Digital Radio Mondiale (DRM)

DRM's Extension to 120 MHz

In XXXX 2005, the DRM consortium announced that they are going to extend DRM to work at frequencies up to 120 MHz. The work is supposedly going to be completed by 2008 - 2010, but I have absolutely no idea why it takes even 3 years to extend an already completed specification, because it really should only take about 1 to 2 years at the very most. In an off-guard moment, someone from the BBC admitted that the DAB broadcasters want the extension to DRM to take as long as possible to give DAB the best chance of being taken-up! This is completely against the best interests of BBC licence-payers, so the BBC should not take this stance and should fully support the extension of DRM in as little time as possible.

As I write this (August 2005), Radioscape (who design DAB receiver modules) have recently launched a new low-cost DAB+DRM module, and combined DAB+DRM receivers will be demonstrated at the IFA consumer electronics show in Berlin next week. Although it is good news that reasonably-priced DRM receivers will become available, the fact that the BBC and other broadcasters want the DRM extension to take as long as possible could have some negative consequences for the UK. The problem is that if a few million DAB+DRM receivers are sold prior to DRM's extension being completed, then that could potentially exclude the UK from using the extended version of DRM, because receivers that have already been sold would not be able to receive the extended modes. We've already seen this problem occur on DAB, where the ultra-short-sighted decision was taken by Digital One to make DAB receivers Band III-only, and now that around 95% or more of the 1.5 million DAB receivers already sold are Band III-only, then this has effectively ruled out UK DAB from using L-band, where about 16 channels will become available in 2007. In comparison, DAB currently only uses 7 Band III channels, and is expected to get another 5 Band III channels at the Regional Radio Conference (RRC-06) in 2006, and that will be all DAB will be getting in the UK, because the last time there was such a major frequency planning conference for broadcasting as RRC-06, was in Stockholm in 1961 -- these things don't come around every day... An alternative way of looking at the decision to limit DAB receivers to being Band III-only would be to consider that GWR (who owned 67% of Digital One at the time) would get all the radio stations they wanted on DAB on Band III spectrum, so disallowing the UK from using L-band for DAB would limit the amount of competition they faced in future. 

Ofcom also announced earlier in 2005 that 31% of all current analogue radio stations will not be able to get onto DAB even after the additional 5 Band III channels are put into use. DRM will, therefore, be the only possibility for these stations to transmit on digital radio. Also, it seems likely that not all of the BBC's local radio stations will get onto DAB, in which case, by deliberately delaying the extension of DRM they are deliberately denying some listeners to their own radio stations. 

It is absolutely crazy to delay DRM, because it completely confuses the situation. And why does it have to be the UK that gets the worst of everything? Low bit rate, low audio quality DAB; 2K DVB-T which rule-out using SFNs (single-frequency networks) and reduces the number of DTT multiplexes we can receive; and now the possibility of adopting DRM before it is extended, which could result in the UK not being able to use the future modes that will allow DRM to work well at higher frequencies. The situation with DAB and DTT were both caused by the UK government's mis-guided desire to be first to adopt the technology, and the BBC trying desperately to cling on to the idea that DAB was still good enough for the 21st century. Other countries will be able to adopt DAB with the new HE AAC audio codec if WorldDAB do go ahead with their plan to adopt the new codec. So, if the UK does sell a lot of DRM receivers before the extension is complete, all the digital terrestrial broadcasting systems in-use in the UK will be using the worst versions of all of these systems!! Digital broadcasting in the UK is turning into a very, very bad joke.

Introduction to DRM

DRM was originally designed to allow digital radio broadcasting in the LW, MW and SW bands, using frequencies below 30 MHz. Radio signals at frequencies below 30 MHz have rather odd propagation characteristics compared to transmissions that use higher frequencies. These strange propagation characteristics are caused by xxxxxx xxxxxxx xxxx

As discussed above, DRM will be extending to work at frequencies up to 120 MHz, where propagation is via xxxxxxxxx xxxxxxxx. Sporadic E

Current DRM Specification

Channel Bandwidths

DRM currently allows for the following channel bandwidths:

Modulation

DRM uses OFDM modulation, which stands for Orthogonal Frequency Division Multiplexing. DVB-T/H, DAB, DMB, IBOC-FM (HD Radio), ISDB-T and various current and future wireless LAN, MAN and PAN standards also use this type of modulation. ADSL also uses OFDM modulation.

