8PSK Modulation: Maximizing Capacity For Broadcast Networks

Mark Weigel
Comtech EF Data Corporation
Tempe, Arizona, U.S.A.

1. ABSTRACT

Digital transmission of high quality compressed video over satellite offers significant opportunities for in-
creasing capacity. For broadcasters, it is the media of choice for increasing the quality and quantity of
program material distributed to their affiliates, and for cable headend delivery it boosts the number of
channels delivered. The implications for increased revenue and reduced cost has spurred interest and
activity throughout the industry. However, even greater returns are achieved when 8PSK modulation is
selected over QPSK for transmission over satellite. The right blend of the elements in the satellite link
make this possible, including:
• Digital Video Compression

• Satellite Transponder Power and Bandwidth

• Receive Station G/T

• High Speed Digital Modem

• Advanced Modulation And Coding Schemes

The satellite transponder is a key resource offering limited power and limited bandwidth. So, any scheme
devised must fit within those bounds. Higher order modulation schemes, like 8PSK and 16QAM, deal with
bandwidth by reducing spectral occupancy compared to lower order modulation methods, like QPSK.
Power is conserved by using concatenated Reed-Solomon coding in tandem with trellis coding. Still,
tradeoffs are required to balance the available power and bandwidth in the transponder with the right
combination of modulation and coding to maximize throughput, and these are explored. Several system
level calculations are presented to enable a transmission link designer to estimate performance versus
the limitations
Other compression schemes for digital video
2. DISTRIBUTION MODEL
include the DVC compression standard deliver-
ing 4:1:1 video at 30 to 36 Mbit/s and the ETSI
The basic system model for a digital video distri-
component video standard producing 4:2:2 qual-
bution over satellite is shown in Figure 1. Video
ity at 34.368 Mbit/s (E3) and 44.736 Mbit/s
compression takes place in the video encoder
(DS3). Advanced Television (ATV) is standardiz-
which reduces the data rate of the source mate-
ing on a format that produces 4:2:2 video at
rial from a typical uncompressed rate of 270
19.39 Mbit/s. Other applications combine sev-
Mbit/s to a lower rate dictated by the level of
eral lower data rate streams in a multiplexer to
video quality, the amount of editing and other
produce a single composite multi-channel data
considerations. These rates range from 2 to 6
stream for transmission as a single carrier.
Mbit/s per individual stream for direct to home
systems and up to 15 Mbit/s for special events
and services using MPEG2 4:2:0 compression.
The video encoder passes the compressed data rier to the input of the demodulator. The antenna
stream to the modulator. This is the initial step in size, ambient noise and the LNB set an impor-
transmission of the digital data stream. The digi- tant figure of merit for the downlink known as the
tal modulator transforms the signal in two ways G/T ratio which is the ratio of the gain of the re-
that are crucial to capacity: modulation and for- ceive antenna to the noise temperature of the
ward error correction (FEC). These two items receive system. The gain of the antenna is pro-
have a major impact on the bandwidth of the portional to its size, so a larger antenna pro-
transmitted signal and the power required for duces a higher G/T ratio and this translates into
transmission. Later, these are examined from a less power required from the satellite trans-
system designers perspective because the deci- ponder.
sions made regarding these ultimately drive the
business case, so careful evaluation of the ap- The demodulator completes the transmission
plication and goals is imperative. link and transforms the carrier back into a digital
data stream. The combination of the modulator
After modulation the signal passes through an and demodulator determine a major part of the
up converter and a high power amplifier (HPA). link performance so any scheme used here that
The up converter translates the signal to the reduces the drain on satellite’s power and
operating frequency of the satellite while the bandwidth resources is worth considering. Fi-
amplifier boosts the signals power level for nally, the data is delivered to the video decoder
transmission. There is a cost tradeoff between where it is transformed back to a video signal in
the HPA and the uplink antenna that determines either a analog or a digital format and then on to
the right balance of size and performance of its ultimate destination.
these two elements.
2.1 LINK PARAMETERS - POWER
A transponder in the satellite receives the trans-
mitted digital video signal, amplifies it and sends One of the tools used to estimate the power re-
it back to earth. The illumination pattern that the quirements for satellite transmission is the link
satellite projects on the ground is an important budget. This is introduced only to show the key
factor in the transmission of the digital signal, items, not to perform a detail analysis which can
and it is specified as the effective isotropic include many other factors that influence per-
radiated power or EIRP. The other satellite formance. As a first cut, the downlink analysis
parameter is its bandwidth which may range for a system is:
from 24 to 36, 54, 72 MHz or some other band-
width. + EIRP dBW Usable Power From
Satellite
At the downlink the receive antenna captures
the video carrier and passes it to an low noise
block converter (LNB) which translates the car-

