WO2006030373A1

WO2006030373A1 - Pulse width modulated noise shaper and related method

Info

Publication number: WO2006030373A1
Application number: PCT/IB2005/052989
Authority: WO
Inventors: Bruno J. G. Putzeys
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2004-09-15
Filing date: 2005-09-13
Publication date: 2006-03-23
Also published as: GB0420508D0

Abstract

The present invention provides for a pulse width modulated noise shaper comprising an input adder having a first input for receiving an input signal and a second input, an output terminal, a filter with a substantially integrating characteristic having an input coupled to receive an output signal from said input adder, a pulse width modulation circuit having an input coupled to receive a signal derived from an output of the filter, an output coupled to the said output terminal, a control loop coupled to receive a signal derived from an output of said filter and to the second input of the adder, the control loop being arranged for generating a feedback signal representative of information conveyed by variable edges of a pulse width modulation signal and for feeding this feedback signal back to the second input of the adder.

Description

DESCRIPTION

PULSE WIDTH MODULATED NOISE SHAPER AND RELATED METHOD

{ The present invention relates to a pulse width modulated noise shaper and related method.

Pulse width modulated noise shapers are commonly used, for example, within a digital amplifier of an audio apparatus for driving a speaker system, and within digital-analogue converters.

A relatively simple method of generating a Pulse Width Modulated

(PWM) noise-shaped signal is known from patent application WO

2004/049561 A. The signals generated by way of such an arrangement are typically used, as noted above, in digital-analogue conversation and in relation to both small-signal and high-power output requirements.

As noted in that earlier application, since a digital signal can only produce a timed PWM signal up to a resolution equivalent to its clock frequency, the PWM operation effectively introduces a quantisation.

The subject matter of the above-mentioned application seeks to create PWM with a low-frequency spectrum essentially equal to that of the input signal, while the quantisation noise is moved to the frequency range above that represented by the input signal.

However, arrangements such as that known from the above-mentioned application suffer limitations due to their distortion performance, and also due to the signal dependency of the associated noise transfer function.

While such limitations do not necessarily affect the use of such PWM noise shapers within, for example, Class D amplifiers, the limitations nevertheless have a notably detrimental effect on the performance of the PWM noise shaper when used, for example, in digital-analogue conversation. Also the signal dependency of the noise transfer function associated with the PWM noise shaper leads to instability if the input signal should exceed the modulation capability of the pulse width modulator. The present invention seeks to provide for a PWM noise shaper, and related method, which exhibit advantages over known such PWM noise shapers and related methods.

According to one aspect of the present invention, there is provided a pulse width modulated noise shaper comprising an input adder having a first input for receiving an input signal and a second input, an output terminal, a filter with a substantially integrating characteristic having an input coupled to receive an output signal, i.e. the error signal, from said input adder, a pulse width modulation circuit having an input coupled to receive a signal derived from an output of the filter, an output coupled to the said output terminal, a control loop coupled to receive a signal derived from an output of said filter and to the second input of the adder, the control loop being arranged for generating a feedback signal representative of information conveyed by variable edges of a pulse width modulation signal and for feeding this feedback signal back to the second input of the adder.

Through generation of the feedback signal as discussed above, the present invention advantageously eliminates the mechanism responsible for the distortion encountered in the prior-art and leads to advantageously improved results as compared with such prior-art.

Also, the present invention produces a completely signal- independent noise transfer function and can serve to place the overloadable element outside of the control loop.

In one embodiment, the feedback is derived from the variable edges of the pulse width modulated signal and can, as a particular advantage, be derived from a combination of such signal and a slope signal.

Advantageously, the feedback signal can be derived prior to the pulse width modulation and directly from the output of the main filter.

In this manner, the pulse width modulator signal can then be derived from the feedback signal, rather than the feedback signal being derived from the pulse width modulated signal. In one arrangement, the means for generating the feedback signal comprises a counter.

In particular, the counter is arranged to be decremented on each clock cycle and to be incremented by an amount equivalent to the cycle count of the pulse width modulated signal once the counter value reaches a predetermined value relative to the input signal, and preferably once the decremented value reaches a value equal to the input signal.

