DVB-T and DAB are accepted ETSI standards for use in Europe and the rest of the world.

Multiple Carriers
COFDM does not rely on the vulnerability of a single carrier but spreads the digital information over many narrow band carriers using Frequency Division Multiplex (FDM). DVB-T uses either 1,705 or 6,817 carriers (referred to as 2,000 & 8,000 respectively). The bandwidth and the data rate on each of these carriers is reduced and therefore the RF robustness is increased.

In digital TV broadcasting, either the 2k or 8k modes can be used. The 2k mode is currently used in the UK, but the 8k mode is preferred in countries operating single-frequency networks or where long-delay echoes are expected (see Guard Interval below).

In microwave links and radio cameras, the 2k mode is the norm to reduce the impact of differential Doppler shift of any echoes and to reduce phase noise in oscillators used in modulating the signal.

Orthogonal Carriers
The carriers are accurately spaced and orthogonal, which means they can be generated and recovered without carrier specific filtering. Indeed even though the spectra of adjacent carriers significantly overlap, each carrier can be demodulated without any crosstalk from its neighbour.

The extremely complex process of modulating and demodulating thousands of carriers is possible by using Discrete Fourier Transform, for which algorithms exist. This means inexpensive, large volume, modulators and demodulators using modern integrated circuits can be achieved.

Symbols
The active symbol period is the sample of digital information accessed by the receiver at any one time, and has a duration of 224us (2k) or 896us (8k). These modulation symbols are arranged to occur simultaneously on each carrier. The number of bits carried in each symbol depends on the choice of modulation i.e.:

Guard Intervals
A guard interval is added to the beginning of each symbol to allow time for echoes to settle before commencing the active symbol period. A wide range of guard interval options are available for DVB-T from 1/32 to ¼, with this fraction representing the ratio between the guard interval and the active symbol period, and reduces the overall data capacity by the same proportion.

Guard intervals will prevent the receiver from being affected by interference, provided the duration of the echoes does not exceed the guard band duration. In 2k mode the guard intervals range from 7 to 56us, and in 8k mode the duration ranges from 28 to 224us. Hence the reason for the 8k mode becomes apparent, i.e. to allow a longer guard interval, thus accommodating echoes up to four times longer than possible when using 2k mode.

Forward Error-correction Coding (FEC)
The purpose of FEC is to improve the bit error rate (BER), increase the threshold of the receiver, and therefore improve recovery of the data stream by the demodulator. It is sometimes quoted in terms of the transmitter power increase in dBs, necessary to achieve the same BER.

The way it works is to add some carefully designed redundant channel coding information at the transmitter, which will provide the receiver with additional information and redundancy to assist in the decision-making process. The greater the additional information added the greater the resilience of the system. This is represented as a fraction eg ½, where one bit of FEC information would be added for each one bit of signal, giving a total of 2 bits of final gross output, in other words, in this case the amount of error correction added reduces the overall system capacity by 50%.

In its simplest form, and taking this example, the encoder would send each symbol twice so that the decoder had two chances of recovering the information. In practice coding systems are not this straightforward and employ complex mathematical computations to correct for errors.

Reed Solomon and Convolutional Coding
In many digital systems the data to be transported undergoes two types of FEC coding algorithms. First Reed Solomon which is a block coding structure, sending blocks of bytes at a time, adding extra bytes to the end of the block for error correction. Secondly, then convolutional coding is added, which multiplies the signal with a pseudorandom sequence running faster than the data, hence adding bits to the data stream for error protection.

Viterbi decoding
Viterbi decoding is a method of recognising, at the receiver, the pseudorandom sequence added at the transmitter by the convolutional encoder. The Viterbi decoder has the ability to recognise the distinctive pattern imposed on the data by the sequence even in the presence of errors. In essence, the Viterbi decoder passes the data through a buffer configured with templates shaped by the pseudorandom sequence and attempts to find the best match between the incoming data, with possible errors, and its templates. The Viterbi decoder outputs a decision based on the best match found.

Hard Decision Decoding
The input to the Viterbi decoder is either 1 or 0 representing which side of a slicing level each demodulated data bit falls. However, it does not take into account the effect any interference or noise will have on the signal amplitude at the instance of decision-making. The Viterbi decoding process then just finds the best match to the incoming data stream.

Soft Decision Decoding
Soft decision coding adds an analysis for the confidence of the template matching process in the Viterbi decoder. The incoming data is not just sliced into 1’s and 0’s, but is converted into a three-bit number representing it’s size (basically a three-bit ADC process). Therefore, a 111 represents a higher confidence than a logical 1 has been received, than a 110. This three bit number is used in the Viterbi decoder’s template matching process a weighting functions when looking for the best match to the data. A data bit that has had its level changed by noise will be given a lower soft decision confidence level and therefore given less weight in the Viterbi decoding process.

Channel State Information (CSI)
Allows a level of confidence to be given to each of the multiple COFDM carriers.

Consider an extreme case of a 0dB (or a 100%) echo which knocks out 1 in 4 of the carriers. It follows that the two adjacent carriers will be unaffected and the fourth carrier will be boosted. From this it can be seen that the information from the nulled carrier is unreliable and should be ignored, but that the information from the fourth carrier is enhanced, will have improved SNR, and hence improved BER. By careful use of FEC in choosing only the information on the three good carriers, the performance is enhanced.

The Viterbi decoder uses the CSI information to lower the soft-decision confidence levels for noisy carriers.

Frequency Interleaving
If an echo is received with a rather shorter duration than the example above, then it would put notches in the channel frequency response, and a number of adjacent carriers will be affected. This would be a problem if the data was carried sequentially on adjacent channels, however if the carrier data is spread out or interleaved then FEC may well be able to recover the data. This Frequency Interleaving is used on both DVB-T and DAB.

Time Interleaving
As the echoes get longer (i.e. in flat fades, Doppler shifts or short term complete loss of signal), then most, or all the carriers will be affected for a period. However if sequential data is spread over a number of carriers with respect to time, then FEC may well be able to recover the data. The longer buffers required to capture Time interleaved information does cause delay.

Time interleaving is not used in DVB-T, as it is primarily designed for distribution to the home, with poor or set top receiving aerials and obstructed paths. DAB does use Time Interleaving to improve mobile operations, car reception, etc; and has a delay approaching one second.

DVB-T Coding
In the DVB-T system, firstly Reed-Solomon parity bit blocks are added, together with carrier frequency interleaving, and then this is followed by Convolutional coding. In the receiver the signal goes through soft decision Viterbi decoding, and then Reed-Solomon for parity checking.
(Turbo coding, a relatively recent innovation (1991), is a parallel-concatenated convolutional coding technique, and claims some success, but is not used in DVB-T)

Overall DVB-T COFDM Parameters

*Options for 6 & 7 MHz by scaling clock, capacity reduced proportionally.

Table of DVB-T non-hierarchical bit rates

Glossary