transmission of tc-8psk digital television signals over ... · transmission of tc-8psk digital...

30 EBU Technical Review Autumn 1996Morello and Visintin

Transmission of TC-8PSK digital televisionsignals over Eurovision satellite linksSystem evaluation and optimizationA. Morello (RAI)

M. Visintin (RAI)

1. Introduction

Over the past few years, the EBU TechnicalDepartment has been studying matters relating tothe digitalization of the Eurovision network [1] [2][3]. These studies resulted from the developmentof the ETSI Standard, ETS 300 174 [4], whichspecifies a digital television codec for contributionlinks at 34.368 Mbit/s (and at 44.736 Mbit/s).

The ETSI digital television codec is directly com-patible with Plesiochronous Digital Hierarchy(PDH) terrestrial networks and with the Intermedi-ate Data Rate (IDR) specification (Intelsat Stan-dard IESS-308 [5]) for satellite modems, based onQPSK modulation and rate 3/4 convolutional cod-ing. Proprietary scaled-down versions of this sys-tem – at 2 x 17 Mbit/s and at 8.448 Mbit/s – areavailable on the market for applications requiringlower picture quality, such as Digital SatelliteNews Gathering (D-SNG).

The Eurovision Network makes use of Ku-bandsatellite transponders with a nominal bandwidth of72 MHz. This is suitable for transmitting two digi-tal signals at 34 Mbit/s in a frequency division

The RAI has carried out a study tooptimize the use of 34 Mbit/s digitaltelevision contribution links within theEurovision satellite network. The aimof the study – which was based oncomputer simulations – was toassess the feasibility of increasingfrom two to three the number of FDMsignals allocated to a 72-MHzsatellite transponder – by using thespectrum-efficient TC-8PSK methodof modulation.

In this article, the Authors report onthe results of the RAI computersimulations.

multiplex (FDM) if QPSK rate 3/4 modulation ata symbol-rate of about 23 Mbaud is adopted. TwoEBU Project Groups, N/DIG (Network Digitaliza-tion) and N/SAT (Utilization and Optimization ofSatellite Capacity), are now evaluating the techni-cal feasibility of increasing from two to three thenumber of digital television signals that can be car-ried within a transponder. In order to achieve this,the QPSK rate 3/4 modulation system is replacedby a more spectrum-efficient method of modula-tion known as Trellis Coded 8PSK (TC-8PSK).

The RAI has carried out a computer simulationstudy to analyse the performance of TC-8PSK mod-

Original language: EnglishManuscript received 2/8/96.

31EBU Technical Review Autumn 1996Morello and Visintin

ulation in a Frequency Division Multiple Access(FDMA) configuration of three signals per 72-MHztransponder. In order to maximize the attenuationmargin on the up-link and on the down-link, thefollowing characteristics were optimized:

– the forward error-correction (FEC) rate;

– the nominal operating point and the “gain-setting” of the on-board high-power travelling-wave tube amplifier (TWTA);

– the frequency spacing of the three signals;

– the modem filter roll-off.

The results of the RAI computer simulation werethen compared with the results obtained by theEBU during its satellite tests on TC-8PSK modu-lation.

2. Model of the TC-8PSKsystem

The digital audio/video signal that was analysed(ETS 300 174 [4]) has a useful bit-rate of34.368 Mbit/s at the output of the encoder. Thevideo components are protected by a Reed-Solo-mon (R-S) code, RS (255,239), which allows theETSI codec to produce good picture quality in thepresence of an input bit-error rate (BER) of 10–4

(after Viterbi decoding). No additional R-S codeis inserted in the modem.

In the computer simulations, the audio/video digi-tal source was replaced by a pseudo-randombinary source (PRBS).

The simulated TC-8PSK system was based onPragmatic Trellis Coding [6] rates 2/3 or 5/6. Inboth cases, a rate 1/2 convolutional encoder wasused, with constraint length 7 (64 trellis states) andgenerator polynomials g1 = 133, g2 = 171 (octalnotation).

Pragmatic Trellis Coding is achieved by associat-ing two coded bits (b0 and b1) with one uncoded bit(b2) to generate an 8PSK symbol (see Fig. 1).Each uncoded bit (b2) identifies two antipodal8PSK constellation points, so that the lack of codeprotection is balanced by the large Euclidean dis-tance in the signal space.

For rate 2/3 Trellis Coding, a group of bits (consist-ing of one bit at the input of the convolutional rate1/2 encoder, plus one uncoded bit) generates threetransmission bits which are modulated into one8PSK symbol. The modulator symbol rate1 (Rs) isequal to half the useful bit-rate (Ru), i.e.:

Rs = 1/2 x 34.368 = 17.184 Mbaud.

For rate 5/6 Trellis Coding, a rate 3/4 puncturedencoder replaces the rate 1/2 encoder of Fig. 1.This time, a group (consisting of three bits at the in-put of the convolutional encoder plus two uncodedbits) generates six transmission bits which aremodulated into two 8PSK symbols. The modula-tor symbol rate (Rs) is given by:

Rs = 2/5 x Ru = 13.747 Mbaud.

In the RAI computer simulations, the TC-8PSKmodems made use of 3-bit “soft-decision” Viterbidecoding, and of raised cosine filtering, equallysplit between the transmitter and receiver, with aroll-off of 35 % unless otherwise specified. With

1. The symbol rate corresponds to the signal bandwidth at–3 dB and the demodulator noise bandwidth, BRX.

TC-8PSK 2/3transmitter a

TC-8PSK 2/3transmitter c

IMUX TWTA OMUX Receiver bTC-8PSK 2/3transmitter b

Useful signal

Interference

Interference

Up-linknoise

Down-linknoise

Satellite

Rate 1/2convolutional

encoder

8PSKmapper

b2

b1

b0

Rs = 17.184 MbaudRu = 34.368 Mbit/s

Figure 1TC-8PSK rate 2/3

modulation andcoding scheme.

