JP5099699B2 - Pulse width position modulation signal generator - Google Patents

Description

  The present invention relates to pulse width modulation used in full digital audio amplifiers, power electronics, and the like.

In a full digital audio amplifier, a pulse width modulation signal that is an input signal to a switching amplifier is generated by a digital circuit. Since the circuit is driven by a clock, the resolution of the pulse width is determined by the period of the clock. Since the pulse period is determined by other requirements, a high frequency clock is required to provide high resolution for pulse width modulation. However, since a clock with a very high frequency cannot be used in reality, the pulse width modulation cannot have a very high resolution. Patent Document 1 proposes a technique for doubling the resolution of pulse width modulation, but higher resolution is desired.
JP 2006-54815 A

  The problem to be solved is to improve the resolution of pulse width modulation for a given clock frequency and pulse period.

  As means for solving the problem, the resolution of modulation is improved by modulating not only the width of the pulse but also the position of the pulse in the pulse period. At that time, the resolution cannot be improved from the viewpoint of the area of the pulse. However, in the evaluation standard in which the second-order or higher filter is applied to the pulse width modulation signal, the filter output value is changed by changing the pulse position. , Resolution can be improved. In a full digital audio amplifier, quantization noise can be reduced when an output error is fed back to a pulse width modulation signal using a second or higher order noise shaping filter.

  FIG. 1 shows an example of the waveform of a pulse modulation output signal. As will be described later, a pulse width modulator that can output positive and negative pulses is used, and the pulse rises or falls in synchronization with the clock timing (broken line in FIG. 1). The pulse width modulator outputs a pulse having a specified pulse width and pulse position. Pulse width modulation is based on symmetric pulse width modulation, and the position of the pulse means how much the output pulse deviates from the center of the pulse in the conventional symmetric pulse width modulation. If it is negative. The pulse width and pulse position are specified in units of clock cycles. Therefore, when the pulse width is an odd number, the pulse position is an odd multiple of 1/2. Although a negative pulse is not shown in the figure, there is a negative pulse, and in this case, the pulse width is expressed by a negative number.

  Such modulation of the pulse width and position is called pulse width position modulation, and a pulse width position modulated signal is called a pulse width position modulation signal.

  The positive and negative pulse width position modulated signals can be generated using two pulse width modulators. FIG. 2 shows an example of a pulse width position modulation signal generator using two pulse width modulators 21 and 22. The difference between the outputs of the two pulse width modulators 21 and 22 is output as a pulse width position modulation signal. In FIG. 2, two pulse width modulation signals w1 (t) and w2 (t) are subtracted, but this may be subtracted as signal processing. Since the load is driven by the differential signal as shown in FIG. 3, the differential signal can be considered as the output signal with respect to the pulse width modulation signals w1 (t) and w2 (t).

FIG. 5 shows an example of the relationship between the two pulse width modulation signals w1 (t) and w2 (t) and the pulse width position modulation signal w (t). y1 and y2 are input signals to the two pulse width modulators and take integer values of positive, negative, and zero. When the value of y1 or y2 is zero, a pulse width modulation signal with a duty ratio of 50% is generated. When the values of y1 and y2 are equal, w1 (t) and w2 (t) are the same signal, and as a result, the pulse width position modulation signal remains constant at zero. When the inputs to the two pulse width modulators are y1 and y2, the pulse width of the pulse width position modulation signal is y, and the pulse position is p, the following relationship exists.
(Equation 1)
y = y1-y2
(Equation 2)
p = (y1 + y2) / 2
In this way, by changing the width and position of the pulse, the pulse width position modulation signal can have many variations. However, in practice, the low frequency region of the pulse width position modulation signal (full digital audio) In the case of an amplifier, it is necessary to reduce the influence of the quantization error in the audible region). Therefore, it is necessary to feedback compensate for the effect of shifting the pulse position.

  FIG. 5 shows the state of feedback compensation. The pulse width modulators 21 and 22 constitute a pulse width position modulator together with a subsequent subtractor. The input signal u [k] is a high-resolution PCM signal whose low frequency component becomes the target value of the low frequency component of the pulse width position modulation signal w (t). The compensator 3 inputs the input signal u [k] and the input signals to the two pulse width modulators 21 and 22, and outputs a scalar command signal z [k] for the pulse width position modulator. In the encoder 1, a combination of command signals y1 [k] and y2 [k] to two pulse width modulators that can output a pulse width position modulation signal with little error with respect to the input command signal z [k]. Output.

