Maximising Efficiency of Band III Spectrum 

September 2004

Contents

1.   Current Digital Radio Situation

Currently there are 7 Band III channels in use in the UK for DAB. These 7 channels consist of 2 national multiplexes operated by the BBC and Digital One, and 5 other channels that are allocated to regional and local multiplexes. The Radio Authority's (now under Ofcom) plan was to provide access to at least 4 multiplexes for the vast majority of people -- 2 x national multiplexes, 1 x regional + 1 x local. London is a special case with access to both nationals plus 3 local multiplexes. Those that receive 4 multiplexes typically receive about 40 stations.

DAB sales have grown sharply since the BBC started their new services accompanied by a great deal of advertising on BBC TV and commercial radio to increase consumer-awareness, along with effort towards getting lower priced radios in the shops.

People welcome the extra choice of stations available compared to FM/MW, although many express concern over the fact that there is significant duplication of formats (for example, multiple pop stations and multiple rock stations), which reduces the breadth of choice available.

Many people are very disappointed with the audio quality on DAB in the UK, especially because it was -- and still is -- promoted as providing a significant improvement over FM in this regard. The problem with the audio quality is predominantly due to the insufficient bit rate levels used. 

The BBC DAB multiplex is full-to-bursting, with either Radio 3 and/or Radio 4 having to have their bit rates cut whenever Radio 5 Sports Extra goes on-air (with Radio 4 being reduced to mono); popular stations like BBC7 being permanently in mono even though a lot of the material was recorded in stereo; and all the music stations apart from Radio 3 using 128kbps, and falling far short of the mark as far as audio quality is concerned.

The commercial music stations virtually all use 128kbps, despite there being significant unused capacity on most commercial multiplexes. And yet the commercial broadcasters use the excuse that they cannot increase bit rates because there isn't space.

To make matters worse, GWR and Emap have both expressed that they would like the minimum bit rate for stereo music stations to be reduced. Nick Piggott of GWR Digital said the following on the subject of DAB bit rates in a speech to the Radio Academy Conference in 2002:

"We may make the case to go down lower again if we can show that the audience can’t tell – that will almost certainly require the broadcasters to keep the audio cleaner for longer."

and someone from Emap admitted to me (just after Emap had reduced some of their music stations to 96kbps mono) that they hoped Ofcom would lower the minimum bit rates for stereo stations so they could change the mono stations to stereo. I kid you not!

Ofcom have given Digital One permission to use the remaining unused space on the multiplex to provide subscription-based data services, which will result in almost 20% of the multiplex being used for data services, and other commercial radio groups are interested in exploiting the data opportunities on DAB. 

2.   Regional Radio Conference (RRC-04/06)

The purpose of this conference is well summed-up in the recent response from the Irish communications regulator, ComReg:

"In May 2004 the International Telecommunication Union held the first (RRC-
04) of two regional planning sessions to address the technical basis for the
planning of digital radio and television services in the VHF and UHF bands.
This is a major programme to bring digital technology to the predominantly
analogue environment in which terrestrial broadcasting currently operates.
RRC-04 also established the work programme for the period between sessions
and the timescales for the work involved.

The work that must take place in the intersessional period requires clear
national broadcast policy to form the framework around which the national
preparations will revolve.

The current plan for analogue broadcasting in the VHF band III dates back to
1961 and has been used successfully for more then 40 years. It can be
expected that any decision made on national broadcast policy following on
from this consultation, will have a lengthy validity."  

The conference will conclude in Geneva in May/June 2006, but the next work to be carried out is:

"On 28 February 05, all administrations will send their requirements (coverage
areas, number of channels, type of service, fixed, mobile, portable…) to the ITU."

and, to this end, Ofcom will be holding a public consultation / review of digital radio some time in Autumn 2004. 

It is generally expected that the UK will apply for, and be allocated, 5 new DAB channels (10A to 11A) in Band III. 