Figure 2BasicTC-8PSK rate2/3 satellite link that

was computer-simulated by the RAI.


this roll-off, the total bandwidth occupied by 8PSKrate 2/3 modulation is 23.2 MHz; in the case ofTC-8PSK rate 5/6, the total bandwidth is18.6 MHz.

3. Model of the satellite linkand frequency plan

The simulated satellite chain is represented inFig. 2. Three TC-8PSK signals (labelled a, b andc) at a bitrate (Ru) of 34.368 Mbit/s were modu-lated and combined in the form of a frequency divi-sion multiplex (FDM). A frequency spacing (�f)of 24 MHz was used (see Fig. 3) so that each chan-nel could exploit 1/3 of the nominal transponderbandwidth of 72 MHz. To simplify the analysis, insome cases only the performance of the centralsignal (b) was studied, while signals a and c weretreated as interference signals.

Fig. 4 gives the AM/AM and AM/PM character-istics of the TWTA, while Fig. 5 shows the fre-quency responses of the input multiplexer (IMUX)and the output multiplexer (OMUX) which wereadopted in the simulations; these responses were atthe limit of the transponder “mask” specified byEutelsat, taking into account also the thermal insta-bilities. The total usable transponder bandwidth

was of the order of 72 MHz (at the –3 dB points)and the total group delay at the edge of this bandwas around 50 to 60 ns.

When multiple signals are transmitted in an FDMwithin a single transponder, the generated TWTAoutput power is split between the signals accordingto (i) their input level and (ii) the TWTA AM/AMnon-linear characteristic. For the three-signal con-figuration and TC-8PSK 2/3 modulation systemdescribed here, Fig. 6 gives OBOb (the outputback-off of signal b with respect to the transponderoutput saturation power) as a function of IBOb (the

Abbreviations

AWGN Additive white Gaussian noise

BRX Receiver noise bandwidth

BER Bit error rate

C/N Carrier-to-noise ratio

Eb/No Noise-margin loss, i.e. the availablecarrier-to-noise ratio (C/N) in a band-width equal to the useful bit-rate of theup-link (Ru)

EIRP Effective (equivalent) isotropic ra-diated power

FDM Frequency division multiplex

FDMA Frequency division multiple access

FEC Forward error correction

G/T Gain/temperature ratio

IBO Input back-off

IDR Intermediate data rate

IMUX Input multiplexer

ISI Inter-symbol interference

Md Down-link margin

Mu Up-link margin

OBO Output back-off

OMUX Output multiplexer

PDH Plesiochronous digital hierarchy

PRBS Pseudo-random binary source

QPSK Quaternary (quadrature) phase shiftkeying

Rs Modulator symbol rate

Ru Useful bit-rate

R-S Reed-Solomon (code)

SNG Satellite news gathering

TC-8PSK Trellis-coded, 8-phase shift keying

TWTA Travelling-wave tube amplifier

XPD Cross-polar antenna discrimination

Bandwidth = 72 MHz

fa = fb – 24 MHz fc = fb + 24 MHzfb

Figure 3Typical FDMconfiguration in thetransponder ofbandwidth 72 MHz.

Figure 4Simulated TWTAAM/AM and AM/PMcharacteristics.


input back-off of signal b with respect to the trans-ponder input saturation power). Different IBOs(between 10 dB and 14 dB) were considered forsignals a and c. The results, obtained by computersimulations, refer to a configuration where signalsa and c were modulated (using 8PSK) but signal bwas unmodulated2.

In the reference configuration adopted duringthe EBU satellite tests on TC-8PSK rate 2/3,namely IBOa = IBOb = IBOc = 12 dB, the out-put back-off of signal b was about 7.8 dB.

Fig. 7 shows the spectral power density of thethree 8PSK rate 2/3 signals on the up-link, assum-ing a quasi-linear transmitting station.

3.1. Satellite link budget

A typical satellite link budget was evaluated for thethree FDM signals and it will be used in the follow-ing Sections as a reference. It relates to clear-skyconditions and involves three classic formulae asgiven below.

L is the free-space loss of the link, expressed in dB,and Eb/No is the available carrier-to-noise ratio(C/N) in a bandwidth equal to the useful bit-rate ofthe up-link (Ru).

L � 20 log4�D�

� 20 log (f)� 4.15

Eb

No,u,cs� EIRPu � 0.3 � Lu �

G�Ts� Ru � K

Eb

No,d,cs� �EIRPd,satur � OBO� � 0.2�

Ld � G�Trx,cs

� Ru � K

whereD = the satellite distance,38.5 x 106 (m)

� = carrier wavelength (m)f = carrier frequency (Hz)u = up-linkd = down-linkcs = clear skys = satelliteRu = reference noise bandwidth for the

link budget, corresponding to auseful bit-rate of 34.368 Mbit/s

2. The power of signal b was measured at the output of anarrow-band filter (200 kHz) which suppressed theinterference from signals a and c.

Figure 5Simulated IMUX and OMUX amplitude and group-delay

characteristics.

Figure 6OBOb versus IBOb for different values of IBOa,c (simulation results).


K = Boltzmann’s constant,–228.6 dB (W/HzK)

satur = saturated TWTAOBO = the output back-off of the TWTArx = receiver

Table 1 gives the reference link budget using theabove formulae, based on a reference IBO of12 dB per signal, which was also adopted duringthe EBU satellite tests on TC-8PSK rate 2/3.