  The compensator 3 feedback compensates for an error including a quantization error in the pulse width position modulation so that the low frequency component of the pulse width position modulation signal w (t) follows the low frequency component of the input signal u [k]. Must be designed to be FIG. 6 shows the concept of the design method of the compensator 3. FIG. 6 shows the concept of the design method, and does not represent the actual configuration of the compensator 3. In order to consider the effect of the pulse width and pulse position on the pulse width position modulation signal w (t), the input signal u [k] is compared as a continuous time signal with a signal multiplied by zero-order hold, and the difference The signal is passed through a continuous time filter having integration and low emphasis characteristics, the output is sampled to calculate a compensation component, and added to the original input signal to generate a command signal z [k]. In this way, feedback compensation is performed on the pulse width position modulation signal output in the past. However, in practice, it is not a good idea to implement a continuous-time compensator. Then, the combination of the compensator 3, the encoder 1, and the pulse width position modulator 2 can be realized as shown in FIG. The parameters in the figure are as follows:

E (y1) and e (y2) are vector functions having the same scalar as an argument, and can be calculated as the following equation.

However, w1 (y1, t) is an output signal w1 (t) when the signal y1 is input to the pulse width modulator 21.

  The encoder 1 outputs a combination of command signals y1 [k] and y2 [k] to an appropriate pulse width modulator with respect to the command signal z [k]. The encoder 1 is designed as follows. First, all permitted combinations of y1 and y2 (referred to as an allowable set) are defined. Then, for each element of the allowable set, all the influence amounts on the compensation signal v [k] after one step when the value is input to the pulse width position modulator are examined. Then, a combination of y1 and y2 in which the absolute value of the compensation signal v [k + 1] after one step becomes as small as possible is found and those values are output. As a method of searching for an appropriate combination of y1 and y2, a table using the command signal z [k] as an argument is used. Another implementation method of the table is a method using a search tree, which is suitable when the encoder 1 is realized by random logic.

In the above, an example realized by two pulse width modulators as a pulse width position modulator has been shown. However, such a configuration is not necessary, and for one pulse width modulator, the width of the output pulse and the pulse It is also possible to modulate the position.

  The first embodiment of the present invention is an application of pulse width position modulation in a full digital audio amplifier having the configuration shown in FIG. Although not shown in FIG. 8, the speaker is driven through a switching amplifier and a low-pass filter as shown in FIG. 3 for the output of the pulse width position modulation signal.

The signal s [i] is a high-resolution PCM signal with a sampling frequency of 44.1 kHz, and the purpose of this apparatus is to cause the low-frequency component of the pulse width position modulation signal to follow the signal s [i]. The oversampler 6 converts the sampling frequency of the signal from 44.1 kHz to 705.6 kHz which is 16 times the sampling frequency. That is, the signal u [k] is a high-resolution signal with a sampling frequency of 705.6 kHz. The compensator 3 compensates for a pulse width position modulation quantization error and signal distortion, and outputs a command signal z [k] for the pulse width position modulation. Details of the compensator 3 will be described later. The command signal z [k] is a PCM signal with a sampling frequency of 705.6 kHz. The encoder 1 generates command values y1 [k] and y2 [k] to the appropriate pulse width position modulator 2 for the command signal z [k]. The pulse width position modulator 2 theoretically has a structure as shown in FIG. 9, and the command values y1 [k] and y2 [k] are command values to the two pulse width modulators 21 and 22. ing. In terms of circuit, the pulse width position modulator 2 has a structure shown in FIG.

  The pulse width modulators 21 and 22 are 64 level pulse width modulators, and the command values y1 [k] and y2 [k] to them are allowed to be 61 levels. As a result, the pulse width position modulator 2 has a resolution of 121 levels with respect to the pulse width. The clock of the pulse width modulators 21 and 22 is 45.1584 MHz which is 128 times the carrier frequency 352.8 kHz, the control frequency is 705.6 kHz, and the carrier frequency of the output signal of the pulse width position modulator 2 is also 705.6 kHz.

  The structure of the compensator 3 is shown in FIG. A, b, and c forming the noise shaping filter have a transfer function

It is set to become. The nonlinear function vectors 35a and 35b are calculated by Equation 5 using a continuous-time state variable expression filter obtained by converting the state variable expression of the noise shaping filter into a continuous-time relationship assuming a zero-order hold. Is. The feedforward compensation element 7 is a cubic function described in Japanese Patent Application Publication No. 2006-54800, and reduces harmonic distortion due to pulse width position modulation.

  The maximum shift amount of the pulse position is 5 clocks (−5 ≦ p ≦ 5). The 512-level quantization is performed on the pulse width position modulation command signal z [k], and a combination of y1 and y2 that minimizes the absolute value of the compensation signal v [k + 1] after one step at each level is obtained. It is calculated in advance and referred to as a table. Since the command signal z [k] has been requantized, an optimal combination of y1 and y2 cannot be obtained, but an almost optimal (suboptimal) combination of y1 and y2 can be obtained. Suboptimal has little impact on quantization noise reduction.

  The effect of the pulse width position modulation is as follows. First, FIG. 10 shows a spectrum in a low frequency region in a pulse width modulation signal when conventional symmetric pulse width modulation is performed. Since the same pulse width modulation clock is 45.1584 MHz, the resolution is 64 levels. The input signal is a 4.13 kHz and 5.51 kHz two-tone signal with a maximum amplitude of 0.7.

