Comparison of the DAB, DMB & DVB-H Systems

Contents

1.  Modulation

1.1  OFDM Modulation

DAB, DMB and DVB-H all use OFDM (orthogonal frequency division multiplexing) modulation. The following table shows the number of OFDM subcarriers that can be used for each of the systems:

DMB uses DAB's transmission system, but adds an additional layer of FEC (forward error correction) coding to improve robustness and spectral-efficiency, and replaces DAB's inefficient MP2 (MPEG Audio Layer 2) audio codec. Both of these technologies will be discussed below.

1.2  Subcarrier Modulation

The following table shows the signal constellations that DAB/DMB and DVB-H can use to modulate the subcarriers, along with the number of bits per symbol that each signal constellations allows to be transmitted on each subcarrier:

Using higher-order signal constellations allows higher multiplex data rates. For example, for a given error correction code rate and number of subcarriers, if DVB-H changed from using QPSK to 16-QAM then the multiplex data rate would double. However, in order to use higher-order signal constellations, because the signal points are closer together than for lower-order signal constellations the transmitter power needs to be increased. DVB-H's intelligent solution to the increase in required transmitter powers is the addition of the MPE-FEC, which is a powerful additional outer-layer of FEC coding, and the MPE-FEC allows DVB-H to use 16-QAM at similar transmitter power levels as are required for transmission of QPSK without the MPE-FEC.

1.3  Coherent / Differential Modulation

The following table shows whether the systems use differential or coherent (phase-synchronised) modulation. Differential modulation means that the data bits cause the subcarrier's phase angle to be changed relative to the phase angle transmitted on the same subcarrier in the previous OFDM symbol, e.g. data bits 01 would cause the subcarrier's phase to shift by 90⁰, whereas coherent modulation means that the data bits are mapped onto a certain constellation point, e.g. for QPSK modulation, data bits 00 would always be mapped to a phase angle of 45^o, data bits 01 mapped to a phase angle of 135^o, and so on.

The advantage of differential modulation is that the receiver only has to calculate the phase difference between the current and previous subcarrier's phase angles and the receiver doesn't need to phase-synchronise. A disadvantage of differential modulation is that it incurs a 3 dB SNR (signal-to-noise ratio) penalty relative to coherent modulation (refer to Digital Communications by J Proakis). 

A further advantage of coherent modulation is that it allows 16-QAM and 64-QAM signal constellations to be used, which enable much higher bit rates and spectral-efficiency.

1.4  Guard Interval

The guard interval (also known as the cyclic prefix) duration determines the maximum distance between transmitters in an SFN (single-frequency network) and the feasible size of SFNs. The guard interval (GI) is added between the useful part of OFDM symbols in order to catch delayed multipaths from the previous OFDM symbol in order to avoid inter-symbol interference (ISI), because ISI destroys the orthogonality between the subcarriers and high levels of ISI will cause reception to fail.

(GI = guard interval, quoted as fractions of  T_u, i.e. GI = 1/4 means that GI = T_u/4)

1.5  Time- & Frequency-Interleaving

Interleaving is the re-arrangement of data symbols over time and/or frequency so that if a signal fades and transmission errors occur then because the interleaver has re-arranged the order in which the data symbols were transmitted then following de-interleaving in the receiver (which puts the data symbols back into the original order) bursts of errors are more spread out, which helps to avoid presenting the Viterbi FEC decoder with burst errors that it cannot correct.

DAB/DMB use both time- and frequency-interleaving. DVB-H uses frequency-interleaving and the 2K and 4K modes can use time-interleaving.

DVB-H needs time-interleaving less than DAB/DMB due to DVB-H using significantly wider channels than DAB because at any instant in time the likelihood that a significant part of the channel bandwidth is in a deep fade is significantly less for DVB-H than for DAB. DVB-H also has 2 outer layers of Reed-Solomon FEC coding, which are inherently very robust against burst errors.

2.  SFNs (Single-Frequency Networks)

As mentioned in the Guard Interval sub-section, OFDM SFNs work so long as the delayed multipaths with significant signal strength arrive with a delay (relative to the first path of significant strength) less than the guard interval duration.