4. Simulation of a singleTC-8PSK signal on a linearand a satellite channel

The two TC-8PSK systems with rates 2/3 and 5/6were simulated on a linear channel affected byadditive white Gaussian noise (AWGN), and onthe satellite channel model described in Section 3.In the latter case, a single 8PSK signal was located

at the transponder central frequency, with theTWTA in saturation and AWGN on the down-linkonly. Fig. 8 shows, for both systems, BER (afterViterbi decoding) versus Eb/No. On the AWGNchannel, the target BER of 10–4 for good servicequality was achieved at Eb/No = 5.6 dB for rate 2/3and at Eb/No = 7.4 dB for rate 5/6. The noise mar-gin losses introduced by the satellite chain (singlesignal, TWTA at saturation) at a BER of 10–4 werearound 0.8 dB for rate 2/3 and 1 dB for rate 5/6.

These losses were mainly due to the TWTA non-linear distortion, since the available transponderbandwidth was very large compared to the signalsymbol rate.

Therefore the 25 % spectrum efficiency gain ofrate 5/6 coding is associated with a loss of about2 dB in terms of satellite power efficiency in thepresence of noise.

Additional simulations were carried out on a moreconventional modulation scheme, QPSK with rate7/8 punctured convolutional coding, based on thesame mother-code as TC-8PSK. At34.368 Mbit/s, the symbol rate is 19.639 Mbaudand the total bandwidth occupation with a roll-offof 35 % is 26.5 MHz, producing a certain degreeof spectrum overlapping in the FDM configurationof Fig. 3. Comparing QPSK rate 7/8 with

Parameter Value

Noise Bandwidth(BRX = Ru) (MHz) 34.368

Reference EIRP ofthe up-link stations (dBW) 66.3

Up-link frequency (MHz) 14,375

Satellite G/T on theup-link footprint contour (dB(K)) +4.5

Reference satellite IBO persignal at reference EIRPu (dB) 12

Reference up-link Eb/Noin clear sky (dB) 16.4

Satellite saturation EIRP(single signal) on the (dBW)down-link footprint contour

46

Down-link frequency (MHz) 11,075

Reference satellite OBO persignal (see Fig. 6) at (dB)reference IBO

7.8

Reference receivestation G/T (dB(K)) 35

Reference down-link Eb/Noin clear sky (dB) 21.1

Figure 7 (upper)Signal powerspectrum on theup-link (simulationresults).

Table 1Reference link budgetfor an FDM of threesignals pertransponder.

Figure 8 (lower)Performance of asingle TC-8PSKsignal (rates 2/3 and5/6) over an AWGNand a saturatedsatellite channel(simulation results).


TC-8PSK rate 2/3 at a BER of 10–4, the computer-simulations indicated an Eb/No loss of 0.2 dB onthe AWGN channel and of 0.3 dB on the saturatedsatellite channel.

Fig. 9 shows the noise margin loss (with respect tothe performance on the AWGN channel) of thesystems at a BER of 10–4, for different values ofIBO. For large IBOs, the transponder linearity im-proves and, hence, the noise margin loss tends tobecome negligible. It should be noted that Ebrefers to the effective signal energy on the down-link at the operational OBO, and not to the nominalenergy at TWTA saturation point. Therefore theeffects of the TWTA power reductions with in-creasing OBOs are not shown.

The above simulations were based on quasi-idealmodems, with near-perfect demodulation carrierphase, filtering and timing recovery. Therefore asuitable implementation margin should be addedto the results to match them to the hardware perfor-mance. A direct comparison between the simula-tion results and the IF-loop performance of com-mercial TC-8PSK modems indicates typicalimplementation losses of the order of 1 dB at aBER of 10–4.

5. Simulation of threeTC-8PSK signals within asatellite transponder

When three signals are transmitted in an FDM onthe same satellite transponder, mutual interfer-ence (i.e. intermodulation) takes place in theTWTA, which degrades the system performance– see Fig. A1-1 in Appendix 1. The presence of anon-linearity makes a theoretical analysis quitedifficult as the principle of effect superimpositioncannot be applied. Also, it is not possible to takemeasurements on each signal alone, because theTWTA working point is modified when the othertwo signals are removed. Consequently, a com-plete simulation of the transmission chain must becarried out for each channel condition (e.g. theIBOs of the three signals, the noise levels).

In order to allow a quick evaluation of the systemperformance under different conditions, a “simpli-fied analysis” method has been developed by theRAI and is described in Appendix 1. This methodis based on splitting the degradation sources intoadditive independent elements as follows:

– the up-link noise;

– the down-link noise;

– the inter-symbol interference (ISI) produced bythe non-linearity and by IMUX/OMUX filter-ing;

– the down-link interference from the adjacentsignals (TWTA intermodulation) and from thecross-polar signals.

Unless otherwise specified, the results reported inthe following Sections refer to complete “errorcounting” simulations, with the three signals(quasi-ideal modems) transmitted in an FDM inthe same transponder, implementing the satellitechain described in Section 3.

The maximum attenuations which could beapplied to the up-link and the down-link signalsbefore the system was driven out-of-service –which is referred to as the available link margins –were taken as a reference for the system optimiza-tion. To evaluate the up-link margin (Mu), signalsa, b and c were initially set at the nominal IBO, andthen the wanted signal was progressively attenu-ated on the up-link (i.e. IBO = 12, 13, 14, . . .20 dB), while the other two signals were kept at thenominal IBO until the service threshold(BER = 10–4) was reached.

The attenuation on the up-link (Mu) can be inter-preted as the rain fade which drives the system out-of-service. To evaluate the down-link margin(Md), the signals a, b and c were set at the nominalIBO, and then the three signals were progressivelyattenuated on the down-link until the servicethreshold (BER = 10–4) was reached.

The attenuation on the down-link (Md) cannot bedirectly interpreted as a rain fade (since the receiv-er noise temperature increase is not considered),but rather as a reduction in the G/T of the receiving

Figure 9Noise margin loss

�Eb/No versus TWTAIBO (simulation

results, singleTC-8PSK signal).


station which drives the system out-of-service inclear-sky conditions.

