Introduction

This article was written in response to an article in the EBU Technical Review that concluded that it is cheaper for Swedish Radio to transmit their radio stations on DAB rather than DVB-H. This article investigates the validity of this claim.

The article was written by a Swedish Radio employee, and was apparently written some time ago as part of Swedish Radio's (failed) attempt to gain funding from the Swedish government to roll-out DAB across the whole of Sweden. Therefore, if you read the article, you should be aware that there was a vested interest in the results favouring DAB.

It should also be noted that the distribution costs are quoted without quoting any of the DVB-H transmission parameters or the transmitter network configurtion. The transmission parameters and network configuration used have a major bearing on the network costs, so they should have been at least mentioned in a technical article, but we will have to simply trust that the parameters were chosen fairly despite the fact that Swedish Radio were trying to make the case for using DAB against some pressure from other parties that wanted Sweden to use DVB-H for digital broadcasting.

1.  EBU article's results

The following figures are copied from the EBU article, and show the estimated distribution costs per radio stations on DVB-H and DAB:

Figure 1 in Economic Analysis of DAB & DVB-H article - Distribution costs per radio station

Figure 2 in Economic Analysis of DAB & DVB-H article (zoomed in to the range where the number of stations is greater than 6)

Looking at both Figures 1 and 2 together, to arrive at the distribution cost per radio station the author of the EBU article has divided the total distribution cost of the DVB-H network by the number of stations, where he has assumed that each radio stations uses 192 kbps. For example, looking at Figure 1, if only one radio station is transmitting on either the DAB or DVB-H multiplexes that radio station has to bear all the distribution costs, and therefore the distribution cost for that station equals the total distribution costs for the whole multiplex. Then looking at the DVB-H (red) curve in Figure 2, the author of the DVB-H article has divided the total distribution cost of the multiplex (€45 million) by the number of services transmitting:

If you cross-check the figures in the third column of the above table with the values on the DVB-H (red) curve in Figure 2 you will see that they correlate precisely, within the accuracy that you can read figures from the graph. Therefore, this method of calculating distribution costs will also be used below.

2.  Incorrect assumptions in the EBU article

The article makes two incorrect assumptions, and these assumptions have a major impact on the results stated in the entire article and, therefore, on the conclusions given. The incorrect assumptions are as follows:

1. On page 2 it says: "The calculations are based on the assumption that a typical radio channel is broadcast at 192 kbit/s."
2. The footnote on page 2 says: "The costs per multiplex for national coverage were calculated by Progira Radio Communication, assuming DAB in the VHF band and DVB-H on UHF with 5.5 Mbit/s per multiplex."

2.1  Assumption 1

The assumption that 192 kbps is used for a radio station assumes that both systems use the same audio codec, which is false: DVB-H uses the AAC and HE AAC (AAC+) audio codecs, whereas DAB uses the MP2 (MPEG Layer 2) audio codec.

Interestingly, however, he does mention audio coding efficiency in his conclusions, which makes you wonder why he didn't include its effect in the results he presented...

In two independent blind listening tests, one carried out by the Canadian Communications Research Centre (CRC), and the other carried out by the BBC, NHK and MIT Media Labs, AAC was found to be twice as efficient as MP2. What this means is that AAC only needs to use half the bit rate that MP2 needs in order to provide a given level of audio quality. The following figure shows the results from one of these blind listening tests:

In the above figure, the curve for MP2 is labelled as 'LII', and you can see that AAC at 96 kbps provides an almost identical audio quality level to MP2 at 192 kbps. 

Therefore, the assumption in the article that 192 kbps should be used for both systems is false, because if 192 kbps needs to be used on DAB then only 96 kbps needs to be used on DVB-H to provide the same level of audio quality. This translates into the distribution cost per radio station on DVB-H being reduced by half, because the fixed total multiplex costs can be spread over twice as many radio stations. I will give more details on this below.

2.1.1  Correction to assumption 1

The correction to this assumption is as follows: twice as many radio stations can be carried on a DVB-H multiplex compared to the figures quoted in the EBU article.

