JP3089477B2 - Quantizer and quantization method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention requantizes the number of bits of each sample of a digital audio signal from a long one (for example, 20 bits) to a short one (for example, 16 bits), The present invention relates to a quantizer and a quantization method which are provided at an output side of a signal processing circuit to which a digital audio signal of bits is supplied and which can be applied to generate a 16-bit output.

[Conventional technology]

Without changing the sampling frequency, set the quantization bit rate to 20.
The following quantization methods are conventionally known as quantization methods for limiting the word length, which is shortened from bits to 16 bits or increased by signal processing.

(1) The input signal is directly rounded down, rounded up or rounded. It is just requantization.

(2) After adding dither to the input signal, rounding down, rounding up or rounding is performed (for example, see Japanese Patent Application Laid-Open No. 56-83).
No. 818). As the dither, for example, a dither having the size of LSB (least significant bit) is used.

(3) The input signal is subjected to noise shaping and then rounded down, rounded up or rounded. Noise shaping is to reduce noise on the low frequency side,
This is for correcting the characteristic of quantization noise, which is white noise, in terms of frequency.

(4) Dither addition and noise shaping processing are performed on the input signal, and then rounded down, rounded up or rounded.

[Problems to be solved by the invention]

The method (1) has a problem that quantization distortion is generated and quantization noise in which quantization noise is modulated by a signal is generated. However, the processing is the simplest of the four methods described above, and according to Lipshitz's paper (AES Tokyo Convention '89 Preliminary Proceedings, pp. 72-75), the audibility is not corrected, that is, Unweighted noise is minimal. Unweighted noise is UN, weighted (modified E
-Weighted) If we represent the noise as WN, according to the above paper,
(UN, WN = (1,0.907).

In the method (2), quantization distortion and quantization noise modulation are reduced, but audible noise is increased because dither is added. (UN, WN = 3,2.721).

In the method (3), quantization distortion is reduced, but quantization noise modulation occurs. (UN, WN = 2,0.295).

The method (4) can reduce the quantization distortion and the quantization noise modulation, and the weighting noise WN is relatively small.
However, among the methods having four circuit scales, the noise is the largest, and the unweighted noise is also the largest. (U
N, WN = 6,0.886).

SUMMARY OF THE INVENTION It is an object of the present invention to reduce quantization distortion and quantization noise modulation by selecting the above-described conventional quantization method adaptively to an input signal, and to reduce a audible noise. And a quantization method.

[Means for solving the problem]

According to a first aspect of the present invention, there is provided a signal processing circuit for performing predetermined signal processing on a digital input signal and outputting a processing content signal indicating the processing content of the signal processing and a digital processing signal as a signal processing result. A detection circuit for detecting a magnitude and a frequency band; a control signal generation circuit for generating a dither control signal and a noise shaping control signal based on a processing content signal and a detection result of the detection circuit; and a dither control signal for converting a dither size to a dither control signal. And a noise shaping circuit that performs a noise shaping process on the digital processing signal based on a noise shaping control signal and adds an output signal of the dither variable device. It is.

A second aspect of the present invention provides a signal processing step of performing predetermined signal processing on a digital input signal and outputting a processing content signal indicating the processing content of the signal processing and a digital signal processing as a signal processing result. A detection step of detecting a signal magnitude and a frequency band; a control signal generation step of generating a dither control signal and a noise shaping control signal based on a processing content signal and a detection result of the detection step; and dither control of dither magnitude. A dither varying step of varying based on a signal; and a noise shaping step of performing a noise shaping process on the digital processing signal based on a noise shaping control signal and adding an output signal of the dither varying step. Method.

[Action]

The process of adding dither and the processing of noise shaping may be necessary when the gain of the input signal or the like is needed, or when noise is increased due to these factors. Therefore, in the present invention, it is regarded that these processes are valid.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings. In FIG. 1, reference numeral 1 denotes a signal processing circuit such as an equalizer and a fader, and a signal after processing is generated from the signal processing circuit 1.

An output signal of the signal processing circuit 1 is supplied to an adder 2 and a parameter detection circuit 3, respectively. Parameter detection circuit 3
Detects the property of the signal to be requantized from the gain of the output signal of the signal processing circuit 1 and the like, and generates a detection circuit Pi. Further, the signal processing circuit 1 generates a detection signal Hi indicating whether the circuit configuration is simple or complicated. These detection signals Pi and Hi are supplied to the determination circuit 4.

The determination circuit 4 controls the control signal based on the detection signals Pi and Hi.
Generate k1 and k2. The control signal k1 is supplied to the multiplier 5, and the control signal k2 is supplied to the multiplier 6. The output signal of the multiplier 5 is supplied to the adder 2, and the output signal of the multiplier 6 is supplied to the adder 7. The output signal of the signal processing circuit 1 is supplied to the adder 2, and the output signal of the adder 2 is supplied to the adder 7. The output signal of the adder 7 is supplied to the quantizer 8, and the quantizer 8 generates an output signal. The quantizer 8 generates an output signal having a predetermined word length, for example, 16 bits by rounding down, rounding up or rounding off.

The control signal k1 is a control signal related to noise shaping. The output signal of the adder 2 and the output signal of the quantizer 8 are supplied to a subtractor 9. A quantization error is detected by the subtracter 9 when the value of the input signal is constant, and the quantization error is supplied to the multiplier 5 via the one-sample delay circuit 10. The feedback loop including the subtractor 9 and the delay circuit 10 forms a first-order noise shaping circuit. For example, when an input signal having a constant value of 7.1 is supplied to the adder 2 and the quantizer 8 generates an output signal having a value of 7, the subtracter 9 detects a quantization error of -0.1. The quantization error is supplied to the adder 2 via the delay circuit 10 and the multiplier 5. Therefore, accumulation of quantization errors can be prevented. Young, (k1
= "0"), the output signal of the multiplier 5 is 0, and the noise shaping process is not performed.