6. Choice of modulation andcoding systems

For TC-8PSK rate 2/3, Fig. 10 shows the resultsof simulations in which the central signal b wasprogressively attenuated on the up-link (i.e.IBOb = 12, 13, 14, . . . 20 dB), while the other twosignals (a and c) were maintained at the referencelevel, IBOa,c = 12 dB. The up-link noise powerdensity level was maintained constant, accordingto the reference link budget of Table 1. WhileIBOb was increased, the power of signal b and theavailable Eb/No progressively decreased on boththe up-link and the down-link, according to thefollowing relationships:

Eb

No,u�

Eb

No,u(reference)� �IBOb (dB)

Eb

No,d�

Eb

No,d(reference)� �OBOb (dB)

where (see Fig. 6):

�IBOb = IBOb – IBOb (reference)

�OBOb = OBOb – OBOb (reference)

System FDMsignal

Mu(dB)

Md(dB)

QPSKrate 7/8

ab

5.96.9

13.113.5

TC-8PSKrate 2/3

ab

6.87.5

14.114.0

TC-8PSKrate 5/6

ab

5.05.0

11.811.0

In addition, the power reduction of signal b on thedown-link degrades its carrier-to-interference ratio(C/I) due to increasing intermodulation effectscaused by signals a and c. Fig. 10 shows curves ofBER versus the available Eb/No (nominal) on thedown-link, for different values of IBOb. Eb/No(nominal) is the available down-link Eb/No at thereference IBO of 12 dB, and is proportional to thereceiving station G/T3.

At high IBOb levels, the BER curves tend to flattenout, because the up-link noise and the intermodula-tion effects in the TWTA lead to a non-negligiblebit-error ratio, even in the absence of down-linknoise. Fig. 10 indicates that, for Eb/No (nominal)= Eb/No (reference) = 21.1 dB on the down-link,the service threshold (BER = 10–4) is achievedwhen IBOb = 19.5 dB, corresponding to an up-link margin (Mu = �IBOb) of about 7.5 dB.

Similar simulations have been carried out on theother two systems under study – TC-8PSK rate 5/6and QPSK rate 7/8 – at the same useful bit-rate of34.368 Mbit/s and with the same FDM spacing of24 MHz within the 72 MHz transponder: Table 2compares the computed up-link and down-linkmargins of the three systems at a BER of 10–4, inthe case where IBO (reference) = 12 dB, both forthe central signal b and for the lateral signal a.

The results of the computer simulation comparefavourably with the real satellite transmission testswhich were carried out by the EBU [1] [2]. Duringthe EBU tests, a complete digital television failurefor TC-8PSK rate 2/3 [2] occurred when the up-link EIRP was reduced to about 57.3 dBW; goodpicture quality (estimated BER of the order of 10–4

before R-S) was available at an EIRPu of 58.3 dB,corresponding to an up-link margin of 8 dB. Thisexperimental result can be compared with the7.5 dB up-link margin indicated by the computersimulations (see Table 2, signal b), which do notinclude the modem implementation losses.

Similar results have been obtained with TC-8PSKrate 5/6 when comparing the EBU satellite tests [1]with the RAI simulations.

Taking into account (i) the uncertainties associatedwith the real satellite characteristics and the linkbudget during the tests and (ii) the lack of directBER measurements, it can be concluded that thesimulation results are in good agreement with thesatellite test results. Since the up-link margin ismainly limited by the up-link noise, an under-estimation of the available Eb/No (reference) by

3. When (G/T)RX = G/T (reference) = 35 dB(K), Eb/No(nominal) = Eb/No (reference) = 21.1 dB (see Table 1).

Figure 10TC-8PSK rate 2/3performance for anEb/No,u of 16.4 dB atIBOb = 12 dB.

Table 2Link margins Mu andMd for IBO (reference)= 12 dB and noiselevels as given inTable 1.


about 1 to 1.5 dB in Table 1 could be the origin ofthe discrepancy between the satellite test resultsand the computer simulations.

The system offering the best performance in termsof up-link and down-link margins is TC-8PSKrate 2/3; the next best system, QPSK 7/8, offers1dB inferior performance figures. This degrada-tion is more significant for the lateral signals (a andc) due to the bandwidth limitations of the satelliteIMUX-OMUX filters. Nevertheless in a realtransmission environment, the satellite filter char-acteristics should be less critical than the simulated“mask”, and the lower implementation margin andphase noise sensitivity of QPSK could furtherreduce the performance difference with respect toTC-8PSK 2/3.The TC-8PSK 5/6 system, whencompared with the other two systems, has an in-trinsically lower power efficiency (see Fig. 8) butan improved resistance to adjacent channel inter-ference. However, on balance, it shows inferiorperformance to the other two systems.

On the basis of these results, only the TC-8PSKrate 2/3 system is further analysed in the followingSections.

7. Optimization of the TWTAoperating point and thegain setting

In order to optimize the nominal TWTA operatingpoint in clear-sky conditions – i.e. the IBO (nomi-nal) values of the three FDM up-link signals –fixed up-link or down-link attenuations were in-serted in the simulations. Also, the residual BERsafter Viterbi decoding were measured for differentvalues of IBO (nominal), around the referenceIBO of 12 dB (see Table 1).

The introduced attenuations were 7.5 dB (up-link)and 14 dB (down-link), corresponding to the com-puted up-link and down-link margins for a BER of10–4 at the reference IBO of 12 dB (see Table 2).For TC-8PSK rate 2/3, Fig. 11 shows the residualBER as a function of the nominal IBOs, for aconstant noise power density which accords withthe link budget of Table 1.

The simulation results indicate that the residualBER is quite constant around the optimum trans-ponder operating point, corresponding to an IBOof 11 to 12 dB for the up-link margin and 9 to 11 dBfor the down-link margin. Tables A2-1 and A2-2 inAppendix 2 report similar results obtained underthe same conditions, using the simplified analysismethod.

According to the simulation results relating to thedown-link margin, an IBO (nominal) of 11 dBcould be considered a slightly better choice thanthe reference one which was adopted during theEBU tests (i.e. IBO = 12 dB, OBO = 7.8 dB).Nevertheless, the down-link margin is not a limit-ing factor of the link under consideration and,therefore, the reference operating point can beconsidered a good choice.

The measured up-link margin is significantly low-er than the down-link margin; an overall improve-ment can be expected by reducing the transpondergain setting by 3 dB, and by increasing the EIRPsof the transmit stations from 66.3 dBW to69.3 dBW. This corresponds to increasing the up-link Eb/No by 3 dB, while keeping constant theTWTA operating point, the signal intermodulationlevels and the down-link Eb/No. Under theseconditions, the computer simulations have indi-

Figure 11 (upper)Residual BER versus

nominal IBOa,b,c foran up-link attenuation

of 7.5 dB and adown-link attenuation

of 14 dB.

Figure 12 (lower)Residual BER versusfrequency spacing for

a down-link attenu-tion of 14 dB.


cated an improvement of about 1.7 dB in the up-link margin (Mu), while the down-link margin(Md) remains quite stable. Also with this gain set-ting, the optimum TWTA operating point, IBO(nominal), remains at around 10 to 11 dB.

8. Optimization of the carrierfrequency spacing and thefilter roll-off

In order to optimize the signal frequency spacing(�f), the three FDM TC-8PSK rate 2/3 signalswere simulated assuming the noise power densitylevels of Table 1 with an additional down-linkattenuation of 14 dB. For signals a and b, Fig. 12shows the relationship between the frequencyspacing and the residual BER after Viterbi de-coding.

The BER curve of signal a shows a “flat” mini-mum around �f = 20 to 23 MHz, since a is mainlyaffected by the IMUX-OMUX distortions for largespacings, and by the interference from signal b forsmall spacings. Conversely, signal b is degradedby the interference from the adjacent signals a andc, and therefore the BER slowly increases as thefrequency spacing (�f) is reduced from 26 to20 MHz.

On the basis of these simulation results, the refer-ence choice of �f = 24 MHz – adopted both for theRAI computer simulations and during the EBUtests – is seen to have been a good choice.

The above simulations were repeated after chang-ing the raised-cosine filter roll-off in the modemfrom 35 to 50 %, in order to evaluate the effect oflarger signal spectra on the frequency-spacingoptimization. The results show that, with a modemroll-off of 50 %, signal a suffered from higherdegradations only for frequency spacings beyond24 MHz (due to larger degradations caused by theIMUX-OMUX filters), while the performance ofsignal b remained unaltered. Therefore the trans-mission of three TC-8PSK rate 2/3 signals, with afrequency spacing of 24 MHz, is not sensitive to

the choice of the modem roll-off in the range 35to 50 %.

9. Effects of cross-polarinterference and alternativeFDM configurations

The link margins have been evaluated under thenoise levels given in Table 1, also in the presenceof cross-polar interference on the up- and down-links, assuming two levels of cross-polar discrimi-nation (XPD):

XPDu = XPDd = 30 dB

XPDu = XPDd = 25 dB.

The up-link margin evaluation was based on theassumption that the rain attenuation only affectsthe wanted signal and not the cross-polar interfer-ence, which is transmitted from a different loca-tion. Conversely, the down-link margin evaluationassumed that both the wanted and the interferingsignals would be attenuated. For simplicity, theXPD variations due to fading were not taken intoaccount.

The results obtained by the simplified analysismethod described in Appendix 1 are given in TableA2.2 of Appendix 2. For a reference IBO of 12 dBand an XPD of 30 dB, the presence of the cross-polar interference produced negligible reductionsin the up-link and down-link margins. For an XPDof 25 dB, the degradation was 0.9 dB on the up-link and 0.4 dB on the down-link. The optimumnominal TWTA operating point remained at anIBO of around 10 to 11 dB.

Finally the up-link and down-link margins wereevaluated for an FDM configuration in which a34-Mbit/s TC-8PSK rate 2/3 signal was replacedby four QPSK signals comprising two Euroradiosignals at 2 Mbit/s (coding rate 1/2) and two SNGsignals at 8 Mbit/s (coding rate 3/4). Fig. 13shows the frequency position of each signal insidethe multiplex, while Table 3 summarizes the main

System Bit-rate(Mbit/s)

�f(MHz)

IBO(dB)

OBO(dB)

Eb/No,u(dB)

Eb/No,d(dB)

QPSK 1/2 2.048 –34.5 21.0 18.7 19.6 22.4

QPSK 1/2 2.048 –31.5 21.0 17.8 19.6 23.3

QPSK 3/4 8.448 –25.5 18.0 14.0 16.5 21.0

QPSK 3/4 8.448 –16.5 18.0 13.9 16.5 21.1

8PSK 2/3 34.368 0.0 12.0 7.5 16.4 21.4

8PSK 2/3 34.368 24.0 12.0 7.6 16.4 21.3

Table 3FDM systemparameters with fourQPSK signalsreplacing one 8PSKsignal.


system and transmission parameters: modulationbit-rate, frequency position relative to the centre ofthe 72-MHz transponder channel, IBO, OBO, Eb/Eo,u and Eb/Eo,d. The Eb/Eo figures include the at-tenuation due to the IMUX and OMUX filters.

The up-link and down-link margins, evaluated bycomplete computer simulations at a BER of 10–4

without modem implementation losses, are givenin Table 4. Comparing the above results with thoseof Table 2 (relevant to three 8PSK rate 2/3 signals),it can be deduced that the four QPSK signals do notdegrade the performance of the remaining two8PSK signals. In addition, the two QPSK rate 3/4signals have similar margins as the 8PSK signals,while the QPSK rate 1/2 signals have significantlylarger margins.