For example, the following graph shows the effect of correcting assumption 1 only:

Looking at Figure 2 in the EBU article (shown above), you can see that the author of the EBU article has indicated that DVB-H is cheaper once the total market demand is greater than or equal to 7,200 kbps. This equates to 38 x 192 kbps radio stations, but the graph above shows that correcting for assumption 1 alone makes DVB-H cheaper than DAB if the number of stations is greater than or equal to 19 — a big difference. This difference is due to the author of the EBU article deciding to build a second DVB-H multiplex once the capacity is greater than 5,500 kbps (refer to the step-change increase of the DVB-H curve on Figure 2), when in fact it is not necessary to build a second multiplex, because 5,500 kbps of capacity can carry 57 x 96 kbps AAC radio stations.

2.2  Assumption 2

The assumption that DVB-H must use multiplexes with a capacity of 5,500 kbps overestimates the transmitter powers that DVB-H must use when the required capacity levels are much lower than 5,500 kbps. 

At multiplex capacity levels much lower than 5,500 kbps DVB-H is expected to use a multiplex where quite a high proportion of its capacity is simply unused. For example, at a total capacity level of, say, 2,500 kbps, the DVB-H multiplex would be 55% empty. In this case, it would obviously be better to use a lower capacity transmission mode for DVB-H in order to reduce the transmitter powers required for this multiplex. In turn, this reduces the transmitter network costs, thus reducing the distribution costs accordingly.

2.2.1  Correction to assumption 2

The correction to this assumption is to use lower capacity DVB-H transmission modes when the number of stations is low in order to reduce the transmitter powers, thus reducing the distribution costs. 

DVB-H Transmission Modes

The following table shows the relevant DVB-H transmission modes, along with additional information, such as C/N (carrier to noise ratio) values (transmission parameters assumed: MPE-FEC is used with code rate of 0.75, guard interval = 1/8, 8K subcarriers, 8 MHz channel):

(All DVB-H C/N and capacity figures  taken from the DVB-H Implementation Guidelines (ETSI TR 102 377))

The closest transmission mode to the 5,500 kbps multiplex capacity quoted in the EBU article is QPSK code rate (CR) 2/3.

Drawing on the assumption made in the "Broadcasting to Handhelds" EBU Technical Review article that the transmitter network cost follows an almost linear relationship with total transmitter power (in this document transmitter powers always refer to ERP (effective radiated power) values), the last two columns in the above table show the difference in transmitter powers required due to the difference in C/N between transmission mode QPSK CR 2/3 and the other transmission modes.

For example, if the total capacity requirement is less than or equal to 4,148 kbps then the QPSK CR 1/2 transmission mode can be used. And because the required C/N for this transmission mode is 3 dB less than for the QPSK CR 2/3 transmission mode assumed in the EBU article, the transmitter powers (and hence the distribution costs) must be multiplied by 0.5 (a reduction of 3 dB equates to halving the power in linear units). Alternatively, if the total capacity requirement is between 5,528 and 8,295 kbps the 16-QAM CR 1/2 transmission mode can be used and the transmitter powers must be multiplied by 1.78, and so on. 

The following graph shows the distribution costs per station for a single DVB-H multiplex where the transmission mode is changed according to required market capacity versus the distribution costs per station values used in the EBU article where the DVB-H multiplex capacity is fixed at 5.5 Mbps and when capacity requirement exceeds 5.5 Mbps a new multiplex is built:

The red curve shows the distribution costs for a single DVB-H multiplex as the transmission mode changes. The far-left of the red curve equates to when the transmission mode is QPSK CR 1/2; then, after the step-change at 22 stations the QPSK CR 2/3 mode is used; then, after the step-change at 29 stations the 16-QAM CR 1/2 mode is used; and finally, after the step-change at 44 stations the 16-QAM CR 2/3 mode is used.

The EBU article simply assumes that DVB-H multiplexes can only use the QPSK CR 2/3 transmission mode, then when the capacity of one multiplex is exceeded a second DVB-H multiplex is built. The EBU article's approach is correct (i.e. cheapest) when the number of stations is greater than or equal to 22, but it is incorrect when the number of stations is less than 22.

The following graph shows the minimum of the two DVB-H curves in the above graph along with the DAB distribution costs curve. 

This graph above is relevant to mobile TV distribution (although the number of mobile TV channels is half the number of stations on the x-axis). The reason this graph is relevant is because both DAB/DMB and DVB-H use the same H.264/AVC video codec, therefore no bit rate correction is needed.