A frequent accuastion made by proponents of DAB who are trying to knock DVB-H is that DVB-H cannot use large-scale SFNs, but, as can be seen in the above table, when a guard interval of 1/4 is used for DVB-H then the guard interval duration is either longer than DAB's (5, 6 and 7 MHz DVB-H channels) or almost as long as DAB's (8 MHz DVB-H channel). Therefore, DVB-H definitely can use large-scale SFNs.

For example, if a country is having difficulty designing a transmitter network for an 8 MHz-wide channel DVB-H SFN due to self-interference (e.g. caused by constraints imposed due to transmitter locations), then rather than changing to a multi-frequency network (MFN, which are significantly less spectrally-efficient than SFNs) they can use a narrower channel in order to increase the guard interval, which will allow an SFN to be used. Having said this, if it was difficult to design a transmitter network for a DVB-H SFN in a 7 MHz channel then it would be just as difficult to design a transmitter network for a DAB SFN because of the similarity in the guard interval durations. 

It might seem to be illogical to transmit, say, a 7 MHz-wide multiplex in an 8 MHz-wide channel, but overall it is significantly more spectrally-efficient than changing to an MFN.

3.  Forward Error Correction (FEC) Coding Schemes

Apart from the modulation used, the FEC coding scheme is the most important part of a wireless digital communication system's transmission scheme, and stronger FEC coding schemes enable higher data rates, lower transmitter powers, and more robust reception.

FEC codes add redundant information that allow errors to be corrected at the receiver, and for a given FEC code, the higher the proportion of redundancy added the more errors can be corrected. An important parameter of FEC codes is the code rate, R_c, which is given by:

R_c = k / n

where k = number of input symbols, and n = number of output symbols from the FEC encoder. The number of redundant symbols added is n-k, and (n-k)/n is the fraction of redundancy added. For example, a convolutional encoder with a code rate of 2/3 has 2 input bits, one redundant (parity) bit is added, 3 bits are output from the encoder, and the fraction of redundancy added is 1/3.

The following figure shows the layers of FEC coding that DAB, DMB and DVB-H use, and each technology will be discussed below:

3.1  Convolutional Coding

DAB, DMB and DVB-H all use convolutional FEC coding with various code rates (implemented by puncturing the mother code).

By modern standards of FEC coding, convolutional coding on its own is an extremely weak form of error correction coding allowing only low data rates and requiring relatively high SNR (signal to noise ratio) at the receiver -- i.e. requiring relatively high transmitter powers for a given bit error rate (BER -- which is the proportion of bit errors out of the total number of bits transmitted).

DAB uses unequal error protection (UEP) for audio where the data that is more likely to cause errors that the listener will be able to perceive are protected more strongly. DMB and DVB-H use equal error protection (EEP), where a fixed code rate is used for all audio data.

The main problem with DAB's FEC coding scheme is that this convolutional code is the only layer of FEC coding that is applied, whereas both DVB-H and DMB have outer layers of FEC coding, as will be described below. This means that any uncorrected errors that get through this single layer of FEC coding are potentially audible, and the audio samples -- which make up about 84% of the total bit rate -- have very weak error protection. The result of too many uncorrectable errors is the infamous "bubbling mud" sound that DAB-owners will likely be accustomed to. 

To show how weak the error protection of the audio samples is, the audio samples are protected by a relatively high code rate of 0.57, and with a BER of 2 x 10^-4 (which is the usual target BER value quoted for DAB) the number of uncorrectable audio sample bit errors per hour for a 160 kbps service is:

Uncorrectable audio sample bit errors = bit rate  x  number of seconds  x  BER  x proportion of audio samples

Uncorrectable audio sample bit errors = 160,000  x  3600  x  2 x 10^-4  x  0.84  = 96,768 uncorrectable errors per hour

or, on average, 27 uncorrectable errors per second.

The obvious solution to the problem of there being so many errors is to simply reduce the number of errors by using stronger error protection. Therefore, for DMB and DVB-H, which both use far stronger error protection schemes, a BER of 10^-6 will be assumed in the rest of this article, and this drastically reduces the total number of errors, as can be seen in the following calculation for a service that uses 48 kbps HE AAC (which provides the same level of audio quality as 160 kbps MP2 as used on DAB):

Total uncorrectable errors = 48,000  x  3600  x  10^-6  =  172 uncorrectable errors per hour

Such a large reduction in the total number of errors (a reduction by a factor of 563) makes DVB-H and DMB reception significantly more robust than DAB reception. Moreover, as well as using standard cyclic redundancy checks (CRC) to detect the presence of errors, MPEG-4 AAC (HE AAC, which DVB-H and DMB use, is a high efficiency version of AAC -- AAC stands for Advanced Audio Coding) allows advanced error concealment and error risilience techniques to be used, which further limit the impact that bit errors cause. 

