JP2006313958A - Pwm signal generator, pwm signal generating apparatus, and digital amplifier - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a full digital amplifier for an audio amplifier or the like, wherein a carrier frequency in pulse width modulation is allowed to be variable to suppress concentration of electromagnetic noise radiation spectrums at a particular frequency and a sampling period variation type digital filter required for realizing the technology above is obtained. <P>SOLUTION: The sampling period variation type digital filter is used for a noise shaping filter in a delta/sigma modulator of the full digital amplifier so as to attain variations in the carrier frequency of the pulse width modulation. Further, the sampling period variation type digital filter is attained by varying a coefficient of the digital filter depending on the sampling period. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a switching amplifier based on pulse width modulation using a digital signal as an input.

  In an audio amplifier that drives a speaker or the like, a digital amplifier using a switching amplifier (class D amplifier) has come to be used because of its power efficiency. This digital amplifier includes a type that inputs an analog signal and a full digital amplifier that is a type that receives a digital signal. In full digital amplifiers, output signals can be generated without going through analog signals, making it possible to reduce the cost of audio systems and to achieve high performance while maintaining high energy efficiency. Has advantages.

  The operation of a general full digital amplifier will be described. An example of the configuration of the full digital amplifier is shown in FIG. The sound source signal r [i] is a pulse code modulation (PCM) signal. For example, when a sound source signal is obtained from a CD, the sampling frequency of the sound source signal r [i] is 44.1 kHz. The sound source signal r [i] is input to the oversampler 4 and converted into a PCM signal u [k] of 705.6 kHz whose sampling frequency is 16 times the sampling frequency of the sound source signal r [i]. The PCM signal u [k] is converted into a coarsely quantized PCM signal y [k] at the same sampling period. The resolution of the PCM signal y [k] is determined by the requantizer 1 and is the same as the resolution of the pulse width modulator 2. The PCM signal y [k] is converted into a PWM signal w (t) by the pulse width modulator 2. The noise shaping filter 3 compensates for the quantization noise generated by the requantizer 1 and the signal distortion generated by the pulse width modulator 2, and the audible range component of the PWM signal w (t) is the PCM signal u [k]. Try to deal with audible range components. Therefore, the audible range component of the PWM signal w (t) corresponds to the sound source signal r [i]. The PWM signal w (t) is supplied to the switching amplifier 5 and converted into a power signal. The PWM signal w (t) is supplied to a speaker as a load after passing through a low-pass filter 6 composed of LC.

  However, in a full digital amplifier, the switching amplifier is driven by a pulse width modulation (PWM) signal generated by a digital circuit. Therefore, when the carrier frequency in PWM is constant, it is strong at a frequency that is an integral multiple of the carrier frequency. There was a risk of radiating peak electromagnetic noise.

  An example of a spectrum of a PWM signal generated by a full digital amplifier in the prior art is shown in FIGS. The method described in Patent Document 3 was used for PWM signal generation. The input signal is a pulse code modulation (PCM) signal with a sampling frequency of 44.1 kHz, and is a sine wave with a frequency of 2.7563 kHz and a modulation rate of 82%. The pulse width modulation uses a 31-level symmetrical PWM, and its carrier frequency is 705.6 kHz. From FIG. 24, it can be seen that quantization noise in the audible range is well suppressed, but from FIG. 25, it can be seen that a large spectrum peak occurs at every integral multiple of the carrier frequency. If this signal is leaked by a method such as electromagnetic radiation, for example, radio interference is given to an AM radio or the like.

As a method for mitigating the influence of the internal operation signal and its harmonic frequency due to electromagnetic noise, a method of dynamically varying the clock frequency to spread the spectrum of electromagnetic noise is known. However, in order to apply the method of spreading the spectrum in the full digital amplifier, it is necessary to dynamically change the PWM carrier frequency in the pulse width modulator 2, which is the sampling of the noise shaping filter 3 which is a digital filter. It becomes necessary to change the period dynamically. However, since a noise shaping filter that maintains desired characteristics even when the sampling frequency is dynamically changed has not been proposed, the spread spectrum technique cannot be used in a full digital amplifier.
Japanese Patent Application No. 2004-298240 Japanese Patent Application No. 2004-047113 JP 2004-236617 A

  It is to realize a full digital amplifier that maintains desired performance while dynamically changing the carrier frequency of pulse width modulation, and to perform spread spectrum on the PWM signal generated by the full digital amplifier.