10. Conclusions

The results of computer simulations carried out atthe RAI Research Centre, within the framework ofthe optimization of the Eurovision network, are ingood agreement with the results of laboratory tests(modem IF-loop tests) and with the results of EBUsatellite test transmissions [1][2]. The differencebetween the computer simulation and the satellitetest results (1 to 1.5 dB, depending on the trans-mission configuration) mainly arises from the un-certainties that were evident on the link budgetduring the satellite tests and on the characteristics(TWTA, IMUX and OMUX) of the real satellitechain.

Three channel coding and modulation systemshave been analysed:

– pragmatic TC-8PSK rate 2/3;

– pragmatic TC-8PSK rate 5/6;

– QPSK with convolutional coding rate 7/8.

The simulations indicate that, in the FDM configu-ration of three signals per 72 MHz transponder, theTC-8PSK rate 2/3 modulation offers the best up-link and down-link margins, followed within 1 dBby QPSK rate 7/8. The optimizations of theTC-8PSK 2/3 signal levels in the TWTA (nominalIBOs under clear-sky conditions) and of the fre-quency spacing (�f) of the signals have confirmedthat the correct choices were made during the EBUtests, namely:

IBO (nominal) = 12 dB�f = 24 MHz

For this frequency spacing, the system perfor-mance is not sensitive to modem roll-off factors inthe range 35 to 50 %.

Additional simulations have been carried out witha different FDM configuration, where one of the8PSK rate 2/3 signals was replaced by two Euro-radio signals (QPSK rate 1/2 at 2 Mbit/s) and twoSNG signals (QPSK rate 3/4 at 8 Mbit/s). Thesesimulations have demonstrated the feasibility ofthe 8PSK rate 2/3 transmission mode.

On the basis of this analysis, an increase from twoto three in the number of digital television signalscarried within each 72-MHz Eurovision trans-ponder seems feasible from a technical point ofview.

The simplified analysis method described in Ap-pendix 1 can be used to estimate the system perfor-mance under different operational conditions,without the need to perform complete simulationsfor each individual configuration.

System �f(MHz)

Mu(dB)

Md(dB)

QPSK 1/2 –34.5 10.4 18.3

QPSK 1/2 –31.5 11.1 19.3

QPSK 3/4 –25.5 7.6 15.6

QPSK 3/4 –16.5 6.9 15.4

8PSK 2/3 0.0 7.4 14.4

8PSK 2/3 24.0 7.1 14.2

Bibliography

[1] N/SAT 014, N/DIG 007: Report of 34 Mbit/s– 8PSK tests on the Eurovision Network17–18.07.95EBU Technical Department, July 1995.

Bandwidth = 72 MHz

–34.5 +24.0 MHz0.0–25.5–16.5–31.5

8PSK 8PSK4 x QPSK

Figure 13FDM configuration

with four QPSKsignals replacingone of the 8PSK

signals in the satellitetransponder.

Table 4Up-link and down-

link margins (withoutmodem implemen-

tation loss).


[2] N/SAT 033, N/DIG 025: Report 8PSK TestSession – 29.01.96 – 31.01.96EBU Technical Department, March 1996

[3] N/SAT 030, N/DIG 022: Investigation of thePerformance of 8 �-PSK Signals over a72 MHz EUTELSAT transponderEBU Technical Department, January 96.

[4] ETS 300 174: Network Aspects (NA): Digi-tal coding of component television signalsfor contribution quality applications in therange 34 – 45 Mbit/sETSI, November 1992.

[5] IESS-308 : Performance characteristics forIntermediate Data rate (IDR) digital carri-ers: Standards A, B, C, E and F earth sta-tionsIntelsat, August 1994.

[6] Viterbi, Wolf, Zehavi, Padovani: A pragmaticapproach to trellis coded modulationIEEE Communication Magazine, July 1989.

Appendix 1

Simplified analysis method

A simplified analysis method has been developedby the RAI in order to allow a first estimation of thesystem performance under different operatingconditions (e.g. up-link EIRP, nominal TWTA in-put back-off, noise power density levels), withoutthe need to perform complete computer simula-tions. The analysis method is focused on the cen-tral signal b of the three-carriers-per-transponderconfiguration which uses TC-8PSK 2/3 modula-tion and a carrier spacing of 24 MHz.

In the simplified analysis method, the followingsources of signal degradation are considered:

g) Gaussian noise

Neglecting the noise compression on the satelliteTWTA, the assumption is made that up-link anddown-link noise of equal power have the sameeffect on the system BER, and their power levelscan be added.

h) Interference (intermodulation) from adjacentsignals in the FDM

Since the channel spacing of 24 MHz is larger thanthe total signal bandwidth, including the roll-off(e.g. 17.184 x 1.35 = 23.198 MHz in the case of8PSK rate 2/3 modulation), the mutual interfer-ence between the three 8PSK signals on a quasi-linear up-link is negligible.

On the down-link, the equivalence and additivityof noise and interference is assumed. In otherwords, an interfering signal of power IBRX (in thereceiver bandwidth) is assumed to produce thesame BER as a Gaussian noise of equal power (i.e.NBRX = IBRX).

This approximation neglects the fact that the inter-modulation amplitude is not Gaussian and that it iscorrelated with the signal itself. It should be notedthat the noise and the interference contributionsmust be evaluated and added after the demodulatorreceiving filter (with noise bandwidth BRX equalto the symbol rate).

A practical problem is how to measure the interfer-ence power (IBRX) in the demodulator filter with-out removing the useful signal and modifying theTWTA working point.

In the RAI computer simulations, while the inter-fering signals a and c were modulated, the wantedsignal b was un-modulated (see Fig. A1-1). The

Figure A1-1 (upper)Measurement of theinterfering power IBRXin the demodulatorfilter.