3.   Audio Quality

3.1   Audio Quality on DAB

If the UK is to be provided with audio quality that is on a par with an average FM signal -- which should be a minimum requirement prior to FM being turned off -- then stereo music stations need to be transmitted at at least 160kbps using DAB's MP2 codec. This has obvious implications for the current situation because as of April 2004 there were 283 instances of stations using 128kbps, so if we are going to have 160kbps as the minimum bit rate for stereo stations on DAB then this alone would require a significant amount of additional capacity just to increase bit rates from current levels. 

I propose there to be a public consultation into what people would like the level of audio quality to be on digital radio -- not what the broadcasters would like to provide us with. And if the general public ask for near CD-quality, then near CD-quality they should be provided with, not sub-FM audio quality but no hiss. This is the 21^st century, after all.

Following the above research and consultation, Ofcom should then come up with a plan to ensure that the audio quality people would like to receive on digital radio be provided. 

The vast majority of DAB receivers in the market today are mono portable radios with a small speaker. Therefore, any "market research" that tries to suggest that the audio quality on DAB is good enough can safely be disregarded, because these devices are incapable of producing hi-fi sound, and the results do not represent an accurate picture of "normal" radio listening. For example, younger people are far more likely to listen to radio at home on a micro/midi/mini or full-size hi-fi system than on a small, mono portable radio, and whereas audio quality on a kitchen radio is not particularly important, audio quality on a hi-fi system is. And how can you tell whether the public accepts the use of mono to transmit music stations on DAB when the vast majority of DAB receivers are mono?

Also, the audio quality of the lower-end hi-fi systems in the market look set to improve markedly due to recent advances in a number of audio signal processing techniques that will be integrated into systems over the next few years, and this will reveal just how low the audio quality of DAB is compared to that of CD. 

The issue of the audio quality on DAB is likely to run for a very long time, because there's never before been a new radio system introduced that provides lower audio quality than its predecessor. It's also a sad indictment of just how low audio quality standards have dropped on DAB that I'm only asking for a 160kbps minimum bit rate for stereo stations, when a good quality FM station would usually be of very similar quality to a 192kbps station on DAB.

It's also interesting to note that MP3 is approximately 25% more efficient at compression than MP2 (as used on DAB), and yet the de facto standard bit rate for MP3 files is 128kbps, which is the equivalent of 160kbps MP2. And yet 98% of all stereo music stations on DAB use 128kbps MP2.

If Ofcom want to be seen to be credible then they have to carry out serious research, undertaken by respected and experienced people with no vested interest in the results. 

3.2   A New Method of Regulating Bit Rates on Digital Radio is Required

The regulation of digital radio by the Radio Authority obviously failed. This is clearly demonstrated by the fact that 98% of stereo stations use the minimum bit rate of 128kbps -- a level only half as much as the 256kbps that the BBC originally intended to use! 

It was extremely naive of the Radio Authority to set a minimum bit rate as low as possible and hope that the commercial radio groups to wouldn't all descend to the lowest they can get away with when there was virtually no listeners. Abusing the trust of the listeners doesn't bother the commercial radio groups one iota, so long as they can get away with it, and so long as the profits keep rolling in. 

Ofcom's proposal of regulating audio quality by means of "diff-grades" is a good idea, but to suggest that a diff-grade of 2 points is acceptable is an absolute disgrace, as can be seen by the other technologies that provide superior audio quality to this level. One example is that the HE AAC codec at 32kbps can even provide significantly higher audio quality than a diff-grade of -2.0.

Suggesting a diff-grade of -2.0 without doing any credible market research into what is the best level of audio quality to use is a sloppy, non-fact-based way of regulating audio, and is letting the general public down very badly indeed. 

4.   Scenario 1 - DAB-only

4.1   Current & Planned Band III Usage

The UK currently has 7 Band III channels:

Click on map to enlarge

Map of Local & Regional DAB Multiplexes

The above map is a bit out of date, because it is dated from November 2002, but it does show the majority of local multiplexes "penciled-in" (for example, Reading), some of which have started transmitting since that map was created, others have not.