  As a means for realizing a noise shaping filter that maintains desired characteristics even when the sampling period varies, a method of changing a coefficient in calculation inside the noise shaping filter in accordance with the sampling period is used. However, it is necessary not only to change the coefficient but also to consider the structure of the noise shaping filter. Further, since there are some differences in the design method of the noise shaping filter depending on the relationship between the sampling period of the input signal and the sampling period of the output signal, description will be made for each case.

  First, a case where the sampling periods of the PCM signal u [k] and the PCM signal y [k] are the same and the sampling period varies will be described. The configuration of the full digital amplifier at that time is the same as in FIG.

  First, in order to determine the target frequency characteristic of the noise shaping filter, a continuous-time filter is designed and expressed as a state variable as follows.

Where u ^* (t) is a continuous time signal corresponding to the PCM signal u [k], w (t) is a PWM signal generated by a pulse width modulator, and v ^* (t) is a correction generated by a noise shaping filter. A continuous-time signal corresponding to the signal v [k], x ^* (t) is a state variable. The continuous time filter is made discrete time with a sampling period τ _k using zero-order hold. However, since the sampling period τ _k changes dynamically, the input signal u ^* (t) is represented by an intermediate time value between sampling points during sampling. This is shown in FIG. However, the correction signal v [k] samples the value at the sampling time. That is,
(Equation 2)
u [k] = u ^* ((t _k + t _{k + 1} ) / 2)
(Equation 3)
v [k] = v ^* (t _k )
And t _k is the k-th sampling time. Then, the following digital filter is obtained.

At this time, the state variable x [k] of the digital filter corresponds to a sample of the state variable x ^* (t) of the continuous time filter. That is, x [k] = x (t _k ), t _{k + 1} −t _k = τ _k . Therefore, even if the sampling period τ _k is different for each sampling, the stability of the discrete time filter is guaranteed, and the input / output transfer characteristics are close to the transfer characteristics of the continuous time filter.

Now consider the case of designing a filter of order n. At this time, since A (τ _k ) in Equation 4 is an n × n matrix, the number of coefficients of the digital filter increases and the amount of calculation of the digital filter operation also increases. Therefore, when designing the continuous-time filter of Formula 1, the matrix A (τ _k ) in Formula 4 is also block-diagonalized by block diagonalization of the matrix A ^* . The number of coefficients can be reduced, and the calculation amount of the digital filter operation can also be reduced. Furthermore, along with the block diagonalization of the matrix A ^* , the digital filter operation calculation is performed by making each element either 1 or 0 for the output vector c of the continuous-time filter of Equation 1. The amount can be further reduced. This is because the output vector does not change due to the discrete time conversion according to Equation 2, as shown in Equation 5.

  The above discrete-time filter assumes a zero-order hold for the input signal and keeps the value between sampling points constant, so this assumption causes signal distortion due to fluctuations in the sampling period. As a method of dealing with this signal distortion, a method of interpolating the value of the input signal between sample points can be considered. There are a number of possible interpolation methods.

First, consider a case where a signal waveform between sample points is approximated by a straight line as an input signal. This is shown in FIG. By performing the first-order approximation, it can be expected that the signal distortion due to the fluctuation of the sampling period is reduced as compared with the case where the zero-order hold is used. A hold element that approximates the response between sample points by connecting them with a straight line is called a triangular hold. Using the triangular hold, the continuous time signal u ^* (t) is interpolated to convert the continuous time filter represented by Equation 1 into discrete time as follows. However, the correction signal v [k] represents the value of the continuous time signal v ^* (t) at the sample time.

This digital filter is not strictly proper and includes a direct term from input to output.

Next, consider a case where a signal waveform between sample points is approximated by a quadratic curve as an input signal. This is shown in FIG. By performing the second-order approximation, it can be expected that the signal distortion due to the fluctuation of the sampling period is smaller than in the case of using the first-order approximation. Here, the value of u ^* (t) in t _k ≦ t ≦ t _{k + 2} is second-order approximated using the values of u (t _k ), u (t _{k + 1} ), and u (t _{k + 2} ) To do. However, the value of k is an odd number. Then, the second-order approximation u ^* (t) is as follows.

Using this interpolated time function, the continuous time filter expressed by Equation 1 is converted into discrete time. However, the correction signal v [k] represents the value of the continuous time signal v ^* (t) at the sample time. The value of k is an odd number.