Figure A1-2 (lower)C/Ib versus IBOb fordifferent values ofIBOa,c (simulationresults).


ÁÁÁÁÁÁ

StepÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

What to evaluateÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Formulae, Constants & Figures to useÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Evaluation of the required C/N+I in BRX

ÁÁÁÁÁÁ

1ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Required Eb/No on AWGN at BER = 10–4 ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

From Fig. 8 (AWGN curve)

ÁÁÁÁÁÁ


Modem implementation loss ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

� 1 dB

ÁÁÁÁÁÁ


IBOtotÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Add TWTA input powersÁÁÁÁÁÁ4ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁISI noise margin loss on TWTA at IBOtot

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁFrom Fig. 9ÁÁÁ

ÁÁÁÁÁÁ

5

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Required Eb/No by satellite at IBOtot

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Combine results from Steps 1, 2, 4

ÁÁÁÁÁÁ


Required C/N+I by satellite in BRX for 8PSK 2/3 ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

C/N+I = Eb/No + 10 Log (Ru/BRX) = Eb/No + 3 dB

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Evaluation of the available C/N+I in BRXÁÁÁÁÁÁ7ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁOBOb versus IBOa, IBOb, IBOc

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁfrom Fig. 6ÁÁÁ

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

8

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Available C/Nu and C/Nd in BRX, at IBOb and OBOb

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

From link budget (Table 1)

Eb/No,u = Eb/No,u (reference) – �IBO

Eb/No,d = Eb/No,d (reference) – �OBO

Eb/No,u (reference) = 16.4 dB

Eb/No,d (reference) = 21.1 dB

IBO (reference) = 12 dB

OBO (reference) = 7.8 dB

C/N= Eb/No,tot + 3 dB

ÁÁÁÁÁÁ


C/Id (b) (intermodulation) in BRX at IBOa,b,c ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

From Fig. A1.2

ÁÁÁÁÁÁ


Available C/(Nu+Nd+Iu+Id) ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

Combine results from Steps 8 and 9

Note: If C/N+I (available) > C/N+I (required), the service quality is guaranteed.