The only local multiplexes that are now planned, that the above map does not show, are as follows:

• Cambridge

• East Lincolnshire

• Kings Lynn

• Luton Town

• Stirling

Filling-in-the-blanks on the above map where there's a multiplex planned, then the following areas are not covered with a local, or regional multiplex, and it's assumed that there's sufficient population to make a multiplex feasible (arbitrarily labeling the new channel number 1):

As can be seen in the above table, one additional channel is required to cover the remaining "white space" area on the map apart from very sparsely-populated areas, and a local multiplex is unlikely to be economically-viable in this regions.

Because one additional Band III channel can cover the remaining "white space" on the map where it is feasible to have local DAB multiplexes, I propose the following use for the 5 new Band III channels:

• 3 Band III channels used for national multiplexes

• 2 Band III channels used for local multiplexes (with priority given to those listed in the above table) across the country

• L-band used for small-scale local multiplexes

4.2   National Multiplexes

All of the large commercial radio groups apart from GWR (GWR own 67% of Digital One and are opposed to any more national multiplexes, despite the fact that it would significantly harm DAB's development) expressed a desire for there to be more national DAB multiplexes in their responses to the Band III public consultation held by Ofcom in January 2004:

• Emap expressed a desire for 4 to 5 additional national Band III DAB multiplexes
• MXR -- representing Chrysalis, Capital, GMG and UBC -- expressed a desire for 2 additional national Band III DAB multiplexes
• SMG expressed a desire for "at least one" additional national Band III DAB multiplex

The reasons cited for expressing a desire for additional national multiplexes included:

• strong consumer demand for national stations
• maximum number of people benefit from the additional choice of stations
• national radio brands could be moved off local multiplexes and on to national multiplexes, thus improving efficiency
• additional capacity can be provided to increase bit rates so that consumers are not disappointed with the audio quality on DAB
• greater advertising revenue from national stations
• strong demand from advertisers for national stations
• BBC could be given additional capacity
• abolishment of GWR's monopoly position whereby they own the majority share of the only national commercial multiplex operator
• public served better by national than by regional or local multiplexes
• local and regional data revenue opportunities are limited

Assuming that 4 Band III channels are allocated for national use in the UK, this allows direct comparison between DAB and DVB-H, with DVB-H using the 7 MHz TV channels. The Band III DAB channel frequencies were allocated so that they could co-exist with 7 MHz TV channels in Band III, and 4 DAB channels fit into the space of one 7 MHz TV channel.

4.3   Analogue Stations on DAB

Using figures from Ofcom's Communications Market 2004 - Radio report, the BBC website and from this website, there are the following number of analogue stations on DAB:

Only 45% of the 325 analogue stations are on DAB at the present time. 

If only 5 Band III channels become available for DAB in the UK then it is obvious that extensive use of L-band will be necessary. However, L-band should be used as sparingly as possible due to the relatively poor RF propagation characteristics in that band, which implies a higher transmitter density to achieve a given coverage area, and therefore higher network infrastructure costs.

Unfortunately, this implies that the smallest radio stations will end up on the most expensive multiplexes to build and operate. If the big commercial radio groups carry on owning the multiplex operators -- as they do now -- this could lead to the undesirable situation where the big commercial radio groups can price the smaller stations off DAB, or force them to be taken-over. 

To avoid the above situation, as many stations should be put on Band III as possible -- without degrading the audio quality -- and carriage charges for L-band multiplexes should be regulated so that the smaller stations do not get ripped-off.

To use L-band as sparingly as possible requires Band III spectrum to used as efficiently as possible, which implies implementing large-scale -- national wherever possible -- SFNs (single-frequency networks).

4.4   Requirements for New National Stations

The following table shows a list of current quasi-national stations, the number of DAB multiplexes they're carried on, and whether they're carried on the Freeview and Sky digital platforms.

Due to all the large commercial radio groups apart from GWR expressing a desire for extra national commercial multiplexes to provide consumers with better choice nationwide, the following stations are the most likely stations to be carried on any new national multiplexes because of the extent of their current coverage.

