Miscellaneous active electrical nonlinear devices – circuits – and – Signal converting – shaping – or generating – Rectangular or pulse waveform width control
Reexamination Certificate
2002-03-18
2003-03-25
Callahan, Timothy P. (Department: 2816)
Miscellaneous active electrical nonlinear devices, circuits, and
Signal converting, shaping, or generating
Rectangular or pulse waveform width control
C327S164000
Reexamination Certificate
active
06538484
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to generating digitally controlled voltage levels and, in particular, to generating width-modulated pulses for obtaining finely controlled voltage levels after low-pass filtering
BACKGROUND OF THE INVENTION
Dynamically-controlled-level direct voltage, or direct current (DC), is required in a wide range of apparatus types, for a great variety of applications and using various technologies. These include, notably, speed control of DC motors, and analog control of the radiation levels of various devices. A common method for generating a direct voltage with dynamically controlled level, using digital circuitry, is Pulse Width Modulation (PWM). This method is considerably less expensive and more power efficient than utilizing analog circuitry, such as digital-to analog converters, especially when high-resolution control is desired. A typical PWM implementation, providing control by a number, n bits long and thus providing voltage level resolution of N=2{circumflex over ( )}n (2 to the power of n) steps, comprises an n-bit counter, an n-bit comparator, and a flip-flop, as shown in FIG.
1
. The counter is fed by a train of clock pulses and whenever it wraps to 0 (i.e. resets itself after reaching maximum count), the flip-flop is set to 1, and when the counter matches the value V of the control number, corresponding to the desired voltage level, the flip-flop is reset to 0. The resulting output waveform is a train of rectangular pulses, whose rate (or frequency), R, is equal, to the clock frequency, C, divided by N and whose controllable pulse width is V clock periods, the average voltage thus equaling the desired output level.
FIG. 5A
shows an example of a waveform for V=17, output from such a prior-art system with n=6 (N=64). The horizontal axis represents time; the units are pulse periods, each being in this case 2{circumflex over ( )}6=64 clock periods long. It is noted that the value of 6 for n in this and subsequent examples has been chosen for clarity of illustration and that in practice the value of n is much higher, typically around 14.
If a smooth analog output is desired, the output signal is passed through a low-pass filter, which removes the high frequency spectral components, and outputs the desired direct-voltage (or DC) level, plus some ripple at the pulse frequency. In many applications, such as when the voltage produces thermal effects, there is no need for a low-pass filter, since the process, or physical effect, itself has an inherent long time constant, thus behaving like a low-pass filter. In the sequel, any such effect will be understood to come under both the general terms “low-pass filter” and “low-pass process”. It will also be understood that any low-pass process has a generally decreasing spectral characteristic, that is—its attenuation is generally increasing with frequency. It is noted that the resolution with which the level is controlled is akin to the resolution (or quantum size) with which the pulse width is determined. The latter is clearly equal to the ratio of a is period of the width-modulated pulses to a clock period, which is 1/N=1/2{circumflex over ( n)}.
A typical distribution of amplitude vs. frequency for a typical PWM pulse train of prior art, before and after a low-pass filter, is presented, respectively, in the logarithmically-scaled spectral graphs of
FIGS. 2A and 2B
.
FIG. 2B
also shows a typical attenuation graph of the filter. The relevant parameters in this example are: Clock-frequency (C)=50 MHz, n=14 and V=1639, whereby the pulse rate is C/N=3 KHz. It is readily seen that the first few harmonics, which are the primary contributors to the ripple, are attenuated by only a little. For a given low-pass filter, the ripple can be reduced by increasing the pulse rate. The latter can, in turn, be achieved either by decreasing the number of control bits, n, or by increasing the clock frequency. Clearly, decreasing n reduces the resolution of level control, whereas increasing the clock frequency requires more complex or expensive circuitry, or may not even be feasible in a given system or under certain circumstances.
There is therefore a clear need for, and it would be advantageous to have, a method for digitally generating pulses that, when applied to a given low-pass filter, result in direct voltage with reduced ripple power, yet whose level is controllable at high resolution.
It is noted that in some current applications the frequency of the pulses is sometimes varied for reasons other than that discussed above. A notable reason is the desire to avoid radio-frequency interference (RFI) from. the circuits. In U.S. Pat. No. 6,204,649 to Roman, for example, there is disclosed a PWM controller which incorporates a varying frequency oscillator, causing the frequency of a switching regulator to vary, thus spreading out the EMI noise generated by it. Also in U.S. Pat. No. 5,537,305 to Colotti there is disclosed a switching power supply, whose switching frequency is varied so as to avoid interference. In all such and other cases, however, the frequency is typically shifted by a relatively small amount, which would be insignificant for ripple reduction; moreover, even when there is a considerable frequency shift, there is no attempt to retain the resolution of the level control.
Dynamic control of the voltage level means that the control number itself varies with time, requiring also the controlled direct voltage level to vary accordingly. As long as this variation is substantially slower than the pulse rate and is well within the pass band of the low-pass filter, the method discussed above is valid, whereby any average of the pulse train is computed over a period equal to a sufficient number of pulse periods, yet short enough for any variation of the control value to be insignificant.
In many cases, the application calls for a plurality of voltage levels to be independently controlled. In prior art systems, this is achieved by simple repetition of the relevant circuits, which may prove to be unduly expensive.
SUMMARY OF THE INVENTION
It is the object of the present invention to provide a method for reducing ripple in a digitally controlled direct voltage level output by a given low-pass filter operating on a train of width-modulated pulses, without reducing resolution, and without requiring a higher clock frequency.
The basic principle underlying the method of the invention is to generate the pulses at a rate R that is much higher than that proscribed by the clock frequency and the length, n, of the control number (as explained above). The rate R is chosen to be equal to, or higher than, the lowest frequency for which the filter's attenuation, is greater than a desired value. The width of the pulses is modulated so that, on the average, it is exactly proportional to the control number, while the width of individual pulses varies cyclically. Preferably the pulse widths are selected from two values—wide and narrow, differing by one clock period, wide pulses being preferably interleaved with narrow pulses. The ratio of wider pulses to the total number of pulses in a cycle is proportional, to the difference between the exact desired level and the level that would be attained if all the pulses were narrow.
Specifically, according to one configuration of the invention, the pulse rate R is chosen to be some integral multiple, G, of the prior-art pulse rate C/N; that is R=G*C/N. The given control value, V, is divided by G, resulting in an integral dividend P and a remainder S. A train of pulses is first generated at rate R, the width of the pulses being exactly proportional to P. These pulses are then modified as follows: The train is logically grouped into successive cycles of G consecutive pulses each. Within each cycle, a certain quantity of pulses is selected, in proportion to the value S and each selected pulse is widened by one clock period. Preferably the selected pulses are distributed evenly over the group so as
Harris Bernard
Rappaport Elchanan
Callahan Timothy P.
Ladas & Parry
Lynx-Photonic Networks Ltd.
Nguyen Linh
LandOfFree
High-frequency PWM voltage control does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with High-frequency PWM voltage control, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and High-frequency PWM voltage control will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3066824