sign in sign up
<<
Home , Digital broadcasting, Digital radio, Radio Data System

International Standards: Voice Coding and Amateur Applications

Here’s a digital-voice standard for broadcast and Amateur Radio—read all about it!

By Cédric Demeure and Pierre-André Laurent

his paper presents a new digi- existing demonstrator to be tested on for much better quality by offering net tal system for use in terrestrial an across-the-Atlantic link, together bit rates compatible with up-to-date Taudio broadcasting in the fre- with the associated and voice audio compression techniques like quency bands below 30 MHz and its coding techniques. AAC-MPEG 4 (near FM quality). potential use for Amateur Radio. The A new coherent COFDM (coded Such a broadcasting scheme may system is based upon a COFDM mo- orthogonal frequency-division multi- even be implemented using existing dem; it was elaborated as a derivation plexed) modem for use in high- power in a transmission of the DRM modem. DRM (Digital quality audio broadcasting in the mode named Simulcast. This mode con- Radio Mondiale) is a worldwide consor- frequency bands below 30 MHz is pro- sists in transmitting a program in a tium proposing through ITU a new posed by DRM as a candidate to replace given frequency channel by using an standard for broadcasting the current AM that is AM compatible analogue waveform and for frequencies below 30 MHz. It is now notoriously perturbed by multipath, the new simultaneously. the only digital standard adopted by Doppler shift and . The main This broadcasting system was ini- ITU for HF digital . characteristics of this system are to tially elaborated within the NADIB A specially adapted version that fits in- be compatible with existing chan- consortium (European Eureka 1559 side a 3-kHz channel was derived for nelization within these bands and with project: NArrow Band DIgital Broad- Amateur Radio. This paper presents the current technology (5-kHz casting) and was adopted and en- spacing). Therefore, it should be a bet- hanced within the DRM consortium as Thales Communications SA ter alternative than satellite broadcast- a candidate for international normal- 66 Rue du Fossé Blanc ing at higher frequencies because it is ization at ITU. Real transmission on 92231 Gennevilliers Cedex cheaper, offers better propagation and a standard 100-kW, shortwave AM France indoor coverage. It also opens the door transmitter was demonstrated at the

Jan/Feb 2003 49 1997 IBC conference in Amsterdam. designed to fit in with the existing AM • Minimum disturbance of AM us- Since then, numerous field trials have broadcast band plan, based on signals ers on the same band: This leads to the been performed. The latest series with 9 or 10-kHz channelization. It design of a signal which, as seen by an (May/June 2002) involved transmis- has modes requiring as little as 4.5- AM receiver, must be as noise-like as sions from Canada, the Caribbean or 5-kHz bandwidth, plus modes that possible. The frequency spectrum of the and Europe to Australia and New can take advantage of wider band- transmitted signal must also be as com- Zealand. widths, such as 18 or 20 kHz. pact as possible to minimize jamming Thales, with its two affiliates Thales • Better audio quality: The aim is to in the adjacent channels. Communications SA and Thales Broad- obtain near-FM quality within a much • Low complexity: This is essential cast and Multimedia SA, was within the narrower frequency bandwidth. The for having receivers of low complexity initial group of companies to work on improvement upon analogue AM is im- and low power consumption, espe- the standard.1 Going for digital broad- mediately noticeable. DRM can be used cially in countries where batteries are casting allows the introduction of new for a range of audio content, including rare and expensive. services such as Program Associated multi-lingual speech and music. • Graceful degradation: If desired, Data (PAD), RDS type alternative fre- • Simple-to-use receivers, especially graceful degradation may be obtained quency switch (AFS) management, when it comes to HF programming by the use of hierarchical coding. This single frequency network (SFN), simul- (frequent frequency changes due to option, although compatible with the taneous data channels for various ser- propagation conditions) system design, is not described in this vices such as picture transmission, and • Low-cost equipment to quickly paper. so forth. This paper describes work