โครสร้าง เอา1

25EBU Technical Review Winter 1993van de Laar, Philips & Olde Dubbelink

General–purpose and application–specificdesign of a DAB channel decoder

F. van de Laar

N. Philips

R. Olde Dubbelink

(Philips Consumer Electronics)

In 1988, the first publicdemonstrations of a new DigitalAudio Broadcasting system weregiven in Geneva. Although thisshowed the feasibility of digitalcompression and modulation fordigital radio, a lot of workremained to be done in the fieldsof standardization, frequencyallocation, promotion andhardware cost and size reduction.This article describes thedevelopment of current DABreceivers, with special emphasison the design of the digital signalprocessor–based channeldecoder which has been used inthe 3rd–generation Eurekareceivers, as well as a prototypedecoder based on an application–specific chip–set.

1. Introduction

Since the introduction of the Compact Disc byPhilips in 1982, people have become more andmore used to the superior quality and user–friend-liness of digital audio. There is every reason, there-fore, to arrange for sound broadcasting to benefitalso from this trend. Digital Audio Broadcasting(DAB) provides the digital communication linkneeded to transfer audio signals (already in digitalform as they leave modern studios), together withadditional services, to the home and mobile receiv-er without loss of quality. To give an impression ofthe performance of the DAB system, Table 1compares DAB with the Digital Satellite Radio(DSR) system [1] which was introduced in Europein 1990, and conventional VHF/FM radio ex-tended with the Radio Data System (RDS). Al-though a DAB receiver is significantly more–com-plex than a conventional VHF/FM receiver, thereare a large number of advantages: CD audio quali-ty, even in a mobile environment [2], operationalcomfort and numerous possibilities for additionalservices such as multi–language programmes,paging, continuous weather and traffic informa-tion and audio–visual advertising.

Original language: EnglishManuscript received 17/1/94

The DAB logo has been registeredby a member of the

Eureka 147 – DAB consortium.

26 EBU Technical Review Winter 1993van de Laar, Philips & Olde Dubbelink

ParametersSystems

ParametersVHF/FM + RDS DSR DAB

Transmission

Reception

RF frequency range

RF sensitivity

Intermediate frequency

Intermediate frequency bandwidth

Modulation method

Multipath resistant

Noise protected

Number of progs per ensemble

Audio bandwidth

Audio signal–to–noise ratio (note 3)

Audio delay

Audio processing

Data processing

Data capacity

Error protection

Decoder dissipation (note 4)

Decoder complexity

Year of introduction

Terrestrial

Home / Mobile

88 – 108 MHz

� 40 dB�V (note 2)

10.7 MHz

300 kHz

FM + DPSK

No

No

1

15 kHz

60 dB

< 1 ms

Analogue

Digital

1.2 kbit/s

CRC

� 200 mW

Low

1987

Satellite

Home

12 GHz

� 48 dB�V

118 / 40 MHz

14 MHz

Diff. QPSK

No

Yes

16

15 kHz

95 dB

4 ms

Digital

Digital

11 kbit/s

Block code

� 500 mW

Medium

1990

Satellite & Terrestrial

Home / Mobile

47 to 230 MHz (note 1)

� 24 dB�V

(36 MHz)

1.536 MHz

OFDM

Yes (allows SFN)

Yes

4 to 16

20 kHz

> 100 dB

< 500 ms

Digital

Digital

� 32 kbit/s

Convolutional code

� 2500 mW

High

1996

Notes: 1 UHF band and 1.5 GHz band possible with DAB Modes II and III2 FM stereo at 60 dB S/N ratio with high–quality receiver3 1 kHz sine–wave reference signal4 Dissipation with 5 V supply, one service selected

Because of world–wide competition, it was impor-tant to set a standard in good time, and to be on themarket within a few years. In Europe, a DAB stan-dard has now been set within Eureka project 147[3], in which most of the European consumerelectronics companies and many European broad-casting organizations participated. During thefinalisation of the standard within Eureka, projectAE–14 was initiated within JESSI (Joint EuropeanSubmicron Silicon Programme), with the aim ofmaking a DAB receiver chip–set with high–leveldesign tools.