OFDM consists of the transmission of many -- usually hundreds or thousands of -- narrow channels in parallel, called subcarriers. The 'orthogonal' property means that the subcarriers do not interfere with any of the other subcarriers.

Signal Constellations

DRM can use the following signal constellations:

The last two columns are related as follows:

bits per subcarrier per OFDM symbol = log₂ (no. signal points in constellation diagram)

Number of Subcarriers

The number of subcarriers used by DRM depends on the Robustness mode used, out of the four options:

The number of subcarriers for other channel bandwidths follows a roughly linear relationship with the above values.

Robustness Mode A is the weakest mode, going to Mode D the strongest. The stronger robustness modes are meant for long-distance international transmissions which use the exotic propagation characteristics of the HF bands. Mode A is the most suitable for normal local, regional and national broadcasting. Stronger robustness modes have lower DRM multiplex capacities than the weaker modes.

Pilot Subcarriers

DRM uses synchronous modulation of the subcarriers. This means that the receiver must synchronise the subcarriers in phase. Also, to allow the higher signal constellation orders such as 16-QAM and 64-QAM, the receiver must perform channel estimation (estimate the frequency response -- both magnitude and phase -- of the wireless channel) and correction. For this reason, pilot subcarriers are transmitted along with the subcarriers that carry data. Robustness Mode A has the least pilot subcarriers and Mode D has the most pilot subcarriers. As pilot subcarriers do not carry data, then the higher the number of pilot subcarriers used the lower the multiplex data capacity will be.

DRM also uses three subcarriers as a frequency reference to allow fast frequency synchronisation.

One of the main drawbacks with the DAB system is the fact that it uses differential modulation as opposed to synchronous modulation, because this limits DAB to using QPSK (so DAB only has a low spectral-efficiency in bps/Hz) and differential modulation incurs a 3dB SNR penalty compared to synchronous modulation. 

OFDM Symbol Durations, Subcarrier Spacing & Guard Interval Durations

The following table shows the OFDM symbol durations (T_u), subcarrier spacing, (1/T_u) and guard interval duation, (T_g):

Subcarrier Spacing

As can be seen in the above table, the subcarrier spacing equals 1/T_u. OFDM subcarriers always have subcarrier spacing inversely proportional to the OFDM useful symbol duration, T_u, because the following equation must be satisfied to maintain orthogonality between subcarriers:

where ω = 2 π / T_u, t = time (integration must take place over an integral number of full sinusoid cycles).

When a receiver moves relative to the transmitter, then the various signal paths (multipath) to the receiver incur Doppler shifts (a shift in frequency). The overall signal the receiver 'sees' is a linear summation of all of these multipaths, and this gives rise to random Doppler shifts of the different subcarriers. These random Doppler shifts of the subcarriers destroys the orthogonality between the subcarriers and gives rise to what is known as 'inter-carrier interference' (ICI). Maximum Doppler shift, f_D,max is given by:

where f = transmission frequency, v = mobile velocity and c = speed of light. Therefore, the higher the speed of the receiver and the higher the transmission frequency, the higher the Doppler shifts will be, and the higher the ICI will be. ICI sets a maximum speed limit at which any given OFDM transmission mode can handle before reception fails.

Another advantage of DRM compared to DAB is the fact that because it uses lower transmission frequencies the ICI will be lower for a given mobile velocity. 

A parameter that is independent of frequency is the 'percentage Doppler shift to subcarrier spacing', and once the Doppler shift reaches a certain percentage of subcarrier spacing, reception will fail. The maximum percentage Doppler shift to subcarrier spacing that a given transmission mode can handle depends on the various transmission parameters, such as signal constellation order, FEC coding type and FEC code rate.

Because of the percentage Doppler shift to subcarrier spacing that a transmission mode can handle, when DRM is extended to work at frequencies up to 120 MHz then the subcarrier spacing will have to increase to allow for the higher ICI at higher frequencies. But because the subcarrier spacing is inversely proportional to the OFDM useful symbol duration, T_u, then the OFDM symbol duration will have to be reduced by the same factor as the increase in the subcarrier spacing.

