60.4.3.1 G3 signal transmission

G3 is a half duplex transmission which uses 2 types of modula­tion/demodulation; the first of 300bit/s for the T.30 protocol and handshaking between terminals. This 300bit/s V.21 modem is a robust FSK transmission. It is easily implemented and does not require training. It is used for the relatively low volume of data establishing machine parameters and providing feedback from re­ceiver to transmitter.

The second is the high speed modem used for the compressed image data transfer. In G3 there is a basic V.27 mode of 4.8kbit/s with 2.4kbit/s fallback where the lines are poor, and optional modes up to 9.6kbit/s and up to 14.4kbit/s for use where terminals and networks permit. It is generally acknowledged that over 80% of transmissions used these higher rates.

Table 60.7 compares the G3 modulation methods, where QAM refers to quadrature amplitude modulation.

60.4.3.2 Modem operation

The data rate modem takes a number of bits of the incoming data (e.g. 4 bits for V.29 9.6kbit/s), and for the symbol time (1/2400 second for V.29 9.6kbit/s) generates the relevant modulation condi­tion for these bits (16 conditions for V.29 9.6kbit/s).

The modulation condition may be generated by a controlled combination of sine and cosine modulated carriers which are then filtered before application to the network to limit damaging out of band signals. The receiving modem tracks the carrier to correct for network frequency shift and recovers the symbol timing. It then studies the signal during this symbol period to decide which of the possible states the received signal most closely matches and decodes that state to provide the digital bit stream.

To establish the symbol timing, and allow the receive modem to set equaliser conditions to compensate for the network impairments, an initial training signal is sent from the transmit modem. In G3 this is followed by a 1.5s period all '0' TCF, so that the machine can decide if it can correctly receive at this rate. If not the 300bit/s handshake reply initiates a new training sequence, which the trans­mitter may choose to send at a lower rate. With ECM it is often found desirable to operate at the higher data rates, accepting that a limited number of errors which can be rapidly corrected may pro­vide an overall increase in transmission efficiency.

60.4.3.3 14.4Kbills option

The CCITT option at 14.4kbit/s uses V.I7 TCM (Trellis Coded Modulation). The symbol rate is the same as V.29 9.6kbit/s which appears optimum for the telephone network, but the bits per symbol increase not to 6 as expected but to 7 to allow FEC (Forward Error Correction) data to be included. The 7th bit, bringing the number of states to be decoded at the receiver to 128, is generated from 2 of the data bits in each symbol interval. This introduced redundancy means that only certain decoded sequences are valid. The receiver knows from the recent signal history when an unlikely condition occurs and can make an educated guess at the correct data. It is this difference in operation which gives a 3dB signal/noise improve­ment in Gaussian noise tests, and noticeable end to end improve­ment over QAM in actual use.