Revert to Section 5.3.3

5.4 Channel coding (error correction coding)

Suitably designed channel coding can be used to reduce errors in both single carrier and multi-carrier modulation systems.

For SCM, a training sequence is usually transmitted to assist adaptive equalizer convergence and system synchronization. For MCM reference signals are usually transmitted to obtain channel state information to assist frequency domain equalization and synchronization.

To achieve adequate performance at an ATV threshold point of 15-16 dB carrier-to-noise ratio, a concatenated coding system attaining a BER of 10-11 in a Gaussian channel is required. In the concatenated coding approach two levels of forward error correction are employed: an "inner" modulation code and an "outer" symbol error correcting code. Interleavers and de-interleavers are also used to fully exploit the error-correction ability of FEC codes.

The presence of various sources of interference generally requires the use of sophisticated error coding strategies containing large depths of interleaving. Single or concatenated codes could be used for this purpose.

Concatenated error correction coding schemes consist of an inner code, an interleaving scheme and an outer code. All parts of the concatenated coding scheme need to be designed together so as to produce an overall coding system that is well matched to use in the terrestrial channel. For the above reason, it is desirable to treat the concatenated code as one entity and not split the inner and outer codes into source and channel subparts.

At this stage of development Trellis codes are the mostly commonly proposed Inner modulation codes. Code rates of 2/3, 3/4 or 7/8 have been suggested. An alternative may be a more complex turbo code which could provide a lower data rate overhead for a given level of error protection.

In the area of the outer error correcting code, there was an emerging consensus on the use of Reed Solomon codes. Although different block lengths and correction distances have been suggested by different system proponents it was thought to be realistic to envisage that a range of different Reed Solomon codes could be processed by a single, appropriately designed, integrated circuit and that this could provide a suitable point for standardization.

As already stated the presence of various channel impairments requires the use of a sophisticated error coding strategy. However an error coding subsystem has already been specified for the European satellite and cable systems. In order to ensure maximum commonality of receivers the European OFDM system has decided to make use of the same error correction as the DVB satellite baseline system with the addition of an inner frequency interleaver. Therefore a concatenated Viterbi Reed-Solomon strategy is proposed with a between codes interleaver.

The inner interleaver interleaves FFT symbols. It operates on one FFT symbol at a time and is therefore a frequency interleaver only. The interleaver works on a bitwise basis and interleaves bits between the modulated symbols on the OFDM carriers. The purpose of the inner interleaver is to improve the system performance when the channel is subject to frequency selective fading or co-channel interference. The interleaver ought to spread clusters of errors caused by carriers with relatively poor S/N or S/I ratios.

The inner code of the error correction is a convolutional code, as specified in the satellite baseline specification, this code can be decoded using the Viterbi decoding algorithm. The inner code may be punctured to increase available data capacity. The puncturing rates and patterns are as defined in the DVB satellite baseline specification. The use of channel state estimation and soft decision information derived from the received data points can significantly improve the transmission performance. The channel state information can be derived in a number of ways, for example using the amplitude equalisation information generated to coherently demodulate each OFDM carrier.

If the capability of the Viterbi algorithm to correct the channel errors is exceeded it will produce bursts of errors. Therefore the outer code must be suited to correcting burst errors. Reed Solomon (RS) codes have been specified for this task. The particular RS code chosen is a (k = 188, n = 204) code. RS codes use symbols of 8 bits (bytes). A codeword of length n containing k data bytes and n-k redundant bytes is used. The code rate R is therefore k/n and the code normally provides the capability to correct t = (n-k)/2 errored bytes which in the RS(204,188) case means that up to 8 errored bytes can be corrected.

Since burst errors at the output of the Viterbi decoder will usually affect more than 1 byte, additional interleaving between the inner and outer codes is employed. This interleaver is again as specified in the DVB Satellite baseline specification. It is a convolutional interleaver which interleaves data bytes.

Single error correcting coding schemes may reduce the scale of the interleaving RAM and lead to savings in the cost of decoders. Some block codes have almost the same performance as that of concatenated codes and appropriate decoding LSIís are available.

5.5 Comparisons of early implementations of single- and multi-carrier systems

In SCM, the information bearing data is used to modulate one carrier which occupies the entire RF channel. In MCM, QAM modulated symbols are used to modulate multiple low data rate carriers which are transmitted concurrently.

There are interesting frequency/time-domain dualities between MCM and SCM. MCM can be thought of as a frequency domain technique and SCM as a time domain technique.

One ramification of frequency-time duality is that, to prevent inter-symbol interference for SCM, one must reserve part of the spectrum for pulse shaping (frequency domain), while for MCM one must insert guard intervals (time domain).

For SCM channels with multipath distortion, a training mechanism is usually transmitted to assist adaptive equalizer convergence and system synchronization. An adaptive equalizer and a high-gain directional antenna can also reduce the impact of co-channel DTTB and analogue TV interferences.

For MCM, pilot carriers are usually transmitted to obtain channel state information for frequency domain equalization and synchronization. SCM and MCM have comparable BER performance when the channel noise is additive white and Gaussian.

For an MCM system the use of a guard interval can almost eliminate the intersymbol interference, but it also reduces data throughput. To minimize the loss of throughput, the size of the FFT must be increased. The size of the FFT, however, is limited by digital signal processing speed, cost and receiver phase noise. To compensate for the frequency selectivity of the channel, a 1-tap frequency domain equalizer can be used in combination with soft decision Viterbi decoding using channel state information. The effectiveness of interleaving is also crucial to the performance of the system. The research on optimum codes for high order QAM-OFDM systems is still ongoing.

The performance of both SCM and MCM under combined impairments of noise, co-channel analogue TV interference and strong multipath distortion is yet to be determined.

5.5.1 Impulse interference

For low power impulse interference, multi-carrier systems are more robust to impulse interference, since interference can be averaged over the entire FFT block. On the other hand, a short but high power burst of interference will be expanded by the OFDM process to cause serious interference for a number of symbol periods equivalent to the duration of the impulse across all carriers. This can correspond to a significant number of errors. However it has been reported that field trial results have shown that, with adequate interleaving and error-correction, this type of interference is not a serious problem.

Single carrier systems are sensitive to time domain impulses such as lightning and car ignition interference.

5.5.2 Multipath distortion

In typical DTTB reception situations, multipath propagation caused by reflections or non-homogenities in the propagation medium will cause intersymbol interference to the unprocessed received data stream. Multipath reception will also manifest itself as frequency selective fading within the channel.

For SCM, intersymbol interference, if uncorrected, will result in eye height closure and an increase in the minimum C/I at which the system can operate.

For practical SCM systems an adaptive equalizer (usually a decision feedback equalizer) is used to minimise the effects of multipath distortion. For its operation it requires a training sequence which will slightly reduce data throughput. An adaptive equalizer can also converge without a training sequence by use of a blind equalization technique. Any adaptive equalizer will however increase the system noise threshold when multipath is present. (Adaptive equalizers may also reduce the impact of co-channel and adjacent channel interference.)

Single carrier systems are inherently rugged against frequency selective fading because the fade will only affect a small portion of the bandwidth in which the signal energy is being received.

Multi-carrier systems can be designed to include a "guard interval" which will allow intersymbol interference (due to multipath reception) to be almost eliminated over a wide range of multipath delay durations.

There are two important cases of the use of guard intervals to reduce intersymbol interference in multipath situations. First, where multipath occurs as a result of reflections or inhomogenities in the transmission media. In this case relatively short multipath delays for example, of up about 50 m s might be encountered. Secondly, if on-channel active repeaters are used as part of a Single Frequency Network (SFN) concept, longer multipath delays may be encountered. (The duration of SFN multipath delays will depend on transmitter spacing.)

The disadvantage of using long guard intervals (which may be required when actual existing transmitter network location requirements are considered) is that, for a fixed overall symbol duration, an increased guard interval will reduce data throughput in proportion to the ratio of guard interval to overall symbol duration. To avoid loss of throughput the size of the FFT used in the MCM system must be increased. This will result in a longer overall symbol duration and an increased number of more closely spaced carriers within the channel. Increasing the FFT size requires the use of processing chips (either DSPs or pipeline processors) that are faster and have greater memory capacity. However from the viewpoint of FFT requirements, implementations with up to 8 000 carriers are within the range of current technology. A more stringent requirement, however, is the receiver phase noise requirements implied by systems with a very large number of carriers. It has been reported that current consumer receiver technologies can provide satisfactory operation of systems with upto 8000 carriers . The design of multi-carrier systems must also consider the effects of frequency selective fading. Even where guard intervals are used to overcome intersymbol interference, in-band fading can still exist which may cause severe amplitude and/or phase distortion to high order QAM signals. For example, if a very strong (0 dB) echo is present on an uncoded OFDM system it can increase the power of 2/3 of the OFDM carriers while decreasing the power of the remainder. However the effect of the carriers suffering a decrease in power outweighs the positive effect of those having the increase and an overall BER in the vicinity of 10-1 would be obtained even though system C/N was 12 dB or more. However the situation changes dramatically for a coded multi-carrier system. If the frequency response of the channel can be measured (for example by using a training sequence) it is possible to effectively assign a signal-to-noise ratio to each OFDM carrier. This channel state information can be communicated to the error correction system, where it can be used to dramatically improve the system performance in the presence of echoes.

