REGENERATION OF BINARY MICROWAVE PULSES 89 



consider noise with a peak level 5 db below slicing level (S/N =11 db) 

 at the input. The loss at this level is 5 db resulting in a noise level 10 db 

 below reference to give a S/N ratio of 16 db. We see that a 4 db improve- 

 ment in S/N ratio at the input results in an 8 db improvement in this 

 ratio at the output. 



Everything which was said above concerning the circulating loop ap- 

 plies equally to a chain of identical repeaters. To set the effective slicing 

 level at half amplitude at each repeater in a chain one would first find 

 two points on the sheer characteristics such as P and P' of Fig. 18. The 

 point P should be in the region of expansion and P' in the limiting region. 

 Also the points should be so chosen that a 6 db increase of input from 

 that at point P results in a 6 db increase in output at the point P'. If 

 now at each repeater we adjust pulse peak amplitude at the sheer input 

 to a value corresponding to that at point P' we will have unity gain 

 from one repeater to the next at levels corresponding to pulse peaks. 

 We will also have unity gain at levels corresponding to one half of pulse 

 amplitude. The effective slicing level is thus set at half amplitude. Ob- 

 viously the procedure for setting the slicing level at some value other 

 than half amplitude would be practically the same. It should be pointed 

 out that although half amplitude is the preferred slicing level for base- 

 band pulses this is not the case for carrier pulses. W. R. Bennett of Bell 

 Telephone Laboratories has shown that for carrier pulses the probability 

 that noise of a given power will reduce signal pulses below half amplitude 

 is less than the probability that this same noise will exceed half ampli- 

 tude. This comes about from the fact that for effective cancellation there 

 must be a 180° phase relationship between noise and pulse carrier. For 

 this reason the slicing level should be set slightly above half amplitude 

 for a carrier pulse system. 



The difference in performance between a perfect sheer and one with 

 characteristics such as shown on Fig. 18 are as follows: For the perfect 

 sheer no effects from noise or other disturbances are passed from one 

 repeater to the next. For the case of the imperfect regenerator some ef- 

 fects are passed on and so tend to accumulate in a chain of repeaters. 

 To prevent this accumulated noise from building up to the breaking 

 point of the system it is necessary to make the signal-to-noise ratio at 

 each repeater somewhat better than that which would be required with 

 the ideal sheer. For the case of random noise the required S/N ratio 

 seems to be about 5 or 6 db above the theoretical value. This is due in 

 part to sheer deficiency and in part to other system imperfections. 



