Full Rate - Technology

Technology

GSM-FR is specified in ETSI 06.10 (ETS 300 961) and is based on RPE-LTP (Regular Pulse Excitation - Long Term Prediction) speech coding paradigm. Like many other speech codecs, linear prediction is used in the synthesis filter. However, unlike most modern speech codecs, the order of the linear prediction is only 8. In modern narrowband speech codecs the order is usually 10 and in wideband speech codecs the order is usually 16.

The speech encoder accepts 13 bit linear PCM at a 8 kHz sample rate. This can be direct from an analog-to-digital converter in a phone or computer, or converted from G.711 8-bit nonlinear A-law or μ-law PCM from the PSTN with a lookup table. In GSM, the encoded speech is passed to the channel encoder specified in GSM 05.03. In the receive direction, the inverse operations take place.

The codec operates on 160 sample frames that span 20 ms, so this is the minimum transcoder delay possible even with infinitely fast CPUs and zero network latency. The operational requirement is that the transcoder delay should be less than 30 ms. The transcoder delay is defined as the time interval between the instant a speech frame of 160 samples has been received at the encoder input and the instant the corresponding 160 reconstructed speech samples have been out-put by the speech decoder at an 8 kHz sample rate.