Probably within three years from now, these devel-opments will lead to the market introduction ofDAB and a car–radio size receiver will be avail-able as an add–on unit for existing car audio equip-

ment. In a later stage, it may be integrated with anFM radio and a Digital Compact Cassette (DCC)player.

2. DAB transmission

2.1. DAB transmitter

A simplified transmiter block diagram is shown inFig. 1. MPEG layer 2 audio coding [4] is used toreduce the linear pulse–code modulation (PCM)audio bit–rate by a factor in the range from 4 to 12,corresponding to audio qualities ranging from CDto normal FM radio. Convolutional coding is ap-plied to add redundancy for error correction at thereceiver. For audio applications, unequal error–protection is applied; this means that the code–rate

m n

Audiocoder

Channelcoder

11 DABensemblemultiplexer

Interleaving andOFDM

modulation

D/A converter&

IF mixerAudioinputs

Configuration

IF / RFoutput

Synchro symbol Localoscillator

Table 1Comparison of maincharacteristics ofVHF/FM + RDS, DSRand DAB soundbroadcasting systems.

Figure 1Simplified DABtransmitter blockdiagram, showing theorder of coding,multiplexing andmodulation.


This is an exampleof text containingerror bursts both in columns and in lines caused by channel selectivefading. After de– interleaving in time and frequencythese errors have been randomized asindicated. After error correction by a Viterbi de– coder the originaltext is recovered.

This is a F FxampFeof text c FnFaini Fgerror bur Ft F Tot F in column F FndTi FTlines cau FeF by F TTannel s Fl Fcti TFfading. A FTFr de F inter Teav FnF in F time an T Fr Fquen Fythe Te err Fr FThav F been rand FmFzed Fsindi TaTedF Ffter F error cor FeFtion F by a Vit TFbF de– F coder the FoFigin Flte Tt is r FcFvere F.

.rirfc FFFrgeltndaoes yr FFFAic cei c ernn FFFdueVet hisn dr FFF it ner TTTTTTTFFFTTTTTTTTT near FFFdtsa onre eief FFFriixrc uil ndhq FFFrtb lodml vnh FFFc sama otyrtoez FFFeec nimtfgeianv FFFnateeea i rniit FFFah.grseat o–eya FFFseuiott b emoi FFFrdni nr b n enn FFFvbl ient rpoix FFFudarbedee

After demodulation After de–interleaving After error correction

Figure 2Illustration of the effect

of de–interleaving inassociation with error

correction in thedecoder.

varies step–wise within a frame [5]. This is appli-cable in the case of audio signal transmissions be-cause the significance of the bits within a frame isdifferent, in contrast to the situation for data trans-mission applications where each bit is equally im-portant.

The encoded bit–stream is then interleaved in timeas well as in frequency [6]. Time interleaving isused to spread out, over a long period of time, therapid channel variations that occur in mobile re-ception. Even if the baseband signal drops outcompletely for a brief period, enough informationmay remain in each frame to permit reconstructionof the original source sequence. This process isillustrated in Fig. 2, which shows a text in whichthe lines correspond to time frames and the col-umns to frequency bands. Errors caused by fre-quency–selective fading are indicated by F anderrors due to time fading are indicated by T. Afterde–interleaving in the receiver, the missing char-acters are dispersed and error correction is possiblebecause of the redundancy within the text; thiswould not have been possible without interleavingbecause several consecutive characters wouldhave been missing.

The transmission equipment includes a multiplex-er which builds a DAB “ensemble” from severalaudio and data streams, having different bit–ratesand code–rates. Information about the ensemble isinserted into the data–stream for service selection(demultiplexing) in the receiver.

For modulation, the orthogonal frequency–division multiplex (OFDM) technique is used[7, 8]. Instead of using one wide–band carrier withthe data changing at a high rate, multiple narrow–

band carriers are used, on each of which the datachanges at a much lower rate. This significantly re-duces the amount of inter–symbol interference(ISI) resulting from a fading mobile channel [9].

