H.l.M) W 11)111 AM) T/<A.\.S.\n.S.SK).\ PEKIURM A S'CE 51U 



We note that the balance disappears if instantaneous sampling is used 

 instead of gating because there is no longer any aperture discrimination. 

 The curve for instantaneous sampling is plotted on Fig. 14. (Calculation of 

 this curve brings out the fact that the amplitude of the interfering compo- 

 nents also depend on the framing phase difference between the two systems. 

 When the framing frequencies are in phase, the two waves have a constant 

 frequency difference, and the interference vanishes. We assume the phase 

 difference as equally likely to fall anywhere within a complete cycle and 

 average the received interference power over all phases. The curve is found 

 to approach an asymptotic ordinate of 9 db at minimum band width as the 

 frequency swings on the two systems approach zero together. The 9 db 

 limit is compounded of 3 db from the marginal ratio between the two 

 carriers, 3 db from averaging over the carrier phase difference, and 3 db from 

 averaging over the framing phase difference. W'hen wide bands and large 

 swings are used, the curve approaches parallelism with the dashed one of 

 Fig. 13 for fluctuation noise but about 15 db lower. Of this difference 9 db is 

 accounted for by the higher marginal mean power level. The remainder is 

 assignable to differences in spectral distribution ; in particular the r-f spec- 

 trum of idle similar system interference is concentrated in half of the band 

 instead of being uniformly spread as in fluctuation noise. 



The curve for gated similar system interference has been estimated by 

 assuming that, with all channels loaded and a wide frequency swing, the 

 performance is like that with fluctuation noise except for a 3 db correction 

 allowed for the more concentrated spectrum. This gives an asymptote on 

 the right parallel to the solid marginal curve of Fig. 13 and 12 db lower. At 

 the left the curve must approach the same asymptote as the one for instan- 

 taneous sampling. It then seems reasonable to assume that the interfer- 

 ence power is directly proportional to the number of active channels and 

 the curv^e for an average of one eighth of the channels loaded is obtained 

 by raising the full load curve 9 db. 



Fig. 15— PCM-AM 



The curves on the left show how the audio signal-to-noise ratio varies 

 with bandwidth, the audio noise being quantizing, or granularity, noise as 

 discussed in Appendix I. The number of digits per code symbol is n and 

 the number of digit values (including zero), i.e., the base, is b. Bandwidth 

 is 2, T where T is the time per pulse and is therefore 16 mc per digit. The 

 curves are, of course, a set of discrete points rather than continuous as shown. 

 The steep rise is to be contrasted with the 3 to 9 db per octave slopes of the 

 curves previously presented. 



The curves on the right plot the niaxinuini values (with a 3 db allowance 

 included) of peak noise or interference, referred to the highest pulse value, 



