analog and digital router matrices in
Step.1
Step 3. Calculate the cost of the stereo analog and digital

routing matrices.
Step 4. Add the cost of stereo A/D and D/A conversion used to
bridge between the two formats. Use one A/D and one
D/A for each stereo crossover.
Step 5. This is the raw cost of the mixed format solution.
Step 6. Add the number of stereo analog inputs and outputs to

the number of digital audio inputs and outputs.
Step 7. Calculate the cost of this digital audio router.
Step 8. Add the cost of stereo A/D and D/A conversion for all
analog equipment.
Step 9. Compare the total in step 9. with that in step 5.
Step 10. This ia the raw cost of the digital solution.
In order to judge which approach is best, try the assessment

technique in Table 7.1. If the all digital approach turns out only
slightly more expensive, bear in mind that it will offer the
following:
86
How do I go digital
a) Any material anywhere in the facility can be accessed without

degradation.
b) Future proof; all future equipment will have AES/EBU interfaces.
c) Low expansion cost: digital crosspoints are cheap.
d) Analog equipment such as VTRs can be replaced by digital

simply by bypassing the converters. No other change is needed.
e) Less power is required; less air conditioning load.
87
*968,)66)%(-2+
The following reading list (Focal Press) builds on the introduction

given here, but covers all the necessary theory and practice of
digital production.
Introduction to Digital Audio

Introduction to Digital Video
The Art of Digital Audio
The Art of Digital Video
The Digital Video Tape Recorder
The Digital Interface Handbook
Audio and Video Compression
References to products with ‘NV’ numbers refer to available

NVISION products. Specific information on these items can be
found in the 1995 NVISION catalog, available from NVISION Inc.
or your nearest NVISION distributor.
NVision2
125 Crown Point Court
Grass Valley, CA 95945
Phone +1 (530) 265 1000
Fax +1 (530) 265 1021
88

Digital Audio White Paper

Uploaded by

Copyright:

Available Formats

Digital Audio White Paper

Uploaded by

Document Information

Original Description:

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

Digital Audio White Paper

Uploaded by

Copyright:

Available Formats

8LI:MHIS)RKMRIIVkW+YMHI

John Watkinson is an independent consultant and author in the

He regularly delivers papers at conferences including AES, NAB,

Our thanks to John Watkinson for agreeing to write this

The most important task in designing a digital audio system

Some of the diagrams contained within this book

Television production equipment is changing over to digital

There are many quality related reasons why digital audio is

Television sound quality is becoming increasingly important.

Against this background, going digital in some way will be essential

1.1 Although there are other approaches, there is one system,

The sampling process seems to be taking no notice of what

In a sense sampling has made audio more like video, which is

Figure 1.1 In pulse code modulation (PCM) the analog waveform is

Fig.1.1 also shows that each sample is also discrete, or represented

Essentially, digital audio carries the original waveform numerically.

at which the measurements were taken and the accuracy of the

Figure 1.2 In a binary number, the digits represent increasing powers of

is determined only by the accuracy of conversion and is

different voltages, magnetically by the reversal of the direction of

Slicing resists noise, but it is powerless against jitter which requires

The combination of a slicer and a phase-locked loop is called a

The reclocker is found everywhere in digital systems. Routers use

engineered digital audio system is independent of the transmission

Someone who is only interested in sound quality will judge the

Digital circuitry uses standardized component outlines and is easier

Once audio is in the digital domain, it becomes data, and as such is

listening at any speed to audio from a memory. The edit can be

Fig.1.5 shows a simple digital system consisting of an analog to

Figure 1.5 When reclocking is used, the routing system is transparent.

2/ The quality is determined by the converters only if there are no

The quality of digital audio is determined by the accuracy of

signals to have a precision far in excess of that required by video.

Figure 2.2 The sampling process requires a constant-amplitude pulse train

2.2 The sampling process originates with a pulse train which is

and below the carrier, sampling also produces sidebands although

The sampled signal can be returned to the continuous-time domain

This filter has a frequency response which prevents the images

If an input is supplied having an excessive bandwidth for the

Figure 2.4 In (a) the sampling is adequate to reconstruct the original

Aliasing can be considered equally well in the time domain. In

The ideal filter with a vertical "brick-wall" cut-off slope is difficult

2.2 Sampling theory is only the beginning of the process which

For broadcast purposes a 15kHz audio bandwidth is considered

Frequency is used in the NICAM 728 stereo TV sound system and

2.3 The sampling clock has to be generated very precisely. Fig.2.5

Figure 2.6 Effects of sample clock jitter on signal-to-noise ratio at different

reduced. The nature of the unwanted signal depends on the

Fig.2.6 shows the effect of differing amounts of random jitter with

Clock jitter is not necessarily random. One source of clock jitter is

If an external clock source is used, it cannot be used directly, but

2.4 Quantizing is the process of expressing some infinitely variable

Fig.2.7a) shows that the process of quantizing divides the voltage

Figure 2.7 Quantizing assigns discrete numbers to variable voltages. All

2.5 Fig.2.7b) shows that the transfer function of an ideal quantizer

The non-linearity of the transfer function results in distortion,

2.6 At high signal levels, quantizing error on program material is

ELW  ELWWUDQVIHUIXQFWLRQ

ELW ELWWUDQVIHUIXQFWLRQ

. . . .

176& $(6(%8 3$/

)6' 1 R6SHF G%X RU G%XG%X

1 RP LQDO 9X G%X RU G%XG%X

+HDGURRP 1 R6SHF G% G% G%

29X G% 29X G% 29X G%

+HDGURRP G% +HDGURRP G% +HDGURRP G%

& + $ 1 1 (/ %, 76

WU D Q VLWLR Q FI3 (

19 19 19 19 19 19

7UDQVPLW Ω Ω 5HFHLYH

)UDPH )UDPH )UDPH

$(6V\QFUHILGHQWLILFDWLRQ ELWV RWKHUZLVHUHVHUYHG

&KDQQHOVWDWXVUHOLDELOLW\IODJV VHH)LJ