VIDEO
LNB DEMODULATOR
DECODER

SATELLITE
DOWNLINK

VIDEO UP VIDEO
MODULATOR HPA LNB DEMODULATOR
ENCODER CONVERTER DECODER

DOWNLINK
VIDEO
DEMODULATOR LNB
DECODER

UPLINK VIDEO
LNB DEMODULATOR
DECODER

DOWNLINK

FIGURE 1. DIGITAL VIDEO DISTRIBUTION MODEL

 - DL_loss dB Downlink Pathloss, DL_Loss In one sense this is fixed once the
Satellite To Rx earth station location and operating
Earth Terminal frequency are established.

Fade dB Additional Loss Due Fade Rain fade at Ku and higher fre-
To Rain Or Other quencies can have quite an impact
Impairment in rainier locations. The amount of
rain fade allocated affects the
+ G/T dB/oK System Gain To availability of service. Making the
Temperature Ratio allowance to large reduces the al-
lowed throughput. Making it to
-k dBW/K-Hz Boltzman’s Con- small permits more frequent out-
stant ages.

+ C/No dBW-Hz Carrier To Noise G/T Increasing this ultimately increases

Ratio C/No and increases the maximum
data rate through the satellite.
- Eb/No_req dB Eb/No required to
achieve BER k Boltzman’s constant, it is fixed.

-Margin dB Desired Margin C/No This the total carrier power to the
system noise in a 1 Hz bandwidth.
DR_dB dB Data Rate From It is the sum of the terms above the
Encoder Into Modu- line.
lator in dB
A satellite with a higher available EIRP has ad-
There is a similar analysis performed for the up- vantages. Every dB higher here increases the
link plus additional calculations for other effects, C/No and throughput. A parameter under direct
but this is a good starting point for getting an user control is the G/T which is increased by
picture performance and limitations. The idea using a larger antenna or reducing the noise
here is not to perfect the link analysis but to ex- temperature. Usually, some improvement in G/T
amine the items that affect the link performance. is possible by reducing the noise temperature
This will aid later in making tradeoffs. with a different LNB but the greatest change is
achieved when the size of the antenna is in-
The top half of the link calculation results in the creased. For the lower half of the link budget:
C/No, a basic quantity of the link. And the bot-
tom half subtracts out two terms to give the data Eb/No The signal to noise ratio measure (en-
rate, in dB, that the link supports. Some com- ergy per bit in 1 Hz noise bandwidth)
ment on each of these terms gives an idea of used to compare the performance of
what they are and some possible changes to modems and the link. The BER or bit
achieve different types of performance: error rate is the number of errors di-
vided by the total number of bits in a
EIRP The more power available from the test. The lower the Eb/No is for a given
satellite transponder the higher BER, the less power is needed to ob-
data rate is supported. If two trans- tain system performance.
ponders are rated with the same
saturated EIRP or EIRPsat, the Margin A system parameter chosen to provide
one with a solid state power a level of safety or to maintain the sys-
amplifier usually provides more tem above the operating point.
available EIRP than the one with a
tube type amplifier. DR_dB This is the remaining quantity once the
Eb/No and Margin are subtracted from
the C/No. Taking the anti-log gives the
data rate.
Eb/No and the Margin are subtracted from the Because of their influence on power, the modu-
C/No to obtain the data rate in dB (DR_dB). No- lation and coding in the modem become crucial.
tice that DR_dB is really a measure of capacity, Today’s modems employ sophisticated coding
and the higher it is the greater capacity or data schemes to correct errors and reduce the
rate the link supports and the more program ma- amount of power needed to operate reliably.
terial it can carry. This also reveals the impor- The techniques used to reduce the Eb/No
tance of Eb/No in the link. The lower it is, for the requirement consist of two coding schemes
same BER, the greater possible capacity operating in tandem, known as concatenated
through a satellite. coding. The first type is Reed-Solomon coding
which is particularly effective at correcting bursts
One way of gauging the maximum possible ca- of errors. Reed-Solomon coding is also used in
pacity through the satellite link is shown in Fig- CD ROM systems where errors tend to be
ure 2. Here the size of the receive antenna is bursty.
allowed to vary permitting the G/T to change. The second type of coding is a convolutional or
The available satellite power and an operating trellis type coding using the Viterbi algorithm.
Eb/No are selected and the other parameters in This error correction scheme is most effective
the link calculation are held constant. The cases when the error environment is Gaussian (white
shown are fairly typical for C-Band and Ku-Band noise) rather than bursty. The satellite channel
satellites. An extra case is presented for Ku- is modeled as a Gaussian channel so the Viterbi
Band to illustrate effects of an 8 dB rain fade decoder is used to correct errors produced by
because this frequency band is much more sus- the demodulator. When the Viterbi decoder itself
ceptible than C-Band to the effects of rain. Some makes an error it tends to do so in bursts so the
caution is warranted using this plot because it Reed-Solomon decoder is placed after it to
assumes a particular set of conditions, so, it clean up the burst errors, a very effective com-
does not apply to every situation. In addition, bination.
this figure does not indicate when the bandwidth
of the transmitted signal may limit the maximum Coding is usually expressed as a fraction (1/2)
data rate through the satellite link. indicating the number of bits into the encoder

