JP2010109402A - DeltaSigma MODULATOR - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem wherein since a distribution of data after noise shaping deviates if a waveform clip is generated in a ΔΣ modulator, an analog output signal decreases in power when the analog output signal is generated through PWM conversion using the output waveform data after the noise shaping. <P>SOLUTION: A re-quantizer 42 re-quantizes data to be processed based upon input waveform data to generate output waveform data. A noise shaping filter 48 performs noise shaping processing on quantization noise generated in the output waveform data to generate feedback noise data to be fed back to the data to be processed. A clip detecting circuit 50 detects a waveform clip possibly occurring when the feedback noise data is fed back to the data to be processed. A selector 52 stops feeding the feedback noise data back when it is detected that a waveform clip is possibly generated. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a ΔΣ modulator that performs requantization on an input digital signal to convert it to a digital signal having a smaller number of bits, and reduces the influence of quantization noise at that time.

  In an electronic device having an audio reproduction function such as a silicon audio player or a CD (Compact Disc) player, a class D amplifier has been used for the purpose of power saving. The class D amplifier operates by receiving a pulse width modulation (PWM) signal or a pulse density (PDM) signal.

  The ΔΣ modulator receives, for example, a pulse code modulation (PCM) signal as an original signal, re-quantizes the original signal into a digital signal having a smaller number of bits, and outputs the digital signal. The delta-sigma modulator performs delta-sigma modulation in the process of requantization, and improves the SN ratio degradation due to quantization noise by noise shaping. The output signal of the ΔΣ modulator is converted into, for example, a PWM signal, amplified in class D, and then converted into an analog signal for driving a load such as a speaker.

FIG. 4 is a block diagram of a conventional ΔΣ modulator 2. The ΔΣ modulator 2 includes a requantizer 4, adders 6 and 8, and a noise shaping filter 10, and processes input waveform data X (z) to generate output waveform data Y (z). For example, when the ΔΣ modulator 2 is a multi-bit ΔΣ system, the output waveform data Y (z) is converted into a PWM signal by the PWM converter. Incidentally, X (z) and Y (z) represent z-transformed input waveform data x (t) and output waveform data y (t) in the time domain. The same applies to W (z) that appears later. The basic characteristic of the ΔΣ modulator 2 is that the transfer function of the noise shaping filter 10 that is a high-pass filter is H (z), and the signal input from the adder 8 to the requantizer 4 is transmitted to the requantizer 4. If the added quantization noise is Q (z), it is expressed by the following equation.
Y (z) = X (z) + (1-H (z)) Q (z) (1)

  The second term on the right side of Equation (1) represents the noise after the noise shaping of the quantization noise Q (z) with the transfer function (1-H (z)). This noise shaping process has the effect of shifting the quantization noise that is present on the average equally regardless of the frequency in the linear PCM method to a high frequency component. Due to this shift of the quantization noise to the high frequency band, the quantization noise is reduced in the low frequency band. Specifically, the Nyquist frequency of the input waveform data X (z) is increased by oversampling, the audible band is included in the low frequency band, and the centroid of the quantization noise is shifted from the audible band to the high frequency band. In addition, for example, by removing a high frequency band using a low-pass filter including a digital filter or an active filter, it is possible to restore the original signal with reduced quantization noise.

The quantization noise synthesized with the input waveform data x (t) using the noise shaping filter 10 can fluctuate at a high frequency, and the signal waveform after the quantization noise is synthesized is synthesized as shown in FIG. A waveform 14 that vibrates at a high frequency around the previous waveform 12 is superimposed. FIG. 6 is a schematic diagram showing the input waveform data x (t) (FIG. 6A) and the waveform data y (t) after noise shaping (FIG. 6B). The axis is the data value. FIG. 6 shows an example in which the input waveform is a sine wave. FIG. 6B conceptually shows the waveform data after the noise shaping process, and shows an approximate distribution range (vibration width of quantization noise) of the waveform data. In FIG. 6B, the upper and lower ends of the distribution range are represented by dotted lines 14u and 14d shown above and below the output (waveform 12 shown by a solid line) when noise shaping is not performed.
JP 2007-6317 A

  FIG. 7 is a schematic diagram showing input waveform data x (t) (FIG. 7A) and waveform data y (t) after noise shaping processing (FIG. 7B), as in FIG. FIG. 6 shows a state in which the fluctuation width of the sine wave as the input waveform is smaller than the data limit width (−1 to +1) and the data distribution after noise shaping is also within the limit width (−1 to +1). FIG. 7 shows a state where the amplitude of the sine wave is larger than 1, the input waveform reaches the upper limit “1” of the absolute value of the data, and a waveform clip is generated.