per- reach the mass market DRM benefits for the listeners are formed by Thales Communications • Text messaging similar to RDS at the following: France to adapt this standard and the FM for simple PAD (Program Associ- • FM-like sound quality with the associated modem to 3-kHz Amateur ated Data) transmission AM reach Radio needs. • Data applications: an open path • Improved reception quality for new applications of this medium • Flexible use of radio, whenever DRM Standards with a large geographical coverage and wherever you want it DRM stands for Digital Radio • Future enhancements: a clear path • No change to existing listening Mondiale, where the spelling of the last for new ideas with an open standard habits: same frequencies, same listen- word is on purpose to show its inter- The DRM standard has been de- ing conditions (fixed, portable and mo- national nature. DRM is a worldwide signed taking into account a number of bile radio), same listening environment consortium established in Geneva, technical constraints, among which are: (indoors, in cities, in dense forests) Switzerland, to promote a unique LF, • Short access time for the receiver: • Low-cost receiver, low energy con- MF and HF audio-broadcasting stan- The listener shall not wait for more sumption dard. See www.drm.org for more than a few seconds before getting ac- • Easy tuning with selection by fre- details. This site contains two reference cess to the desired program, and shall quency, station name or programming ITU papers on the subject.2, 3 obtain radio broadcasting • More diverse program content, DRM’s members include: broad- even faster. using the full capabilities of new digi- casters and broadcasting associations; • Maximum quality (objective and tal features network operators; research insti- subjective): In the allowed transmis- • Wide receiver range with more tutes; component, receiver and trans- sion bandwidth, a maximum useful bit and better features mitter manufacturers; regulatory and rate must be conveyed. This implies a that will give you programs standardization authorities. high-spectral-efficiency modulation with associated text information, sta- There are presently more than 80 scheme—more than 2 bit/s/Hz. tion name, record title, singer’s name. members in DRM, corresponding to 27 • Robustness against distortions countries. In the USA for example, (multipath, Doppler, noise): This is man- DRM System Description members include: Harris Broadcast, datory, especially in the shortwave With the limited bit rate available, the International Broadcasting bands (which are often severely affected it is important to strike the right bal- Bureau, Continental Electronics by propagation disturbances and inter- ance between flexibility and efficiency Corporation, Sangean America Inc, ference) or in medium waves during the while protecting each bit of information Technology for Communications Inter- transition between day and night. to an appropriate degree. A distinction national. Some other active members • Flexibility: According to the cur- is therefore made between main pay- include: the British Broadcasting Cor- rent broadcasting frequency band, the load data and the various types of data poration (BBC), Sony, Bosch, Thales, frequency separation of different that the receiver needs to help it find RFI, DW, JVC, Telefunken and so on. transmissions, the bandwidth of the and decode the desired program. The standardization process involves transmitter and the total available The main payload is called the mainly the ITU (International Telecom- bandwidth can be adjusted to the Main Service Channel (MSC). Two munication Union) based in Geneva, needs of the broadcasters. In the same subsidiary channels are also provided the European body ETSI (European way, the required protection level is namely the Fast Access Channel (FAC) Telecommunications Standard Insti- not the same in LW, MW and SW and and the Service Description Channel tute) and the ISO (International Stan- in a given band. It can vary according (SDC). These two are key to ensure dard Office). Other bodies such as IEC to the time of day. Moreover, the sys- simplicity of receiver operation and and ARIB are also involved. tem should include operating modes are therefore designed to be reliably Key features of the standard include: that can be used in the transition received in adverse conditions, with • A worldwide standard to allow for phase, where simultaneous broadcast- different forward error-correction unique replacement of the AM format ing (simulcast) of compatible AM is schemes from the MSC. • Compatibility with existing required. Notice that a change in any The FAC is intended to be decoded channelization: The DRM signal is parameter needs no intervention from quickly by the receiver on first acquir- the listener, since the receiver is re- ing the signal (at switch-on, or during 1Notes appear on page 56. mote controlled by the transmitter. scanning). It carries a minimum of