It can be seen that it is cheaper to distribute mobile TV on DVB-H than on DAB apart from a small range in the centre of the graph where DAB has slightly lower distribution costs. Importantly, however, when the number of radio stations or mobile TV channels is low, the distribution costs with DVB-H are cheaper than the distribution costs using DAB. Therefore, correction of assumption 2 alone disproves the EBU article's main claim that DAB is cheaper to distribute when the number of stations or number of mobile TV channels is low. 

3.  DAB vs DVB-H distribution costs per radio station after correction of both assumptions

Correction of both assumptions simultaneously produces the following graph of DVB-H versus DAB distribution costs per station: 

Correcting the first incorrect assumption in the EBU article (that both systems must use a bit rate of 192 kbps per radio station) results in a doubling of the number of stations that can be carried by a given DVB-H transmission mode. This has the effect of "stretching" the DVB-H curve from the previous graph to twice it's "length", because the distribution costs per station are halved. Distribution costs are halved because the cost of transmitting a certain DVB-H transmission mode is fixed, and this fixed total transmission cost can be spread over twice as many stations.

Another result of the correction to assumption 1 is that instead of DVB-H having to use two multiplexes as claimed in the EBU article, now only one multiplex is needed to carry all of the radio stations. 

Correcting the second assumption in the EBU article results in lowering the distribution costs when the number of stations is low, as explained previously.

The above graph clearly shows that DVB-H is, in fact, far cheaper for digital radio distribution than DAB. This is hardly a surprise result when you consider the technological differences between the systems:

4.  Alternative options for digital radio transmission

4.1  Transmission parameters' effect on transmitter powers

In order to analyse some alternative options for digital radio it is necessary to evaluate the difference in transmitter powers that result when certain transmission parameters vary.

Transmission frequency's effect on transmitter powers

The relationship between transmission frequency and required transmitter powers can be found from Friis' equation for free space propagation (the power loss resulting from transmitting over line-of-sight paths):

where P_t and P_r are the transmitter and receiver powers, G_t and G_r are the transmitter and receiver aerial gains, d is the transmitter–receiver separation, L is a loss factor, and λ is the wavelength. The relationship between wavelength and transmission frequency is given by: λ = c / f, where c is the speed of light (3 x 10⁸ m/s) and f is the transmission frequency.

Friis' equation can be restated in decibel form as a free space path loss as follows:

Free space loss (dB) = 32.44 + 20 log f (MHz) + 20 log d (km)

For a derivation of the above equation from Friis' equation, see here.

Therefore, for a given location the first and third terms on the right-hand-side of the equation remain constant, and the required transmitter power will change as the transmission frequency changes according to the 2nd term only:

Power difference between transmitting at frequencies f₂ and f₁ (dB) = 20 log (f₂ / f₁)

Antenna length's effect on transmtter power

The antenna gain for antennas with a length of up to half the wavelength of the signal is directly proportional to the length of the antenna. Therefore, the power difference due to antenna length for a dipole antenna that is shorter than half a wavelength is given as follows:

Power difference due to antenna length (dB) = 10 log (antenna length / half-wave dipole length)

Signal bandwidth's relationship with transmitter powers

The total power of a signal when measured in the frequency domain is equal to the area under the curve of the graph. For example, on a spectrum analyser the x-axis's units are Hz and the y-axis is power per Hz, thus the total power equals the integral of the curve, i.e. the area under the curve.

DAB, DVB-H and DRM+ all use the OFDM transmission scheme, which has a spectrum that is flat across the signal bandwidth (when transmitted). Therefore, the bandwidth of a signal is directly-proportional to the transmitted power, which can be written mathematically as follows:

Power difference between transmitting with bandwidths B₂ and B₁ (dB) = 10 log (B₂ / B₁)

C/N's (Carrier-to-noise ratio) relationship with transmitter powers

Carrier-to-noise ratio is similar to signal-to-noise ratio (SNR), with the only difference being the location at which these parameters are measured: C/N is measured at the input of the RF section of the receiver, whereas SNR is measured at the input of the baseband section of the receiver -- i.e. after the RF section.

C/N's effect on transmitter powers is simply that an increase/decrease in C/N requires an equivalent increase/decrease in the transmitter powers to compensate for the change. 

