1062 THE BELL SYSTEM TECHNICAL JOURNAL, SEPTEMBER 1956 



ferred to as "zero wander." In a regenerative repeater the trigger poten- 

 tial is tied to the zero level by a constant bias. Zero wander then will 

 produce a changing bias which reduces the signal to noise margins of 

 the repeater, or in some cases even prevents regeneration. Suppose, for 

 example, a transmission line is equalized so the ideal pulse train shown 

 on Fig. 1(a) will appear as Fig. 1(c) after being transmitted over the 

 line. The individual pulses have widened until the envelope of a sequence 

 of consecutive pulses shows as a ripple with a much smaller amplitude 

 than the individual pulse. If the pulse train distorted by low frequency 

 cutoff shown on Fig. 1(b) is transmitted over this line its output will 

 appear similar to that shown on Fig. 1(d). The portion of the signal where 

 the peak amplitude Hes below the trigger threshold will not be regener- 

 ated. 



1.2 Compensatio7i for Low-Frequency Distortion 



In the past many circuits have been devised to prevent zero wander, 

 but none have been completely satisfactory. The repeater described in 

 this paper effectively eliminates zero wander in a string of consecutive 

 repeaters by means of a new and simple method. This may be better un- 

 derstood by referring to Fig. 2. Here are represented two successive re- 

 peaters of a transmission system. These repeaters have what appears as 

 a conventional negative feedback loop consisting of a pair of resistors, R. 

 The function performed by this feedback loop bears little if any resem- 

 blance to the negative feedback of linear amplifiers and is referred to as 

 "Quantized feedback" in this paper.* 



Suppose an isolated pulse of amplitude P,„ is regenerated in repeater 

 M and is applied to the line through its output transformer. The low 

 freciuency cutoff" of this transformer will produce a transient response to 

 the regenerated pulse as given in (1). A spectrum analysis of the transient 

 tail shows that most of its energy occurs in the lower portion of the pass 

 band of the equalized line. Consequently, it will be transmitted over the 

 line to the next repeater with little if any frequency or phase distortion, 

 but will be attenuated by a factor a. This transient at the input of the 

 following repeater may be expressed as 



Tm - akMPMe~'' (2) 



where t is again measured from the end of the pulse. Suppose the re- 

 generation of the pulse at the output of repeater N is delayed by time ti 



* A paper by Rajko Tomovich entitled "Quantized Feedback" was published 

 in the I.R.E. Transactions on Circuit Theory. There are some fundamental dilTer- 

 ences in the meaning of the term, quantized feedback, as used in these papers. 