Since the continuous time signal u ^* (t) is approximated by interpolation using the values of the PCM signal u [k] at three consecutive sampling times, the filter operation is performed at odd sample times and odd number samples. It differs from that at the sample time. Further, since causality does not hold for this digital filter, it can be used only when the input signal can be prefetched one sample or when the output signal can be delayed by one sample.

As another method, the continuous-time signal u ^* (t) is subjected to quadratic interpolation. However, quadratic interpolation is performed between sampling points using the sampling time of the filter operation and the value of one input signal therebetween. Think about the case. This is shown in FIG. Since the sampling interval of the input signal is narrower than the above-described method, it can be expected that the signal distortion with respect to the fluctuation of the sampling period is reduced.

Now, it is assumed that the PCM signal u [k] is the value of u ^* (t) at the sampling time of the loop shaping filter, and the input signal between the sampling points is newly sampled and input at every sampling point. Shall. The sampling time between the sampling points is arbitrary, but here, sampling is performed at the time of the middle point of the sampling, and the signal is u _c [k]. Then, the value of u ^* (t) in t _k ≦ t ≦ t _{k + 1} can be second-order approximated using the values of u [k], u [k + 1] and u _c [k] as follows: .

Using this interpolated time function, the continuous time filter expressed by Equation 1 is converted into discrete time. However, the correction signal v [k] represents the value of the continuous time signal v ^* (t) at the sample time.

This digital filter is not strictly proper and includes a direct term from input to output.

  Next, consider a case where the sampling period of the PCM signal u [h] is constant and the sampling period of the PCM signal y [k] varies. The noise shaping filter is operated in synchronization with the sampling of the PCM signal y [k]. The configuration of the full digital amplifier at this time is shown in FIG. By making the sampling period of the PCM signal u [h] constant, signal distortion due to dynamic fluctuation of the sampling period of the noise shaping filter can be suppressed. Even in this case, it is assumed that the target of the dynamic characteristic of the noise shaping filter is given by the continuous-time filter expressed by Equation 1.

It is assumed that the sampling of the PCM signal u [h] occurs at most once between each sample point of the PCM signal y [k]. Zero-order hold is applied to the PCM signal u [h]. This is shown in FIG. For the output signal v [k], consider the case where the value of the signal at the sample time is represented. Such a digital filter is as follows. Let t ^u _h be the time when the input u [h] is sampled, and t _k be the time when the output y [k] is sampled.
(Equation 17)
t ^u _h ≤ t _k <t _{k + 1} <t ^u _{h + 1}
In other words, when the input signal is not sampled between the output sampling points, the following occurs.

Also,
(Equation 20)
t ^u _h ≦ t _k <t ^u _{h + 1} ≦ t _{k + 1} <t ^u _{h + 2}
In other words, when the input signal is sampled once between output sample points, the following occurs.

  In the above-described method, the zero-order hold is applied to the PCM signal u [h]. However, a noise shaping filter can be similarly designed even when a triangular hold is used or when another interpolation method is used.

  Next, consider a case where a noise shaping filter is calculated in accordance with the sampling period of the PCM signal u [h]. Since the sampling period of the PCM signal v [k] dynamically changes, the output of the PCM signal v [k] does not necessarily synchronize with the calculation period of the noise shaping filter. For the PCM signal u [h], a zero-order hold is used to convert the continuous time filter expressed by Equation 1 into discrete time. In this case, if the PWM signal w (t) is used as it is, the table necessary for the calculation of the noise shaping filter becomes large, so the PWM signal w (t) is approximated by the PCM signal y [k].

Here, the desired noise shaping filter is as follows. When τ ^u = t ^u _{h + 1} -t ^u _h , the noise shaping filter to be obtained is as follows.
(Equation 23)
t _k-1 ≦ t ^u _h ≦ t _k <t ^u _{h + 1} ≦ t _{k + 1}
If it is,

(Equation 25)
t _k-1 ≦ t ^u _h ≦ t _k <t _{k + 1} <t ^u _{h + 1} t _{k + 2}
If it is,

It becomes. Although the above method assumes that zero-order hold is used for input, a digital filter can be similarly designed even when no hold is used or when a triangular hold is used.

  By using the PWM signal generator of the present invention, it is possible to dynamically change the PWM carrier frequency without causing much signal distortion. As a result, the electromagnetic noise generated from the full digital amplifier has a specific frequency. The conventional problem of having a strong peak in can be overcome.

  The best mode for carrying out the present invention will be described through examples.

The first embodiment of the present invention is a full digital audio amplifier using a PWM signal whose carrier period dynamically varies. The configuration is as shown in FIG. 1 and FIG. The sound source signal r [i] is a PCM signal with a sampling frequency of 44.1 kHz, and is input to the oversampler 4. Although the over-sampler 4 is converted to the PCM signal u [k] of the sampling interval tau _k, rather than the sampling interval tau _k constant, 1/16 of the sampling interval of the sound source signal r [i] (approximately 1.472Myuesu) or One of 15/64 times (about 1.329μs) is taken, and which one is taken is determined by pseudo random numbers at almost the same rate. In the noise shaping filter 3, the filter operations shown in Equations 4 and 5 using 0th-order interpolation are performed, and the frequency shaping of the quantization noise in the PCM signal y [k] that is the output signal of the requantizer 1 is performed. Made and suppresses audible range components. The number of quantization steps of the requantizer 1 is 31 steps when the sampling interval τ _k is 1/16 times the sampling interval of the sound source signal r [i], and 29 steps when it is 15/64 times. The pulse width modulator 2 generates a PWM signal according to the PCM signal y [k]. The carrier signal period at this time is the same as the sampling interval τ _k and dynamically changes. The switching amplifier 5 is driven by the generated PWM signal, and the switching amplifier 5 drives a speaker as a load through the low-pass filter 6.

  Several spectra of the PWM signal w (t) in this fully digital audio amplifier are shown. FIG. 9 is an example of the spectrum of the PWM signal w (t) in the vicinity of the audible range, and the sound source signal is a sine wave having a frequency of 2.7563 kHz and a modulation rate of 80%. Although the second harmonic is slightly observed, it can be seen that the quantization noise in the audible range is suppressed. However, the noise floor in the audible range is slightly increased due to the influence of the interpolation error of the signal u [k] in the noise shaping filter as compared with the case where the sampling period of u [k] is constant. FIG. 10 shows a wide spectrum of the PWM signal w (t). It can be seen that the spectrum is spread by dynamically changing the carrier frequency of the PWM signal. Although the total amount of spectrum does not change much, it is possible to avoid the concentration of spectrum at a specific frequency, so that electromagnetic noise countermeasures are provided.

  FIG. 11 is a spectrum of the PWM signal w (t) in the vicinity of the audible range when the sound source signal is a sine wave having a frequency of 16.5378 kHz and a modulation rate of 80%. Since the frequency of the sound source signal has increased, the floor noise in the audible range has increased. This is due to the influence of the interpolation error of the signal u [k] in the noise shaping filter.

  One of the advantages of the first embodiment of the present invention is that the amount of calculation can be suppressed because the concept of zero-order interpolation is used for the calculation of the noise shaping filter.

The second embodiment of the present invention is a full digital audio amplifier using a PWM signal whose carrier period dynamically changes. The configuration is as shown in FIG. The sound source signal r [i] is a PCM signal with a sampling frequency of 44.1 kHz, and is input to the oversampler 4. Although the over-sampler 4 is converted to the PCM signal u [k] of the sampling interval tau _k, rather than the sampling interval tau _k constant, 1/16 of the sampling interval of the sound source signal r [i] (approximately 1.472Myuesu) or One of 15/64 times (about 1.329μs) is taken, and which one is taken is determined by pseudo random numbers at almost the same rate. In the noise shaping filter 3, filter operations shown in Equations 6 and 7 using primary interpolation are performed, and frequency shaping of quantization noise in the PCM signal y [k] that is an output signal of the requantizer 1 is performed. Made and suppresses audible range components. The number of quantization steps of the requantizer 1 is 31 steps when the sampling interval τ _k is 1/16 times the sampling interval of the sound source signal r [i], and 29 steps when it is 15/64 times. The pulse width modulator 2 generates a PWM signal according to the PCM signal y [k]. The carrier signal period at this time is the same as the sampling interval τ _k and dynamically changes. The switching amplifier 5 is driven by the generated PWM signal, and the switching amplifier 5 drives a speaker as a load through the low-pass filter 6.