Maximum Data Rate (vs) Antenna Diameter

700
Ku-Band
600

500
Ku-Band, 50 dBW EIRP
400 C-Band, 36 dBW EIRP
Eb/No = 10 dB
300

C-Band
200
Data Rate (Mbit/s)
100 Ku, 8 dB Fade

0
1 2 3 4 5 6 7 8 9 10
Antenna Diameter (Meters)

FIGURE 2. MAXIMUM COMPOSITE DATA RATE (VS)

ANTENNA DIAMETER
verses the number of coded bits exiting the tremendous improvement all these coded sys-
coder. Common code rates for Viterbi coding tems have over uncoded systems.
are 1/2, 2/3, 3/4, 7/8 an others. Reed-Solomon
codes are based upon appending error correc- 2.2 BANDWIDTH
tion code words to the end of a block of data
and have code rates like 188/204. The principle If power was the only consideration, then it is
is the same. only necessary to select the modulation and
coding technique that produces the minimum
Besides the coding there is also the modulation Eb/No at the performance point. But, bandwidth
to consider. Among the common modulation is also a key resource, so, tradeoffs affecting it
techniques employed are BPSK, QPSK, 8PSK are also considered.
and 16QAM. In QPSK modulation the incoming
bit stream is mapped to one of four phase The occupied or radiated 3 dB bandwidth for this
states, while 8PSK maps them to one of eight class of digital modulation is symbol rate (SR).
phase states. 16QAM modulation maps the in- This quantity is easily calculated and is based
coming stream in to one of 16 states, but in this upon both the order of the modulation and the
case the states consist of both phase and ampli- code rate. BPSK, QPSK, 8PSK and 16QAM
tude. One characteristic of these is that as the modulation types all produce 2m output states,
modulation is altered from QPSK to 8PSK to where ‘m’ is the number of bits mapped into a
16QAM each one requires progressively more symbol. A larger value of ‘m,’ corresponds to a
power (Eb/No) to produce the same BER per- higher order of the modulation, and the band-
formance. Although this appears as a drawback, width of the transmitted signal is reduced by a
there are benefits to the reduced bandwidth that 1
make the tradeoff worthwhile. factor of .
m
Several modulation and coding schemes are
compared in Figure 3. This shows the Eb/No Higher order modulation schemes are often re-
required to produce a BER of 10-10 for each ferred to as bandwidth efficient. For several
modulation and coding combination. This illus- common forms of modulation:
trates the effect that higher order 8PSK and Number
16QAM systems require progressively more Modulation Of States Relative
power than QPSK. What is not evident is the Type m (2m) Bandwidth

11.0
Eb/No (vs) Modulation & Coding Type
10.0 (BER = 10 -10 )
9.0

8.0

7.0
Eb/No (dB)
6.0

5.0

4.0
QPSK QPSK QPSK 8PSK 8PSK 16QAM 16QAM
1/2 V + 3/4 V + 7/8 V + 2/3 V + 5/6 V + 3/4 V + 7/8 V+
188/204 188/204 188/204 188/204 188/204 188/204 188/204
RS RS RS RS RS RS RS