Of course, if GWR are so opposed to any new national multiplexes then they have the option not to put the 2 stations (Classic Gold and The Storm) they own in the table on any further national multiplexes.

Also, if GWR continue to throw tantrums about other groups being allowed to transmit nationally, then a degree of innovative frequency planning would allow those groups that wish to transmit their stations nationally to do so, although not on a single, national SFN. For example, if, say, Emap became the multiplex operator of one of these quasi-national multiplexes, then they could use one frequency for England, Wales and mainland Scotland, and use a different frequency in the Outer Hebrides. This would then not be a national SFN, because it's on 2 frequencies.

4.5   Requirements for BBC Stations

The BBC should be given adequate capacity to transmit all of their stations at bit rates at least -- and in some cases higher -- as high as they are on Freeview and digital satellite:

* usually not transmitting, so bit rates not included in table 

4.6   Requirements for Digital One Stations

The following additional bit rates are required to provide a minimum bit rate of 160kbps for the stereo stations and to return Classic FM and Virgin Radio to the bit rate they previously transmitted at:

4.7   Overall Requirements

Almost all UK DAB multiplexes use error protection level 3 (PL3), which has an absolute maximum capacity of 1184 kbps. Therefore, the number of national multiplexes that would be needed to provide 5504 kbps would be:

5504 / 1184 = 4.65 multiplexes

Therefore, 5 additional national multiplexes would be required.

However, a beneficial side-effect of changing quasi-national stations to truly national stations is that it frees up the spectrum used for the regional multiplexes, as well as a large proportion of local multiplex capacity. Therefore, if 5 Band III DAB channels are all allocated for national-use in the UK, local multiplexes can use the spectrum in areas currently used by the regional multiplexes on channels 11D, 12A and 12C. 

Two very significant drawbacks of the above scheme, though, are that:

• even using 5 national DAB multiplexes only provides about 400 kbps for future national stations or further data services, which is totally insufficient
• L-band will still have to be used extensively to accommodate the 55% of analogue stations not yet transmitting on DAB

4.8   Scenario 1 - DAB-Only Conclusions

If only 5 new Band III channels were allocated for DAB then this capacity could be taken-up by quasi-national stations taking-up places on national multiplexes alone. The space made available as a result of this move would still require heavy-use of L-band, which is undesirable.

Due to the low initial cost of starting new digital radio stations (I remember reading that someone from Emap said that the initial cost was as low as £100,000 per station), then it should be expected that we will see a significant number of new start-up stations in the coming years as the number of digital radio listeners continues to grow -- especially as we move towards analogue radio switch-off. Because the allocation of 5 Band III channels to DAB provides only a very limited scope to expand the number of national stations transmitting (beyond the quasi-national stations changing to national stations), it is highly likely that a repeat of the current bandwidth shortage looms at some point in the future -- leading to reduced bit rates and audio quality significantly lower than FM. Therefore, alternative, higher-efficiency solutions, like DVB-H, should be investigated seriously with a view to the long-term success of digital radio.

If a DAB-only system is to be used in the UK, then extensive use of L-band will be required, which will require approximately 95% or more of current DAB-owners to purchase new DAB receivers. L-band requires higher investment in network infrastructure due to higher transmitter density, which will lead to higher transmission and maintenance costs for all stations on such multiplexes. Again, it makes sense to investigate the feasibility of using higher-efficiency alternative solutions, like DVB-H, to minimize the use of L-band; minimize total investment required, and minimize the transmission and maintenance costs.

5.   Scenario 2 - New DVB-H Multiplex

5.1.1   DVB-H System Requirements

The new DVB-H ('H' stands for handheld) specification was designed specifically to provide suitable transmission modes for handheld devices such as mobile phones and PDAs. The DVB-H specification provides additional modes on top of the modes available for DVB-T, and DVB-H can use any of the standard DVB-T modes. Some of the objectives that it was designed to address were as follows:

• highly robust reception at high speed in a mobile environment
• reception with a single, small antenna
• low power requirements suitable for mobile phones

Because the requirements are more severe for mobile phones than for typical digital radio use, this makes DVB-H highly-suited for digital radio use.

