Pulse or digital communications – Pulse width modulation
Reexamination Certificate
2000-09-19
2002-08-06
Tse, Young T. (Department: 2634)
Pulse or digital communications
Pulse width modulation
C375S254000, C375S296000, C330S010000, C332S109000
Reexamination Certificate
active
06430220
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to digital audio amplifiers and more particularly to a method and apparatus for reducing harmonic distortion resulting from uniform pulse width modulation.
BACKGROUND OF THE INVENTION
Digital audio amplifiers typically implement one of several pulse width modulation schemes to effect audio signal modulation. In these amplifiers, uniform pulse width modulation, hereinafter UPWM, is a standard method of pulse width modulation used in some known audio amplifiers wherein the amplitude of each input sample is converted to a corresponding output pulse width. The beginning and ending points of a pulse can be defined by points of intersection between a saw-tooth modulation waveform and a quantized input signal waveform. UPWM is easily implemented but is known to produce significant detrimental harmonic distortion resulting in a nonlinear response. The frequency spectrum of each pulse varies as a function of the pulse width. This dynamic variation is the cause of undesirable harmonic distortion.
Another method of modulating audio signals known as natural pulse width modulation, hereinafter NPWM, is inherently distortion free. The beginning and ending points of a pulse are defined by points of intersection between a saw-tooth modulating waveform and an analog input waveform. However, NPWM cannot be implemented in a quantized system with digital input because it operates on an analog input signal. Various techniques are known to approximate NPWM in fully digital systems and thereby reduce harmonic distortion.
Other methods are known that emulate NPWM by using linearizing algorithms which operate on information from more than one input sample to determine each output pulse width. Such methods are often referred to as linearized pulse width modulation, hereinafter LPWM. Prior art demonstrates linear interpolation between two adjacent input samples to achieve the output pulse width. For example, Mellor, P. H. et al, “Reduction of Spectral Distortion in Class D Amplifiers by Enhanced Pulse Width Modulation Sampling Process” IEEE Proceedings-G, Vol. 138, No. 4, August 1991 pp. 441-448 teaches an enhanced sampling process wherein an analog input signal f(t) is transformed into an enhanced signal f
n
(t) where f
n
(t)=f((1−&egr;)(t−nT)+nT) and 0≦&egr;≦1 before it is combined with a saw-tooth modulation wave to define output pulse widths. T is the sampling period, n is the sample index and t is the independent variable. As variable &egr; approaches 0, the interpolation approaches natural sampling. As &egr; approaches 1, the interpolation approaches uniform sampling. Other interpolation methods are known using increasing numbers of samples to define each output pulse. Such methods more closely approach natural sampling but require more complex implementation.
Other methods of distortion reduction of UPWM in the prior art provide improvements by introducing several filters to correct for non-linearity. For example, Hawksford, M. O. J., “Dynamic Model Based Linearization of Quantized Pulse Width Modulation for Applications in Digital to Analog Conversion and Digital Power Amplifier Systems” Journal of Audio Engineering Society, Vol. 40, NO. 4, April 1992, pp. 235-250 teaches a method of compensating for the frequency dependant harmonic distortion over a limited bandwidth. Multiple finite impulse response (FIR) filters are interleaved between over-sampled pulse code modulated (PCM) data and a pulse width modulator wherein frequency responses are individually matched to the inverse square of each corresponding pulse in a pulse width modulated sequence. A dynamic filtering process is thereby provided. The extra filtering components used in these methods require an undesirably large number of storage elements and calculation cycles.
Still other methods of distortion reduction in pulse width modulated systems involve much additional complexity and require significant ROM storage. For example, Craven, Peter, “Toward the 24 bit DAC: Novel Noise-Shaping Topologies incorporating Correction for the Non-linearity in a PWM Output Stage.” Journal Audio Engineering Society, Vol. 41, No. 5, May 1993, pp. 291-311 teaches a complex method of nonlinear noise shaping which digitally simulates the intrinsic non-linearity of a pulse width modulation system and provides corrections through feedback and feed-forward paths.
U.S. Pat. No. 5,617,058 to Adrian et al. which is incorporated herein by reference teaches implementation of techniques for ternary modulation wherein a compensating pulse is added to an output pulse to compensate the linearity of a power switch. The Adrian et al. patent also teaches signal flow for digital amplification.
All prior mentions of linear interpolation in digital pulse width modulation either do not include an algorithm to determine the pulse width, or they disadvantageously use a division operation to compute the pulse width. Division operations are extremely computationally inefficient in digital signal processing and require many more computation steps than addition or multiplication operations, for example. Additional silicon area is therefore required to implement techniques involving division. Generally known implementations do not provide computationally efficient methods to reduce harmonic distortion in pulse width modulated systems. None of the prior art methods completely eliminate harmonic distortion.
SUMMARY OF THE INVENTION
The present invention substantially reduces harmonic distortion in a pulse width modulated system by approximating natural pulse width modulation (NPWM) using a method of linear interpolation that requires only additions and multiplications of data in a manner which is computationally efficient for digital signal processing.
According to the invention, a linearization algorithm is implemented between an interpolation, such as may be performed using a finite impulse response filter, (FIR), and a noise shaper. The linearization algorithm iteratively estimates the intersection between a natural input and the modulation wave thereby significantly reducing certain harmonic distortion that is inherent in uniform pulse width modulation (UPWM).
The present invention implements a frequency detector algorithm which disables the linearization algorithm at high frequencies. The frequency detector in an exemplary embodiment consists of a state machine that selects either to implement or bypass the linearization algorithm. The frequency detection algorithm is performed once per input sample before the interpolation filter. Where ternary pulse width modulation is used, a limiting frequency is chosen at the sampling frequency divided by six because the third harmonic is the dominant distortion product of compensated ternary pulse width modulation. Where binary pulse width modulation is used, a limiting frequency is chosen at the sampling frequency divided by four because the second harmonic is the dominant distortion product for compensated binary pulse width modulation. Above these frequencies, the dominant distortion product will be above the audio frequency band.
The output of the linearization algorithm in the illustrative embodiment is input to the noise-shaper, which reduces the number of bits and thereby allows the PWM output clocking to run at an attainable speed. The noise shaper output is then encoded into ternary PWM timing information. The ternary PWM output comprises three states: positive, negative and damped. Each of these states is entered every PWM cycle. The values generated for the output timer are, compensation, pulse, delay and sign. These values are loaded into the counter and determine the length of time the output remains in each state for each PWM cycle. For binary pulse width modulation, the output of the noise shaper indicates the sole pulse value. The application of the linearization algorithm practically necessitates the use of single-sided modulation because using double sided modulation with linearized pulse width modulation severely increases the com
Apogee Technology Inc.
Brown Rudnick Berlack & Israels LLP
Michaelis Brian L.
Quinn Joseph P.
Tse Young T.
LandOfFree
Distortion reduction method and apparatus for linearization... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Distortion reduction method and apparatus for linearization..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Distortion reduction method and apparatus for linearization... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2934466