FIGURE 3. Eb/No (VS) MODULATION & CODING TYPE At BER = 10-10

 BPSK 1 2 100% CRv = Code Rate, Viterbi
QPSK 2 4 50 CRr = Code Rate, Reed-Solomon
8PSK 3 8 33
The effect on bandwidth is more dramatically
16PSK, 4 16 25
illustrated in Figure 4, which shows the relative
16QAM
bandwidth for each modulation and code type.
Previously, the beneficial effects of error correc- The figure is plotted for the same data rate, in all
tion coding were examined. There ability to re- cases, with the modulation and coding combina-
duce the required power from the satellite is not tions normalized to the QPSK R1/2 case which
without some penalty. There is extra information is set to 100%. Observe that a QPSK R3/4 car-
added to the original data stream that aids the rier occupies approximately 65% of the band-
error correction process, and this increases the width required by a QPSK R1/2, while for the
radiated bandwidth of the video carrier. To same data rate the required bandwidth shrinks
quantify this, the Viterbi code rate is referred to to a mere 28% with 16QAM R7/8. For this illus-
as CRv while the Reed-Solomon code rate is tration the Reed-Solomon code rate is held con-
designated CRr. The impact coding has on stant.
bandwidth, then, is a proportional to
It is important to reiterate that the improvement
1 in bandwidth efficiency for the higher order
. modulation schemes is offset by an increase in
CRv × CRr power to achieve a given level of performance.
Conversely, the improvement in power efficiency
There is now enough information to calculate afforded by forward error coding is at the ex-
the symbol rate: pense of bandwidth. Within this context system
technical tradeoffs are made to meet the needs
DR of the business.
SR = ,
m × CRv × CRr
2.3 SYSTEM TRADEOFFS
Where:
Two applications are reviewed briefly to show
SR = Symbol Rate how different goals drive system choices in con-
DR = Data Rate trasting directions. The first system is distribu-
m= modulation factor tion of broadcast programming to network affili-

Bandwidth For Various Modulation And Coding Types

16QAM 7/8
16QAM 3/4
8PSK 5/6
8PSK 2/3
QPSK 7/8
QPSK 3/4
QPSK 1/2
LISTED IN ORDER
QPSK 1/2 = 100% TOP TO BOTTOM

-110 -100 -90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90 100 110
Relative Bandwidth (%) - For Same Data Rate

FIGURE 4. RELATIVE BANDWIDTH (VS) MODULATION & CODING TYPE

ates, and the second is direct to home television home is 9 degrees instead of the usual 2 de-
distribution. First, a comparison of the goals: grees. To avoid interference from the satellites
already in orbit at 2 degrees, the frequency plan
Network Direct of the direct to home satellites is also somewhat
Item Broadcast To Home higher than the existing satellites.
Rx Antenna 4.5 to 7.3 me- 0.5 meter
ter By contrast, the broadcast distribution model
Material 4:2:2 4:2:0 with its substantially higher G/T supports a
Contribution Distribution higher data rate but only if the bandwidth of the
Quality Very High Good signal does not overfill the transponder. If the
System/ Professional Consumer same QSPK modulation selected for the direct
Customer to home application is attempted then the maxi-
Quantity 100s > 10 million mum data rate is only 37 Mbit/s through the 36
Satellite Band C-Band Ku-Band MHz transponder. This is roughly half the 66
(DBS) Mbit/s throughput possible by taking advantage
Availability Very High Moderate of the larger antenna and the narrower band-
Program Edit Pre and Post Insert width consumed by 8PSK. The reasons for mak-
Eb/No + 11.7 9.8 ing the system tradeoffs are compelling.
Margin
2.4 TEST RESULT
In both cases maximum capacity is desired, but
the available elements in each system are dif- Although bench testing of modems back to back
ferent. A summary of some key system tradeoffs is necessary as starting point, nothing tests out
is shown below: the proof of concept like testing over a satellite
Network Direct To transponder. A satellite test includes all of the
Parameter Broadcast Home transmission impairments due to bandwidth re-
Rx G/T 24 dB/K 12.5 dB/K strictions and non-linearities in the ground
Satellite 36 MHz 24 MHz equipment and the satellite. It also includes the
Transponder degradation caused by adjacent channel inter-
ference (ACI) and cross-polarization interfer-
Satellite EIRP 35 dBW 50 dBW
ence (CPI), and in the case of multi-carrier op-
Data Rate 66 Mbit/s 24.5 Mbit/s eration the degradation due to intermodulation
Modulation, 8PSK, R5/6 QPSK R2/3 between the carriers.
Coding & RS & RS
Tx Bandwidth 28.6 MHz 20.0 MHz 2.4.1 TWO 34 MBIT/S CARRIERS
(3 dB)
Note: RS = Reed-Solomon, CRr = 188/204 Tests were conducted to verify whether 2 digital
component video signals could be successfully
In direct to home applications a small antenna is transferred over a 36 MHz transponder. In this
needed to meet cost goals and comply with local application the video encoder / decoder were
ordinances. Combing the small size with the based upon ETS 300 174 standard which is ca-
available EIRP (50 dBW) leads to the selection pable of operating at either 34.368 Mbit/s (E3) or
of two systems parameters, the use of Ku-Band 44.736 Mbit/s (T3/DS3). In this application the
frequencies for a smaller antenna and a lower video programming originates from a number of
data rate (24.5 Mbit/s) because of a lower G/T. sites that share a single transponder so combin-
The lower data rate consumes less bandwidth ing the streams into a single composite was not
than a standard 36 MHz transponder, and this a consideration. Figure 5 shows a block dia-
permits the deployment of a narrower 24 MHz gram of the setup for the test. In the test both
bandwidth transponder. Although the capacity carriers were originated from the same earth
per transponder is lower, some capacity is re- station. Additional studio equipment including
claimed because more 24 MHz transponders fit taped source material, digital tape recorders,
within 500 MHz than 36 MHz transponders. In studio monitors and other equipment were on
this application the receive antenna is so small site to evaluate the video quality but these are
that the spacing between satellites for direct to not shown as part of the transmission diagram.
 VIDEO
MODULATOR
ENCODER