For those not interested in the comparison of the technology used on DAB and DVB-H then skip the following 2 sections and read from the Audio Codec section onwards.

5.1.2   Low Power Consumption

With the same level of spectral efficiency (bit rate / bandwidth), DVB-T already significantly outperforms DAB in the mobile environment. The problem with DVB-T though is that its power consumption was deemed to be too high for mobile phone batteries, so DVB-H has introduced "time-slicing", where transmissions for one service occur over a short (burst) period at high bit rate, then that channel ceases transmitting (the other channels get their "time slice"), which allows the DVB-H receiver to switch off until it is time to receive the next time-slice. This achieves power consumption reduction of greater than 90% compared to DVB-T receivers, and is the main reason why DVB-H is feasible for inclusion in mobile phones.

5.1.3   Forward Error Correction (FEC) Coding Scheme

From the point of view of comparing systems, the main attraction of DVB-H is that it allows at least double the spectral efficiency that DAB provides, and does so at a very high level of robustness. The reason it is able to provide very high robustness at such high spectral efficiencies is due to DVB-H's use of a very powerful 3-layer FEC (forward error correction) coding scheme compared to DAB's weak single-layer FEC coding scheme. 

It's instructive to compare DAB's, DMB's and DVB-H's FEC coding schemes:

The inner convolutional coding schemes used for DAB, DMB and DVB-T/H are very similar, thus very similar error correction performance will be achieved when the same FEC code rate is used on DAB or DVB-T/H, i.e. if a code rate of 1/2 is used on both DAB and DVB-T/H then you could expect the error correction performance to be almost identical. The code rate is the ratio of useful (e.g. actual MPEG data) bits to the number of output bits from the coder. For example, for a code rate of 2/3, 2 useful bits are encoded into 3 output bits, therefore one redundant bit is added to aid error correction at the receiver.

DVB-T, DVB-H and DMB then use a Reed-Solomon (RS) FEC coding scheme whose job is to catch uncorrected errors which the convolutional coding was unable to correct. RS coding is very powerful for correcting bursts of errors, and errors typically occur in bursts on wireless digital communication systems. RS coding is a form of "block coding" (as opposed to convolutional coding, which is the other main type of FEC coding), where the encoder takes a block of data and encodes it all at once. The inner RS coding scheme used for DVB-T/H takes 188 bytes of data and adds 16 parity bytes, so the code rate is 188/204. Despite the fact that the percentage of redundancy (the parity bytes) added is only 8%, the RS coding can correct up to 8 bytes of data out of each 204 byte codeword, irrespective of whether 1 bit or all 8 bits of a byte are in error. For example, if there is a very high concentration of errors that are spread over 8 consecutive bytes, then all the 8 errored bytes can be corrected. Hence the reason why it is deemed to be powerful for correcting bursts of errors.

To show how powerful this RS coding is, the worst bit error rate (BER = number of bit errors / total number of transmitted bits) at the output of the convolutional decoder that DVB-T can handle before the signal "fails" is 2 x 10^-4, which is equivalent to 1 bit error out of every 5000 bits, on average. At this bit error rate, the RS coding can reduce the overall bit error rate to between 10^-10 to 10^-12, which is called quasi-error free (QEF) reception. At a typical MPEG-2 video bit rate of 4 Mbps, a bit error rate of 10^-10 equates to just one uncorrectable bit error about every 42 minutes, and in that time the RS FEC decoder will have corrected over 2 million errors (800 errors per second)!

Clearly, it's a shame that DAB doesn't use RS coding, because if it had then reception would have been far more robust than it has actually turned out to be, and for the sake of just 8% redundancy it certainly would have been a price worth paying. Certainly, reception problems like the infamous "bubblind mud" sounds and poor reception on personal DAB radios would have been vastly improved if RS coding had been used.