  Several spectra of the PWM signal w (t) in this fully digital audio amplifier are shown. FIG. 12 is an example of the spectrum of the PWM signal w (t) in the vicinity of the audible range, and the sound source signal is a sine wave having a frequency of 2.7563 kHz and a modulation rate of 80%. Although the second harmonic is slightly observed, it can be seen that the quantization noise in the audible range is suppressed. The noise floor in the audible range is slightly higher than the case where the sampling period of u [k] is constant due to the influence of the interpolation error of the signal u [k] in the noise shaping filter. The increase is suppressed as compared with FIG. 9 where the 0th-order interpolation is performed. FIG. 13 shows a wide spectrum of the PWM signal w (t). It can be seen that the spectrum is spread by dynamically changing the carrier frequency of the PWM signal. Although the total amount of spectrum does not change much, it is possible to avoid the concentration of spectrum at a specific frequency, so that electromagnetic noise countermeasures are provided.

  FIG. 14 shows a spectrum of the PWM signal w (t) in the vicinity of the audible range when the sound source signal is a sine wave having a frequency of 16.5378 kHz and a modulation rate of 80%. Since the frequency of the sound source signal has increased, the floor noise in the audible range has increased. This is due to the influence of the interpolation error of the signal u [k] in the noise shaping filter. Floor noise of almost the same size as when the 0th-order interpolation is performed is generated.

  One of the advantages of the second embodiment of the present invention is that the first-order interpolation concept is used for the calculation of the noise shaping filter. Therefore, when the frequency of the sound source signal is low without greatly increasing the amount of calculation, the zero-order The floor noise due to the interpolation error can be reduced as compared with the case of interpolation.

The third embodiment of the present invention is a full digital audio amplifier using a PWM signal whose carrier period dynamically changes. The configuration is as shown in FIG. The sound source signal r [i] is a PCM signal with a sampling frequency of 44.1 kHz, and is input to the oversampler 4. Although the over-sampler 4 is converted to the PCM signal u [k] of the sampling interval tau _k, rather than the sampling interval tau _k constant, 1/16 of the sampling interval of the sound source signal r [i] (approximately 1.472Myuesu) or One of 15/64 times (about 1.329μs) is taken, and which one is taken is determined by pseudo random numbers at almost the same rate. In the noise shaping filter 3, filter operations shown in Expressions 10 to 13 based on the concept of quadratic interpolation are performed, and frequency shaping of quantization noise in the PCM signal y [k] that is an output signal of the requantizer 1 is performed. To suppress audible range components. The number of quantization steps of the requantizer 1 is 31 steps when the sampling interval τ _k is 1/16 times the sampling interval of the sound source signal r [i], and 29 steps when it is 15/64 times. The pulse width modulator 2 generates a PWM signal according to the PCM signal y [k]. The carrier signal period at this time is the same as the sampling interval τ _k and dynamically changes. The switching amplifier 5 is driven by the generated PWM signal, and the switching amplifier 5 drives a speaker as a load through the low-pass filter 6.

  Several spectra of the PWM signal w (t) in this fully digital audio amplifier are shown. FIG. 15 is an example of the spectrum of the PWM signal w (t) in the vicinity of the audible range, and the sound source signal is a sine wave having a frequency of 2.7563 kHz and a modulation rate of 80%. It can be seen that the second harmonic is well suppressed and the quantization noise in the audible range is also suppressed. The influence of the interpolation error of the signal u [k] in the noise shaping filter is also so small that it is hidden by the quantization error. FIG. 16 shows a wide spectrum of the PWM signal w (t). It can be seen that the spectrum is spread by dynamically changing the carrier frequency of the PWM signal. Although the total amount of spectrum does not change much, it is possible to avoid the concentration of spectrum at a specific frequency, so that electromagnetic noise countermeasures are provided.

  FIG. 17 is a spectrum of the PWM signal w (t) in the vicinity of the audible range when the sound source signal is a sine wave having a frequency of 16.5378 kHz and a modulation rate of 80%. Since the frequency of the sound source signal has increased, the floor noise in the audible range has increased. This is due to the influence of the interpolation error of the signal u [k] in the noise shaping filter. Floor noise of almost the same size as when the 0th-order interpolation is performed is generated.

  One of the advantages of the third embodiment of the present invention is that the second-order interpolation concept is used for the calculation of the noise shaping filter. Therefore, when the frequency of the sound source signal is low without increasing the amount of calculation much, The floor noise due to the interpolation error can be reduced as compared with the case of interpolation.