Guard Interval Duration

A guard interval is insterted between the useful parts of OFDM symbols to avoid inter-symbol interference (ISI). Multipaths, whether they're naturally delayed (e.g. bouncing off hills) or artificially delayed (e.g.multipath that arrives from distant transmitters in the SFN), must arrive with a delay (relative to the first path that arrives) that is less than the guard interval duration, or otherwise ISI results. Some ISI can be handled, but too much ISI will cause reception to fail.

The guard interval durations currently specified are extremely long due to the huge difference in distances between different signal paths. These very long guard interval durations are only necessary for the very long distance transmissions that are possible in the lower frequency bands. At frequencies above 30 MHz, transmissions will be for local, regional or national services, and the guard interval duration can be reduced significantly. For example, the guard interval duration used by DAB is only 246 μs compared to the 2.66 ms used by DRM in Robustness Mode A (i.e. DRM's guard interval is 10 times higher than DAB's, which is unnecessary for transmissions at frequencies above 30 MHz).

Typical multipath delays that occur naturally by signals bouncing off hills or buildings are usually below 20 μs, and longer guard interval durations are specified for systems such as DAB and DVB-T/H etc to allow larger distances between transmitters in an SFN. So it is likely that the higher-frequency modes for DRM will use similar length guard interval durations as DAB and DVB-T/H after it has been extended.

Single-Frequency Networks (SFNs)

OFDM allows the use of SFNs, where a single frequency is used to cover a large area. SFNs are more spectrally-efficient than MFNs (multi-frequency networks), so if a large area is to be covered then SFNs should be used wherever possible. 

SFNs are made possible by the guard interval between OFDM symbols, and the issues surrounding SFNs and the guard interval are discussed above. 

Forward Error Correction (FEC) Coding

DRM uses a multi-level coded modulation (MLCM, or simply multi-level coding (MLC)) scheme for error correction. Quoting from the DRM specification:

"The channel encoding process is based on a multilevel coding scheme. The principle of multilevel coding is the joint optimization of coding and modulation to reach the best transmission performance. This denotes that more error prone bit positions in the QAM mapping get a higher protection. The different levels of protection are reached with different component codes which are realized with punctured convolutional codes, derived from the same mother code." 

However, if the convolutional codes used in DRM are replaced by turbo codes, to produce a turbo-multi-level coded modulation (T-MLCM) FEC coding scheme, the increase in performance (reduction in required SNR at the receiver) is large (the figure is for an 8-PSK constellation, but a similar, or possibly even better, increase in performance could be expected for 16-QAM constellations):

What this means in practical terms is that by using T-MLCM, instead of the standard MLCM DRM uses now, the required SNR at the receiver can be reduced significantly, which translates to lower required transmission powers and more robust reception. 

Hierarchical Modulation

Interleaving

Audio Coding

Scalable AAC

For more information about DRM visit the following web sites:

• http://www.drm.org/indexdeuz.htm (this website has DRM audio samples that you can download)

• http://www.wohnort.demon.co.uk/DAB/rxdrm.html (shows all available DRM receivers)

• http://www.drmradio.co.uk/

• http://www.drmrx.org/

and a good overview of the technology that comprises the DRM system see this research paper:

• http://www.drm.org/pdfs/newsevents/drm_trans_on_broadcasting.pdf

In Ofcom's 'Spectrum Framework Review: Implementation Plan' document, which was concerned with spectrum that could be freed-up between 2005 - 2008, it says this about spectrum that could potentially be used for DRM:

2.16 There are two other bands not discussed in this document where spectrum is
potentially available for award. These are:
• Spectrum between 47 and 68 MHz (Band I). In the UK the allocation is
mobile services, while in Europe it is used for television broadcasting.
• VHF low band (68 – 83 MHz). In this band there is approximately 2 x
1.325MHz plus 12 simplex channels (12.5kHz).
2.17 These have not been included as Ofcom judged from past discussions that there
was no interest in these bands. However if stakeholders are interested in these
bands please respond accordingly to the consultation.