3.2  Reed-Solomon (RS) Coding

Reed-Solomon coding is used as the outer layer of FEC coding for DMB and as the middle layer of FEC coding for DVB-H. RS coding fixes the weakness of DAB's single layer of convolutional coding.

The RS code used by both DVB-H and DMB uses packets (codewords) of 204 bytes, which consist of 188 input bytes and 16 parity bytes (code rate = 188 / 204, i.e. the amount of redundancy added is only 16/204 = 7.8%, which is very low), and can correct any 8 bytes that contain bit errors in a packet irrespective of the number of bit errors in each byte, i.e. all 8 bits can be in error, but that only counts as one of the 8 bytes that can be corrected. 

The use of Reed-Solomon codes as the outer layer along with a convolutional code for the inner layer of an FEC coding scheme is a very good combination for wireless systems. This is because errors tend to occur in short, concentrated bursts in wireless systems which then overwhelm the Viterbi convolutional decoder, and the benefit of using an outer layer of RS coding is that when the Viterbi decoder becomes overwhelmed due to there being too many errors, the Viterbi decoder itself produces a short burst of errors, which is exactly what RS codes are good at correcting. 

As is shown in the DVB-T specification, this RS code reduces the BER (bit error rate) at the output of the convolutional (Viterbi) decoder from 2 x 10^-4 down to 10^-11 (see DVB-T specification), as is required for video reception, which means that this RS code reduces the number of bit errors by a factor of 20 million! 

The fact that the outer RS code reduces the BER by such an large factor allows DMB to use a higher inner-layer convolutional code rate (less redundancy added) whilst still using the same transmitter power and achieving the same level of signal robustness. This allows the multiplex data capacity to increase significantly, and hence the spectral-efficiency for DMB is greater than for DAB. This also applies to DVB-H, although because DVB-H uses 3 layers of FEC coding then its spectral-efficiency is significantly higher than DMB's, as will be discussed in the next sub-section. 

The vastly superior performance gained from combining an outer layer of Reed-Solomon coding with an inner layer of convolutional coding for OFDM was shown in an article in 1987, which was 4 years before the FEC coding scheme for DAB was finalised, so the fact that they decided not to use an outer layer of RS coding for DAB was a big mistake.

In fact, it is interesting to note that there are plans currently underway to add this outer layer of Reed-Solomon coding to the DAB specification. Unfortunately, none of the current DAB receivers can decode this new outer layer or RS coding, and the huge gains in efficiency that will be shown below can only be attained by changing to the modern high-efficiency audio codecs, like HE AAC, which DAB cannot use and there are no plans to add the HE AAC audio codec to the DAB specification.

3.3  MPE-FEC

DVB-H also uses a very strong outer layer of FEC coding called the MPE-FEC (multi-protocol encapsulation forward error correction), which allows DVB-H to use 16-QAM at reasonable C/N values at the receiver, and so allows reception of high data rates at high speed with a single-antenna.

The MPE-FEC code is also a Reed-Solomon code, but can correct any 64 bytes out of a 255-byte packet (codeword). Of the 255 bytes, 191 are input bytes and 64 are parity bytes (the code rate of this code = 0.75). The reason that the MPE-FEC code can correct so much errored data is due to its use of erasures, which means bytes that are known, or thought to be, in error are flagged as being unreliable (erasures), and because half of the error correction capability of an FEC code is consumed by locating the errors, the use of erasures to inform the error correction decoder where the errors are allows the decoder to correct twice as many errors as when erasures are not used.

The MPE-FEC is an extremely powerful FEC code and is the key technology which allows DVB-H's spectral-efficiency to be so much greater than DAB's or DMB's.

4.  Multiplex Capacities & Required C/N

4.1  DAB Multiplex Capacities & Required C/N

DAB uses so-called error protection levels (PL) rather than code rates due to the use of UEP. The following table shows the multiplex data capacities and required C/N (carrier to noise ratio) for the protection levels of interest. Bizarrely, the higher the protection level the lower the strength of the error protection is for the data. There are 5 protection levels which go from strongest error protection at PL1 to weakest error protection at PL5. At the moment in the UK, about 98 - 99% of all radio stations on DAB use PL3, along with some stations on the DRg London local multiplex using PL4, and a couple of radio stations using PL2. 