The fourth embodiment of the present invention is a full digital audio amplifier using a PWM signal whose carrier period dynamically changes. The configuration is similar to that of FIG. 2 except that the oversampler 4 outputs not only the PCM signal u [k] but also u _c [k], and u _c [k] is added to the input of the noise shaping filter 3. . The sound source signal r [i] is a PCM signal with a sampling frequency of 44.1 kHz, and is input to the oversampler 4. Although the over-sampler 4 is converted to the PCM signal u [k] of the sampling interval tau _k, rather than the sampling interval tau _k constant, 1/16 of the sampling interval of the sound source signal r [i] (approximately 1.472Myuesu) or One of 15/64 times (about 1.329μs) is taken, and which one is taken is determined by pseudo random numbers at almost the same rate. In the noise shaping filter 3, filter operations shown in Equations 15 and 16 based on the concept of quadratic interpolation are performed, and frequency shaping of quantization noise in the PCM signal y [k] that is the output signal of the requantizer 1 is performed. To suppress audible range components. The number of quantization steps of the requantizer 1 is 31 steps when the sampling interval τ _k is 1/16 times the sampling interval of the sound source signal r [i], and 29 steps when it is 15/64 times. The pulse width modulator 2 generates a PWM signal according to the PCM signal y [k]. The carrier signal period at this time is the same as the sampling interval τ _k and dynamically changes. The switching amplifier 5 is driven by the generated PWM signal, and the switching amplifier 5 drives a speaker as a load through the low-pass filter 6.

  Several spectra of the PWM signal w (t) in this fully digital audio amplifier are shown. FIG. 18 is an example of the spectrum of the PWM signal w (t) in the vicinity of the audible range, and the sound source signal is a sine wave having a frequency of 2.7563 kHz and a modulation rate of 80%. It can be seen that the second harmonic is well suppressed and the quantization noise in the audible range is also suppressed. The influence of the interpolation error of the signal u [k] in the noise shaping filter is also so small that it is hidden by the quantization error. FIG. 19 shows a wide spectrum of the PWM signal w (t). It can be seen that the spectrum is spread by dynamically changing the carrier frequency of the PWM signal. Although the total amount of spectrum does not change much, it is possible to avoid the concentration of spectrum at a specific frequency, so that electromagnetic noise countermeasures are provided.

FIG. 20 shows a spectrum of the PWM signal w (t) in the vicinity of the audible range when the sound source signal is a sine wave having a frequency of 16.5378 kHz and a modulation rate of 80%. It can be seen that the floor noise in the audible range is not increased even when the frequency of the sound source signal is increased. This is because the interpolation error is reduced by using the value u _c [k] between the sampling points in the interpolation of the signal u [k] in the noise shaping filter.

  One of the advantages of the fourth embodiment of the present invention is that the floor noise caused by the interpolation error can be reduced because the value between sample points and the concept of quadratic interpolation are used for the calculation of the noise shaping filter. However, the calculation amount in the oversampler 4 and the like increases accordingly.

The fifth embodiment of the present invention is a full digital audio amplifier using a PWM signal whose carrier period dynamically changes. The configuration is as shown in FIG. The sound source signal r [i] is a PCM signal having a sampling frequency of 44.1 kHz and is input to the oversampler 4. The oversampler 4 outputs a 705.6 kHz PCM signal u [h] whose sampling frequency is 16 times. Noise-shaping filter 3 and the quantizer 1 are outputs a PCM signal y sampling interval τ _{k [k],} rather than the sampling interval tau _k constant, 1-fold sampling interval of the PCM signal u [h] (approximately 1.472 μs) or 13/16 times (approximately 1.152 μs), and either value is determined by pseudorandom numbers at almost the same rate. In the noise shaping filter 3, filter operations shown in Expressions 17 to 22 based on the concept of 0th-order interpolation are performed, and frequency shaping of quantization noise in the PCM signal y [k] that is the output signal of the requantizer 1 is performed. To suppress audible range components. The number of quantization steps of the requantizer 1 is 31 steps when the sampling interval τ _k is 1/16 times the sampling interval of the sound source signal r [i], and 25 steps when it is 13/64 times. The pulse width modulator 2 generates a PWM signal according to the PCM signal y [k]. The carrier signal period at this time is the same as the sampling interval τ _k and dynamically changes. The switching amplifier 5 is driven by the generated PWM signal, and the switching amplifier 5 drives a speaker as a load through the low-pass filter 6.

