A Non-Linear Encoder Using a Binary Counter for Pulse-Code Modulation

Abstract

A review of pulse-code modulation has shown that interest is again being taken in pulse-count encoders. On these and other p.c.m. systems, logarithmic companding is desirable but it is usually applied in a separate stage. This is an investigation into methods of incorporating the companding action in a pulse-count coder. A theoretical appraisal has revealed at least five possible methods of non-linear pulse-count coding. The first may be instrumented simply only at the encoder. Three methods may not instrumented conveniently at either end. The fifth method may be applied to p.c.m. because the same type of coder is required at both ends. The first method was checked experimentally by testing the transfer characteristic of a monostable multivibrator which had transistor resistance modulators as timing elements. A new timing circuit has been proposed which should give a symmetrical characteristic for positive and negative modulation. Experimentally, a small asymmetry was given, and it is concluded that a non-linear encoder could not be based upon a monostable multivibrator. To test the fifth method, low frequency models were developed for the two coders. The main component stage was a novel voltage-to-—freaguency convertor driven by a distorted triangular sweep waveform. The junction f.e.t. has a certain advantage as 5 single-ended compandor and has been used in the sweep generator. The test upon the encoder gave a symmetrical characteristic but on the overall test, errors due to a synchronism occurred. It is concluded that these errors are inherent in a system where the clock pulse frequency is smoothly varied. Finally, an alternative encoder giving a segmented characteristic, which should be error-free, is proposed.