ÁÁÁÁÁÁÁÁÁÁÁÁ

IBOa,b,cnominal


XPD = �ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

XPD = 30 dBÁÁÁÁÁÁÁÁÁÁÁÁ

XPD = 25 dB

ÁÁÁÁÁÁÁÁ

nom inalÁÁÁÁ

MuÁÁÁÁÁÁ

MdÁÁÁÁÁÁ

MuÁÁÁÁÁÁ

Md ÁÁÁÁ

MuÁÁÁÁÁÁ

Md

ÁÁÁÁÁÁÁÁ

5 ÁÁÁÁ

6.0ÁÁÁÁÁÁ

13.6ÁÁÁÁÁÁ

5.5ÁÁÁÁÁÁ

13.5ÁÁÁÁ

4.8ÁÁÁÁÁÁ

13.1ÁÁÁÁÁÁÁÁÁÁÁÁ

6ÁÁÁÁÁÁ

6.7ÁÁÁÁÁÁÁÁÁ



13.8ÁÁÁÁÁÁ


13.5

ÁÁÁÁÁÁÁÁ

7 ÁÁÁÁ

7.2ÁÁÁÁÁÁ

14.4ÁÁÁÁÁÁ

6.9ÁÁÁÁÁÁ

14.3ÁÁÁÁ

6.2ÁÁÁÁÁÁ

14.0

ÁÁÁÁÁÁÁÁ

8 ÁÁÁÁ

7.8ÁÁÁÁÁÁ

14.5ÁÁÁÁÁÁ

7.4ÁÁÁÁÁÁ

14.4ÁÁÁÁ

6.7ÁÁÁÁÁÁ

14.1

ÁÁÁÁÁÁÁÁ

9ÁÁÁÁ

8.2ÁÁÁÁÁÁ

14.6ÁÁÁÁÁÁ

7.8ÁÁÁÁÁÁ

14.5ÁÁÁÁ

7.1ÁÁÁÁÁÁ


10ÁÁÁÁÁÁ




14.5ÁÁÁÁÁÁ


14.2

ÁÁÁÁÁÁÁÁ

11 ÁÁÁÁ

8.2ÁÁÁÁÁÁ

14.4ÁÁÁÁÁÁ

7.9ÁÁÁÁÁÁ

14.3ÁÁÁÁ

7.3ÁÁÁÁÁÁ

14.1

ÁÁÁÁÁÁÁÁ

12 ÁÁÁÁ

7.9ÁÁÁÁÁÁ

14.0ÁÁÁÁÁÁ

7.7ÁÁÁÁÁÁ

13.9ÁÁÁÁ

7.1ÁÁÁÁÁÁ


13ÁÁÁÁÁÁ




13.5ÁÁÁÁÁÁ


13.3

ÁÁÁÁÁÁÁÁ

14 ÁÁÁÁ

6.9ÁÁÁÁÁÁ

13.0ÁÁÁÁÁÁ

6.7ÁÁÁÁÁÁ

12.9ÁÁÁÁ

6.3ÁÁÁÁÁÁ

12.7


IBOa,b,cnominal


XPD = �ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ

XPD = 30 dBÁÁÁÁÁÁÁÁÁÁÁÁ

XPD = 25 dB

ÁÁÁÁÁÁÁÁ

nom inalÁÁÁÁ

MuÁÁÁÁÁÁ

MdÁÁÁÁÁÁ

MuÁÁÁÁÁÁ

Md ÁÁÁÁ

MuÁÁÁÁÁÁ

Md

ÁÁÁÁÁÁÁÁ

5 ÁÁÁÁ

3.7ÁÁÁÁÁÁ

11.3ÁÁÁÁÁÁ

3.4ÁÁÁÁÁÁ

11.1 ÁÁÁÁ

2.8ÁÁÁÁÁÁ


6ÁÁÁÁÁÁ




11.8ÁÁÁÁÁÁ


11.2

ÁÁÁÁÁÁÁÁ

7 ÁÁÁÁ

5.5ÁÁÁÁÁÁ

12.6ÁÁÁÁÁÁ

5.2ÁÁÁÁÁÁ

12.4 ÁÁÁÁ

4.4ÁÁÁÁÁÁ

12.0

ÁÁÁÁÁÁÁÁ

8 ÁÁÁÁ

6.3ÁÁÁÁÁÁ

12.9ÁÁÁÁÁÁ

5.9ÁÁÁÁÁÁ

12.7 ÁÁÁÁ

5.2ÁÁÁÁÁÁ

12.3

ÁÁÁÁÁÁÁÁ

9ÁÁÁÁ

6.8ÁÁÁÁÁÁ

13.1ÁÁÁÁÁÁ

6.4ÁÁÁÁÁÁ

12.9ÁÁÁÁ

5.6ÁÁÁÁÁÁ


10ÁÁÁÁÁÁ




13.0ÁÁÁÁÁÁ


12.7

ÁÁÁÁÁÁÁÁ

11 ÁÁÁÁ

7.0ÁÁÁÁÁÁ

13.1ÁÁÁÁÁÁ

6.7ÁÁÁÁÁÁ

12.9 ÁÁÁÁ

6.0ÁÁÁÁÁÁ

12.7

ÁÁÁÁÁÁÁÁ

12 ÁÁÁÁ

6.7ÁÁÁÁÁÁ

12.6ÁÁÁÁÁÁ

6.5ÁÁÁÁÁÁ

12.5 ÁÁÁÁ

5.8ÁÁÁÁÁÁ


13ÁÁÁÁÁÁ




12.2ÁÁÁÁÁÁ


11.9

ÁÁÁÁÁÁÁÁ

14 ÁÁÁÁ

5.8ÁÁÁÁÁÁ

11.6ÁÁÁÁÁÁ

5.6ÁÁÁÁÁÁ

11.5 ÁÁÁÁ

5.1ÁÁÁÁÁÁ

11.2

Table A1-1Simplified analysis

method.

Table A2-1 (left)Up-link and down-

link margins (dB) ofsignal b versus

nominal IBOs(without modem

implementation loss).

Table A2-2 (right)Up-link and down-

link margins (dB) ofsignal b versus

nominal IBOs(including modem

implementation lossof 1 dB).


parameter Cb corresponds to the TWTA saturationpower attenuated by OBOb. Fig. A1-2, obtainedby computer simulations, shows the C/Ib curvesfor the central signal (b) versus IBOb, for differentvalues of IBOa,c. A nominal channel spacing of24 MHz was adopted according to Fig. 3.

i) Inter-symbol interference

The inter-symbol interference (ISI) produced bythe TWTA non-linearity depends on the workingpoint (IBO) that is determined not only by the sig-nal itself but also by the other signals that are multi-plexed in the transponder. For a single carrier, thenoise margin loss with respect to the AWGN chan-nel ranges from 0 dB (high back-offs, quasi-linearTWTA) to about 0.8 dB (saturated TWTA, singlecarrier) (see Fig. 9). In the simplified analysis de-veloped by the RAI, it is assumed that the ISI deg-radation in the FDM configuration is the same asfor the single signal configuration – for the sametotal IBO. IBOtot refers to the power sum of thethree input signals.

j) Cross-polar co-channel digital interference

It is assumed that cross-polar co-channel digitalinterference has similar effects on the system BERas Gaussian noise of equal power in the receivingfilter. When the cross-polar transponder carriesthe same signal configuration as the wanted trans-ponder, the interference power (I) is equal to the

wanted power (C) attenuated by the cross-polarantenna discrimination (XPD): i.e. C/I = XPD.

The simplified analysis method adds all the inter-ference, intermodulation and noise power con-tributions (measured within the receiver filterbandwidth) in order to evaluate the availableC/N+I ratio on the satellite links, and compares itwith the C/N+I ratio required by the modulationsystem to deliver the target BER of 10–4. The ser-vice continuity and quality is assured whenC/N+I (available) > C/N+I (required).

The detailed computation procedure is summa-rized in Table A1-1.

Appendix 2

Link margin evaluation by means ofthe simplified analysis method

The link margins, as defined in Section 5, have beenevaluated with the simplified analysis method,without cross-polar interference (i.e. XPD = �)and with cross-polar co-channel interference of25 dB and 30 dB, both on the up-link and the down-link. The noise power density levels correspond tothe reference link budget of Table 1.

Tables A2-1 and A2-2, respectively, show the re-sults without and with modem implementationlosses.

Dr. Alberto Morello graduated in Electronic Engineering from the Turin Polytechnic in 1982 and took hisdoctorate degree in 1987. He joined the Research Centre of RAI–Radiotelevisione Italiana in 1984 andis now Head of the Digital Communication Laboratory. He is engaged in research on digital modulationand coding techniques for television, audio and data transmission and broadcasting, via terrestrial andsatellite channels.

Dr. Morello is a member of several international EBU and ITU-R groups and has participated in variousEUREKA and RACE projects. He was Chairman of EBU Sub-group V4/MOD which defined the technicalspecification for the DVB-S system. He is the author of various technical and scientific articles and haspresented numerous papers, relevant to his studies and researches, at national and international events.

Mr. Michele Visintin graduated in Electronic Engineering from the Turin Polytechnic in 1987. He joinedthe Research Centre of RAI–Radiotelevisione Italiana in 1988 where he is involved in research on digitalmodulation and coding techniques for television broadcasting via terrestrial and satellite channels.

transmission of tc-8psk digital television signals over ... · transmission of tc-8psk digital...

Documents