Of course, the counter can be arranged to be incremented on each clock cycle and to be decremented once the counter value reaches the predetermined relative value.

According to another aspect of the present invention, there is provided a method of pulse width modulation noise shaping comprising the steps of the filtering an output signal from an input adder, pulse width modulating a signal derived from an output of said filter and producing a pulse width modulated signal at an output terminal of the pulse width modulating noise shaper, and including the step of generating a feedback signal representative of information conveyed by the variable edges of the pulse width modulated signal and feeding back this feedback signal to an input of the said input adder. The above-mentioned method can advantageously be adapted to provide for further advantageous features such as those discussed above in relation to the pulse width modulated noise shaper.

The invention is described further hereinafter, by way of example only, with reference to the accompanying drawings in which:

Fig. 1 is a simplified circuit diagram of a pulse width modulated noise shaper known from the prior-art;

Figs 2a and 2b are representations of the signals arising in relation to the circuit representation of Fig. 1 ; Fig. 3 is a timing diagram of signals arising in accordance with the present invention; Fig. 4 is a timing diagram illustrating the generation of a feedback signal in accordance with an embodiment of the present invention; Fig. 5 is a simplified circuit of a pulse width modulated noise shaper embodying the present invention; and Figs. 6a and 6b comprise representations of signals arising within the embodiment of the present invention illustrated in Fig. 5 and as compared with the representations of Figs. 2a and 2b.

Turning now to Fig. 1 there is a basic schematic representation of a pulse width modulated noise shaper comprising an upsampler and as known from patent application WO 2004/049561 A.

Within the noise shaper 10, a digital audio input signal 12 is first up- sampled at up-sampler 14 so as to increase its sampling rate to a value equal to that applied to the pulse width modulated waveform. The up-sampled input is then developed to a control loop by way of an adder 16 having an output for delivering a signal to a main filter 18, the output 20 of which is delivered to a pulse width modulating generator 22 which also receives a pulse width modulating sampling signal 24.

The output pulse width modulated signal 26 is then also feedback as feedback signal 28 to the adder 16.

The control loop represented within Fig. 1 effectively shifts the quantisation noise created by the pulse width modulating operation to a higher frequency range that arising initially.

As noted above, the distortion performance of the known arrangement such as that illustrated in Fig. 1 is a disadvantageously limiting factor. Such distortion generally arises since the main filter 18 of the loop, while high in order, by necessity exhibits only one less zero than it has "poles". As a result, its rejection of the pulse width modulated carrier and the related sidebands is relatively limited. Referring now to Figs. 2a and 2b there are shown, superimposed, the pulse width modulating sampling signals 24, the output signal 20 from the main filter 18 and the output pulse width modulated signal 26 derived from the pulse width modulation generator 22 illustrated in Fig. 1.

As appreciated within the present invention, the pulse width modulation operation effectively samples its own residual signal at a frequency, which is the same as its own. Such sampling produces a DC alias product and, since the residual changes phase and amplitude as the low-frequency input changes, the sampling instants change likewise and this DC alias is therefore also modulated. The modulation sidebands of this DC alias then appear as distortion. It should of course be appreciated that, for the purpose of clarity, Figs.

2a and 2b illustrate a single-sided pulse width modulated operation whereas known pulse width modulation generators such as known from application WO 2004/049561 A tend to operate employing double-sided pulse width modulation thereby reducing the impact of, but not eliminating, the distortion mechanism.

A particularly important feature of the present invention is the realisation that the problem of sampling a residual of varying amplitude and phase at a varying sampling instant can be overcome by ensuring that the amplitude of the residual is kept constant while the phase with respect to the sampling instant is also kept constant.

This generally leads to a requirement that the output signal of the main filter 18 of the loop is different from that arising within the known arrangement and this, in turn, requires that the input signal to the filter and thus the feedback signal be modified. From the above realisation that the amplitude of the residual should generally be kept constant, while the phase with respect to the sampling instant is also kept constant, there separation of the falling and arising edges of the pulse width modulated square wave signal becomes a relevant issue.

