Speech coding and channel coding

The speech coder is a regular pulse excited linear predictive coder (RPE-LPC) with long term prediction. This provides a net bit rate of 13kbit/s. It is a block based coder where the input samples are analysed in blocks with a 20ms duration. Work is also being carried out to specify a half rate speech coder which will effectively double the system capacity of GSM.

Before being assembled into the timeslots and frames, the digital speech and signalling data is encoded and interleaved. The speech coder output is divided up into three classes of bits and the most sensitive bits are encoded by adding parity check bits followed by a convolutional coder. Signalling data is encoded using a FIRE code. A process of interleaving is then used to spread the data blocks over a number of bursts.

For speech, an interleaving degree of 8 is used, i.e the speech block is spread over 8 bursts, whilst an interleaving degree of 4 is used for signalling. This overall process is shown in Figure 47.11, and the combined use of coding and interleaving provides good protection of channel data from the fading, dispersion and inter­ference effects on the radio path. With the addition of frequency hopping and diversity techniques, the GSM air interface is particu­larly robust.

One of the penalties to be paid for this is the overall transmission delay. The speech coder contributes about 25ms and the channel coding and interleaving a further 37ms. The rest of the transmission delay budget allows for analogue to digital conversions, 16kbit/s transmission and switching in various parts of the network. The overall one way transmission delay thus amounts to around 90ms. Such a delay means that echo control is necessary even on short national calls.

Table 47.4 GSM logical channels

Figure 47.11 GSM channel coding and interleaving

Дата добавления: 2015-11-16