When comparing DAB with DVB-H there are published values of required C/N for both systems. But for DRM+ there are no published required C/N figures, so a C/N figure will have to be estimated. 

4.2  DVB-H in Band III

Another assumption made in the EBU article is that DVB-H must transmit at UHF frequencies. Although DVB-H is more likely to be used at UHF than in Band III, DVB-H is specified to use the same spectrum as DVB-T, and DVB-T is commonly transmitted in continental Europe in Band III. Therefore, it is instructive to investigate the effect on distribution costs when DVB-H is used at Band III frequencies. And as Swedish Radio's request for additional funding for DAB was turned down by the Swedish government, Sweden will therefore have some Band III spectrum going spare after DAB has vacated the band, and this spectrum could be used to carry DVB-H.

Using typical frequency values for Band III and UHF (Bands IV and V) of 220 MHz and 600 MHz, respectively, the difference in power due to transmitting in Band III after transmitting at UHF is:

Applying this result to the data for DVB-H distribution costs at UHF frequencies produces the following graph of distribution costs per radio station:

Therefore, it is far cheaper to distribute DVB-H than DAB when both systems use Band III. It is easy to see why the EBU article didn't want to allow DVB-H to be transmitted in Band III, when you consider that Swedish Radio were trying to make the case for using DAB.

4.3  DRM+

The main thing that struck me when reading the EBU article was the absence of DRM+, because as DRM+ uses lower frequencies than both systems and each station requires a much narrower bandwidth, it is trivial to see that the DRM+ transmitter powers per radio station will be far lower than for either DVB-H at UHF or DAB, hence the distribution costs will also be far lower.

Whereas both the DVB-H and DAB systems are inherently multiplex-based with wide bandwidth multiplexes, DRM+ is similar to FM in that a narrow bandwidth channel is used to carry one radio station only. This allows DRM+ to use two different transmission scenarios:

1. a group of stations can transmit together sharing transmission resources — better for medium to very large coverage areas;

2. a station transmits individually without sharing resources with other stations — good for small stations with a small coverage areas.

As this article is interested in distribution costs for a medium to large number of stations, only scenario 1 will be investigated.

4.3.1  Changes to transmission parameters

Power difference due to transmission frequency

DRM+ will be specified to use frequencies up to 120 MHz. However, the FM band (87 - 108 MHz) is obviously currently in use, and the most likely spectrum for the system to use at the current time seems to be Band I. And, indeed, Australia has recently reserved the 45-50 MHz and 56-70 MHz bands for DRM+ use. Therefore, a transmission frequency of 60 MHz will be assumed here for DRM+.

Power difference due to antenna length

Band I antennas are very long (a half-wave dipole at 60 MHz would be 2.5m in length), so it will be assumed that FM aerials will be used, and half-wave dipoles are 1.5m in length (the material-dependent velocity factor cancels when calculating relative antenna loss, so will be ignored). 

Power difference due to bandwidth

The channel bandwidths that have been proposed for DRM+ use are 50 kHz and 100 kHz. Last year I wrote a short Matlab program to estimate the capacity that these channel bandwidths would be able to carry.

The assumptions I made were that the maximum speed of the receiver was 120 mph, transmission frequency was 120 MHz, DRM Mode A and (typical) FEC code rate of 0.6 were used. As a starting point the figures used the capacity values quoted in the DRM specification for 20 kHz channels with the appropriate transmission mode parameters (modulation order and FEC code rate). Appropriate OFDM subcarrier separation were used to allow for the increased Doppler spread at higher frequencies and receiver speeds, and a 300 μs guard interval was used to allow for very large SFN cell sizes. 

The following table shows the estimated capacity values produced by the program:

Like DVB-H, DRM/DRM+ uses the AAC and AAC+ audio codecs, so a 50 kHz channel will easily be able to carry a 96 kbps AAC radio station stream using a typical transmission mode (16-QAM CR 0.6). Therefore, a 50 kHz DRM+ channel will be able to carry a radio station at the same level of audio quality as a 192 kbps MP2 station on DAB. In comparison, a DAB multiplex has a bandwidth of 1,710 kHz.

