1478 THE BELL SYSTEM TECHNICAL JOURNAL, NOVEMBER 1957 



bility to random noise. The small effect which does occur is a disadvan- 

 tage which can run up to some 3 or 4 db for ten or twelve channels, for 

 the multichannel as compared to the single channel system. 



2. Channel subdivision is advantageous over single channel use, with 

 regard to vulnerability to impulse noise. 



3. Channel subdivision is disadvantageous compared to single channel 

 use, with regard to vulnerability to single frequency noise. 



3.3.2 Noise Effects in Vestigial Sideband System 



Some brief discussion of the noise problems in the vestigial sideband 

 system under consideration has already been presented in Sections 2.1.3 

 and 3.1. That such problems may be important in the use of actual 

 telephone facilities has generally been checked by some unpublished tests 

 carried out by J. Mallett, of Bell Telephone Laboratories. 



The noise problem is most serious in the application of data transmis- 

 sion over Nl carrier. It is particularly significant for the Nl installations 

 previous to the most recent. The most recent installations are engineered 

 with distinctly more conservatism in regard to noise performance. As is 

 noted in Section 3.1, the principal characteristic of the Nl system that 

 affects this application is that it uses a compandor and that its design 

 for telephone use assumes a reduction of the noise by this compandor. 

 The reduction then is not realized for data signals. A second characteris- 

 tic is that Nl channels are exposed to impulse noise. The channels are of 

 course designed to limit such noise to the extent that it will not sensibly 

 impair telephone speech. But short data pulses are more vulnerable to 

 the impulse noise than speech. The 2B noise meter, which is normally 

 used for telephone noise measurements, does not read sharp noise im- 

 pulses according to their effect on data signals, and other methods of 

 measurement have been explored. 



A summary of some of Mallett's results is plotted in Fig. 6. Here the 

 reading on the 2B meter is compared with that on a level distribution 

 recorder which records peaks of 1 millisecond or longer. The "one per 

 cent" point is noted, which means that one per cent of the one second 

 intervals in the period of measurement contained one or more peaks (of 

 1 millisecond or longer duration) of the amplitude indicated. This is 

 found convenient as a measure of the error frecjuency tolerated in the 

 system considered (one in 10^ bits). 



The heavy dots indicate the correlation between the two sets of noise 

 readings on idle tested channels of an operating Nl system (other chan- 

 nels in the system being busy due to normal use) . All channels had com- 

 pandors in. Dots that mark the boundaries of a group of tests (usually a 

 single system in the group) are connected by straight lines. The open 



