82 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1956 



broad-band random noise an rms signal to noise ratio of 20 dl) was meas- 

 ured.* This compares Avith a ratio of 18 db as measured by Messrs. 

 Meacham and Peterson for a system employing complete regeneration 

 and a single repeater, f 



Recently, A. F. Dietrich repeated the circulating loop experiment at 

 a radio frequency of 11 kmc. His determinations of required signal-to- 

 noise ratios are substantially the same as those reported here. From the 

 various experiments we conclude that for a long chain of properly func- 

 tioning regenerative repeaters of i-he type discussed here practically 

 perfect transmission is obtained as long as the signal-to-noise ratio at 

 the input to each repeater is 20 db or better on an rms basis. In an operat- 

 ing system it might be desirable to increase this ratio to 23 db to take 

 care of deficiencies in automatic gain controls, power changes, etc. 



From the experiments we also conclude that the price we pay for using 

 partial instead of complete regeneration is about 3 to 4 db increase in 

 the required signal-to-noise ratio. In a radio system which provides a 

 fading margin this penalty would be less since the probability that two 

 or more adjacent links will reach maximum fades simultaneously is very ' 

 small. Under these conditions only one repeater at a time would be near 

 the breaking point and the system would behave much as though the 

 repeater provided complete regeneration. 



TIMING 



Although we have considered the problem of retiming of signal pulses 

 up to now we have not discussed the problem of obtaining the necessary ' 

 timing pulses to perform this function, but have simpl}^ assumed that a 

 source of such pulses was available. As w^as mentioned earlier timing I 

 pulses would probably be derived from the signal pulses in a practical »^ 

 system. These pulses would be fed into some narrow band amplifier 

 tuned to pulse repetition frequency. The output of this circuit could be 

 made to be a sine wave at repetition frequency if gaps between the input 

 pulses were not too great. Timing pulses could be derived from this sine 

 wave. This timing equipment could be similar to that used in these ex- 

 periments and described earlier. Further study of the problems of ob- 

 taining timing information is being made. 



* For Gaussian noise it is not possible to specif.y a theoretical value of minimum 

 S/N ratio without specifying the tolerable percentage of errors. For the number of 

 errors detectable on the oscilloscope it seems rasonable to assume a 12 db peak 

 factor for the noise. The peak factor for the signal is 3 db. The 6 db peak S/N 

 which would be required for an ideal regenerator then becomes 15 db on an rms 

 basis. 



t L. A. Meacham and E. Peterson, B. S. T. J., p. 43, Jan., 1948. 



" 