DVB-H has a further layer of RS coding, called MPE-FEC (multi-protocol encapsulation - forward error correction), where typically 25% of redundancy (parity bytes) is added to provide very highly-robust reception even at very high speeds in a mobile environment and using high-order modulation, such as 16-QAM. The RS code used takes 191 bytes and adds 64 parity bytes to make a 255-byte codeword. A cyclic redundancy check (CRC) is used at the receiver to check if MPE-FEC packets contain any errors. If so, then the MPE-FEC packet is marked as unreliable. Using this reliability information allows the RS code to correct double the number of bytes that it would have been able to correct ordinarily and, in this case, any 64 errored-bytes out of a 255-byte codeword can be corrected.

The only feature that DAB's transmission scheme has which DVB-T doesn't -- time-interleaving -- has also been added to the DVB-H spec. Overall though, it is the vastly more powerful FEC coding scheme that DVB-H provides that makes it so superior to the DAB transmission scheme.

5.1.4   Audio Codec

DVB-H can use the modern audio codecs AAC and HE AAC. The following table shows bit rates for AAC and HE AAC that provide approximately the same level of audio quality as DAB's codec, MP2:

(HE AAC is only used at low bit rates)

5.2   Multiplex Capacity

DAB multiplexes have a bandwidth of approximately 1.7 MHz, thus 3 contiguous DAB channels has a bandwidth of approximately 5.1 MHz. Therefore, a 5 MHz DVB-H multiplex would fit into 3 contiguous DAB channels in Band III.

The capacity for a DVB-H multiplex using 16-QAM modulation, MPE-FEC (25% redundancy) in a 5 MHz channel with a guard interval of Tu/8 = 179µs (to allow a national SFN) and convolutional coding code rate of 3/4 is 7.776 Mbps.

The maximum capacity for 3 DAB multiplexes using error protection level 3 -- as used on virtually all UK multiplexes -- is 3 x 1.184 Mbps = 3.552 Mbps.

5.3   Possible Number of Stereo Stations

Using the above information to calculate the number of stereo stations that can be fit into 5.1 MHz of spectrum:

1 -  'same quality' denotes approximately equivalent audio quality to 160kbps MP2

2 -  'higher quality' denotes approximately equivalent audio quality to 192kbps MP2

Just to put those figures into perspective, 'same quality' equates to significantly better quality radio stations than the 128kbps stations on DAB we have today.

5.4   Scenario 2 - New DVB-H Multiplex Conclusions

The most significant conclusion of using a national DVB-H SFN multiplex is that it can carry over 5 times as many stereo stations (at the same level of audio quality) as 3 national DAB SFN multiplexes can carry when occupying the same total bandwidth. This would allow the number of national digital radio stations to significantly increase in years to come and would allow a significant amount of bandwidth to be used for data services. Comparing this with the situation if 3 national DAB multiplexes were used, there is almost no opportunity to increase the number of national stations once the quasi-national stations are moved on to truly-national stations, and nor will there be any opportunity to increase the bandwidth used for data services as calculated above. 

Another significant advantage of using DVB-H rather than DAB, is that the network infrastructure costs for one national DVB-H multiplex will obviously be significantly lower than the costs required for 3 national DAB multiplexes. This would obviously lower the carriage charges, which would encourage more stations onto the national DVB-H multiplex, thus freeing-up more capacity on local multiplexes to allow more stations to get onto DAB. This has obvious benefits in that less L-band channels would have to be used, thus lowering the overall investment in network infrastructure required to allow all current analogue stations to get onto DAB.

One obvious result of using a national DVB-H SFN multiplex in Band III would be that people that already own a DAB receiver would not be able to receive any DVB-H signals. However, if there is only 5 available Band III channels then L-band would have to be used extensively to provide sufficient capacity, which would also require people to replace their current Band III-only DAB receivers. Therefore, any arguments against the use of DVB-H, due to current receivers not being able to receive DVB-H signals are unfair on DVB-H, because 95% of current DAB receivers will not be able to receive any L-band signals either. 