Σ UP
CONVERTER
HPA 7.3 METER

VIDEO
MODULATOR
ENCODER
UPLINK

VIDEO
DEMODULATOR
DECODER
4.5 METER
LNB

VIDEO
DEMODULATOR
DECODER
DOWNLINK

FIGURE 5. MULTICARRIER TEST SETUP

provides no improvement in performance.

The tests conducted over the satellite demon-
Without the Reed-Solomon coding stage the
strated a viable link using 8PSK R5/6 modula-
radiated bandwidth is smaller.
tion. Below are some of the link parameters and
results:
3. See text.
Parameter Description
Required EIRP per 32 dBW (C-Band) Carrier spacing is a key issue, and the general
Carrier problem to address is shown in Figure 6. To
Total EIRP For 35 dBW minimize the interference caused by spectral
2 Carriers overlap the carriers are moved apart.
Satellite Backoff From 13 dB
1 dB Compression
(38 dBW) ADJACENT CARRIERS AND TRANSPONDER
Down Link G/T 24 dB/K
Operating 11.8 dB / 1.9 CARRIER
Eb/No/ Margin TRANSPONDER
SPACING

Modulation And Coding 28PSK R5/6,

no Reed-Solomon
Symbol Rate Band- 13.7 MHz
width (3 dB)
Carrier Spacing 316 MHz
Video Quality/ Stan- 4:2:2 / ETS 300 174
dard
Program 3 excellent quality, TST2-5
0 1 2 3 4
Concatenations 5 minimal FREQUENCY
degradation
Notes: FIGURE 6. CARRIER SPACING

1. Separate testing on the satellite simulator

indicated no degradation at a backoff of 1.25 As this continues the edges of the transmitted
dB for 8PSK and 2 dB for 16QAM. spectrum experience distortion as they encroach
on the band edges of the transponder. Combat-
2. Reed-Solomon coding is implemented in the ing transponder edges by moving the carriers
video decoder. 2 levels of Reed-Solomon closer together will eventually return them to the
point where they interfere with one another
again. Empirical testing was undertaken to lo-
cate the optimal spot or the best compromise
between these sources of degradation. The best
spacing between the carriers was found at 17
MHz separation.

Additional bench testing of the same system

with a satellite simulator (flight hardware in a
ground test station) was performed with 16QAM
at R3/4 + Reed-Solomon. Carrier spacing is not
an issue because the symbol rate bandwidth of
the 16QAM carrier (12.4 MHz) is sufficiently nar-
row so that both carriers fit comfortably in the
transponder. Performance of the 16QAM system
was very good and margin, with this setup, was
virtually identical to the 8PSK case.

2.4.2 TWO 44.736 MBIT/S CARRIERS

Testing was performed over satellite with two

44.736 Mbit/s (T3) carriers using 16QAM and
R7/8 + Reed-Solomon coding. The test results
were quite good, however, the receive antenna
diameter was a very large 13 meter device. Cal-
culations indicate a 7.3 Meter antenna will sup-
port 16QAM operation when the power per car-
rier is 27 dBW.

Table 1 indicates the range of data rates possi-

ble for compressed video schemes. These
range from lower range 2 to 6 Mbit/s per individ-
ual stream for direct to home systems up to 45
Mbit/s and beyond for digital component com-
pression. Other applications combine a several
lower data rate streams in a multiplexer to pro-
duce a single composite multi-channel data

TABLE 1. POSSIBLE DATA RATES

Video Data Rate
Type Quality (Mbit/s) Application
MPEG2, ML at MP 4:2:0 2 to 6 Direct To Home
MPEG2, ML at MP 4:2:0 6 to 15 Satellite News Gathering (SNG)
(ETS 300 421) Special Events
DVC 4:1:1 30, 36 SNG, Field Editing
ETS 300 174 4:2:2 34.368 (E3) , Digital Component
44.736 (T3)
ATV 4:2:2 19.39 High Definition TV
stream.