As described above, the waveform data after the noise shaping process is distributed up and down around the value corresponding to the input waveform data. Here, when the absolute value of the input waveform data is clipped to the upper limit or is close to the upper limit, a portion near the upper limit in the distribution of data generated corresponding to the quantization noise after noise shaping is clipped. Therefore, due to the clip (period P _{CL +} ) at the upper end (dotted line 14u) of the data distribution range, the data distribution caused by the quantization noise after noise shaping is biased to a smaller value than the original, and conversely the data distribution Due to the clip (period P _CL− ) at the lower end of the range (dotted line 14d), the distribution is biased toward a larger value than the original. Due to this distribution deviation, the analog signal reproduced based on the output waveform data is a value whose amplitude is defined by the data limit width (−1 to +1) in the periods P _{CL +} and P _{CL− in which} waveform clipping occurs. There was a problem that it became smaller and the power was lower than the original.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a ΔΣ modulator in which power-down of an output signal is suppressed when input waveform data is clipped or becomes close to it. And

  The delta-sigma modulator according to the present invention re-quantizes processing target data based on n-bit (n is a natural number) input waveform data, and outputs m-bit output waveform data (m is a natural number satisfying m <n). A requantizer to generate, a feedback circuit for performing noise shaping processing by ΔΣ modulation on quantization noise generated in the output waveform data, and generating feedback noise data fed back to the processing target data; and the processing target Clip detecting means for detecting that a waveform clip can occur when the feedback noise data is fed back to data, and feedback for stopping feedback of the feedback noise data when it is detected that the waveform clip can occur Stopping means.

  According to the present invention, since the noise shaping process is stopped during a period when the waveform clip can occur after the noise shaping process, the data distribution according to the quantization noise after the noise shaping does not occur. Therefore, the phenomenon that the distribution of the data is biased by the clip (hereinafter, sometimes referred to as a bias of the data distribution) does not occur, and the power down of the output signal due to the bias is avoided.

  Hereinafter, embodiments of the present invention (hereinafter referred to as embodiments) will be described with reference to the drawings.

  FIG. 1 is a schematic block diagram of a digital audio device 20 using a ΔΣ modulator according to an embodiment of the present invention. In the digital audio device 20, a PCM signal reproduced from a digital recording medium such as a CD or DVD is converted into a signal suitable for a class D amplifier that performs an amplification operation by an ON / OFF switching operation. First, a PCM signal regenerator 24 reads a PCM signal with high accuracy (16 to 24 bits) recorded on a digital recording medium 22 such as a CD or a DVD. The oversampler 26 performs processing such as interpolation on the digital signal read from the digital recording medium 22 to increase the sampling frequency and generate a high-rate PCM signal.

  The noise shaper 28 is configured using the ΔΣ modulator according to the present invention, re-quantizes the high-precision high-rate PCM signal input from the oversampler 26, and converts it into a low-precision (about 4 to 9 bits) PCM signal. . At the same time, the noise shaper 28 biases the frequency spectrum of the quantization noise generated at this time to a high frequency, and reduces noise in the audio band.

  The PWM converter 30 converts a low-precision PCM signal that is an output signal of the noise shaper 28 into a PWM signal. The class D amplifier 32 has, for example, an H bridge circuit, and switches the two push-pull circuit transistors that constitute the H bridge circuit on and off based on the PWM signal from the PWM converter 30. As a PWM signal amplified at the output terminal of the push-pull circuit, a normal phase PWM output and a negative phase PWM output are obtained. This output signal is demodulated into an analog signal by a low-pass filter (LPF) 34 composed of a second-order LC filter, and the speaker 36 is driven.

  FIG. 2 is a block diagram showing a schematic configuration of the ΔΣ modulator 40 constituting the noise shaper 28. The ΔΣ modulator 40 includes a requantizer 42, adders 44 and 46, a noise shaping filter 48, a clip detection circuit 50, and a selector 52. The ΔΣ modulator 40 uses the high-accuracy high-rate PCM signal from the oversampler 26 as input waveform data X (z), and the requantizer 42 basically performs re-quantization on X (z) to achieve low-accuracy. Output waveform data Y (z) which is a high-rate PCM signal is generated, and this output waveform data Y (z) is input to the PWM converter 30.