The control signal k2 is a control signal for controlling the process of adding dither. When (k2 = “1”), the dither is supplied to the adder 7, and the dither is added before the quantizer 8. When (k2 = "0"), no dither is added. The dither is, for example, a random data sequence having a value of 1 LSB.

FIG. 2 is a table showing control operations performed by the control signals k1 and k2. FIG. 2A shows a case where a 20-bit input signal is supplied to the signal processing circuit 1 and a 16-bit output signal is output from the quantizer 8. FIG. 2B is a table showing a control operation when obtaining a signal, and FIG. 2B is a table showing a control operation when supplying a 16-bit input signal to the signal processing circuit 1 and obtaining a 16-bit output signal from the quantizer 8. It is.

In this table, three types a, b, and c are prepared as parameters indicating the properties of the input signal, and the detection signal Pi corresponding to each of them is prepared.
Is generated by the parameter detection circuit 3.

a: 0 level input signal b: small level signal, direct current (DC) or low frequency input signal c: other input signal Also, depending on the complexity of the configuration of the signal processing circuit 1, The following d, e and f are identified by the detection signal Hi generated from the signal processing circuit 1.

d: The ratio of input to output gain is 1
E: a circuit having a simple configuration such as a fader or a quantizer f: a complicated signal processing circuit. As shown in FIG. 3, one-sample delay circuits 11 and 12 to which an input signal is supplied, coefficient multipliers 21, 22, and 23 to which input signals and output signals of the delay circuits 11 and 12 are respectively supplied, and output signals An equalizer composed of one-sample delay circuits 13 and 14 to which signals are supplied and coefficient multipliers 24 and 25 to which output signals of the delay circuits 13 and 14 are respectively supplied is an example of a complicated signal processing circuit.

Among these input signal and circuit configuration types, type a
Does not require noise shaping and dither addition regardless of the circuit configuration, and (k1, k2 =
“0”).

Regarding the input signal of the type b, when the word length is shortened (FIG. 2A), the control signal k1 can be set regardless of the circuit configuration.
And / or k2 is set to “1”. When the word length is shortened, quantization distortion occurs. As a countermeasure, noise shaping processing or addition of dither is performed.

If the word length does not change (FIG. 2B), noise shaping and dither addition processing are not performed when the circuit configuration is d, and these processings are performed when the circuit configuration is e and f.

With respect to the input signal of the type c, when the word length is shortened (FIG. 2A), the circuit configuration is the type of d and e, the control signal k1 and / or k2 is "1", and the circuit configuration is f When the type is set, both k1 and k2 are set to "0".

When the word length does not change (FIG. 2B), noise shaping and dither addition processing are performed only when the circuit configuration is e, and these processings are not performed when d and f.

The equalizer shown in FIG.
This is an example. In this circuit, since a large number of signals are added by the adder 26, the lower bits of the large number of signals perform a function similar to dither. Therefore, even when the type of the input signal is c and the circuit configuration is f, the noise shaping process and the addition of dither are not performed. As a result, unweighted noise can be minimized. Here, the weighted noise may be minimized by setting (k1 = 1, k2 = 0).

In the above-described embodiment, a primary noise shaping circuit is shown. However, various configurations such as a secondary or higher noise shaping circuit may be used.

〔The invention's effect〕

According to the present invention, the processing of adding dither or noise shaping is performed according to the input signal and the signal processing. That is, when the input signal is absent or when the input / output transfer function of the preceding signal processing circuit is 1, generation of noise in the quantizer can be prevented. Further, when the signal processing circuit performs complicated processing and the input signal to the quantizer is not small or not a signal of a very low frequency, the increase in quantization noise is minimized. Further, in the cases other than the cases described above, the occurrence of quantization distortion and the occurrence of quantization modulation noise are minimized.

[Brief description of the drawings]

FIG. 1 is a block diagram of one embodiment of the present invention, FIG. 2 is a schematic diagram showing a control operation of this embodiment, and FIG. 3 is a block diagram of an equalizer which is an example of a signal processing circuit. Explanation of main reference numerals in the drawings 1: signal processing circuit, 2: adder for noise shaping, 3: circuit for detecting the type of input signal, 4: judgment circuit, 7: adder for adding dither .

Continuation of front page (56) References JP-A-2-25116 (JP, A) JP-A-2-143713 (JP, A) JP-A-60-5691 (JP, A) JP-A-61-146021 (JP) , A) JP-A-61-158217 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) H03M 7/36

Claims (2)

(57) [Claims] A signal processing circuit for performing predetermined signal processing on a digital input signal and outputting a processing content signal indicating the processing content of the signal processing and a digital processing signal as a result of the signal processing; A detection circuit that detects a magnitude and a frequency band of a control signal generation circuit that generates a dither control signal and a noise shaping control signal based on the processing content signal and a detection result of the detection circuit; A dither variable device that varies based on a dither control signal; and a noise shaping circuit that performs noise shaping processing on the digital processing signal based on the noise shaping control signal and adds an output signal of the dither variable device. A quantizer characterized by the above. 2. A signal processing step of performing predetermined signal processing on a digital input signal and outputting a processing content signal indicating the processing content of the signal processing and a digital signal processing as a result of the signal processing; A detection step of detecting a magnitude and a frequency band of the control signal; a control signal generation step of generating a dither control signal and a noise shaping control signal based on the processing content signal and a detection result of the detection step; A dither varying step of varying based on a dither control signal; and a noise shaping step of performing noise shaping processing on the digital processing signal based on the noise shaping control signal and adding an output signal of the dither varying step. A quantization method characterized by the above-mentioned.