Power difference due to change in C/N

As mentioned earlier, there are no published C/N figures for DRM+ yet, so a change in C/N will have to be estimated. As a starting figure I will use the published C/N figure of 15.1 dB for DVB-H using 16-QAM with an inner code rate of 1/2 and using the MPE-FEC. So, with both systems using the same modulation order, the difference in required C/N will be due to their different FEC coding schemes. 

DVB-H uses three layers of FEC coding consisting of an inner layer of convolutional coding, a middle layer of RS coding and an outer layer of RS coding (the MPE-FEC). DRM+, on the other hand, uses a single layer of multi-level coding (MLC). 

Only considering the layers of FEC coding used in each system, DVB-H's FEC coding is far more powerful than DRM+'s. However, whereas DVB-H uses equal error protection (EEP) to protect the audio data, DRM+ uses unequal error protection (UEP), which provides some gain relative to EEP.

Overall, I estimate that DRM+ will need 3 dB higher C/N for 16-QAM CR 0.6 than DVB-H requires for the 16-QAM CR 1/2 transmission mode. And as the C/N required for 16-QAM CR 1/2 is 15.1 dB, the C/N for DRM+ is therefore estimated to be 18.1 dB.

4.3.2  DRM+ vs DVB-H vs DAB distribution costs

Scenario 1 - A group of DRM+ stations share transmission resources

A simple way to estimate the distribution costs for groups of DRM+ stations is to group the same number of stations as are grouped on a DAB multiplex (six stations). Then it is possible to apply an overall correction factor to the DAB distribution costs per station curve given in the EBU article, because this includes the cost savings enabled by a medium to large number of stations sharing resources on a number of multiplexes. 

The following table calculates the overall correction factor to be applied to the DAB distribution costs curve:

The following graph shows the resulting distribution costs per radio station for all of the system options for a low number of stations:

The following graph shows the distribution costs per radio station for all of the system options for a typical number of stations:

The above two graphs clearly show that DRM+ and DVB-H at Band III frequencies are by far the cheapest options to use for distributing digital radio.

4.4  Which system has the lowest distribution costs?

Eliminating DAB and DVB-H at UHF, the following graph shows the distribution costs per radio station for DRM+ and DVB-H using Band III frequencies for a low number of stations: 

The following graph shows the distribution costs per radio station for DRM+ and DVB-H using Band III frequencies for a typical number of radio stations: 

The distribution costs per station for DRM+ are lower than for DVB-H in both graphs apart from when there are around 40 - 42 stations transmitting. Overall, therefore, DRM+ is cheaper to transmit per radio station than DVB-H.

The two graphs above show the benefit gained from grouping a larger number of radio stations together, because when there is just one radio station DVB-H is three times as expensive to transmit, but the difference narrows until at around 40 - 42 radio stations DVB-H actually becomes cheaper to transmit. However, the only reason six DRM+ stations were grouped together was to aid analysis, because a factor could be applied directly to the DAB distribution values. In reality, because DRM+ channels are so narrow and the transmitter power of each station is so low, a very large number of DRM+ stations could easily share the same transmitter and digital hardware. This would allow the distribution costs per radio station to fall further for DRM+ as more stations shared the same hardware resources, so it is very likely that DRM+ will actually be significantly cheaper than DVB-H at Band III for any number of stations. 

4.5  Are distribution costs really linearly related to transmitter powers?

All of the above analysis draws on the assumption suggested in the "Broadcasting to Handhelds" EBU Technical Review article that transmission costs are linearly-related to total transmitter powers. At medium transmitter power levels this is likely to be a fair assumption, because, for example, if you buy two transmitters that have the same power rating then you can expect to pay double what you'd pay for one (ignoring discounts). But at very low (and very high) transmitter power levels the relationship is likely to become non-linear. For example, at very low transmitter power levels the cost of the equipment and installation costs will be disproportionate to the actual power level, and the distribution costs will be higher than that predicted by a linear relationship. At very high transmitter powers the cost of the transmitter and installation costs will likely not be double that of a transmitter half its power, and a linear relationship would overestimate the distribution costs. Therefore, the relationship between transmitter powers and distribution costs is likely to be more S-shaped than linear, and this would mean that the DRM+ and DVB-H at Band III distribution costs given above will be an underestimate of the costs that you could expect in practice. 

