JPH0144052B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a nonlinear analog-to-digital converter, and particularly has the advantage of being able to select and output a polygonal compressed digital signal based on the μ-law or the A-law. For signals such as audio signals, where the probability of occurrence of small amplitudes is high and the probability of occurrence of large amplitudes is low, the linear analog-to-digital conversion method that performs uniform quantization is not efficient, so large amplitudes are relatively coarse. It is well known that a nonlinear analog-to-digital conversion method is used in which small amplitudes are quantized relatively finely, that is, the quantization step is changed depending on the input amplitude. One method of non-linear quantization is to first compress the input signal through a compressor and then uniformly quantize it to obtain a substantially non-linear quantized signal, which is a pulse code for speech. In the modulation communication method (PCM for audio), the input analog signal compressed according to a logarithmic function is further compressed by 15 or 13
A non-linear quantized signal obtained by uniformly quantizing after approximation using polygonal lines (segments) is used. The former is usually called a quantization method based on the 15-fold μ law, and the latter is called a quantization method based on the 13-fold A law.
Both are international standard non-linear analog-to-digital conversion methods based on the recommendations of the Consultative Committee on International Telegraph and Telephones (CCITT). Therefore, hereinafter, the former will be simply referred to as conversion based on the μ-law, and the latter will be simply referred to as conversion based on the A-law. Here, μ or A is a numerical value related to the compression ratio of the input signal, and the larger the value, the higher the compression ratio. μ and A in the μ-law conversion and A-law conversion are 255 and 87.6, respectively. On the receiving side, the original signal can be demodulated by passing the uniform quantization decoder through a decompressor whose nonlinear characteristics are opposite to those of the compressor. Compression and decompression are collectively called companding. ing. When this term is used here, the companding characteristic F(x) according to the above μ law (μ = 255) or A law (A = 87.6) is expressed by the formula
Given by (1) and (2). F(x)=sgn(x)ln(1+μ|x|)/ln(1+μ
)(1) F(x)=sgn(x)A|x|/1+lnA, (|x|1/
A) =sgn(x)1+lnA|x|/1+lnA, (1/A|x|
1) (2) However, μ=255, A=87.6, or sgn(x)
indicates the positive and negative polarities of the input analog signal x. Figures 1a and 2b and 2a and 2b are diagrams showing the relationship between analog signals and digital signals in μ-law conversion and A-law conversion, respectively. In b, the receiving side is shown using an 8-bit digital code each. In the conversion using the μ-law shown in Figures 1a and b, the compression characteristics according to the μ characteristic (μ = 255) are approximated by 15 broken lines (segments), including the positive and negative values, and in the conversion using the A-law shown in Figures 2a and b, , the compression characteristic according to the A characteristic (A=87.6) is approximated by 13 polygonal lines (segments) including the positive and negative by making the four positive and negative sections near the origin into one long polygonal line (segment). Therefore, the contact points of each broken line (segment) on the μ-characteristic or A-characteristic line naturally satisfy the respective compression characteristics. Here, these two
Comparing the polygonal line characteristics according to the two transformation rules, the ratio of the quantization steps in adjacent polygonal lines (segments) is all equal to 2, but when it comes to the quantization steps themselves, the transformation according to one of the μ-laws is 2, 4, and 8. ,16,
32, 64, 128, and 256, whereas in the case of the conversion according to the other A law, 2, 2, 4, 8,
16, 32, 64, and 128, so the difference near the small amplitude and large amplitude of the input analog signal x is quite large. By the way, no matter which conversion rule is used, it consists of 8 bits: one polarity bit S and seven signal bits e ₁ , e ₂ , e ₃ , e ₄ , e ₅ , e ₆ , e _7. In the quantized signal, the next 3 bits e ₁ , e ₂ , e ₃ excluding the first polarity bit S indicate a polygonal region where the sample value is included, and the latter 4 bits e ₄ , e ₅ , e ₆ , e ₇ Since it indicates the number of quantization steps within the broken line, the level difference between the quantization codes near the small amplitude and large amplitude of the input analog signal x is also quite large. However, these two
The most significant difference between the two conversion rules is the content of the output code of this quantized signal. In other words, according to international standards based on CCITT recommendations:
In the μ-law conversion, the full-scale value of the sample value is encoded as “0000000” and this quantization code is sent as an output code, whereas in the A-law conversion, the full-scale value of the sample value is The scale value is coded as “1111111”,
Further, for an 8-bit quantized code including the polarity bit S, it is specified that the output code is obtained by inverting the even-numbered bit codes. Therefore, there will be a significant difference between the digital conversion output codes under these two conversion rules. However, from the technical point of view, the selection of which code to use for the full-scale value is arbitrary, so as shown in Figure 2 a and b, the full-scale value of the sample value can be converted in the same way as the μ-law conversion. ,
The digital conversion output code of the A-law conversion when encoded with “0000000” is the polarity bit S.
For an 8-bit quantization code containing
The codes of the fifth and seventh odd-numbered bits are respectively inverted. That is, on the vertical axis of FIG. 2a and the horizontal axis of FIG. 2b, the quantized code before bit inversion is shown in brackets together with the digital conversion output code. The following table shows the above figures 1 a, b and 2 a, b.
Based on the μ-law conversion and A when both full-scale values are encoded as “0000000”,
The digital conversion output code of the conversion according to the law,
It is compiled into a list regardless of level. However, only the positive side of the sample value was displayed.

【table】

Claims (1)

[Scope of Claims] 1. A successive approximation type analog-to-digital conversion circuit comprising a successive approximation register and a digital-to-analog converter having μ-characteristic or A-characteristic compression characteristics, and a polarity bit final storage stage of the successive approximation register. A bit sign inversion circuit is selectively connected and inserted between the master stage and slave stage of each stage except for when the compression characteristic of the digital-to-analog converter is set to μ characteristic or A characteristic. , the successive approximation type analog-to-digital conversion circuit converts the bit sign of the μ-law or A-law analog-to-digital conversion code, which is stored in the master stage of the successive approximation register one bit at a time, through the bit sign inversion circuit to the first one. A non-linear analog-to-digital converter characterized in that the odd or even bits are simultaneously inverted and output except for the polarity bit. 2. A step quantization circuit of a weighting resistor that sets the compression characteristic of the digital-to-analog converter to μ characteristic, and a weighting resistor corresponding to the minimum step of the quantizing circuit having a resistance value the same as that of the weighting resistor at both ends thereof. resistance elements are inserted in parallel to make the compression characteristic A
2. A nonlinear analog-to-digital converter according to claim 1, further comprising mechanical switching means for switching characteristics.