Coded data generation or conversion – Analog to or from digital conversion – Digital to analog conversion
Patent
1996-03-01
1998-01-13
Young, Brian K.
Coded data generation or conversion
Analog to or from digital conversion
Digital to analog conversion
375 23, H03M 166
Patent
active
057084335
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND TO THE INVENTION
An attractive way of producing high quality digital to analogue converters (DACs) for use in Audio and other applications is to produce a pulse-width-modulated waveform in which pulses are of constant height but in which the timings of one or both edges of each pulse is modulated in response to the incoming digital signal.
In this field of use, the pulse repetition frequency is typically of order 3 MHz (64 times the original sampling rate), and the edge timings are typically determined by counting the cycles of a crystal clock (the bitclock) with frequency up to 50 Mhz. The resulting waveform contains the desired audio modulation plus strong signals at 3 MHz and its harmonics, which are substantially removed by a low-pass filter.
A closely related field is the use of pulse-width-modulation (PWM) in power amplifiers, in particular in "Digital" power amplifiers. The Digital Power Amplifier can be regarded as a DAC operating at a high power level. In this case considerations of power dissipation and switching speed in the output devices impose strong practical reasons for using a pulse repetition frequency (switching frequency) as low as possible. Typically the switching frequency is no more than 8 to 16 times the original sampling frequency. For reasons to be explained, this reduction in switching frequency entails an increase in the bitclock frequency if the same signal-to-noise ratio is to be maintained in the audio band, and bitclock frequencies of 100 MHz or more have been contemplated in this field.
In general it is necessary to use an oversampler, or interpolator, to produce digital samples representing the original audio waveform but at a higher sampling rate than is normally used for storage or transmission. Having done this, there are several ways in which these (oversampled) digital samples may be used to modulate the positions of the pulse edges (see FIG. 1):
(a) Leading-edge modulation: each digital sample modulates the timing of the leading-edge of the corresponding pulse
(b) Trailing-edge modulation: ditto, trailing-edge
(c) Double edge modulation: the two edges of the pulse are shifted in time in opposing directions, in response to the digital sample value
(d) Consecutive edge modulation (CEM): the trailing-edge of a pulse is modulated in response to one sample value, and the trailing-edge in response to the following sample value.
(a) and (b) are not in fact fundamentally different from each other, and one can transform from one to the other by adopting the opposite convention about which electrical polarity is labelled "positive".
It will be seen that in modulation types (a), (b) and (c) the pulse repetition frequency is the same as the oversampled sampling frequency, but in consecutive edge modulation it is one half of this frequency.
It is recognised in the prior art that the PWM process is fundamentally non-linear. For example if a small high-frequency modulation and a large low-frequency signal are simultaneously applied to leading-edge modulation, the low-frequency signal will move the edge backwards and forwards in time, thus giving a time-varying phase-shift to the dealing with this problem at least partially.
Any of the above modulation schemes can in principle be implemented using circuits in push-pull mode, though this is more practical in the case of DACs than in the case of power amplifiers. Push-pull operation cancels even order distortions (such as the phase-shift mentioned above) and has application number PCT/GB 92/00312.
It will also be appreciated that the resolution implied by the above figures is totally inadequate for high quality audio. With a 3 MHz switching frequency and a bitclock rate of 48 MHz the pulse length is quantised to only 16 possible lengths giving a 4 bit resolution. Even with the possible sqrt(n) advantage of oversampling, we have only 7 bits or 42 dB signal-to-noise ratio.
It is widely recognised that the solution to this problem is to use noise-shaping, a technique that allows noise in a quantised system to be substantially
REFERENCES:
patent: 4095218 (1978-06-01), Crouse
patent: 4947171 (1990-08-01), Pfeifer et al.
LandOfFree
Digital converter does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Digital converter, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Digital converter will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-330013