  Several spectra of the PWM signal w (t) in this fully digital audio amplifier are shown. FIG. 21 shows an example of the spectrum of the PWM signal w (t) in the vicinity of the audible range. The sound source signal is a sine wave having a frequency of 2.7563 kHz and a modulation rate of 82%. It can be seen that the second harmonic is well suppressed and the quantization noise in the audible range is also suppressed. Since the sampling period of the PCM signal u [h] is fixed, no floor noise is generated due to the influence of the interpolation error. FIG. 22 shows a wide spectrum of the PWM signal w (t). It can be seen that the spectrum is spread by dynamically changing the carrier frequency of the PWM signal. Although the total amount of spectrum does not change much, it is possible to avoid the concentration of spectrum at a specific frequency, so that electromagnetic noise countermeasures are provided. Since the variation of the carrier frequency is large, the spread of the spectrum is also large.

  FIG. 23 shows a spectrum of the PWM signal w (t) in the vicinity of the audible range when the sound source signal is a sine wave having a frequency of 16.5378 kHz and a modulation rate of 82%. It can be seen that even when the frequency of the sound source signal is increased, no floor noise is generated in the audible range.

  One of the advantages of the fifth embodiment of the present invention is that since the sampling period of the PCM signal u [h] is constant, no floor noise is generated due to the interpolation error of the PCM signal u [h]. However, there is a problem that the calculation in the noise shaping filter 3 becomes complicated. However, the required number of sets of FIR filter coefficients in the oversampler 4 can be reduced.

  One of the other advantages of the fifth embodiment of the present invention is that the fluctuation interval of the sampling interval of the PCM signal y [k], that is, the carrier period of the PWM signal w (t) can be increased. Can be sufficiently diffused.

  In the fifth embodiment of the present invention, the filter operation shown in Expression 17 to Expression 22 is performed for each sampling period of the PCM signal y [k] in the noise shaping filter 3, but for each sampling period of u [h]. Alternatively, the filter calculation represented by Equation 23 to Equation 27 may be performed. At that time, the obtained result is the same except for the influence of rounding error in the calculation.

  By using the PWM signal generator and digital amplifier of the present invention, it is possible to realize a full digital amplifier that overcomes the conventional problem of generating electromagnetic noise having a strong peak at a specific frequency.

The block diagram which shows the structure of the 1st Example of this invention. The block diagram which shows the structure of a general full digital amplifier. The figure which shows the relationship between the signals of a 1st variable sample noise shaping filter. The figure which shows the relationship between the signals of a 2nd variable sample noise shaping filter. The figure which shows the relationship between the signals of a 3rd variable sample noise shaping filter. The figure which shows the relationship between the signals of a 4th variable sample noise shaping filter. The block diagram which shows the other structure of a full digital amplifier. The figure which shows the relationship between the signals of a 5th variable sample noise shaping filter. The spectrum in the vicinity of the audible range of the pulse width signal generated by the first embodiment of the present invention. FIG. 3 is a broadband spectrum of a pulse width signal generated according to the first embodiment of the present invention. FIG. The spectrum in the vicinity of the audible range of the pulse width signal for the high-frequency sound source signal generated by the first embodiment of the present invention. The spectrum in the vicinity of the audible range of the pulse width signal generated by the second embodiment of the present invention. FIG. 5 is a wideband spectrum of a pulse width signal generated by the second embodiment of the present invention. FIG. The spectrum in the audible range of the pulse width signal with respect to the high frequency sound source signal produced | generated by the 2nd Example of this invention. The spectrum in the vicinity of the audible range of the pulse width signal generated by the third embodiment of the present invention. FIG. 6 is a broadband spectrum of a pulse width signal generated according to the third embodiment of the present invention. FIG. The spectrum in the audible range of the pulse width signal with respect to the high frequency sound source signal produced | generated by the 3rd Example of this invention. The spectrum in the vicinity of the audible range of the pulse width signal generated by the fourth embodiment of the present invention. FIG. 6 is a wideband spectrum of a pulse width signal generated by the fourth embodiment of the present invention. FIG. The spectrum in the vicinity of the audible range of the pulse width signal for the high frequency sound source signal generated by the fourth embodiment of the present invention. The spectrum in the vicinity of the audible range of the pulse width signal generated by the fifth embodiment of the present invention. FIG. 6 is a wideband spectrum of a pulse width signal generated by the fifth embodiment of the present invention. FIG. The spectrum in the vicinity of the audible range of the pulse width signal with respect to the high frequency sound source signal generated by the fifth embodiment of the present invention. Spectrum near the audible range of a pulse width signal generated by conventional techniques. Broadband spectrum of a pulse width signal generated by conventional techniques.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Requantizer 2 ... Pulse width modulator 3 ... Noise shaping filter 31 ... Delay element 32 ... Square matrix 33 ... Output vector 34 ... Input vector 35 ...・ Nonlinear function vector 36 ・ ・ ・ Nonlinear element 4 ・ ・ ・ Oversampler 5 ・ ・ ・ Switching amplifier 6 ・ ・ ・ Low pass filter

