Abstract: An analog signal to be converted into a high resolution digital signal is first converted to a coarsely quantized signal by a direct feedback coder. A group of successive coarsely quantized signals are then processed by a converter which produces the high resolution digital signal. In the converter, the processing is performed by an arrangement of two accumulators. The first accumulator provides an output signal indicative of the running sum of each group of coarsely quantized signals. The second accumulator has the capability of providing an accumulation of either the negative or the positive of the output signal of the first accumulator. A negative accumulation is used for a first portion of the group and a positive accumulation is used for the second portion of the group to provide a net sum. The net sum is a triangularly weighted accumulation of the successive signals in the group which occasions self-cancelling of the quantizing noise inherent to the direct feedback coder.

Abstract: Digital character signals are split into at least two bit groups of different binary significance. The first and most significant of the bit groups is registered and periodically sampled for conversion in a resistor network to an analog format. The bits of the second, or least significant, bit group are utilized to produce a signal train including in a character time a number of pulses determined by the value of the second bit group and which pulses are approximately evenly located in the interval of the character. Each pulse of the train, as it occurs, momentarily augments the most significant bit group.One embodiment is shown in which the signal train produced is successive overflow carries from a high speed accumulation operation initially utilizing the least significant bit group and a starting character and thereafter using the least significant bit group and the immediately preceding accumulation sum.

Abstract: In each character time of a pulse code modulation signal stream, the most significant bit group of the PCM character is reversibly incremented and decremented a plurality of times. The incrementing and decrementing is carried out in a selectable sequence for each discrete pair of subintervals in the character time. The selectable sequences are determined by the value of the least significant bit group of the character whereby the average value, over the character time of an analog signal derived from the changing most significant bit group information in each subinterval, is equal to the value represented by the character.

Abstract: A shift-companded, or n:m, pulse code, representing on a time-amplitude basis the information of a conventional, amplitude companded, pulse code, is converted to a linear, digital, amplitude, pulse code by first converting the shift-companded amplitude information into 2s-complement form with a number of bits sufficient to yield substantially the same resolution of the encoded information as did the shift-companded code form. Thereafter half of the 2s-complement code words are loaded into an accumulator to be averaged. The resulting digital word representing the average is in 2s-complement, linear PCM format and is thus readily useful for data processing or for further conversion to a commercial, companded, transmission code.