Transmission

COFDM

Coded Orthogonal Frequency Division Multiplexing (COFDM) was largely developed into a practical proposition by the CCETT in France in the late 1980s, although the idea of OFDM was originally propounded elsewhere. It found its first major application in Digital Audio Broadcasting (DAB), developed in the Eureka 147 project. Many organisations and collaborative projects subsequently took up the idea of COFDM for application to digital terrestrial television, including CCETT ("Sterne"), the Nordic HD-Divine, HDTVT in Germany, ITC ("Spectre"), RACE dTTb project, Thomson, BBC, etc.

COFDM is a system of modulation well-suited to the propagation and interference environment of terrestrial broadcasting in the VHF (used in UK for DAB) and UHF bands (used in UK for DVB-T). It uses a large number of carriers, each carrying a small part of the total coded data rate. The frequency spacing is carefully chosen to ensure "orthogonality" - the carriers do not crosstalk one to the other, despite having spectra which overlap in the frequency domain.

The low data rate per carrier itself reduces the susceptibility to multipath, because the symbol period is quite long compared with the likely echo delay. And it makes it feasible to enhance the multipath immunity substantially by the addition of a guard interval. Each symbol is transmitted for a period which is longer (by the guard-interval duration) than the period during which the receiver examines it. The guard-interval duration is chosen to be just greater than the expected spread of multipath delays for any particular application. The immunity to multipath reflections is a particularly important property which can also be exploited in so-called single frequency networks, where a chain of transmitters can all use the same frequency for transmission - the delayed signals from distant transmitters just appear to the receiver as echoes within the guard interval duration.

The use of coding (the "C" of OFDM) is an important factor in coping with both co-channel analogue TV interferers and the frequency-selective fading caused by multipath. The predominant effect of co-channel analogue TV interfererence is to corrupt only a few carriers (those lying around the vision carrier, and to a lesser extent around the sound carrier and colour sub-carrier). There is plenty of information in the many unaffected carriers for the coding to repair this damage. Thus DVB-T is robust against co-channel interference from analogue TV. It is also fairly benign as an interferer to analogue TV, as it is similar in character to flat noise and will also be transmitted at a much lower power level than analogue TV, thanks to its smaller signal-to-noise ratio requirement. These two factors explain how it has been possible to fit new digital services into a UHF spectrum already crowded with analogue transmissions.

The development of the DVB-T system specification was taken on by a collaborative group called the Task Force on System Comparison (TFSC), comprising representatives from BBC, CCETT, CNS, DLR, IRT, ITC, Philips, RAI, Tele Danmark, Teracom and Thomson, and was adopted by DVB and ETSI.

The Plan for the UK

Estimates use a 1km sq grid. The figures quoted are the populations within grid squares where it is predicted that at least 90% of viewers will be served.

It should be noted that reception of digital signals is a bistable event - it either works or it doesn't - and the predictions have assumed an ideal ITU standard aerial installation. Common sense suggests that people often have very sub-standard installations and are content to watch truly awful pictures. It thus raises the debate as to whether people should be allowed to buy digital terrestrial receivers as a "carry away" purchase, or whether the sale price should include a professional installation or checkup.

Number of Carriers

To take advantage of OFDMs capability for single frequency networks, the DVB originally defined a system using 8,000 carriers, the so-called 8k system. However, there were concerns about the additional complexity of an 8k system, so DVB allowed the use of 2000 carriers (2k) as an alternative for early implementation and the ITC have specified 2k carriers in their requirements for digital terrestrial in the UK. 2k gives very restricted capability for single frequency networks but, in the UK, the plan is to use 81 of the existing main transmitter sites such that a s.f.n. would not have been appropriate. In the event, it now appears that most STB / receiver implementations will offer 2k / 8k switchable, such that a single implementation can be used across Europe.

Datarates and Bit Budgets

Within COFDM transmission, there are a number of parameters that directly affect the datarate of the signal. The method of modulation (QPSK, 16QAM, 64QAM), the code rate and the guard interval can be varied to achieve an optimum compromise between datarate and robustness of signal. QPSK is most robust but achieves datarates only up to 10.6Mb/sec., 16QAM can achieve a maximum of 21Mb/sec and with 64 QAM, over 30Mb/sec can be achieved but with increased liability to interference.

In the UK situation with 2k carriers, the NTL have predicted interference limited coverage, using three different systems:

Modulation Code Rate Failure Point Data Rate
16 QAM 1/2 12dB 12Mb/s
16 QAM 3/4 16.5dB 18Mb/s
64 QAM 2/3 20dB 24Mb/s

The failure point figures are approx 3db above the theoretical figures given in prETS 300 744 but have still to be confirmed by field trials. The coresponding coverage figures are as follows:

Failure Point Mux 1 Mux 2 Mux 3 Mux 4
12dB 96% 95% 93% 92%
16.5dB 93% 91% 90% 88%
20dB 89% 85% 85% 82%

So, moving to 64QAM increases the possible datarate to 24Mb/sec but reduces the predicted service areas by a significant amount. There is concern, too, that the coverage may become patchy and unpredictable with 64QAM, giving problems to retailers at the sharp end of selling receivers.

Coverage Predictions

A word is in order on how population coverage predictions are made and what they mean. The steps in the process are as follows:

(i) Firstly a field strength prediction is made from ERP and terrain profile information, on a 1km grid, assuming an omni-directional transmit antenna.

(ii) The field strength prediction is then modified by the designed radiation pattern of the antenna, taking account of any needs to reduce interference in adjacent areas.

(iii) The field strengths of co-channel and adjacent-channel interfering signals are then compared and grid squares where there is significant interference removed from the coverage profile. This has to consider separately both analogue and digital transmissions.

Three different coverage predictions may be quoted - the so called 90% cut-off method, the proportional method, and the composite coverage method. In all cases, a log-normal distribution about a mean signal strength is assumed; in other words, if calculations show a certain mean signal strength, then a proportion of households are assumed to have greater and less signal strength according to a log-normal curve. (The BBC and NTL have their own special algorithms in areas of urban clutter which can give rise to differences)

In the 90% cut-of method, the population of a grid square is only counted if it is calculated that 90% of households will be able to receive an acceptable signal. Any less than 90% and the reception is considered doggy and not counted.

In the proportional method, the percentage of housholds in each square is counted, usually down to a 50% cut-off, even though reception may be a bit hit-and-miss.

Lastly, a composite coverage figure may be quoted in overlap areas. If transmitter A is calculated to serve 80% of the households in a grid square and transmitter B is calculated to serve 60% of households, the the composite coverage is calculated as 80 + 60% of the remaining 20% = 92%.

The number of households in each grid square is derived from a Post Office address file which gives the number of households for each 6 digit postcode. Digital predictions have been based on 1993 figures. The number of viewers is calculated using a figure of 2.39 viewers per residential address, with the exception of N. Ireland, where a figure of 3.0 is used.

Limitations of Coverage Predictions

Coverage predictions use a standard ITU defined aerial which has a 15dB front-to-back ratio. Thus, whilst signal strength on a set-top aerial will be considerably worse (How much?) it is also possible to install a significantly better aerial (how much) in cases of marginal reception prediction. Digital reception is calculated to be largely interference limited, so a more directional aerial will be of more help than a head-amp.

The methods of prediction used are inherently conservative, so coverage is likely to be greater than prediction, though it cannot be guaranteed. Because digital signals will will fail suddenly, rather than degrade as with analogue, a 3dB margin above cut-off has been assumed, so it is said that digital will usually work where analogue reception is poor. However, without statistically sound evidence of practical installations, that is unsubstantiated.