50 Jan/Feb 2003 constantly repeated data that might hardware and software level. that starts with the SDC special group be essential at this stage: informing These definitions are used hereafter: of COFDM symbols and contains the receiver what bandwidth option is The system conveys symbols that are exactly three transmission frames. As in use, what modulation is used for the located at known instants and frequen- decoding the SDC is necessary to start SDC and MSC, which length of cies. A carrier is the set of symbols that decoding the whole bit stream, one can interleaver is used for the MSC, and are located at the same frequency. A see that a standard delay to start re- so forth. COFDM symbol is the set of symbols ception is on the order of a super-frame. The FAC is naturally concentrated that are synchronously transmitted on Short access time is obtained by in the narrowest frequency bandwidth all the used carriers. Hence, the num- means of a few dedicated carriers that available (namely 4.5 kHz) so that the ber of symbols in a COFDM symbol is the receiver looks for in a first step for receiver needs only to receive and de- the number of used carriers. The fast frequency synchronization. In ad- code this bandwidth to know exactly COFDM symbols are grouped in a com- dition, another set of carriers always what is transmitted. (See Fig 1.) plete transmission frame that appears contains the same information at a The SDC contains more data, also periodically with the same format. That given instant—this corresponds to a sent repeatedly but in a longer cycle to is, the same repartition of reference and constant known waveform/pattern maintain efficiency. It contains an - useful symbols. The notion of a trans- which is used for pattern synchroniza- tification of the services available in the mission super-frame is also used to de- tion. MSC, together with further information note the group of transmission frames Maximum quality is obtained by the to instruct the receiver how to decode each service. Here, lists of alternative frequencies and frequency schedules would be transmitted if appropriate. Finally, the bulk of the signal con- veys the MSC. With the limited bit- rate available within one 9- or 10-kHz channel, this would normally be used to carry one audio program, together with a modest stream of data. Never- theless, there is a degree of flexibility, so the MSC may contain between one and four streams of data. Streams and services are distinguished as follows. An audio service consists of one stream carrying audio, and optionally one stream carrying data. A data ser- vice consists of one stream carrying data. The proposed system is based upon a multi-carrier modulation. It can be seen as a regular juxtaposition in the frequency domain of k elementary nar- row-band , each conveying a bit rate of d/k if d is the bit rate of Fig 1—DRM time-frequency structure showing the various channel positions for 4.5 and the overall system. This choice comes 10-kHz bandwidth. from the robustness of such a waveform when a high is nec- essary, while at the same time severe propagation conditions must be en- dured. Similar choices were made for higher-frequency terrestrial radio broadcasting (DAB, see Note 2) and TV broadcasting (DVB-T, see Note 3) in Europe, or more recently for the indoor high-data-rate communica- tions standard IEEE 802.11a at 5 GHz. This leads to an optimum occu- pancy of the available bandwidth since the frequency spectrum of the signal is (almost) rectangular. This is possible because the signal that is conveyed by each carrier is orthogonal to the oth- ers. This property enables the signals conveyed by the different carriers to be separable, even if the narrow-band carrier spectra overlap. The adjust- ment of the system bandwidth is obtained via the modification of the number k of subcarriers, without any other change, especially at the receiver Fig 2—Symbol, frame and carrier definitions (example of 10-kHz overall bandwidth).

Jan/Feb 2003 51 use of multilevel quadrature amplitude Low complexity is inherent in the although it may also be useful in some modulation (QAM). The proposed system. Since the signal can be seen as longer-distance applications. The guard QAMs have levels of 4, 8, 16 or 64, the a number of elementary carriers uni- interval is sufficient for SFN operation. number of states being chosen as a func- formly spaced in frequency, the main Mode B can be described as intended tion of the desired level of robustness. digital signal processing is done by for “Time and frequency selective chan- For this type of modulation, coherent means of several fast Fourier trans- nels, with longer delay spread.” It has 1 is required; at each in- forms or inverse fast Fourier transforms a guard interval of 5 /3 ms, with a car- 7 stant and at each frequency, it is neces- (FFTs, IFFTs) which are known for rier spacing of 46 /8 Hz and a higher sary to estimate the complex gain of the their very efficient implementations. At density of pilots. The longer guard in- transmission channel. To evaluate the the receiver, the FFT is equivalent to a terval is intended to cope with greater channel response, some of the sym- large filter bank, each filter selecting multipath spread, as can be caused bols—at predefined frequencies and only one . The complexity during multimode, multihop sky-wave instants—are sent with predefined level is only proportional to the occu- propagation, while the greater carrier amplitude and phase references so that pied bandwidth, and it is independent spacing and pilot density give greater the gain of the channel can be evalu- of the channel quality. tolerance to Doppler spread. ated at any instant (using time Finally, ease of use of the system is Mode C is devoted to “Fast varying interpolation) and any frequency (us- obtained by automatic remote control time and frequency selective channels, ing frequency interpolation). of the receiver. Dedicated, highly pro- with longer delay spread.” It has a 1 Robustness is achieved by the use tected symbols convey all the necessary guard interval of 5 /3 ms, with a carrier 2 of multilevel coding (MLC). An asso- configuration parameters: Any change spacing of 68 /11 Hz and the highest ciation of convolutional encoding and in the transmission characteristics density of pilots. The long guard-inter- interleaving leading to an optimiza- (bandwidth, coding, interleaving) is val is intended to cope with greater tion of the transmission efficiency.4, 5 automatically taken into account by the multipath spread, as can be caused In conjunction with coding, time and receiver. These symbols are grouped in during multimode, multihop sky-wave frequency interleaving are used, which the FAC and the SDC channels. propagation, while the greatest carrier have the effect of spreading perturba- DRM currently contains 4 modes: spacing and pilot density gives maxi- tions (frequency-selective fading, flat • Mode A (ground wave): Gaussian mum tolerance to Doppler spread. fading, interference) on distant sym- channels, with minor fading adapted Each symbol is modulated indepen- bols, so that decoding is more efficient. to LW and MW during daytime. dently from its neighbors by choosing “Parametrability” is obtained by an • Mode B (sky wave): Time and fre- a given point in a predefined constel- incremental design of the broadcast quency-selective channels, with longer lation according to the value on an signal. The signal contains a standard delay spread for SW and MW nighttime. information word (2-6 bits). The ref- kernel group of carriers (occupying • Mode C (robust): Time and fre- erence symbols are transmitted with 4.5 kHz) that is common to all ver- quency-selective channels, with greater a predefined amplitude and phase. A sions. The required bandwidth/bit rate Doppler spread for bad SW channels. reference symbol is used either for is obtained by adding additional • Mode D (extreme): Very robust synchronization purposes or for chan- groups of subcarriers on either side of mode, but with a reduced net bit rate nel complex-response estimation. A the kernel group. (See Fig 3.) (approximately half of the bit rate of useful symbol must be demodulated During the analog-to-digital tran- mode A). and decoded in order to recover the sition phase, a small number of addi- The first three modes are expected original information it conveys. tional groups may be used. The unused to cover most applications. The subcarriers are located at offsets 2 spectrum will be occupied by an AM- Mode A has a guard interval of 2 /3 from the center frequency that are ∆ 2 compatible waveform, which can be ms, together with a carrier spacing of multiples of f = 1/Tu, that is 41 /3 Hz 2 7 received by classical AM receivers 41 /3 Hz. This is described as intended (system for ground wave) or 46 /8 Hz 2 without any modification. for “Gaussian channels, with minor fad- (system for sky wave) or 68 /11 (robust). There is flexibility because the bit ing,” and is thus particularly suitable By convention, in the complex baseband stream is divided into a main stream for local or national coverage at LF/MF, representation, the kth carrier is at an for conveying standard audio or audio plus still pictures or data and a system data stream with a much lower bit rate conveying additional data, such as pro- gram-associated data (PAD similar to RDS). With up-to-date audio encoders/ decoders, the ratio of audio to pictures may be instantaneously variable in the main stream, and the significance of the data bit stream can vary as desired. There is minimum disturbance to traditional AM users because the digi- tal waveform has a flat spectrum so that, as received by an AM receiver, the digital signal sounds almost like white Gaussian noise. A pulse shap- ing of the transmitted symbols (that is, time windowing of the COFDM symbol) and/or additional output sig- nal filtering can further decrease the disturbance of AM receivers in the adjacent channels. Fig 3—Various bandwidths possible.

52 Jan/Feb 2003 offset of k∆f from the dc component; k can be positive (to the “right” of the ref- erence carrier) or negative (to the “left” of the reference carrier). The 0th car- rier is dc. In the first two parameter sets the transmission frame contains 15 COFDM symbols; there are 20 COFDM symbols in the last one, leading to a constant common duration of 400 ms. As already mentioned above, the car- riers are grouped in a kernel group of carriers, which is common to all trans- mission modes. The group conveys all the reference symbols necessary for time and frequency synchronization, as well as the FAC symbols that describe the current mode. The kernel group is immediately above the carrier at fre- quency F0 and does not contain it. Its bandwidth is exactly 4.5 kHz between the zeros of its frequency spectrum. Additional groups of subcarriers (possibly none, in the 4.5-kHz version), the number of which is defined accord- ing to the desired bit rate and available bandwidth. An additional group does not contain any synchronization or FAC symbol. If there are additional groups below the carrier, their number is al- ways such that the frequency spectrum of the signal is symmetrical around the carrier. The bandwidth of each addi- tional group is exactly 1.5 kHz between the zeros of its frequency spectrum. The source coding requirements are directly derived from the channel ca- pacity. The capacity available for au- dio within a single 9- or 10-kHz chan- nel is distinctly limited—at 20 to 25 kbps and perhaps as little as 10 kbps for some extremely unfavorable HF paths. This clearly represents a seri- ous source-coding challenge for DRM, which expects to deliver good audio quality for both speech and music. DRM has incorporated work and key technologies already done in the development of source coding else- where and fine-tuned it to this par- ticular application. For coding most broadcast program material, an audio coder is needed to cope with the arbi- trary mix of speech, music and inci- dental background sounds. For this purpose, DRM uses (AAC) from the ISO MPEG-4 standard, supplemented by spectral band replication (SBR). The SBR technique synthesizes the sounds that fall within the highest fre- quency octave. Sounds in this range are usually either: 1. noise-like (sibilance, percussion instruments such as shakers, brushed cymbals and so on), or 2. periodic and related to what

Fig 4—Complete DRM architecture for transmit side.

Jan/Feb 2003 53 appears lower in the spectrum (over- • To simplify the multiplex scheme transmission frame, or the end-of- tones of instruments or voiced sounds). and its associated description to reduce message (EOM) flag. At the sender, the highest-frequency the bit rate for signals and have a mini- • The positions of the FAC cells and band of the audio signal is examined to mum transmission delay (see above). some frame-synchronization cells determine the spectral distribution and • To simplify intellectual-property have been modified to fit in the re- whether it falls into category 1 or 2 issues. duced bandwidth. above. A small amount of side informa- The main similarities between the • The whole original signal is shifted tion is then prepared for transmission proposed system and the original towards dc to fit in a bandwidth of 300- to help the decoder. The highest-fre- DRM standard are the following. 3000 Hz. The 3-kHz bound is always quency band is then removed before the respected. The 300-Hz bound is only remaining main band of the audio sig- Unmodified Features approximately respected because of the nal is passed to the AAC coder, which • Symbol duration and guard times spacing between tones (333 Hz, 281 Hz codes it in the conventional way. • The positions of the three un- and 273 Hz for modes A, B and C). At the receiver, the AAC decoder equally spaced unmodulated carriers • Finally, each vocoder frame and first decodes the main band of the au- (pure tones), which help find rapidly each free data block is completed by dio signal. The SBR decoder then adds the actual frequency shift of the re- an eight-bit CRC (parity check) to de- the synthetic upper band, helped by ceived signal. tect errors and minimize their impact the instructions sent in the side infor- • The repartition and positions of on the audio quality. mation. Overtones are derived from the gain-reference symbols In practice, the signal-processing the output of the AAC decoder, while • The modulation schemes (constel- software for transmission and recep- noise-like