Overall, the assumption is reasonable for providing rough-and-ready estimates when comparing options at a system level. 

Therefore, because the differences in transmitter power levels are so large between the two most expensive options, DAB and DVB-H at UHF, and the two cheapest systems, DRM+ and DVB-H, it is safe to say that DRM+ and DVB-H at Band III will be far cheaper to distribute radio than using DAB or DVB-H at UHF. For example, comparing DAB with DRM+, one DRM+ station requires 23-times lower transmitter powers than one DAB station, so errors caused by the non-linear relationship between distribution costs and transmitter powers will be insignificant in comparison to this huge difference in power.

5.  Other issues

5.1  Frequency planning issues

As well as the distribution costs, frequency planning issues are an important factor when considering the adoption of a digital radio system:

As you can see in the table above, DRM+ is also better than DAB and DVB-H in terms of its flexibility for frequency planning and its suitability to carrying the full range of stations from the largest to the very smallest. The flexibility for frequency planning results from the major difference between DRM+ and the other two systems: the fact that it doesn't multiplex stations together. For example, a small local radio station with a small coverage area can simply transmit a very low power DRM+ signal without having to be carried on a multiplex that must be transmitted over a far larger coverage area in order to be financially viable. This issue also has serious consequences for total spectrum consumption, because with DAB using 1.7 MHz-wide multiplexes and DVB-H using 7 or 8 MHz-wide multiplexes it simply isn't realistic for these systems to carry one or two radio stations with a very small coverage area, because if you repeated this for all such small radio stations across a country the amount of spectrum required would be unfeasibly large and wasteful. 

5.2  Spectral efficiency

Spectrum is a limited resource for which there is a high demand. Therefore, it is important that a digital radio system be as spectrally efficient as possible in order to minimise the amount of spectrum consumed. The following table shows how much spectrum one station consumes at the same level of audio quality on the different systems:

1 - PL3 stands for Protection Level 3, and is the most commonly used DAB multiplex error protection level

The above table shows that there is a huge difference in spectral efficiency between DAB and DRM+, and ignoring the fact that not all DAB multiplexes will be full, you can expect that DAB would need almost 6-times as much spectrum as DRM+ would need in total. 

For example, the UK currently has 7 Band III DAB channels, which consumes a total of 12 MHz of spectrum. As DRM+ is 5.7-times as efficient, DRM+ would only require 2.1 MHz of spectrum to carry all that is currently transmitted on DAB in the UK.

5.3  Spectrum availability

In many countries, there are spectrum availability problems for both DAB and DVB-H due to the transmission of TV in the bands that these systems use. Also, demand for TV spectrum will increase in some countries in the next few years as HDTV services launch across Europe.

Digital radio does require far less spectrum than TV, but due to DAB being very spectrally inefficient, the amount of spectrum DAB requires is far from being insignificant. And if mobile TV is transmitted in Band III, this spectrum requirement increases rapidly.

DRM+, on the other hand, could use Band I spectrum (47 - 68 MHz), which is allocated to broadcasting for primary use in Region 1 (Europe) in internationally-agreed frequency allocations. Band I is used in continental Europe for TV broadcasting, whereas in the UK this band is allocated to mobile services. But Ofcom wrote in its Framework Spectrum Review last year that there was spectrum in Band I that could be freed-up, but this band had "not been included [in the Framework Spectrum Review] as Ofcom judged from past discussions that there was no interest in these bands." 

Therefore, as there seems to be little demand for Band I spectrum from other systems, and as DRM+ only requires a very small amount of spectrum in order to carry a large number of services, Band I would seem to be ideal spectrum for DRM+ to use.

Alternatively, DRM+ could also be transmitted within the FM band (87 - 108 MHz) in between currently transmitting FM radio stations. 

5.4  DAB Version 2 (DABv2) & Combined DAB / DRM+ Receivers

Prototypes of combined DAB / DRM receivers have already been displayed at the IFA in Berlin last September, and are expected to be released in the coming months (as of time of writing, January 2006). 

Also, new DAB standards — dubbed DAB version 2, or DABv2 — are expected to be released in the combing months, and it is expected that DABv2 will incorporate the AAC and AAC+ audio codecs that DVB-H and DRM/DRM+ already use. 

