A-law algorithm

The A- law method is primarily used in Europe Companding for analog audio signals in the telecommunications sector, which is standardized in Recommendation G.711 ITU -T. In this case, the signal for digitization is a non-linear characteristic, the so-called A- characteristic curve, quantized. The aim is to achieve the same resolution a higher dynamic range and a larger signal -to-noise ratio by large signal excursions coarser and small signal swing a finer resolution.

In North America and Japan, the μ -law -like process is used similar to the A-law, but is not compatible. For communication, for example, in a telephone conversation between Europe and USA, the digital data must be implemented by appropriate converter.

The A characteristic curve

The companding of the analog signals carried on a logarithmic quantization, the A- characteristic. It is defined as a continuous function:

The inverse function is:

Where:

The A curve has a high dynamics and high levels a little momentum at low signal levels.

The 13-segment characteristic

Recommendation G.711 describes further a piecewise linear approximation of the characteristic curve A, which is substantially better suited for use in digitally operated systems the so-called 13-segment characteristic.

The proposed PCM encoding with 8- bit words would result in use of a linear characteristic to a level-independent resolution of 1/ 128 of the maximum amplitude. According to 13-segment characteristic small signal levels are now accurately quantized at 1/2048 of Maximalampitude, high signal levels, however, only with a resolution of 1 / 32nd The signals are first digitized to 12-bit accuracy. The division into segments was such that compression and expansion can be very easily done by a 12 - bit to 8 -bit conversion and an 8- bit to 12 -bit conversion.

The audio signal is first digitized linear 13 -bit value in the form of an integer with sign. This 13 -bit value is shown in the following table as an 8- bit value:

It is this coding as 1.3.4 floating point number 8 bits (so-called mini- float ) and an integer value consider (1 bit sign, 3 exponent bit, 4-bit mantissa). The sign- bit in the result is inversely used ( n = not s ).

To avoid the continuous transmission of zeros in silence, the 8- bit words are in the data stream with 0x55, ie 01010101b exclusive - ORed so that bits 2, 4, 6 and 8 (from the perspective of the ITU standards ) or 6, 4, 2, 0 (from the perspective of the binary positioning system ) is inverted.

10293
de