Also, there should be no concerns over the power consumption requirements of a dual-mode DVB-H/DAB receiver, because most of the hardware/software can be shared by both DVB-H and DAB demodulators, because they contain common blocks, such as an OFDM demodulator and a Viterbi convolutional decoder. Also, as described above, DVB-H has been specifically designed to be as power-efficient as possible, and DVB-H receivers consume less power than DAB receivers, which would actually increase the battery-life of portable digital radios.

One migration plan would be to wait until DVB-H/DAB dual-mode receivers were sufficiently cheap (say, once portable dual-mode radios cost £25) before the Digital One and BBC stations were moved onto the national DVB-H multiplex, and then the frequencies for the Digital One and BBC multiplexes could be re-used for local stations.

If it sounds like it might take a long time before dual-mode DVB-H/DAB receivers could be made cheaply, then you have to consider that because of the extremely high volumes that mobile phones sell at, as soon as DVB-H is added to mobile phones, DVB-H chips will be extremely cheap, so DVB-H/DAB dual-mode receivers could be manufactured just as cheaply as DAB receivers.

A significant benefit to the digital radio industry from using dual-mode DVB-H and DAB receivers is that this would vastly increase the likelihood of mobile phone manufacturers including DVB-H/DAB chips in their phones (because the mobile phone manufacturers are all very keen on DVB-H), which would unlock the huge potential of people listening to digital radio stations on their mobile phones. 

6.  Scenario 3 - DMB

Digital Multimedia Broadcasting (DMB) is a system that has recently been designed in Korea, and is based on the Eureka 147 DAB system. The significant differences between DMB and DAB are as follows:

• allows use of the new BSAC (bit-sliced arithmetic coding) codec which provides efficiency equivalent to AAC
• adds outer-layer of Reed-Solomon FEC coding to improve signal robustness

The outer-layer of Reed-Solomon FEC coding is identical to that used on DVB-T and DVB-H. Therefore, DMB solves 2 of the main problems with the DAB system: it replaces the inefficient MP2 codec and adds an outer-layer of Reed-Solomon FEC coding, which significantly improves robustness compared to the weak, single-layer FEC coding used on DAB.

However, because it is based on the DAB system, it is still limited to using QPSK, which only allows 2 bits to be carried per subcarrier, whereas DVB-H typically uses 16-QAM, which allows 4 bits to be carried per subcarrier. Therefore, the spectral efficiency of DMB is virtually identical to that of DAB, and therefore only about half as spectrally efficient as DVB-H.

DMB typically uses DAB's equal error protection (EEP) level 3A (PL3A), which allows the same maximum capacity as DAB does using its unequal error protection (UEP) level 3 (PL3). However, because of the outer-layer of FEC coding, the useful capacity is slightly lower than for a DAB multiplex using PL3. A DMB multiplex using PL3A has a maximum capacity of:

Max capacity = 1184kbps x (188 / 204) = 1091 kbps

An advantage of using DMB is that current DAB transmitters could still be used, and only the multiplexing hardware / software would need to be changed. 

DMB also has a frequency-planning advantage for local multiplexes compared to DVB-H due to the narrower 1.7 MHz channels.

7.   Comparison of DAB, DVB-H and DMB

As shown previously, the systems will be compared to see how many stereo stations they can carry in 5 MHz of Band III spectrum, which equates to 3 x DAB channels, or 1 x 5 MHz channel for DVB-H:

1 - 'same quality' denotes equivalent audio quality to 160kbps MP2 (using 64kbps HE AAC)

2 - 'higher quality' denotes equivalent audio quality to 192kbps MP2 (using 96kbps AAC)

3 - DMB uses BSAC (bit sliced audio coding), which provides performance equivalent to the AAC codec

An alternative way of looking at these figures is to see how many DAB or DMB multiplexes are required to deliver the same number of stereo stations at an audio quality equivalent to 160kbps MP2 as the 153 stations that can be carried on one DVB-H multiplex:

8.   Overall Conclusions

DAB was designed in the late-1980s and early-1990s, and despite there being no DAB receivers available until 1999 (which, incidentally, cost £800), the specification was not changed, and therefore ignored the vast improvements in digital communications and audio compression that were going on around it. Additionally, the 1990s saw a huge growth in the number of commercial radio stations on-air in the UK, and with the introduction of digital radio in the late-1990s there has been a further expansion in the number of radio stations transmitting. Also, due to the relatively small cost of starting new digital radio stations, it can be expected that a significant number of new stations will be created in the coming years as the number of listeners to digital radio grows.