  By re-quantization by the re-quantizer 42, quantization noise Q (z) is added to the output waveform data Y (z). The ΔΣ modulator 40 performs noise shaping processing by ΔΣ modulation on the quantization noise Q (z), and feedback noise data fed back to data W (z) input as processing target data to the requantizer 42. A feedback circuit for generating

This feedback circuit includes an adder 46, a noise shaping filter 48, a selector 52 and an adder 44. The adder 46 extracts the quantization noise Q (z) from the output waveform data Y (z) of the requantizer 42. In the present embodiment, the adder 46 subtracts the output waveform data Y (z) from the input data W (z) to the requantizer 42 and outputs -Q (z). That is, the adder 46 performs an operation represented by the following equation.
W (z) -Y (z) =-Q (z) (2)

  The noise shaping filter 48 is a high-pass filter having a transfer function H (z), receives −Q (z) from the adder 46, converts the power spectrum of the quantization noise Q (z), and returns feedback noise data. (-H (z) Q (z)) is generated and output. The feedback noise data (−H (z) Q (z)) output from the noise shaping filter 48 can be input to the adder 44 via the selector 52.

  The adder 44 adds and synthesizes the output data of the selector 52 to the input waveform data X (z), and generates processing target data W (z) of the requantizer 42.

  The selector 52 is controlled by the output of the clip detection circuit 50 and selectively selects one of the feedback noise data (−H (z) Q (z)) input from the noise shaping filter 48 and the value “0”. Output.

The clip detection circuit 50 feeds back the feedback noise data (−H (z) Q (z)) to the processing target data W (z), that is, the adder 44 uses the input waveform data X (z) and the feedback noise data. When (-H (z) Q (z)) is added and combined, the waveform clip limited by the upper limit value “+1” or the lower limit value “−1” is added to the processing target data w (t) obtained thereby. Detect what may happen. Clip detection circuit 50 in this embodiment determines that the inputted input waveform data x (t), the waveform clips may occur when the absolute value of x (t) exceeds a predetermined threshold value x _TH. The threshold value x _TH is the feedback noise data (−H (−) with respect to the input waveform data x (t) defined in the range (−1 to +1) or the upper limit value “+1” of the absolute value of the processing target data w (t). z) A value having a difference corresponding to the fluctuation range of Q (z)) is set. The fluctuation range of the feedback noise data (-H (z) Q (z)) basically takes into consideration the distribution form of the feedback noise data (-H (z) Q (z)) having a peak in the vicinity of 0. It is set and can be selected from various parameters representing the spread of the distribution. For example, the threshold value x _TH is set so that the distribution deviation, the maximum amplitude, and the like are (1−x _TH ).

  When the clip detection circuit 50 determines that the waveform clip does not occur, the clip 52 controls the selector 52 so that the selector 52 outputs the feedback noise data (−H (z) Q (z)) from the noise shaping filter 48. To do. On the other hand, when it is determined that a waveform clip occurs, control is performed so that the value “0” is output from the selector 52.

When the clip detection circuit 50 determines that a waveform clip does not occur, the processing target data W (z) of the requantizer 42 is given by the following equation.
W (z) = X (z) -H (z) Q (z) (3)

In this case, from the equations (2) and (3), the characteristic of the ΔΣ modulator 40 is
Y (z) = X (z) + (1−H (z)) Q (z) (4)
It becomes. That is, the characteristic of the ΔΣ modulator 40 in this case is the same as that of the conventional ΔΣ modulator 2 having the characteristic of the expression (1).

On the other hand, when the clip detection circuit 50 determines that a waveform clip can occur, the processing target data W (z) of the requantizer 42 is given by the following equation.
W (z) = X (z) (5)

In this case, from the equations (2) and (5), the basic characteristic of the ΔΣ modulator 40 is
Y (z) = X (z) + Q (z) (6)
It becomes. That is, in this case, the vibration of the quantization noise due to noise shaping is not superimposed on the signal waveform.

FIG. 3 is a schematic diagram showing input waveform data x (t) (FIG. 3A) and processing target data w (t) or output waveform data y (t) (FIG. 3B). Is the time, and the vertical axis is the data value. FIG. 3 shows a state where the input waveform is a sine wave and its amplitude is larger than 1 and a waveform clip is generated, as in FIG. FIG. 3B conceptually shows the waveform data after the noise shaping process, like FIG. 6B and FIG. 7B, and shows an approximate distribution range of the waveform data (quantum after the noise shaping). Vibration width of the noise). Upper and lower ends of the distribution range in Fig. 3 (b), respectively, from the output when no noise shaping (waveform 60 shown in solid _lines) (1-x TH) dotted positioned vertically away 62u, represented by 62d Yes. Based on the input waveform data x (t) and the threshold value x _TH , the clip detection circuit 50 detects periods P _{CL +} and P _{CL− in} which the distribution range of the data after noise shaping causes a waveform clip. Then, during the periods P _{CL +} and P _CL−, the noise shaping is stopped by controlling the selector 52. As a result, in the periods P _{CL +} and P _CL− , the output waveform data y (t) becomes a waveform that does not have quantization noise oscillation due to noise shaping, that is, the waveform 60. As a result, power down of the output signal due to the bias of the data distribution is avoided in the periods P _{CL +} and P _CL− .