Interleaving in the frequency domain is obtainedby re–arranging the carrier order within a symbol,according to a fixed scheme. This will cause thefrequency–selective fading that occurs in recep-tion to be spread out across the full transmissionbandwidth.

After digital–to–analogue conversion, the data aremodulated in quadrature onto a radio–frequencysignal. This means that the transmitted DAB signalis the sum of two RF signals of the same frequencyand having a mutual phase shift of 90°. These sig-nals are modulated with an in–phase (I) and aquadrature (Q) component of the baseband signal.

2.2. Baseband signal

The complex baseband signal is sub–divided intoframes, which contain several services. Everyframe consists of a fixed number of symbols, eachone composed of a large number of orthogonal car-riers with an equal frequency spacing. The band-width is 1536 kHz, containing for example 384carriers with 4 kHz spacing (Fig. 3). Every symbolis preceded by a guard interval, in which part of thetime signal is cyclicly repeated. As long as the timedifference between the first and last reflections oc-curing in the mobile channel does not exceed theduration of this interval, no ISI will occur and nochannel equalization is needed. The receptionquality in an urban or hilly area where there aremany reflections will be substantially improvedcompared with other demodulation schemes.


1.2

1.0

0.8

0.6

0.4

0.2

0.0

–0.2R

elat

ive

ampl

itude

Frequency

0 4 8 12 16 20 kHz

Figure 3Carrier spectrum showingorthogonality: each carrier hasa peak position where allothers have zero–crossings.

24

Three operational modes have been defined fordifferent DAB broadcast options. Mode I, whichhas a long guard interval for the absorption ofmulti–path reflections, is specially intended forsingle–frequency networks (SFN), in which sever-al transmitters may share the same band for thesame programmes. Modes II and III have a widercarrier spacing to allow higher transmission fre-quencies to be used without the risk of losing carri-er orthogonality at the receiver. An overview of themodes is given in Table 2. An example of the framestructure for Mode II is given in Fig. 4. The syn-chro channel (two symbols) is reserved for receiv-er synchronization. It is followed by the fast in-formation channel (FIC – three symbols) whichcontains service information such as the multiplexconfiguration, service labels, paging codes andtraffic message control. The data field consists ofthe remaining symbols and contains data for a vari-able number of service components in a time mul-tiplex. To simplify service component selectionand mode conversion, the data field is sub–divided, in a mode–independent manner, into 864capacity units (CUs) of 64 gross bits each.

Sync.Data

field 71

FastInformation

Block 2

Datafield 72

FastInformation

Block 3

FastInformation

Block 1

Datafield 1

Datafield 2

Time, Date, Types, Labels,Multiplex configuration,Traffic Message Control

CRC

Guardinterval

Symbol384 carriers � 2 bits

12 CU � 62 bits

Sync.

24 ms DAB baseband frame (Mode II)

Information channel: 32 kbit/s Data field: 864 CU � 64 bits: 2304 kbit/s

62.5 �s 250 �s256 net bits

Figure 4Frame structure of baseband signal in Mode II.

Table 2DAB system parameters for Modes I, II and III.


3. Basic functional modulesof a DAB receiver.

A functional block diagram of a DAB receiver isgiven in Fig. 5. The channel decoder which han-dles the digital demodulation and decoding of thebaseband signal is indicated within the shadedarea. A description of each block is given in thissection.

3.1. Front–end

The front–end of a DAB receiver can be derivedfrom a television tuner for bands I and III, corre-sponding to a frequency range of 50 to 250 MHz.For satellite reception in Modes II or III a 1.5–GHzdown–converter is also needed.

In order to maintain carrier orthogonality, it is re-quired that the frequency difference between thetransmitter and receiver shall be less than 2.5% ofthe carrier spacing. For Mode I (SFN) this corre-sponds to 0.1 ppm at a transmission frequency of250 MHz. To guarantee adequate selectivity, the IFoutput signal is passed through a surface acousticwave (SAW) filter before being applied to thequadrature demodulator.