Reference is now made to the timing diagram of Fig. 3, which illustrates a variety of signal traces useful for explaining one particular aspect of the present invention. It should be appreciated that, for a single-sided pulse width modulated signal, there is only one sampling instant per cycle, i.e. the variable edge. In accordance with the present invention, the variable edge comprise the rising edge, and the falling edges, generally known as stationary edges, have a perfectly constant interval and so carry no information. However, it is appreciated within the present invention that both the variable and stationary edges contribute equally to the generation of the residual signal. However, it is identified that the dividing of the residual into two separate portions corresponding to variable and stationary edges of the pulse width modulated signal would lead to a variable residual which is actually relative constant in amplitude and is quite readily phase-locked to the variable edge. The constant value of the amplitude arises from the fact that, as long as the frequency of the modulation is well below the pulse width modulation reputation rate, the interval does not vary in an extreme manner. As noted, the invention is based on the realisation that it would be particularly advantageous to produce a feedback signal that conveys exactly the same information as the variable edges of the pulse width modulated signal and does not have falling edges.

Fig. 3, comprises an illustration as to how the pulse width modulated signal can be viewed as consisting of two signals, each containing only the rising or falling edges respectively. For clarity purposes, the input signal is simplified as not containing any ripple. Within the invention, an upsampler such as that of Fig. 1 is generally employed.

Illustrated at the top of the timing diagram is an input signal 120 which is sampled through comparison with a sawtooth pulse width modulating sampling signal 124 so as to provide for a resulting pulse width modulated signal 126.

Through the aforementioned division of the falling and rising edges of the pulse width modulated signal 126, there is then derived two monotonic staircase signals 130, 132 of which the ascending staircase signal 130 steps at variable intervals coinciding with the rising edges of the pulse width modulated square wave signal 126 and thereby serves to convey the actual information from the modulation process. The other staircase signal 132, i.e. the descending staircase signal, is found to be perfectly regular since it is derived from the falling edges of the pulse width modulated signal which convey no information.

Thus, it will be appreciated that the combination of the staircase signals 130 and 132 will effectively lead to the pulse width modulated square wave signal 126.

Since, in the illustrated embodiment, it is noted that the descending staircase trace 132 does not convey any information, no information loss is incurred by replacing the descending staircase signal trace 132 with a smooth continuous slope signal 134.

Through combination of the ascending staircase signal 130 and the continuous slope 134, a feedback signal 136 comprising a sawtooth trace extending on a gradual slope is produced.

It should be appreciated that, spectrally, the pulse width modulated square wave signal 126 and the feedback signal are identical with the exception of the pulse width modulated signal frequency and its harmonics.

As should be appreciated, the feedback signal 136, i.e. a signal composed of the combination of the variable edges of the pulse width modulated signal, i.e. signal 130, and the continuous slope 134, conforms to particular requirements of the present invention in that it convey exactly the same information as the variable edges of the pulse width modulated signal and that it does not have any falling edges.

While, from the above, it is noted that the feedback signal can be derived from the pulse width modulated signal, it should be noted that, from the relationship between the feedback signal and the input signal, it can prove more practical to generate the feedback signal 136 directly from the output of the main filter, and then derive the pulse width modulated signal 126 from the feedback signal.

Turning now to Fig. 4, there is provided a timing diagram illustrating such direct generation of the feedback signal.

The timing diagram illustrates signals such as those arising in Fig. 3 and which comprises primarily the sawtooth pulse width modulating sampling signal 136 extending along a slope, which is consistent with that of the input signal 120.

The pulse width modulated output signal 126 is generated on the basis of the change in zero crossing of the pulse width modulating sampling signal 136 which is of course dependent upon the slope of the input signal 120.

According to a particular embodiment, the feedback signal is generated employing a counter such as that discussed further below and as illustrated with reference to Fig. 5.

Within the PCM noise shaper embodying the present invention such as in Fig. 5 there is provided an input adder 166 again providing an output signal to the main filter 118, which in turn provides an output signal 120. A pulse width modulation generator 122 is provided at the output of the knowledge shaper to provide a pulse width modulated signal 126. However, as illustrated, and with particular relevance to the direct generation of the feedback signal; the control loop includes a feedback path 128 linking an input of the pulse width modulating generator 122 to the second input of the input adder 116.