sounds are synthesized us- lations) tion are essentially the same as for the ing a noise generator with suitable • The interleaving principles (short DRM system. The only changes are spectral shaping. and long interleaving) software simplifications and different An alternative for speech-only pro- • The convolutional codes—in prac- primary data (constants or data ar- gramming is to use a coder designed tice, we use only the simplest one, for rays). The demonstrator was devel- expressly for speech, in which case the both vocoded audio and free user data. oped to show the highest quality bit-rate can be reduced much more Its performance is, of course, the same achievable with current voice-com- than with a waveform audio coder, as with other more complex encoding pression techniques and the possibil- while retaining the same speech qual- schemes. ity to add new services. ity. Although this offers broadcasters The voice codecs used are part of further flexibility, there is, however, Modified Features the Thales portfolio of coders (the HSX some doubt whether this approach • The SDC stream has been com- family for Harmonic Stochastic would be used much in practice (apart pletely suppressed since it is no longer eXcitation). Thales Communications from multilingual ). necessary. All the configuration infor- has been active in low-bit-rate voice Even speech-only broadcast material mation is located in the FAC only, so codecs since 1968. In particular, the contains jingles, background sounds in that the super-frame of DRM (three 800 bits/s Stanag (Standard NATO interviews and so on, all of which can frames) now only contains one frame Agreement) and the ETSI Tetra cod- cause serious problems to speech cod- (latency reduction, software simplifi- ers (for Professional Mobile Radio, the ers. Figure 3 summarizes the various cation). European equivalent of mode-25 ra- basic channel-content possibilities. • The FAC contains only 40 symbols dios) came out of these laboratories. Thus the complete system is given in (instead of 65), 26 useful bits (instead The advantage of having coders from Fig 4 (transmit side). of 64) and is more protected (code rate the same family is that they share a lot 0.5 instead of 0.6). It contains the full of common parts, allowing for simple Adaptation to Amateur Radio Use set of parameters of the current trans- switching of coders, reduced code size From the beginning, Thales has pro- mission, as well as other information and thus simple maintenance or even posed to limit the kernel group of carri- like the frame contents: A flag indicat- hierarchical error-correcting schemes. ers to 3 kHz, thus enabling easily a ing the number of vocoder frames in the Such schemes correspond to transmis- 3-kHz mode for use in radio amateur HF or military applications.6 So far, this mode has not been retained by the DRM consortium to keep the basic set of modes as limited as possible, thus low- ering the development cost of the first version of the receiver chipsets. Sensing the importance of such a mode, a demonstrator was developed anyway. Further modifications of the DRM system were applied while keep- ing the same overall structure: • To limit the kernel group of carri- ers to less than 3 kHz. • To adapt to push-to-talk mode, and other features of amateur transceivers. • To include lower bit-rate vocoders (waveform coders are not usable at such low bit rates) imposed by the further reduction in available bandwidth (1200, 2400, 3200 bits/s voice coders are imple- mented in the demonstrator). Fig 5—Evolution of voice coders quality through time.

54 Jan/Feb 2003 sions where the additional information synchronization) and the estimated The DRM consortium normalized necessary to go from a given quality to channel impulse response (here two laboratory test conditions to get a fair a better one is less protected. So qual- paths are present with around 90° comparison between various propo- ity at reception varies depending on the shift between them). nents. The channel model selected is ability of the modem to decode part or Other information are given such described hereafter. all the information, thus insuring at as the text message received, the mode The approach is to use stochastic least the minimum quality and at best, detected, the estimation of the Dop- time-varying models with stationary the best quality when the channel per- pler shift, the estimated instanta- statistics and define models for good, mits it. Such a scheme is not imple- neous SNR. moderate