Because higher protection levels (i.e. lower protection strength) allow higher data capacities, the only PLs of interest for use are PL3, PL4 and PL5.

1 - C/N values taken from Table 2.5 in Digital Audio Broadcasting, Principles & Applications, edited by Hoeg & Lauterbach, using an RF front-end noise figure of 5 dB (a noise figure of 5 dB is used for the RF front-end in the DVB-H Implementation Guidelines)

4.2  DVB-H Multiplex Capacities & Required C/N

DVB-H has primarily been designed to allow reception of mobile TV transmissions (DVB-H is obviously perfectly suited to audio reception as well), and all available C/N data assume reception of video streams. Video streams require far lower BERs (bit error rates) than audio streams (for example, the MPEG Layer 2 audio codec requires a BER of 2 x 10^-4, whereas the MPEG-2 video codec requires a BER of 10^-11). Therefore, all quoted figures for DVB-H significantly overestimate the required C/N for audio stream reception, so it is necessary to estimate the C/N from available data in the DVB-H Implementation Guidelines.

The following table contains the quoted required C/N figures from the DVB-H Implementation Guidelines for video stream reception (the MPE-FEC code rate is 0.75 in all cases on this page):

To estimate the required C/N for audio reception the following assumptions were made and curves were then extrapolated from the known data:

• the required BER for video stream reception on mobile phones was chosen to be 10^-9 rather than the 10^-11. 10^-9 is a figure that has previously been quoted in articles about mobile TV;
• the correct C/N values for the modes in the above table were used as the end-points for each curve;
• the start-points of each curve were a BER of 0.5 and a C/N of 0 dB. A BER of 0.5 is the worst possible BER value;
• each curve has a "waterfall" shape, which all BER vs C/N curves have where the curves are all monotonic (i.e. they always decrease with an increase in C/N), start off reducing slowly and become steeper at higher C/N values.

The estimated C/N values from the above graph for DVB-H for audio reception at a BER of 10^-6 are given in the table below, along with multiplex capacities for 7 and 8 MHz channels:

4.3  DMB Multiplex Capacities & Required C/N

Similar to the situation with DVB-H, DMB was designed to enable TV reception on mobile phones, so there are no required C/N figures available for audio reception and it is necessary to estimate these figures. 

The most often quoted mode that DMB uses for video reception is the equal error protection (EEP) level 3A (PL3A), where the outer layer of Reed-Solomon coding reduces the BER from 10^-4 for audio down to 10^-11 required for reception of video streams.

The following table shows the required C/N values for relevant EEP levels:

1 - C/N values taken from Table 2.6 in Digital Audio Broadcasting, Principles & Applications, edited by Hoeg & Lauterbach, using an RF front-end noise figure of 5 dB (a noise figure of 5 dB is used for the RF front-end in the DVB-H Implementation Guidelines)

2- DMB multiplex capacities are calculated by multiplying the DAB multiplex capacity for the respective EEP protection level by the code rate of the RS code (code rate = 188/204)

Using the above C/N figures for video reception, the required C/N figures for audio reception can be obtained by interpolation from the estimated C/N figures for DVB-H. PL4A won't be calculated, because the required C/N is far higher than the other modes and allows only a modest increase in multiplex capacity compared to PL3B:

5.  Audio Codecs

5.1  DAB Audio Codec

DAB uses the MPEG Layer 2 (MP2) audio codec. This audio codec requires a bit rate of 192 kbps to provide good audio quality on stereo stations.

However, it is debatable whether even a high bit rate such as 192 kbps can provide the same level of audio quality as FM can, because it is certainly the case that current 192 kbps transmissions on Freeview and digital satellite do not sound as good as the same stations on FM.

Unfortunately, in the UK 98% of all stereo stations use the insufficient bit rate of 128 kbps, with the result being that the 128 kbps radio stations sound far worse than the same stations do on FM (obviously this does not apply to digital-only stations which don't have an FM version).

5.2  DVB-H Audio Codec

DVB-H uses the HE AAC (High-Efficiency Advanced Audio Coding) audio codec, which is currently the most efficient audio codec in existence (efficiency is measured by the bit rate required to provide a given level of audio quality -- the lower the required bit rate, the higher the efficiency), and requires a bit rate of 64 kbps to provide good audio quality on stereo stations (roughly equivalent to 192 kbps MP2). The same caveat applies to 64 kbps HE AAC as applied to 192kbps MP2 in that although it can be described as providing "good audio quality" it does not provide the same level of audio quality as FM can and does. This does, of course, depend on having a strong FM signal, but digital radio was supposed to improve audio quality, so digital radio broadcasts should at least match the audio quality that FM provides.

5.3  DMB Audio Codec

DMB was originally designed to use the BSAC (bit-sliced audio coding) audio codec, which is less efficient than the HE AAC audio codec, but DMB now also uses the HE AAC audio codec. 

6.  Chipset Power Consumption

6.1  Time-Slicing

Time-slicing is a power-saving technique that takes advantage of the fact that the service that the user wants to watch or listen to is only transmitted for a fraction of the time due to the fact that there are multiple services carried in one multiplex. This allows the RF front-end to be turned off when the desired service is not being transmitted. This allows a significant amount of power to be saved, because the RF front-end's amplifiers are relatively inefficient in terms of power consumption due to the fact that OFDM reception requires highly-linear RF amplifiers, and the higher the required linearity of the amplifier the lower the power efficiency.

The problem of RF amplifier linearity -- and hence power efficiency -- is exacerbated by the fact that OFDM signals have a high peak to average power ratio (PAPR), which means that the amplifiers must have a relatively high dynamic range so as to avoid/minimise intermodulation distortion and damage to the amplifier. 

The percentage of power saved increases with higher multiplex capacities, because with higher multiplex bit rates the receiver has to turn on its RF front-end for a smaller fraction of the time than for lower multiplex bit rates. This can be seen in the following graph, where the burst bit rate is equal to the multiplex capacity.

6.1.1  Time-Slicing for DAB/DMB & DVB-H

DAB, DMB and DVB-H can all take advantage of time-slicing to reduce power consumption, but the percentage of power saving is different for each system due to the large difference in multiplex capacities. For example, looking at the 100 kbps curve (green) in the figure above, for DAB/DMB with a multiplex capacity of about 1.1 Mbps this achieves a power saving of about 87%, whereas for a DVB-H multiplex with its high multiplex bit rate, the power saving is approximately 97%.

6.2  DAB/DMB & DVB-H Chipset Power Consumption

Chipset power consumption is an important parameter because it determines how long a battery will last between re-charges for portable and mobile receivers.

The following table shows the lowest power consumption requirements for currently available chipsets:

7.  Comparison of DAB, DMB & DVB-H

7.1  Requirements of a Modern Digital Radio System

Before comparisons between the systems are made, it is instructive to look at what the fundamental requirements of a digital radio system for the 21^st century are:

• the digital radio system should provide a wide range of radio stations;
• the audio quality on the wide range of radio stations should be at least as good as on FM and preferably better;
• due to spectrum scarcity the digital radio system must be spectrally-efficient -- i.e. it should carry as many radio stations in a given amount of bandwidth as possible.

7.2  Parameters for Comparison

The efficiency of DAB, DMB and DVB-H can be fairly compared if the following variables are held constant for all systems:

• fixed bandwidth for all systems and all systems use Band III frequencies
• same level of audio quality for all systems
• systems use approximately equal total transmission powers

7.2.1  Fixed Bandwidth for All Systems

Band III is used for TV in some countries and DAB in other countries, and the international frequency plan is such that 4 DAB/DMB channels fit into 1 x 7 MHz TV channel, as can be seen in the following table (e.g. 5A, 5B, 5C and 5D can each be used for DAB/DMB or the same channels can be used for one 7 MHz TV channel for DVB-H):

Therefore, comparison of the systems in a fixed bandwidth will assume 1 x 7 MHz channel for DVB-H and 4 x DAB or DMB channels.

7.2.2  Same Level of Audio Quality

To compare the systems with equivalent levels of audio quality it is necessary to use different bit rates for the systems to reflect the different efficiencies of the audio codecs. The following table shows the bit rates required for given levels of audio quality:

The systems will be compared for all 3 levels of audio quality.

For those that have not yet heard the HE AAC codec in action the following website carries numerous internet radio streams at various bit rates (ignore the exaggerated descriptions on the following website of what each bit rate equates to):

http://www.tuner2.com/   (to listen to the streams from this above website you will need to have a recent version (at least version 5) of Winamp)

The bit rates chosen to be equivalent in the above table were based on comparing the audio quality of 128 kbps DAB stations and 160 and 192 kbps stations on DTT (which use the same MP2 codec as DAB) with the HE AAC internet radio streams on tuner2.com as well as numerous HE AAC encodings of CDs using Nero 6's HE AAC encoder.  

7.2.3  Similar Total Transmission Power Levels

The total transmission power for a multiplex is equal to the integration of the power spectrum over the signal bandwidth. Because the spectra of OFDM signals have a flat top and a sharp roll-off (the roll-off at the edges in the figure below is exaggerated for clarity) then the integration of power over the signal bandwidth can be approximated by multiplication of the OFDM signal bandwidth on the horizontal axis by the level of the power spectral density (measured in dB/Hz) on the vertical axis -- which is equivalent to using C/N figures.

The OFDM signal bandwidth, B, is equal to the subcarrier separation multiplied by the number of subcarriers. For DAB/DMB this is

B_DAB/DMB = 1 kHz x 1,536 = 1.536 MHz

For a 7 MHz DVB-H channel using 8K-subcarriers mode (8K mode uses 6,817 subcarriers) the same calculation is:

B_DVB-H = 977 Hz x 6,817 = 6.657 MHz

The difference between using 1 x 7 MHz DVB-H channel and 4 x DAB channels in dB is:

10 log (6.657 / (4 x 1.536)) = 0.3 dB

The DVB-H and DMB transmission modes chosen are the modes with the closest required C/N to DAB's required C/N using PL3, as shown in the following table:

Total transmission power is also a reasonable way to compare total transmitter network costs, because although DVB-H's peak transmission power will be higher than for DAB or DMB due to its wider bandwidth -- and hence each DVB-H transmitter will cost more than a DAB/DMB transmitter -- DAB/DMB requires 4 times as many transmitters and multiplex hardware as DVB-H. So, DVB-H transmitters would either be allowed to cost 4 times as much as DAB/DMB transmitters or a DVB-H network could use 4 times as many lower power transmitters and take advantage of SFN network gain. So differences in transmitter network costs are really a case of swings and roundabouts when comparisons are made with respect to transmitter powers. 

7.3  Number of Radio Stations in 7 MHz of Spectrum

The following tables show the number of radio stations at a given level of audio quality that can be carried in 7 MHz of spectrum, and shows the relative efficiencies of the systems with respect to DAB:

7.3.1  Number of Good Audio Quality Radio Stations in 7 MHz of Spectrum

For good audio quality, 192 kbps is required using MP2 and 64 kbps is required using HE AAC.

7.3.2  Number of Average (Minimum Acceptable) Audio Quality Radio Stations in 7 MHz of Spectrum

For average (minimum acceptable) audio quality, 160 kbps is required using MP2 and 48 kbps is required using HE AAC.

7.3.3  Number of Very Poor Audio Quality Radio Stations in 7 MHz of Spectrum

For very poor audio quality, 128 kbps is required using MP2 and 40 kbps is required using HE AAC.

7.4  Number of Radio Stations on Equivalent-Cost Transmitter Networks

As mentioned above, DVB-H transmitters require higher peak transmission powers due to the wider bandwidth of DVB-H multiplexes, whereas DAB and DMB require 4 times as many transmitters. Therefore, an alternative way of comparing the systems is by keeping the transmission powers of a single multiplex equal, because then an identical number of transmitters with identical (or virtually identical) transmitter powers can be used for each system, thus the transmitter network costs for each system will also be virtually identical.

For equivalent transmitter powers for each transmitter it is necessary to calculate a bandwidth correction factor for the bandwidth of one 7 MHz DVB-H multiplex (OFDM bandwidth of 6.657 MHz) relative to the bandwidth of one DAB/DMB multiplex:

Bandwidth correction factor (dB) = 10 log (6.657 / 1.536) = 6.37 dB

The following table shows the transmission modes with the closest total transmitter powers to a DAB multiplex using PL3:

The following table shows the number of radio stations that can be carried in one multiplex with approximately equivalent transmitter powers:

7.5  Comparison of Transmission Costs Per Service

Judging systems based only on the total investment required to roll-out a single multiplex does not tell the whole story, because such comparisons fail to take into consideration the number of services that different systems' multiplexes can carry. Therefore, to compare the relative costs of different systems it is better to do so on a transmission cost per service basis.