Claims (6)

  In the PWM signal generator that receives the first PCM signal and outputs the PWM signal, the low-frequency component of the PWM signal corresponds to the low-frequency component of the first PCM signal. The PCM signal has a first sampling period, the first sampling period is constant, and the PWM signal is generated by digital means based on a second PCM signal having a second sampling period, and the second sampling period The period may vary for each sampling period according to an external command or a predetermined sequence, and the first PCM signal is converted into a second PCM signal by a delta-sigma modulator, The sigma modulator has a filter and a quantizer, and the filter inputs a first PCM signal and a second PCM signal and inputs a third PM signal. M signal is output, the third PCM signal is converted to the second PCM signal through the quantizer, and the gain of the quantizer dynamically changes in inverse proportion to the value of the second sampling period. And the coefficient and function in the internal calculation of the filter are relative to the second sampling period or the second sampling period and the timing of sampling of the first PCM signal and the timing of sampling of the second PCM signal. A PWM signal generator characterized in that it is determined by the relationship and may change dynamically.   In the PWM signal generation apparatus that receives the fourth PCM signal and outputs the PWM signal, the low-frequency component of the PWM signal corresponds to the fourth PCM signal, and the fourth PCM signal is the second PCM signal. The PWM signal generation device having three sampling periods, the third sampling period being constant, the fourth PCM signal as an input and the first PCM signal as an output and the first PCM signal as an input The PWM signal generator according to claim 1, wherein the third sampling period is longer than the first sampling period.   In the PWM signal generator that receives the first PCM signal and outputs the PWM signal, the low-frequency component of the PWM signal corresponds to the low-frequency component of the first PCM signal. The PCM signal has a first sampling period, and the PWM signal is generated by digital means based on a second PCM signal having a second sampling period, and the second sampling period is determined by an external command or predetermined. The first sampling period is equal to the second sampling period, and the resolution of the second PCM signal is coarser than the resolution of the first PCM signal. A first PCM signal is converted to a second PCM signal by a delta-sigma modulator, and the delta-sigma modulator is a filter And a quantizer, wherein the filter inputs the first PCM signal and the second PCM signal and outputs a third PCM signal, and the third PCM signal passes through the quantizer and is supplied with the second PCM signal. The gain of the quantizer may change dynamically in inverse proportion to the value of the second sampling period, and the coefficients and functions in the internal calculation of the filter depend on the second sampling period. PWM signal generator, characterized in that it is determined and may change dynamically.   In the PWM signal generator that receives the first PCM signal and outputs the PWM signal, the low-frequency component of the PWM signal corresponds to the low-frequency component of the first PCM signal. The PCM signal has a first sampling period, and the PWM signal is generated by digital means based on the second PCM signal having the second sampling period, and the second sampling period is determined by an external command or predetermined. The sampling timing of the first PCM signal is obtained by adding the timing between the sampling of the second PCM signal to the sampling timing of the second PCM signal. The resolution of the second PCM signal is coarser than the resolution of the first PCM signal, and the first PCM The signal is converted to a second PCM signal by a delta sigma modulator, the delta sigma modulator has a filter and a quantizer, and the filter inputs the first PCM signal and the second PCM signal. A third PCM signal is output, the third PCM signal is converted to a second PCM signal through the quantizer, and the gain of the quantizer is dynamically inversely proportional to the value of the second sampling period. A PWM signal generator, characterized in that the coefficients and functions in the internal operation of the filter are subject to change and may change dynamically as determined by the second sampling period.   In the PWM signal generation apparatus that receives the fourth PCM signal and outputs the PWM signal, the low-frequency component of the PWM signal corresponds to the fourth PCM signal, and the fourth PCM signal is the second PCM signal. The sampling period is 3, the third sampling period is constant, the third sampling period is longer than the first sampling period, and the fourth PCM signal is input and the first PCM signal is output. 5. A PWM signal generation device comprising the PWM signal generator according to claim 3 or 4, wherein a sampler and a first PCM signal are input and an output signal of the PWM signal generation device is output.   A digital amplifier realized by driving a switching amplifier with a PWM signal generated by the PWM signal generation device according to claim 2.