As noted above, in a particularly advantageous embodiment of the present invention, the feedback signal is generated by means of a counter 138 which receives, as input signal, the output signal 120 of the filter 118, and produces an output signal 136 which is fed back by way of the feedback path 128 of the control loop to the second input of the input adder 116.

Returning to Fig. 4, it will be appreciated that the input signal 120 can be offset as illustrated at 120' by an amount equal to half the step size of the sore tooth pulse width modulating sampling waveform 136. While the ratio of the slope to step size sets the pulse width modulating frequency, the feedback signal is actually generated by way of the counter 138 which is arranged to be decremented by a value of one for each clock cycle and incremented by an amount equal to the step size of the pulse width modulating sampling signal 136 once the decremented value has reached the value of the input signal. Such comparison of signals is illustrated within Fig. 5 for the counter

138. In this manner, the slope of the input signal advantageously serves to maintain the appropriate profile of the feedback signal 136.

Turning lastly to Figs. 6a and 6b, there is illustrated, for comparison in particular with Figs. 2a and 2b, waveforms arising from the embodiment of the present invention of Figs. 4 and 5 and for similar situations as those illustrated by way of Figs. 2a and 2b.

As noted, the illustrated signals comprise the pulse width modulating sampling signal 140, the residual signal 120 and the output pulse width modulated signal 126. As can be appreciated, the effect of the residual signal which leads to noise within the known arrangement illustrated in Figs. 2a and 2b is quite clearly absent from the arrangement of the present invention as illustrated in Figs. 6a and 6b.

Claims

1. A pulse width modulated noise shaper comprising an input adder having a first input for receiving an input signal and a second input, an output terminal, a filter with a substantially integrating characteristic having an input coupled to receive an output signal from said input adder, a pulse width modulation circuit having an input coupled to receive a signal derived from an output of the filter, an output coupled to the said output terminal, a control loop coupled to receive a signal derived from an output of said filter and to the second input of the adder, the control loop being arranged for generating a feedback signal representative of information conveyed by variable edges of a pulse width modulation signal and for feeding this feedback signal back to the second input of the adder.

2. A pulse width modulated noise shaper as claimed in Claim 1 , wherein the control loop is arranged such that the feedback is derived from the variable edges of the pulse width modulated signal.

3. A pulse width modulated noise shaper as claimed in Claim 2, wherein the control loop is arranged such that the feedback signal is derived from a combination of the variable edges of the pulse width modulated signal and a slope signal.

4. A pulse width modulated noise shaper as claimed in any one or more of Claims 1 to 3, wherein the feedback signal is derived prior to the pulse width modulation and from the output of the filter.

5. A pulse width modulated noise shaper as claimed in any one or more of Claims 1 to 4, having means for generating the feedback signal comprising a counter.

6. A pulse width modulated noise shaper as claimed in Claim 5, wherein the counter is arranged to be decremented on each clock circle while maintaining comparison between its decremented signal and the said input signal.

7. A pulse width modulated noise shaper as claimed in Claim 6, wherein the counter is arranged to be incremented by an amount equivalent to the step size of the pulse width-modulating signal once the decremented value reaches a predetermined value relative to the input signal.

8. A method of pulse width modulation noise shaping comprising the steps of filtering an output signal from an input adder, pulse width modulated a signal derived from an output of said filter and producing a pulse width modulation signal at an output terminal of the pulse width modulating noise shaper, and including the step of generating a feedback signal representative of information conveyed by the variable edges of the pulse width modulated signal and feeding back this feedback signal to an input of the said input adder.

9. A method as claimed in Claim 8, and including the step of deriving the feedback signal from the variable edges of the pulse width modulated signal.

10. A method as claimed in Claim 9, and including the step of deriving the feedback signal from a combination of the said variable edges and a slope signal.

11. A method as claimed in Claim 8, 9 or 10, wherein the feedback signal is derived prior to the pulse width modulation and from the output of the filter.

12. A method as claimed in Claim 8, 9, 10 or 11 , and wherein the feedback signal is generated by means of a counter producing a decremented signal to be compared with an input signal.