  Next, FIG. 11 shows a spectrum in a low frequency region in a pulse width modulation signal when pseudo-symmetric pulse width modulation that allows a pulse width of an odd number of clocks is performed. The resolution of pulse width modulation is 128 levels. It can be seen that the quantization noise in the high band is reduced by about 6 dB compared to the case of symmetric pulse width modulation.

  FIG. 12 shows the spectrum in the low frequency region of the pulse width position modulation signal when the pulse width position modulation is used in the first embodiment of the present invention. It can be seen that the quantization noise in the high band can be reduced by about 9 dB as compared with the case where symmetric pulse width modulation is used, and by about 3 dB as compared with the case where pseudo symmetric pulse width modulation which allows an odd pulse width is used.

  Further, FIG. 13 shows a spectrum in the low frequency region of the pulse width position modulation signal when the pulse width position modulation in the first embodiment of the present invention is used and the level of the input signal is lowered. The quantization noise in the high band is further reduced by about 5 dB, about 14 dB compared to the case using symmetric pulse width modulation, and about 8 dB reduced compared to the case using pseudo symmetric pulse width modulation that allows an odd pulse width. is made of. Thus, the degree of reduction in quantization noise due to the introduction of pulse width position modulation depends on the signal level. The reason is that when the pulse width is narrow, the effect of shifting the position of the pulse is small, so high resolution can be realized, but when the pulse width is wide, the effect of shifting the pulse is large, and the effect of shifting the pulse by one clock. However, it becomes close to the effect of changing the pulse width by one clock, so that it is impossible to expect improvement in resolution by pulse shift. The degree of such a phenomenon varies depending on the setting of Expression 6, which is a transfer function of the noise shaping filter.

  In the first embodiment of the present invention, the encoder 1 is implemented as a table, but other implementation methods may be used, and combinations of optimum command values y1 [k] and y2 [k] All the boundary values of the area of the command signal z [k] are stored, and the optimum command values y1 [k] and y2 [k] are obtained by comparing these values with the value of the command signal z [k]. It may be calculated.

  In the first embodiment of the present invention, a structure that uses two pulse width modulators to output a differential signal as a pulse width position modulator is used, but without using a differential signal, A non-differential pulse width position modulator may be used.

  In the first embodiment of the present invention, the pulse width position modulator is applied to a full digital audio amplifier. However, the pulse width position modulator may be applied to a digital-analog converter or may be applied to other products. .

  When the pulse width position modulation method of the present invention is used, it is possible to provide a higher resolution than the conventional pulse width modulation method, so that it is included in the output signal in the case of application to a full digital audio amplifier, for example. Quantization noise can be reduced.

An example of a waveform of a pulse width position modulation signal. The block diagram which shows the example of the pulse width position modulator implement | achieved by two pulse width modulators. The block diagram which shows the implementation example of the pulse width position modulator in a full digital audio amplifier. The figure which shows the relationship between the pulse width modulation signal and pulse width position modulation signal in the pulse width position modulator shown in FIG. The block diagram which shows the structural example of the pulse width position modulator containing a compensator. The block diagram for demonstrating the design method of a compensator. The block diagram explaining the structure of a compensator. The block diagram explaining the structural example of the full digital audio amplifier by pulse width position modulation. The block diagram explaining the 1st Embodiment of this invention. The example of the spectrum of the pulse width modulation signal at the time of using the conventional symmetrical pulse width modulation. The example of the spectrum of the pulse width modulation signal at the time of using the conventional pseudo symmetric pulse width modulation. The example of the spectrum of the pulse width position modulation signal with respect to the high level input in the 1st Embodiment of this invention. The example of the spectrum of the pulse width position modulation signal with respect to the low level input in the 1st Embodiment of this invention. .

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Encoder 2 ... Pulse width position modulator 21, 22 ... Pulse width modulator 3 ... Compensator 31 ... Delay device 32 ... System matrix 33 ... Output vector 34 ... Input vector 35a, 35b ... Nonlinear function vector 36 ... Zero order hold 37 ... Sampler 38 ... Integrator 41,42 ... Switching amplifier 5 ... Speaker 6 ... Over Sampler 7 ... Feed forward compensation element

Claims (2)

A pulse width position modulator for outputting a rectangular signal having a pulse width and a pulse position corresponding to the pulse command signal; and a target signal for the pulse command signal to the pulse width position modulator and an output signal of the pulse width position modulator A compensator that inputs an input signal and outputs a modulation command signal; and an encoder that inputs the command signal and outputs the pulse command signal. there is a pulse width but include those specifying the different said pulse positions, wherein the compensator internal state of the compensator be the pulse width is the same for the pulse command signal information of the pulse position is different An apparatus for generating a pulse width position modulation signal, wherein the influences on the pulse width are different. The pulse width according to claim 1, wherein the pulse width position modulator includes two pulse width modulators, and the pulse command signal includes respective pulse width command values for the two pulse width modulators. Position modulation signal generator.