and bad conditions by tak- mented yet, but it could be added in a The demonstrator is able to provide ing appropriate parameter values of future version of the system. various transmission applications: the general model with stationary sta- Fig 5 shows the evolution through • Real-time voice transmissions, tistics. One of those models with time of the quality of voice coders. The using a microphone/speaker and a adaptable parameters is the wide- middle-range quality lying between sound card in the PC sense stationary uncorrelated scatter- MOS (mean opinion scores) of 3.5 and • Transmit digital data (hand-writ- ing model (WSSUS model). The justi- 4 are considered as the range accept- ten text or even complete files). The fication for the stationary approach able for low- to medium-quality com- maximum available digital data rate with different parameter sets is that mercial applications. The latest is directly related to the vocoder bit results on real channels lead to BER 2400-bits/s coders have equivalent or rate and the current error correcting curves between best and worst cases better quality than older 4800-bits/s code and modulation. found in the simulation. ACELP coders (adaptive code-excited • Test the modem performance by A tapped-delay-line model is then linear prediction), while-1200 bit/s sending a test sequence for bit-error- used for multipath generation.7, 8 The coders become of acceptable quality. rate estimation at the receiver time-variant tap weights are zero- New research in further rate reduc- To test the demonstrator in a con- mean, complex-valued stationary tion allows hope for 600 bits/s in a few trolled environment, a real-time SW Gaussian random processes. The mag- years. Competition for a high-quality propagation simulator was included to nitudes are Rayleigh-distributed and 4-kbits/s codec has been ongoing for show the insensitivity of the digital part the phases are uniformly distributed. several years without anything yet of the system to multipath, whereas at For each weight, there is one stochastic satisfying all the very stringent con- the same time, an analogue AM equiva- process, characterized by its variance straints (such as a very low delay). All lent is seriously perturbed. The simu- and power density spectrum (PDS). The the G.xxx codecs are standardized by lator may generate a maximum of four variance is a measure for the average the ITU, while the STG coders are paths. All the characteristics (ampli- signal power, which can be received via military NATO standards—also tude, average power, frequency offset, this path and the PDS determines the named Stanags. Doppler spread, time spread) are ad- average speed of variation in time. This Other potential coders could have justable in real time using a natural type of channel model is known as the been: man/machine interface. Narrow-band “Watterson” model. • DVSI: proprietary format (IMBE interference may be added for system A number quantifies the width of and AMBE: 2 to 9.6 kbits/s) mainly tests. All these impairments are added the PDS and this quantity itself (the chosen for satellite communications at the transmit side. width) is commonly referred to as the systems (Inmarsat), Iridium and APCO 25. • MELP and EMELP: basis for the STG 4591 Stanag The main reason for not choosing them is that the first is a proprietary, not easily available format, and that both come with IPR licensing fees and conditions, without any important gain in quality. Our demonstrator software is 100% PC based for simple integration with standard off-the-shelf transceivers. A simple and intuitive man/machine in- terface (MMI) based upon the use of LabWindows software has been devel- oped. The demonstrator has been fully integrated with Ten-Tec off-the-shelf transceivers. Transatlantic tests with ARRL are underway. A screen dump or the MMI in receive mode is given in Fig 6. The six sub-windows show (left to right, top row first): the time values of the received signal, the spectrum of the received signal (filtered within 3 kHz), the reception level (estimation of the signal-to-noise ratio, SNR), the audio output signal, the received con- stellation (after time and frequency Fig 6—Demonstrator MMI while in receive mode.