3.2. Quadrature demodulator

The quadrature demodulator can be realised eitherin digital form (A/D convertor at the input) or inanalogue form (A/D convertor at the output). Aquadrature oscillator is used to regenerate thebaseband I and Q components, with 768 kHz band-width each, from the analogue IF signal. This IFoscillator is adjusted by an automatic frequencycontrol (AFC) signal, which is generated by thesynchronization processor.

The SAW filter, together with low–pass filters be-fore the A/D converter(s), provide sideband sup-pression of about 70 dB. The envelope of the out-

put signal is used for automatic gain control (AGC)and detection of the null symbol which indicatesthe start of a frame.

3.3. Synchronization processor

After detection of the null symbol, the fast Fouriertransform (FFT) of the reference symbol is ac-quired by the synchronization processor. It is usedto calculate the frequency offset of the incomingsignal, and delivers an AFC value for the IF oscil-lator. Frequency offsets as large as ±15 carriers canbe detected and adjusted.

Next, the FFT of the reference symbol is correlatedwith the original reference symbol. Applying aninverse FFT to the result gives the impulse re-sponse of the channel.

From analysis of this impulse response, controlsignals are derived which serve to synchronize thereceiver clock generator (voltage–controlled crys-tal oscillator – VCXO) to the transmitter and to ad-just the symbol selection window to avoid interfer-ence (removal of the guard interval).

3.4. Demodulation processor

In the receiver, the orthogonal carriers, onto whichthe data is differentially modulated by quadraturephase–shift keying (QPSK), are derived from thetime signal on a symbol basis by means of an FFT.After this FFT the phase difference with respect tothe previous symbol is determined for each carrier(Fig. 6). Quasi–stationary deviations that occurwithin the channel as a result of displacement ef-fects are eliminated this way. The reference sym-bol determines the initial carrier phase value foreach frame.

For each carrier, the phase difference is repre-sented as a complex number with two components,called metrics. Each metric separately correspondsto a gross bit, associated with a 3–bit amplitude–dependent reliability indication.

Audiooutput

RF IF

AFC PAD

Audiodecoder

D/Aconverter

Quadraturedemodulator

De–interleaving

FFT &differential

demodulation

Selection&

Decoding

FADIC SIVIC

DSP �C

SAA2500

DRAM

Systemcontrol

Channeldecoder

PCMdata

Sync.processor

CIR User interface

Front–end

AGC

Figure 5Functional block

diagram of a DABreceiver.

ASIC partitioning for thechannel coder is also

shown (FADIC, SIVIC,etc.).


01

11

00

10

Guard Symbol x Guard Symbol (x+1)

Time

1

0

–1

� ��/4

The phase difference between successive symbols determines the “di–bit” combination.In the example shown, Cx+1 – Cx = (3�/4) – � = (–�/4) , corresponding to a “10” bit combination.

Figure 6Differential demodulationof one carrier.

Cx Cx+1

3.5. Decoder

Error bursts caused by fluctuations in the mobilechannel are pseudo–randomly distributed by de–interleaving in the receiver. Since this operationsis complementary to the interleaving in the trans-mitter, the output sequence is equal to the input se-quence, but with a delay of 15 frames. About2 kbits of memory is needed for the de–interleav-ing of one capacity unit (CU).

The de–interleaving process removes undesiredmetric correlation at the input of a Viterbi decoderwhich serves for error correction [9]. The Viterbidecoder uses redundancy added at the encoder toreconstruct the originally transmitted sequencewith maximum likelihood. Soft–decision logic isused, which means that it takes into account thereliability indication available from each metric.

The amount of redundancy added at the encodermay be different for each programme or each ap-plication, and it should be matched to the channelcharacteristics. This is obtained by puncturing thecoded bit–stream according to a known scheme.For cable distribution a significantly smalleramount of redundancy is needed compared to thatfor wireless communication in a hilly terrain. Fordata applications one can choose a code rate R (ra-tio between net bits and gross bits) of 1/4, 3/8, 1/2or 3/4.