The system is most easily implemented using convolutional codes and a soft decision Viterbi decoder.

As an example of the improvement possible, an uncoded system was considered to fail (Viterbi decoder suffering a BER of 10-4) in the presence of a -4.5 dB echo but with the addition of rate 3/4 convolutional coding (k=7) with channel state estimation, the system was able to operate at an echo level of 0 dB. Research into optimum QAM-OFDM codes is ongoing. Areas of study include the appropriate code rates, and determination of appropriate interleaving factors. One of the distinct advantages of MCM over SCM with an adaptive equalizer is that MCM is less sensitive to variations in delay, as long as the multipath falls within the guard interval and the interleaver can effectively decorrelate the faded signal. Adaptive equalization performs better on short delay multipaths and is less effective on long delay multipaths. Therefore, MCM may be a better candidate for single-frequency networks (SFN).

5.5.3 Co-channel interference from analogue TV

Single carrier systems are robust to tone interference since signal power is spread over the entire spectrum.

For a single carrier system, an adaptive equalization can be used to reduce the severity of co-channel analogue television interference.

Another approach, for single carrier systems, is to use comb filtering to create notches in the spectrum at the receiver which align with the frequencies of the unwanted interfering carriers.

Multi-carrier systems can be sensitive to co-channel interference because of the very low power in each carrier. An MCM system is especially vulnerable to the non-flat spectrum of co-channel analogue TV as carriers located near the luminance, chrominance and audio carrier frequencies may suffer from strong interference.

One approach to avoiding this problem is to delete from the multi-carrier ensemble those carriers likely to suffer interference. However the disadvantage of this approach is that the data carrying capacity of the deleted carriers is lost at all points in the DTTB coverage area, even those locations where co-channel or adjacent channel interference would not have been a problem. This approach should perhaps not be rejected out-of-hand, particularly for difficult co-channel cases, as careful selection of a small number of carriers (principally around the interfering vision carrier) for deletion might produce a benefit of up-to (about) 10 dB with only a small proportion of data loss.

A second approach that avoids this disadvantage is to apply error coding to the multi-carrier system. As with the case of coding to improve the performance of the multi-carrier system in the presence of multipath, it is necessary to estimate the state of the channel - the amount of interference on each carrier. One way of achieving this is to switch the OFDM off for short periods and measure the interference power. An interleaver and channel estimator combined with a soft-decision decoding algorithm could be used to combat co-channel analogue TV interference. Using this technique on a real OFDM system it has been reported that in an extensive field trial, protection ratios of better than 0 dB were easily achievable. It is also worth noting that in locations where there is no co-channel or adjacent channel interferer, the error coding provides a residual level of error-correction capability which will improve the system's resilience against other interference.

5.5.4 Peak and Average Power Ratio Issues

Both single and multi-carrier modulations have an essentially noise-like spectrum. For single carrier modulation, the peak-to-average power ratio depends on the filter roll off. Faster roll off (which will have higher spectral efficiency) will result in a higher peak to average power ratio. It has been reported that for 99.99% of the time, the peak-to-average power ratio of a simulated 8 VSB single carrier system is 6.9 dB or less. (Lower values may be obtained if peak limiting is applied but in this case an increased level of adjacent channel energy will be produced which may well require additional transmitter filtering.) Some ATV systems may be able to take advantage of the asymmetric shape of analogue television receiver input filters to transmit more power or to implement a pilot carrier (which would improve the system ruggedness at low C/N) without increasing co-channel interference.

It is also noted that multi-carrier systems with a flat spectrum and a large number of carriers can be modelled as Gaussian distributions. Table 14 provides measured data on the peak to average power ratios of a typical COFDM signal.

Table 14

Peak to Average Ratio Measurements

Peak to Average Ratio

dB

99%

6.5

99.5 %

7.0

99.9 %

8.2

99.99 %

9.5

99.999 %

10.3

Desired Signal Level : -10 dBm

If clipped to the 95% value an Es/No penalty of less than 0.25 dB applies for BERs of 10-3. However the effects on adjacent channel filtering requirements need further consideration. If spectrum shaping is used in the multi-carrier system, opening holes in the spectrum, a few more dB gain may be achieved. But of course this will reduce the effective data rate of the multi-carrier system.

 

Continue to Section 5.6

Return to DTTB Tutorial Table Of Contents

Return to Tutorial Index Page