JPH0295017A - Non-linear quantizing device - Google Patents

Abstract

PURPOSE:To increase the quality of a reproducing signal by feeding back quantizing noise through a delaying device. CONSTITUTION:The title device is composed of an adder 2, a non-linear decoder 7, a quantizing noise detector 9 and a delaying device 11. A non-linear decoding is executed based on the non-linear quantized quantizing segment and quantizing level, the difference between the decoded signal and the inputted signal is detected, next, added to the input of the non-linear quantizing device 4 and the quantizing noise is decreased. Thus, while the system of the non-linear quantization is obtained, a non-linear quantizing device to decrease the quantizing noise and improve the quality can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a nonlinear quantization device that converts an analog signal into a digital signal by nonlinear quantization.

Conventional technology Converting an analog signal to a digital signal was usually performed by linear quantization, but in order to reduce the amount of digital data, nonlinear quantization, which is easy to configure and inexpensive, has been used. became.

FIG. 2 shows a block diagram of this conventional nonlinear quantizer. In the conventional nonlinear quantization device configured as described above, the number of quantization bits consisting of quantization segments and quantization levels is reduced, and the number of quantization levels within the quantization segment interval is also reduced. As a result, the size of the quantization step, which is the minimum unit of quantization noise, increases.

Problems to be Solved by the Invention However, with the above configuration, up to now, for example, 16
Linear quantization of bits is now divided into segments representing the quantization interval and bits representing the level within that interval, each with 4 bits and 12 bits, respectively, resulting in increased noise due to quantization errors. This has the problem that the quality of the reproduced signal is sacrificed.

In view of this, an object of the present invention is to provide a nonlinear quantization device that reduces quantization noise and improves quality while using a nonlinear quantization method.

Means for Solving the Problems In order to solve the above problems, the present invention performs nonlinear decoding based on nonlinearly quantized quantization segments and quantization levels, and decodes the decoded signal and input signal. The difference is detected and then added to the input of the nonlinear quantizer to reduce quantization noise.

Effect: With the above-described configuration, the present invention can reduce quantization noise compared to nonlinear quantization without feedback of quantization noise, and improve the signal-to-quantization noise, that is, the S/N ratio. By doing so, the quality of the reproduced signal can be improved.

Embodiments Hereinafter, embodiments of the nonlinear quantization device of the present invention will be described with reference to the drawings.

FIG. 1 shows a block diagram of a nonlinear quantization device in an embodiment of the present invention. In Fig. 1, 1 is the input analog signal, 2 is the output signal 12 from the delay device.
3 is the output signal from adder 2, 4 is a nonlinear quantizer that converts the output signal 3 from adder 2 into digital, and 5 is the output signal from nonlinear quantizer 4. quantization segment signal indicating the quantization interval, 6
is a quantization level signal indicating the quantization level output from the nonlinear quantizer 4, and 7 decodes the quantization segment signal 5 and quantization level signal 6 to set 1/2 of the sampling frequency as the power/soto-off frequency. A nonlinear decoder including a low-pass filter; 8 is the output from the nonlinear decoder 7; 9 is a quantization noise detector that takes the difference between the output signal 8 from the nonlinear decoder 7 and the output signal 3 from the adder 2; 10 is a quantum A delay device 11 delays the output signal from the quantization noise detector 9 and the output signal 10 from the quantization noise detector 9 by one sample time of the time sampled by the nonlinear quantizer 4. 12 is a delayed output signal from the delay device 11; FIG. 3 is a diagram schematically showing signal and noise spectra in analog/digital conversion. Analog/digital conversion inevitably introduces errors because it converts a continuous analog signal into a discrete digital signal. This error is
This is generally called quantization noise. Furthermore, it is known that in order to demodulate the original signal by digital/analog conversion, it is sufficient to perform sampling at twice or more the frequency spectrum of the original signal according to the sampling theorem. FIG. 3(a) shows the spectrum when the original signal is sampled at the sampling frequency f, and aliasing noise accompanying the sampling of the original signal appears every f5. Quantization noise appears at a constant level determined by the resolution of the analog-to-digital converter. Therefore, in order to extract the original signal during demodulation, it is necessary to remove frequency components higher than the Nyquist frequency, which is 1/2 of the sampling frequency. At the time of demodulation after the low-pass filter, quantization noise is distributed along with the spectrum of the original signal from the DC to the Nyquist frequency. FIG. 3(b) shows the spectral distribution when the feedback loop of the present invention is configured. Since the phase of the feedback signal is delayed by the delay device in the feedback loop, it becomes a negative feed signal in some frequency bands and a positive feed signal in other frequency bands, resulting in a band where quantization noise decreases and a band where it increases. This shows that they appear alternately. Figure 4 shows the slot-soto configuration and input and output quantization when a quantization segment S = 3 bits and a quantization level R = 3 bits are allocated in a slot of quantization bits N = 6 bits. FIG. 2 is a diagram showing the configuration of quantization segments and quantization levels. Here, the fact that both the segment and the level are 3 bits means that each has only 8 levels of gradation, and the input signal first has a segment indicating which of the 8 levels it falls into. separated,
Next, the segment is divided into levels representing which of the eight levels within that segment it is closest to. The value becomes one sample data consisting of 3 bits each as a quantization segment and a quantization level. FIG. 5 is a symbolic representation of the block diagram of FIG. 1, where 53 represents a delay device that represents a delay of only one sampling period, 51 represents a signal adder, and 52 represents a signal subtracter. Also, X(n) + Y
(n), U (n), Q (n), and R(n) treat the signals as time-series signals that change in synchronization with the sampling period. Q (n) here represents the quantization noise. G represents an amplifier, and if G=1, it will output with a gain of 1 in positive phase, and if G=-1, it will output with a gain of 1 in reverse phase. Here, considering the negative feedback state, we will deal with case G--1.