3.6. Audio decoder

The audio decoder takes the compressed sourcesignal as its input and performs an inverse quantiz-ing operation on 32 equally–spaced audio sub-bands, to obtain a PCM digital audio signal with

full 20–bit accuracy. The operation is based uponthe psychoacoustic properties of the human ear,which masks weak frequencies delivered simulta-neously with loud frequencies. Depending on therequired audio quality, the input bit–stream is fixedfrom 32 kbit/s (speech, mono) to 384 kbit/s (studioquality, stereo). A programme–associated datastream (PAD) may be included in the audio bit–stream; this could contain, for example the Interac-tive Text Transmission System (ITTS) giving in-formation such as artists’ names, titles or lyrics.

3.7. System control

The receiver should first select a DAB ensembleand then a service within it. A service may containany combination of audio and data components.Information is provided to select a service by eithernumber, name or type. Since the amount of serviceinformation within DAB is almost unlimited, theuser will have to make a choice between the facili-ties offered. A display will present ensemble andservice information [10].

Service selection by the user will be processed bya micro–controller, which first interrogates the fastinformation channel (FIC) to find out where the se-lected service components are allocated in the datafield. A cyclic redundancy check is performed tocheck the validity of the FIC data. If it is valid, theinformation is transferred to the decoder which ex-tracts and processes the service components. ThePAD data will be passed back to the micro–con-troller for display. To reduce the demands on pro-cessing capability and power dissipation, the re-ceiver will not process service components whichare not selected.


EPROM32k � 8

RAM64k � 16

Switches &LEDs

Viterbidecoder

EPROM32k � 8

RAM64k � 16

EPROM32k � 8

RAM64k � 16

8–bitparallel I/O

PCinterface

AnalogueI/O

Dividers

EPROM32k � 8

RAM64k � 16

DAB3DSP56156

DSP56156

DSP56156

DSP56156

VCXO12.288 kHz

AFCTuner

Baseband

Host

Audiodecoder

PC

Codec

Test

Interleaving& error prot.

Selection &Sync.

FFT &Diff. demod.

FFT &Diff. demod.

Figure 7Hardware block

diagram of theDSP–based channel

decoder used in Eureka3rd generation DAB

equipment.

4. DAB receiver prototypes

4.1. DSP–based prototypeimplementation

The development of a prototype channel encoder/decoder took place in parallel with standardizationwithin EUREKA–147. Although the basic func-tionality was known and verified at the start of thishardware development, the DAB system parame-ters were still changing and the overall systemcomplexity was gradually increasing. Therefore achannel decoder was designed which was basedupon a general–purpose multi–processor architec-ture; this approach provided maximum flexibility,required minimum glue logic and uses a single pro-gramming language. Four digital signal processors(DSP) were used (DSP56156), each one havingtwo identical serial interfaces allowing commu-nication in either a serial, parallel or ring configu-ration. For the implementation of the audio decod-er one DSP56001 was used, this having the 24–bitprecision required for audio channels [11].

A block diagram of the channel decoder DSPboard is depicted in Fig. 7 and the printed–circuitboard (PCB) is shown in Fig.8. The functionalityof each major system block, based upon the pro-cessor capability, is roughly indicated in this dia-gram. Since the DSPs are too slow to do a real–time FFT on a symbol basis, the symbolsassociated with the selected service component arestored and processed on a double–frame basis. Aseparate memory–mapped Viterbi decoder is

needed since the DSPs do not have the perfor-mance needed for this function.

All DAB–specific interfaces as well as the DSPaddress decoders and clock dividers are imple-mented with simple gate–array logic (GAL) de-vices (type 16V8). This allowed the correction ofmistakes and adaption to changing interface

Figure 8DSP–based Philips

channel decoder boardused in Eureka 3rd

generation equipment.


CharacteristicsIntegrated circuit

CharacteristicsFADIC SIVIC SAA2500

Function

Algorithm

DAB modes

Design level

Technology

Size (mm2)

Package

On–chip memory

Off–chip memory

Data capacity

Processing unit

Control interface

Dissipation

Demodulation

FFT (2084 points)

Mode I

Behaviour

CMOS 1.0 micron

>100

CLCC–68

DRAM 160 k

None

3 Mbyte/s

Symbol

16–bit DSP

500 mW

Channel decoding

Viterbi (64 states)

Modes I, II, III

Register transfer

CMOS 0.8 micron

<50

QFP–64

SRAM 14 k

DRAM 256 k � 4

384 kbyte/s

Capacity unit (CU)

L3/�C

300 mW

Audio decoding

MPEG layer II

n/a

Register transfer

CMOS 0.8 micron

<50

QFP–44

SRAM

None

384 kbyte/s

Frame

L3/�C

100 mW

formats without PCB modifications. The inter-faces were designed in such a way that their direc-tion could be reversed; by re–programming theDSPs and the GALs, it was possible to realise apartial channel encoder, which allows simpleback–to–back verification and realisation of a testtransmitter.

The decoder board is able to process the synchro-nization channel and one service component (4

symbols in Mode I, 16 in Mode II and 32 inMode III). The maximum output bit–rate is256 kbit/s.

The general–purpose approach proved to offerenough flexibility for the implementation of a sys-tem which was not yet fixed. The major design ef-fort was expended on software rather than on hard-ware, especially for the implementation of Mode Iwith its distinct frame structure which took several

Frank van de Laar received his degree in computer engineering at Breda, the Netherlands, in 1984.He joined the data transmission group of the Philips Research Laboratory in Eindhoven, where heworked on adaptive filters for acoustic echo cancellation.

Since 1990 he has been working at the Philips Consumer Electronics Advanced DevelopmentCentre on Digital Audio Broadcasting. He is involved in the design of DAB channel decoders, ASICspecifications and standardization of coding and modulation..

Norbert Philips is senior project manager at the Philips In-ternational Technology Centre in Leuven (ITCL), Belgium.He received his master of science degree in electrical engi-neering at the Kathilieke Universiteit Leuven in 1976, afterwhich he stayed at the university as research assistant in thefield of switched–mode power supplies.

In 1983 he joined N.V. Philips Industrial Activities where hewas involved in the development of analogue ICs for audioapplications. Since 1989 he has been working on DAB; heis involved in standardization issues and the realisation ofDAB receivers, with special focus on control algorithms andtheir implementation.

Rob Olde Dubbelink studied electrical engineering at the University of Twente in Enscede, in the Neth-erlands. He obtained his degree in 1990 and in 1991 joined the Advanced Development Centre of Phil-ips Consumer Electronics. He is involved in simulation programs for channel decoder ASICs, the test-ing of DAB receivers and the delivery of DAB test systems.

Table 3Main characteristics ofthe ASICs developedby Philips for DAB.


months of effort. Although the algorithms are quitecomplex, their implementation was relatively sim-ple. More attention had to be given to proper inter-rupt handling, to ensure that data was not lost frommemory–mapped I/O and serial interfaces. Thesecomplications are typical in real implementationand do not occur in simulation programs. It wasdiscovered that the non–orthogonality of the DSPinstruction set limited the effectiveness of both thedesign and execution time of the DSP algorithms.Reduced instruction sets and automatic contextswitching would have been more effective in real–time applications like these.

4.2. Design of an ASIC set

The DSP–based prototypes were built as a meansof gaining a complete understanding of the system,for verification of the standard and to prove its vi-ability by field tests. For an economic implementa-tion in a car, however, a significant reduction insize, price and power consumption is needed.Therefore Philips Consumer Electronics has de-veloped application–specific integrated circuits(ASIC) for the channel decoder and the audio de-coder within the JESSI AE14 project. The parti-tioning of this chipset corresponds to the function-al partitioning shown in Fig. 5. A summary of themain ASIC characteristics is given in Table 3. OneDSP is reserved for synchronization tasks and dis-play of the channel impulse response (CIR); thisallows further optimization of the algorithms be-fore integration.

With the DSP solution it was necessary to use thecomplete 24 ms frame duration for demodulationof one service component. With the ASIC solutionthe choice was made to do real–time processing ona symbol basis [12]. This offers the following ad-vantages:

– Processing delay and buffer sizes are reducedfrom one frame to one symbol, at the expense ofa more complex arithmetic unit.

– More regularity and thus simplified control.