Jan/Feb 2003 55 Doppler spread of that path. There might be also a non-zero center fre- GLOSSARY quency of the PDS, which can be in- AAC Advanced Audio Coding (MPEG-2/4) terpreted as an average frequency ACELP Adaptive Code Excited Linear Prediction shift (or Doppler shift). AFS Alternative Frequency Switching It is common to assume a Gaussian AM amplitude statistic for the process, ARIB Association of Radio Industries and Businesses which is a reasonable assumption for AWGN Additive White Gaussian Noise an ionospheric channel. The stochas- BER Bit-Error Rate tic processes belonging to every indi- COFDM Coded Orthogonal Frequency Division Multiplex vidual path then become Rayleigh pro- DAB Digital Audio Broadcasting cesses. WSSUS does not define the DC Direct Current shape of the PDS. For ionospheric DRM channels, Watterson has shown a DVB Digital Broadcasting Gaussian shape to be a good assump- 9, 10 DVB-T DVB - Terrestrial tion. The one-sided Doppler spread EBU European Broadcasting Union is then defined as the standard devia- ETSI European Telecommunications Standard Institute tion (s) of the shape of the PDS. FAC Fast Access Channel Five channels are currently used by FFT Fast Fourier Transform DRM: • FM AWGN (additive white Gaussian HF High-Frequency noise): one path of constant amplitude IBC International Broadcasting Conference (for ground-wave propagation). • IEC International Electrotechnical Commission Ricean with delay: one constant IFFT Inverse Fast Fourier Transforms path of unit amplitude, a second with ISO International Standard Office half amplitude delayed 1 ms, with ITU International Telecommunication Union 0.1-Hz Doppler spread (for flat fading ITU-R ITU - Radiocommunication Sector at MW and SW). • LF Low-Frequency US Consortium: four paths of am- LW Long-Wave plitudes (1, 0.7, 0.5, 0.25), delay (0, 0.7 MF ms, 1.5 ms, 2.2 ms) Doppler spread (0.1 MLC Multi-Level Coding Hz, 0.5 Hz, 1 Hz, 2 Hz) and frequency MMI Man Machine Interface shifts (0.1 Hz, 0.2 Hz, 0.5 Hz, 1 Hz). • MOS Mean Opinion Scores CCIR poor: two paths of equal MPEG Moving Picture Experts Group amplitude delayed 2 ms and equal 1- MSC Main Service Channel Hz Doppler spread (SW propagation). • MW Similar to , but with 4- NADIB Narrow Band ms delay and 2-Hz Doppler spread OFDM Orthogonal Frequency Division Multiplex (bad SW propagation). PAD Program Associated Data Conclusion PDS Power Density Spectrum QAM Quadrature Amplitude Modulation This paper presents a complete sys- RDS tem proposal for Amateur Radio digi- RF Radio Fequency tal voice transmission derived from SBR Spectral Band Replication the DRM standard. It shows the cur- SDC Service Description Channel rent state of the art in COFDM mo- SFN Single Frequency Network dem technology. COFDM was chosen SNR Signal-to-Noise Ratio because of its very good robustness SSB Single Side Band when a high spectral efficiency is de- SW Short Wave sired, while at the same time severe UEP Unequal Error Protection propagation conditions must be en- dured. Such a system should start wide interest in digital voice, as the 2J. Stott, “DRM—Key Technical Features,” Le Floch, “Thomson—CCETT Common necessary computing power required EBU Technical Review #286, March 2001. Proposal for a Digital Audio Broadcasting is well within current PC technology 3M. Cronk, “DRM—Implementation Issues,” System at Frequencies below 30 MHz,” EBU Technical Review #286, March 2001. NADIB Report, June 1998. reach (or DSP for future integration 4 9 inside transceivers). H. IMAI and S. HIRAKAWA, “A New Multi- J. Lindner, D. Castelain, F. Nicolas, “Speci- This system proposal is currently level Coding Method Using Error Correcting fication of an Ionospheric Channel Model Codes,” IEEE Transactions on Information for AM Radio Broadcasting Bands,” being tested (real-time demonstrator Theory, Vol. IT-23, pp 371-377, May 1977. NADIB Report , March 1998. measurements) using a transatlantic 5U. Wachsmann, R. Fisher, and J. Huber, 10Annex B to STANAG #4285, “Evaluation of link. “Multilevel Codes: Theoretical Concepts Employing the Stanag 4285 and Practical Design Rules,” IEEE Trans- Waveform.” actions on Information Theory, Vol. IT-45, References Notes pp 1361-1391, July 1999. C. C. Watterson, J. R. Juposher and W. B. 1C. J. Demeure, P.A. Laurent, “A new Mo- 6DAB system: ETSI Norm, ETS 300 401 ed. Bensema, “Experimental Verification of an dem for High Quality Sound Broadcasting 2, May 1997 Ionospheric Mode,” ESSA Technical Re- at Short Waves,” IEEE 4441, 7th Confer- 7DVB-T system: ETSI Norm, ETS 300 744, port, ERL 112-ITS, 1969. ence on Radio Systems and Techniques, March 1997 CCIR Recommendation 520, Use of High Fre- Nottingham, UK; pp 50-54, July 1997. 8P. A. Laurent, C. Demeure, D. Castelain, B. quency Ionospheric Channel Simulators. ††

56 Jan/Feb 2003