Note that the clip detection circuit 50 does not estimate the occurrence of the waveform clip after noise shaping from the input waveform data x (t) based on the threshold set to x _TH <1, but instead of the input waveform data x (t). It is also possible to detect that a waveform clip is actually generated by setting a threshold value at x _TH = 1. On the other hand, if x _TH <1, the output signal power down due to the bias of the data distribution can be avoided in the periods P _B and P _A before and after the period in which the waveform 60 clips (period P _{C in} FIG. 3). .

Incidentally, the periods P _B and P _A can be expanded by decreasing x _TH , but since the distribution density of the data after noise shaping decreases as the distance from the waveform 60 increases, the amount of decrease in x _TH or the period P _B , P _As the amount of _A increases, the power-down suppression effect does not increase. Further, in the periods P _{CL +} and P _CL− including the periods P _B and P _A , the noise shaping effect is not obtained, and the quantization noise Q (z) of the requantizer 42 appears as it is. The ratio is lower than during noise shaping. Therefore, it is possible to select how much the threshold value x _TH or the periods P _B and P _A are set in consideration of those viewpoints.

  In a state where waveform clipping does not occur, the clip detection circuit 50 controls the selector 52 to pass the output of the noise shaping filter 48 to the adder 44. The operation of the present embodiment with respect to fluctuations in input waveform data in a range where no waveform clipping occurs is the same as that of the prior art described with reference to FIG.

  In the above-described embodiment, the selector 52 is used to stop noise shaping, but other configurations are possible. For example, data input from the noise shaping filter 48 to the adder 44 may be simply cut off, or the noise shaping filter 48 may be stopped.

  In the above-described embodiment, the case where the ΔΣ modulator 40 is a multi-bit ΔΣ system and a PWM signal is generated based on the output has been described. However, in the present invention, the ΔΣ modulator 40 is a 1-bit ΔΣ system. The present invention is also applicable to cases where a PDM signal is generated based on the output of the ΔΣ modulator 40.

1 is a schematic block diagram of a digital audio device using a ΔΣ modulator according to an embodiment of the present invention. It is a block diagram which shows the schematic structure of the delta-sigma modulator which comprises the noise shaper in embodiment of this invention. FIG. 6 is a schematic diagram showing input waveform data x (t) and processing target data w (t) or output waveform data y (t) when a waveform clip occurrence is detected in the embodiment of the present invention. It is a block block diagram of the conventional delta-sigma modulator. It is a schematic diagram which shows the signal waveform after synthesize | combining the quantization noise produced | generated using a noise shaping filter. It is a schematic diagram which shows the input waveform data x (t) and output waveform data y (t) when a waveform clip does not arise. It is a schematic diagram showing input waveform data x (t) and output waveform data y (t) in a state where waveform clipping occurs in a conventional ΔΣ modulator.

Explanation of symbols

  20 digital audio equipment, 22 digital recording medium, 24 PCM signal regenerator, 26 oversampler, 28 noise shaper, 30 PWM converter, 32 class D amplifier, 34 LPF, 36 speaker, 40 ΔΣ modulator, 42 requantizer , 44, 46 Adder, 48 Noise shaping filter, 50 clip detection circuit, 52 selector.

Claims (3)

a requantizer that re-quantizes data to be processed based on n-bit (n is a natural number) input waveform data to generate m-bit (m is a natural number satisfying m <n) output waveform data;
A feedback circuit that performs noise shaping processing by ΔΣ modulation on quantization noise generated in the output waveform data, and generates feedback noise data fed back to the processing target data;
Clip detection means for detecting that a waveform clip may occur when the feedback noise data is fed back to the processing target data;
Feedback stop means for stopping feedback of the feedback noise data when it is detected that the waveform clip may occur;
A ΔΣ modulator comprising: The ΔΣ modulator according to claim 1,
The feedback circuit is
A noise shaping filter that converts the power spectrum of the quantization noise to generate the feedback noise data;
An adder that combines the feedback noise data with the input waveform data to generate the processing target data;
Have
The feedback stop means is provided between the noise shaping filter and the adder, and when it is detected that the waveform clip may occur, the feedback noise data from the noise shaping filter to the adder is detected. Blocking input,
A ΔΣ modulator characterized by the following. The ΔΣ modulator according to claim 1 or 2,
The clip detection means is set with a threshold value having a difference corresponding to a fluctuation range of the feedback noise data with respect to an upper limit value of the absolute value of the input waveform data, and the absolute value of the input waveform data exceeds the threshold value A delta-sigma modulator characterized in that it is determined that said waveform clip can occur.

Images