To compare transmission costs per service for the different systems it is easiest to make comparisons for the case when the transmitter networks cost the same for each system, so figures from section 7.4 above will be used.

The total costs incurred over the depreciation period of a network for equal-cost transmitter networks should be very similar, because very similar costs would be incurred for all of the fixed costs, and the only variable cost would be for the data collection network, whose costs are very small relative to the initial investment costs of the network itself along with all the multiplex hardware, premises, maintenance, staff costs etc. Therefore, it will be assumed that the total costs incurred for all of the systems will be the same.

Using the above assumption of equal total costs for all the systems over the depreciation period of the network, then the relative cost per service for each of the systems can be compared simply by comparing the relative efficiencies of the systems. As can be seen in the mathematical derivation in this pdf file, the relationship between the transmission costs per service for each of the systems is inversely-proportional to the efficiencies of each system. For example, using the figure from the above table, DVB-H is 12.9 times as efficient as DAB, and 1 / 12.9 = 0.08, or in other words, the transmission cost per service for DVB-H using QPSK with a code rate of 2/3 is 12.9 times cheaper than the cost per service for DAB. The following table shows the transmission costs per service relative to the cost per service for DAB:

For example, if you compare DAB with DMB, then to provide, say, 56 stereo radio stations in a given location at an average level of audio quality (160 kbps MP2 for DAB or 48 kbps HE AAC for DVB-H and DMB), then for DAB 8 multiplexes would be needed, whereas if DMB were used only 2 multiplexes would be required. So it stands to reason that the total cost for DAB will be 4 times that for DMB. 

8.  DMB over DAB Networks

The following figure shows how DMB services can be carried over existing DAB transmitter networks:

As can be seen from the above figure, the existing ensemble DAB multiplexer at the multiplex operator's offices, the COFDM modulators and the transmitters themselves do not need to be replaced, and only a single piece of DMB-processing equipment is required per service, as well as the replacement of the MPEG Layer 2 audio codecs by HE AAC audio codecs. This makes the transition from DAB to DMB a very cheap and low-risk option for broadcasters that have existing DAB transmitter networks, with the obvious benefit being that each multiplex will be able to carry 4 times as many radio stations as DAB can carry, and the transmission cost per service is also reduced by a factor of 4.

Changing from DAB to DMB thus hugely benefits both listeners and broadcasters.

9.  Conclusions

• DAB is not suitable for use as a digital radio system for the 21^st century because it fails to meet the fundamental requirements placed upon it -- it cannot provide a wide range of radio stations at a good level of audio quality -- and there is no longer any justification for any country to choose to use DAB because DMB and DVB-H are both vastly superior to DAB;
• DVB-H is the most efficient system, and because it is the most efficient system it is also the cheapest system to implement in terms of total cost per service;
• DMB is less efficient than DVB-H, but DMB can still carry almost 4 times as many radio stations as DAB, and is therefore almost 4 times cheaper to implement in terms of cost per service;
• The main problem with DAB is its use of the extremely inefficient and outdated MPEG Layer 2 (MP2) audio codec, which provides very poor audio quality at a bit rate of 128 kbps, and cannot even match the audio quality on FM at higher bit rates such as 192 kbps. Both DVB-H and DMB solve this problem by using modern and far more efficient audio codecs;
• The inefficiency of DAB's audio codec is compounded by its use of an outdated FEC coding scheme, which results in DAB being spectrally-inneficient. Again, both DVB-H and DMB solve this problem by using more modern technology.

10.  Recommendations

10.1  DMB Should Replace DAB

DMB can carry almost 4 times as many radio stations per multiplex as DAB can and, therefore, DMB is almost 4 times cheaper to transmit in terms of cost per service. Obviously, with spectrum being as scarce as it is, then using DMB instead of DAB vastly reduces the problem of having to reduce audio quality to very low levels -- like has happened in the UK -- if a wide range of stations are to be transmitted on digital radio.