In other words, the quantization noise Q (n) is the error between the input X (n) and the output Y (n) of the nonlinear decoder, so Q (n
) =U (n) -Y (n) or Y (n) =U (n) +Q (n) where N n
=(L L 2. ...) The quantization noise Q (n) is delayed by one sampling time and its phase is inverted by the delay device Z-1, and then added to X (n) to form a nonlinear quantum quantizer, i.e., the input of the nonlinear quantizer is U (n) = X (n) − Q (n
-1). Eliminating U(n) from both equations above, we get Y(n
), Y (n) = X (n) + (Q (n) −Q
(n-1)), and if the quantization noise from the delay device is not fed back, Y (n) = X (n) 10 (n), which is the state of the nonlinear quantizer in the conventional case. represents.

This quantization noise component (Q (n) −Q (n −1
)) means that the quantization noise one sampling time ago is subtracted from the current quantization noise, which means a differential operation. Also, expressing the delay device as a complex number function, Z-'=e-no T w=2πf1 j2=-1, T
=1/fs, and this quantization noise component is.

Q (n) X (1-Z-') = Q (n) The quantization noise of the low frequency component approaches O. Here, W is the angular frequency, fS is the sampling frequency, T is the delay time of the delay device, and π is 3.14.
shows. Further, when the gain frequency characteristic is determined, G (f) = 2X I S in (πf/fs). Here, 11 means taking the absolute value, (f)
means that it is a function of frequency. Quantization noise Q
If (n) is white noise, the power spectral density of the conventional quantization noise is a constant value γ0 that does not depend on the frequency, and the power spectral density of the quantization noise of the present invention γ(f)
From the above formula, γ (f) = l G (f) + 2Xγθ
=4 (Sin (πf/fs))2X7e. This noise component is attenuated by a low-pass filter in the nonlinear decoder, and the residual noise power NQ' between DC and the cutoff frequency fc of the low-pass filter is NQ'=f'O1lγ(f)df, where 1 .l'fcOdf means integrating a function of frequency f from 0 to frequency fc.

Therefore, if πf<<fs, S in Cyr f/ fs)'=yr f/
fs, and the power spectrum density of the quantization noise can be expressed as 7 (f)44 (πf/fs)2Xγ++. Therefore, the residual noise power between DC and the cutoff frequency fc of the low-pass filter is NQ' = J ''94
(πf/fs) 2d f=4π27efc3/31
's2. Figure 6 shows the frequency spectrum of the quantization noise expressed by the above equation, and the shaded part NQ' in Figure 6 is the quantization noise after the low-pass filter with the cutoff frequency fc. It represents electric power.

As described above, according to this embodiment, by sampling at a frequency higher than the spectral distribution of the original signal and providing feedback of the quantization noise via the delay device, the cutoff frequency fc of the low-pass filter of the nonlinear decoder is lower than or equal to According to the present invention, according to the present invention, the quantization noise power included in the original signal band can be drastically reduced while the original signal remains unchanged. Although the number of quantization bits is small, nonlinear quantization has the advantage of greatly improving the S/N ratio and quality, and its practical effects are great.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a nonlinear quantizer in the first embodiment of the present invention, FIG. 2 is a block diagram of a conventional nonlinear quantizer, and FIG. 3(a) is a block diagram of a conventional nonlinear quantizer. Fig. 3(b) is a spectrum diagram of signals and noise in the analog/digital conversion of the present invention; Fig. 4 shows quantization bits N = 6 bits, quantization segment S = 3 bits, FIG. 5 is a circuit diagram in which the block diagram of FIG. 1 is represented by symbols, and FIG. 6 is a frequency spectrum diagram of quantization noise. 1... Analog signal to be input, 2... Adder, 3
... Output signal from adder 2, 4... Nonlinear quantizer, 5... Quantized segment output signal from nonlinear quantizer 4, 6... Quantization from nonlinear quantizer 4 Level output signal, 7... Nonlinear decoder, 8... Output signal from nonlinear decoder 7, 9... Quantization noise detector, 10... Output signal from quantization noise detector 9, 11
. . . Delay device, 12 . . . Output signal from the delay device 11.

Claims (1)

[Claims] means for adding the analog input and the output from the delay device; means for nonlinear quantization for converting the output from the adder into a quantization segment and a quantization level; and a low-pass filter for performing nonlinear quantization from the quantization segment and quantization level. and quantization noise detection means for subtracting the decoded signal and the output signal from the adder, the delay device delaying the output of the quantization noise detection means. A nonlinear quantizer featuring: