1084 THE BELL SYSTEM TECHNICAL JOURNAL, SEPTEMBER 1956 



of errors counted. Random code was used in this case and the repeater 

 bias was adjusted to provide equal omission and insertion thresholds. 

 The threshold for this particular case occurred when the peak to peak 

 sinusoidal interference was 63 per cent of the signal amplitude. This is 

 lower than the theoretical maximum which with a constant bias centered 

 at the half amplitude point, would be 100 per cent of the peak to peak 

 signal amplitude. For the bias conditions illustrated on Fig. 12, this 

 percentage would be 86 per cent for the positive insertion threshold and 

 88 per cent for the minus omission. This becomes apparent when the 

 negative and positive excursions of the interfering sine wave are con- 

 sidered as minus and positive impulses respectively. The remaining loss 

 in the interference margins can easily be due to maladjustments of tim- 

 ing, quantized feedback or inhibiting. 



When the frequency of the sinusoidal interference is varied, the 

 number of errors for a constant interference voltage at the blocking os- 

 cillator emitter does not change appreciably. However, the input trans- 

 former and condenser coupling introduce a substantial frequency charac- 

 teristic. This reduces considerably the errors caused by power line 

 crosstalk. One of the striking things about the sinusoidal interference 

 errors is the rate at which they increase above the threshold. For ex- 

 ample, a change of 1 per cent of the interference amplitude can triple 

 or quadruple the total number of errors. 



5.0 SUMMARY 



New techniques and devices now" make it possible to build practical 

 regenerative repeaters for use in digital transmission. Such a repeater 

 which is suitable for a 12-channel, 7-digit PCM system, is discussed. 

 Simple, inexpensive devices are used to eliminate the effects of distortion 

 due to low frequency cutoff and to provide self timing for the circuit. 

 Experimental evidence is presented which shows the repeater to func- 

 tion as expected. 



ACKNOWLEDGEMENTS 



I am deeply indebted to J. V. Scattaglia for his aid in tliis project and 

 to the pioneering work of A. J. Rack on quantized feedback which was 

 of great help in the development of this regenerative repeater. I also 

 wish to thank W. R. Bennett, C. B. Feldman and Gordon Raisbeck for 

 their aid and many valuable suggestions. 