– Demodulation of a complete ensemble is pos-sible. Power dissipation will be reduced if onlya sub–set of the symbols are selected for de-modulation.

For the decoder part it is desirable to be able to pro-cess at least three service components simulta-neously, for example FIC data (32 kbit/s), audio(�256 kbit/s) and other data (�64 kbit/s).

The total net data rate for the three channels corre-sponds to a requirement for a decoding capability

of at least 352 kbit/s. Considering a 12.288–MHzsystem clock and a 64–state Viterbi decoder, anoutput data–rate of 384 kbit/s can be achieved bythe sequential processing of two combined statesin each clock cycle. For the time de–interleavingof one audio service plus a data service compo-nent, and frame buffering for all components,about 256k�4 of RAM is needed. An externalDRAM is used for this, with the controller inte-grated on the decoder. By carefully taking into ac-count the frame structure, no refresh cycles areneeded.

All ASICs operate on a single 5V supply with a12.288 MHz system clock. A boundary scan test(BST) is implemented for the testing of printed cir-cuit board assembly. Recently a prototype receiverhas been constructed based on these ICs. First re-sults show that this receiver is working correctlyfor stationary reception in Mode I. The next devel-opment step is the design of a demodulation ASIC,which will support all DAB transmission modes.Fig. 9 gives an impression of how a next genera-tion of receiver might look.

4.3. Evaluation

A baseband simulation program was written be-fore the hardware implementations took place.The program was used for building up systemknowledge, for studying the effects of coding andinterleaving on different channel characteristicsand as a reference for implementation. The resultsof simulations in a Gaussian and a Rayleigh chan-nel are given in Fig. 10, which shows the bit–errorrate (BER) of a decoded sequence as a function ofthe signal–to–noise ratio Eb/No where Eb is the en-ergy per received bit and No the variance of thenoise. It clearly shows the effect of using interleav-ing in a Rayleigh channel.

Figure 9Impression of a DAB

test and measurementreceiver of the next

generation.


The curve for the Gaussian channel was comparedwith results of measurements on the DSP decoder,with a signal applied at the IF level and without us-ing AFC. The carrier–to–noise ratio (C/N) wasmeasured with a spectrum analyser and convertedto a Eb/No value. The same IF oscillator was usedfor the transmitter and receiver. No significant dif-ferences were found between the simulations andthe measurement results on the DSP decoder.

5. Conclusions

Prototype DAB receivers based upon DSPs andfunctioning according to the Eureka–147 standardare available and are being used world–wide forfield tests, for measurement and for verification ofthe standard. They show the robustness of the Eu-reka–147 DAB system.

Prototype ASICs for channel decoding and audiodecoding are also available now, and these will al-low the construction of small size and reasonably–priced car–radio DAB receivers for the marketintroduction of DAB.

DSP/GAL design is very effective for makingprototype decoders while the standardization pro-cess is under way. Although the design effort need-ed for an ASIC set is significant greater than for aDSP solution, it is impossible to make a feasibleconsumer product without it.

The chip size is mainly determined by the Mode Iimplementation, which took a lot of design effortfor both DSP and ASIC approaches. From a re-

ceiver point of view, Modes II and III are more at-tractive for low–cost and small–size IC imple-mentation.

Additional receiver design efforts should concen-trate on price reduction of IF oscillators, the imple-mentation of useful data services, the reduction ofelectromagnetic radiation to allow portable receiv-ers to be offered, the use of a 3V power supply andcombination with FM radio.

For the successful introduction of DAB it is alsoimportant that broadcasters and receiver manufac-turers concur on the choice of same initial features,chosen from the numerous possibilities offered bythe standard.

Acknowledgement

The authors wish to thank A. Delaruelle and J.Huisken of the Philips Research Laboratory fordiscussions on partitioning and design of the first–generation channel decoder ASICs.

Bibliography

[1] Digital Satellite Radio (DSR). Soundbroadcasting via satellite. Specificationfor the transmission methodTechnische Richtlinie ARD/ZDF No. 3 R1,Ausgabe 3, November 1989.

