20.11.2 Multichannel load increase

Companding gives an increase in the peak multichannel load. The increase is mainly determined by the compression ratio, but the unaffected level and both the attack and decay times all impact on the overall loading.

Considering the main contribution from the compression ratio, the power in an single average talker is given in Table 20.6 where

P = -12.9dBmO, g = 5.8 and t = 0.25 to derive L = -15dBmO.

When compression is applied to the signal then the parameters above are modified by the compression ratio to give

P_vo = -12.9/2dBmO, g = 5.8/2 and t = 0.25 to derive L_q = -lL48dBmO.

Using the new value for L (compressed) of-11.48dBmO new values for P (compressed) can be calculated from Equation 30.24. For example a 4MHz 960 channel system has an overload require­ment of +26.62dBmO with normal —15dBmO single channel load­ing. With compression the overload requirement is +29.73dBmO or 3.2dB higher.

It will therefore normally be necessary to drop the transmission level by 3.2dB (WdB) and to raise it again before expansion to avoid overloading the system (see Figure 20.17).

This adjustment to the transmission level will reduce some of the companding advantage by increasing the thermal noise contribution at the output of the expandor where the extra WdB gain is applied.

20.11.3 Compandor noise advantage

From Figure 20.16 it can be seen that while no speech is transmitted the noise power will be subject to an attenuation of 25dB at the expandor output.

The theoretical noise advantage will therefore be given by Equa­tion 20.38.

However, the circuit noise while speech power is transmitted is virtually unaffected by the compression/expansion process and there is no advantage. During talking, therefore, the conversation is subject to the full effect of the system noise. The effect of this noise produces an impairment to speech intelligibility that detracts from the companding advantage by an estimated 5dB.

Overall advantage is therefore given by Equation 20.39.

20.11.4 Attack and decay time

The compression and expansion process is speech power activated and it therefore takes time to recognise the increase (attack time) or decrease (recovery time) in speech power. Attack time is normally set at 5ms and recovery time 22.3ms. (CCITT, 1985, Recommenda­tion G.162.)

20.11.5 Usage of companders

Companders are recommended for use with speech transmission if the mean noise power of a circuit in any hour is greater than 40000pWOp (-44dBmOp). (CCITT, 1985, Recommendation G.143.) The comparable limit for telegraphy is 80000 pWOp (-41dBmOp). (CCITT, 1985, Recommendation H.21.)

20.12 Through connections

Within the network it may be required that a band of channels (typically group or supergroup) are allocated as a through connec­tion path without translation to voice band. This could be at a drop and insert point for instance as shown in Figure 20.18.

20.12.1 Through connection filter

In this situation the band of interest has to be cleaned before passed back into the modulation process. From Figure 20.1 and Figure 20.19 a typical demodulator filter Fb selects the required band of frequencies (band N) from the high order broadband of channels. Because of the selectivity of the filter Fb, the selected band of frequencies contain some of the adjacent channels from N+l and N-l.

Normally these are removed by subsequent filtering and demodu­lation to voice level. If presented for re-translation however the adjacent channels must be removed to prevent crosstalk. Sup­pression of all such possible crosstalk paths is less than 80dBmO.

Such filters are large and complex and are usually separate equip­ments.