DMB offers an attractive and very low risk option for replacing DAB for countries that already have a DAB transmitter network, because:

• DMB can use existing DAB network hardware without any modifications to the current multiplexers, modulators or transmitters;
• changing from DAB to DMB is an extremely cheap option;
• changing from DAB to DMB reduces the transmission costs per service by a factor of 4;
• using DMB will be far more profitable for commercial broadcasters than using DAB due to the higher number of radio stations that it can carry and the increased bandwidth available for data services;
• DMB offers multiplex operators far higher levels of profitability and shorter payback-period than DAB due to the far higher number of digital radio stations that DMB can carry;
• DMB allows broadcasters to provide a wide range of radio stations at a good level of audio quality, whereas DAB only allows a wide range of radio stations at low audio quality, or very few radio stations at good audio quality -- not both, unless there is a huge amount of Band III spectrum available;
• it will be far easier to sell DMB to the general public because it can provide a wide range of radio stations all of which can be at a good level of audio quality, and therefore there it is not necessary to be dishonest about the audio quality levels -- like the UK broadcasters are;
• T-DMB is already operational in Korea;
• numerous highly-integrated and low component count DMB modules and chipsets already exist, including from the same companies that have developed DAB modules and chipsets (e.g. Frontier-Silicon and Radioscape), and could be integrated into current DAB receiver designs very easily without, or with very minimal, changes to the operation of the devices, thus devices could be available very quickly -- Radioscape's DMB modules are pin-compatible with their DAB modules, therefore all DAB receivers that use Radioscape modules can be immediately changed to DMB use without any design modifications;
• DMB uses the same bandwidth as DAB, therefore DMB is flexible in terms of frequency planning, especially for local and regional multiplexes;
• because DMB is carried over DAB networks, then any data services developed for DAB can still be used on a DMB multiplex;
• global mobile phone sales were over 600 million units in 2004, so even if a small percentage of mobile phones include DMB receiver chips in future the cost of DMB receiver chips will be significantly lower than the cost of DAB receiver chips (because the cost of ICs is highly-dependent on volume), thus the bill of materials for DMB receivers will be lower than that for DAB, so DMB radio prices would be lower than DAB radio prices;

The one and only supposed problem with changing to DMB is the current lack of DMB receivers. But to see how quickly new receiver prices drop  you just have to consider that after DAB was launched (i.e. first advertised on BBC TV) in the UK in 2002 it only took 1.5 years for DAB radio prices to fall from £100 (~150 €) to £50 (~75 €) with sales of only about 250,000 per annum, which shows that affordable DMB receivers could be available very quickly.

DMB offers huge benefits with minimal drawbacks, and changing to DMB from DAB is very low risk and cheap to implement.

Because broadcast radio systems have historically been in operation for decades, then it is imperative that the system adopted works well. DAB no longer meets the fundamental requirements placed upon it in the 21^st century, whereas changing to DMB solves all of DAB's problems while providing listeners with a far better service while being far more profitable for broadcasters.

10.2  DVB-H Should be used for National SFNs Wherever Possible

As can be seen in the preceding sections, DVB-H is easily the most efficient system of the three, being approximately 50% more efficient than DMB and approximately 500% more efficient than DAB. Also, because of its significantly higher efficiency than the other two systems, its transmission costs per service are also significantly lower.

Its only drawback is that its channel bandwidth is significantly wider than DAB/DMB channels, which makes DVB-H less suitable for local and regional multiplexes which must be planned on a multi-frequency network (MFN) basis. DVB-H is, however, very well suited to being used for national or large-scale SFNs, and because DVB-H multiplexes offer very large capacities then a single national SFN DVB-H multiplex could carry all of a country's national radio stations as well as leaving a large amount of capacity left over, which could be used for data services. For example, as can be seen in section 7.3.1, a 7 MHz DVB-H multiplex using 16-QAM with a code rate of 2/3 can carry 136 stereo radio stations all with a good level of audio quality (181 radio stations can be carried if the audio quality is average), which is more national radio stations than any country would need. Therefore, the remaining space on the multiplex could be used for data services to enhance the service for listeners and provide the commercial broadcasters with higher profits.

As was mentioned in section 1.4, the guard interval for a 7 MHz DVB-H multiplex using a guard interval of T_u/4 is 256 μs, which is higher than DAB's guard interval (using transmission mode 1) of 246 μs, so national SFNs are obviously possible.

Introduction to Wi-Fi Internet radios