Therefore, as all combined DAB/DRM receivers will contain AAC and AAC+ decoders inside them (because AAC and AAC+ are the standard codecs used by DRM), and as it is expected that the new DAB standards will incorporate the AAC and AAC+ codecs, it makes sense for DAB to move away from using the current MP2 codec and towards using AAC and AAC+. This will improve the efficiency and reduce the distribution costs for DAB, although it will never be able to approach the efficiency or low distribution costs of DRM+ or DVB-H using Band III.

The UK communications regulator Ofcom recently made a statement in which it encouraged the UK broadcasters and DAB receiver manufacturers to adopt the new DAB standards and to use combined DAB/DRM receivers so that in a number of years, once the vast majority of DAB receivers in the market can receive AAC/AAC+ transmissions, it would be possible for the UK to change to using these new audio formats. 

This course of action would be beneficial because DABv2/DRM+ receivers could be sold in all countries that adopt one or both of the standards, and thus receiver prices would be lower due to economies of scale than if DABv1 and DABv2 receivers were made separetely.

5.5  Mobile TV on DRM+

The following table is the same table that was presented earlier, and contains the estimated DRM+ channel capacities:

It can be seen that a 100 kHz DRM+ channel has sufficient capacity to carry one mobile TV channel. And as DRM+ requires low transmitter powers, it would cost far less to distribute mobile TV over DRM+ than via either DAB or DVB-H at UHF frequencies.

The issue of DRM+ requiring relatively long antennas isn't a problem so long as the headphone lead is used as the antenna — DAB personal radios also use the headphone lead for their antenna. And the issue of receiver power consumption would be improved by the fact that the RF front-end power consumption is related to bandwidth, and 100 kHz is very narrow compared to DAB and DVB-H multiplexes.

6.  Conclusions

• Digital radio distribution via DRM+ and DVB-H using Band III were both overlooked by the EBU article, but the distribution costs per radio station are far cheaper using these options than either DVB-H at UHF and especially DAB, which were the two options investigated in the EBU article.

• DRM+ is the cheapest system to distribute digital radio out of all of the possible options.

• Two incorrect assumptions made in the EBU article had a major impact on the results presented in that article, and contrary to the EBU article's primary claim, once the assumptions were corrected the distributions costs per station on DVB-H are actually significantly lower than they are on DAB, so long as the number of stations is greater than six.

• All of the results presented in the EBU article — both for digital radio and mobile TV — are incorrect, because all of the results include the effects of the incorrect assumptions made.

• All of the conclusions made in the EBU article are incorrect because the conclusions were drawn from the incorrect results. In particular, virtually all of the conclusions suggested that DAB was the cheapest system, whereas DVB-H is in fact much cheaper.

• Similar to the case for digital radio, DRM+ and DVB-H using Band III will also be the cheapest options for mobile TV distribution.

• As would be expected, the component technologies (e.g. the audio codec, FEC coding and modulation type) of which the digital systems are comprised have a major impact on the spectral efficiency and power efficiency figures, and have a similarly large effect on the distribution costs per radio station. 

• The effect of transmission frequency — which seems to be somewhat overlooked — has a large impact on required transmitter powers, and hence on distribution costs.

• DRM+'s greater flexibility in terms of frequency planning and its suitability to carry radio stations from the largest nationals down to the very smallest local stations makes DRM+ better-suited to digital radio use than both DAB and DVB-H. DRM+ is therefore the best option for digital radio transmission.

• DRM+ is almost 6-times as spectrally efficient as DAB, which implies that DRM+ would need approximately 6-times less spectrum allocated to it to carry the same number of stations.

• Unlike respected engineering research journals (such as those published by the IEEE), articles published in the EBU Technical Review are not peer-reviewed, and the authors therefore have a free reign to write whatever they like irrespective of its accuracy; authors have articles published irrespective of their level of understanding of the relevant technologies; and articles are published irrespective of authors' vested interests. The EBU Technical Department needs to address these problems because the EBU Technical Review is a trusted source of broadcast engineering information and is read by a large number of people who will think that the information published in articles under its name will be correct. Indeed, it is possible that decision-makers will now read the "An Economic Analysis of DAB & DVB-H" article and form the opinion that DAB is indeed the cheapest system for digital radio use, when, in fact, this cannot be further from the truth.