Currently, only 45% of all analogue radio stations are transmitting on DAB. Therefore, a significant amount of spectrum is needed to allow the remaining analogue-only radio stations to migrate to digital radio. If DAB continues to be the only system used for mobile terrestrial digital radio then a large amount of L-band spectrum will have to be used to accommodate the stations not currently on DAB.

Using L-band will significantly increase the overall required investment in network infrastructure due to the poor RF propagation characteristics in that band, and the resulting increase in transmitter-density to cover a given area compared to Band III. Therefore, L-band use should be avoided if at all possible.

New systems such as DVB-H and DMB provide far superior efficiency compared to DAB. DVB-H is the most efficient system of the 3 due to its combination of high modulation order (16-QAM), modern FEC coding and ability to use the new, highly-efficient audio codecs AAC and HE AAC. DVB-H's only drawback for digital radio is that it can only use relatively wide channels (5, 6, 7 or 8 MHz bandwidth -- 5 MHz was added to the DVB-T/H specification in June 2004). DMB is significantly more efficient than DAB due to its ability to use the new, highly-efficient audio codecs AAC and HE AAC, but it also provides higher-robustness than DAB due to its additional layer of Reed-Solomon FEC coding.

If a new digital radio system were to be implemented from scratch today the best combination of systems to use would be:

• DVB-H multiplex for national services
• DMB used for local services

DAB is simply far too inefficient by modern digital broadcast system standards.

The UK is in the position of being the most "advanced" DAB market in the world. Tessa Jowell is very proud of this fact. Ofcom are very proud of this fact. The BBC Digital Radio department is very proud of this fact. Ironically, it is precisely because we are the most "advanced" DAB market that it is looking increasingly likely that we will end up with one of the most backward digital radio networks in the developed-world.

Some other European countries have significant DAB area coverage, but they don't have high numbers of DAB receiver owners. Therefore, although changing system would mean writing off the investment in some transmitter hardware and annoying a few of the earliest adopters, the opportunity-cost foregone by not moving to one of the far more efficient systems drastically outweighs the loss incurred on DAB transmitter investment, that country's administrations and broadcasters should analyze carefully whether it is best to stick with DAB at all. The UK is the only European country that doesn't have this option to quit DAB now and start afresh, and in the long-term we will pay the price for foolish, politically-motivated, desires to be first.

We've already got the lowest audio quality on DAB in the entire world. Even the IBOC FM system in the US -- which has been much-maligned by advocates of DAB -- is providing far superior audio quality to that provided in the UK, or, indeed, are ever likely to get by the way things seem to be going. 

Ofcom only has this single opportunity to significantly improve the UK's terrestrial digital radio system, and any poorly-made decisions now are likely to condemn the UK to having a significantly sub-standard system for decades to come.

9.   Recommendations

1. 3 Band III channels should be allocated to allow 1 x 5 MHz national DVB-H SFN multiplex to be implemented

2. If DVB-H is not used, then 3 Band III channels should be used for either DAB or DMB -- preferably DMB, due to its superior spectral-efficiency compared to DAB

3. The BBC should be allowed additional capacity on the national multiplex to allow increased bit rates for their services

4. 2 remaining Band III channels should be allocated for local use

5. DMB should be used for all future local multiplexes in L-band

6. Use Band III spectrum as spectrally-efficiently as possible to minimize the use of L-band

7. Consultation should be held asking the public what level of audio quality they would like to be provided with on digital radio

8. Research commissioned into audio quality provided by digital radio codecs and, drawing from results of aforementioned consultation, minimum bit rates set accordingly

Introduction to Wi-Fi Internet radios