[2] Dosch, Ch., Ratliff, P.A., Pommier, D.: FirstPublic Demonstrations of COFDM/MASC-AM. A milestone for the future of radiobroadcastingEBU Technical Review No. 232 (December1988).

[3] Proceedings of First International Sympo-sium on Digital Audio Broadcasting, Mon-treux, 8–9 June 1992EBU, Geneva.

[4] Stoll G.: Source coding for DAB and theevaluation of its performance: A major ap-plication of the new ISO audio coding stan-dard.Proceedings of First International Symposiumon Digital Audio Broadcasting, Montreux, 8–9June 1992, EBU, Geneva.

[5] Plenge, G.: DAB – A new sound broadcast-ing system. Status of the development –routes to its introductionEBU Technical Review No. 246 (April 1991).

[6] Pommier, D., Wu, Y.: Interleaving or spec-trum–spreading in digital radio intendedfor vehiclesEBU Technical Review No. 217 (June 1986).

cUrban,ode rate 1/2,

no int.

1 2 3 4 5 6 7 8 9 10 11 120

100

10–1

10–2

10–3

10–4

10–5

Bit–

erro

r ra

tio (

BE

R)

Eb/No

dB

Urban,code rate 1/2,

with int.Urban,code rate 1/3,

with int.

code rate 1/2,Gaussian

Urban,no coding

Figure 10Simulated bit–errorrate as a function ofthe signal–to–noiseratio Eb/No.


[7] Le Floch, B., Halbert–Lasalle, R., Castelain,D.: Digital Sound Broadcasting to mobilereceiversIEEE transactions on Consumer Electronics,Vol. 35, No. 3, August 1989.

[8] Alard, M., Lassalle, R.: Principles of modu-lation and channel coding for digitalbroadcasting for mobile receiversEBU Review – Technical, No. 224 (August1987).

[9] Proakis, J.G.: Digital Communications (2ndedition)McGraw Hill, 1989.

[10] EBU Technical Statement D74–1992: EBUexpectations for first–generation consum-er DAB receivers

[11] Dehery, Y–F.: Real–time software proces-sing approach for digital sound broad-castingAdvanced digital techniques for UHF satellitesound broadcasting – Collected papers onconcepts for sound broadcasting into the 21stcenturyEBU, Geneva, 1988.

[12] Delaruelle, A. et al.: A channel demodulatorIC for Digital Audio BroadcastingAccepted for CICC–94, San Diego.

List of VHF/UHF television stations

The 38th annual edition of the EBU List of VHF/UHF television stations was published in January 1994.The document gives details of 42,650 transmitters in Bands I, III and IV/V, in service in the EuropeanBroadcasting Area on 1 September 1993.

For each country, the transmitters are listed in order of frequency; the list includes an alphabetical indexof all station names.

The 575–page List is typeset directly from the EBU computer files. The following information is given foreach transmitter: organization and programme service, transmission channel, carrier offset, geographicalcoordinates, height above sea–level, real and effective antenna heights, ERP in the main beam(s) andpolarization. An indication is included to show whether each transmitter is a main transmitter or a rebroad-cast transmitter. Data for transmitters operated by EBU Members are taken from official sources.

Orders for the EBU List of VHF/UHF television stations should be sent to EBU Publications, using theorder form on page 41. Price category “D” (70 Swiss francs, including surface mail).

The EBU also publishes lists of LF/MF and VHF/FM broadcasting stations, as follows:

List of LF/MF broadcasting stations (situation in May).

LF/MF List no. 46 covers the period May 1993 to May 1994 (available now)LF/MF List no. 47 covers the period May 1994 to May 1995 (available autumn 1994)

Price category: “D”Please state clearly on your order form the number of the List you require (LF/MF–46 or LF/MF–47).

List of VHF sound broadcasting stations in the European Broadcasting Area (situation in January).

VHF/FM List no. 38 covers the period January 1993 to January 1994 (available now)VHF/FM List no. 39 covers the period January 1994 to January 1995 (available Spring 1994)

Price category “D”Please state clearly on your order form the number of the List you require (VHF/FM–38 or VHF/FM–39).

Documents

โครสร้าง เอา1