SELF-TIMING REGENERATIVE REPEATERS 893 



I REGENERATION AND RETIMING 



1.0 General 



In an ideal regenerative repeater the received pulse train is sampled 

 at proper fixed intervals, to determine whether a pulse is present. The 

 regenerated pulses transmitted into the next repeater section are all 

 of the same shape and amplitude, independent of the shape of the input 

 pulses. Thus pulse distortion from noise and other system imperfec- 

 tions is removed, provided the maximum distortion is held within proper 

 limits. Errors in the form of pulses in place of spaces, or conversely, are 

 encountered when these limits are exceeded. In a repeater chain there 

 will be cumulation of errors in proportion to the number of repeater 

 sections in tandem. However, the rate of errors in each section and thus 

 in the w^hole chain can be limited by a relatively small increase in the 

 signal-to-noise ratio of each section as the number of repeaters in tandem 

 is increased. This increase in signal-to-noise ratio with increasing length 

 of the repeater chain is much less than with ordinary nonregenerative 

 repeaters. For this reason regenerative rather than ordinary repeaters 

 are desirable, though not essential for systems employing binary codes. 



An ideal regenerative repeater with the above features would entail 

 rather complicated instrumentation for precise timing, sampling and 

 pulse regeneration. With partial rather than complete exact retiming 

 the repeaters can be simplified, in exchange for some sacrifice in per- 

 formance, as shown later. 



1.1 Regeneration Without Retiming 



It would be possible to have a repeater in which pulses would be re- 

 generated in amplitude and shape, but without retiming. Pulses would 

 in this case be regenerated when the amplitude of the pulse exceeded a 

 certain triggering level L. If the pulse shape is given by P(t), this would 

 occur at a time to such that 



P(to) = L. (i.n 



This would permit simple instrumentation, since regenerated pulses 

 would be triggered without separate sampling of the received pulse 

 train. With this method, however, timing deviations in the regenerated 

 I pulses would result from transmission distortion of the received pulses 

 I by noise and other system imperfections. These timing deviations would 

 I cumulate in a repeater chain and cause a reduction in the tolerance of 

 I the repeaters to noise, such that the signal-to-noise ratio would hsixe to 



