Precision pulse placement to form a binary pulse signal

#The present invention relates to an improvement of precision pulse placement, in particular for precision pulse placement representing a pulse width modulation (PWM) or pulse position modulation (PPM) for precision analog signal generation using a waveform generator.

#Pulse width modulation (PWM) may be used for analog signal generation. While the pulse rate has to be kept constant, signal processing calculates the ideal pulse width. The ideal pulse width depends upon the analog signal to be generated and may be found using natural sampling, click modulation, zero position coding (ZePoC) or uniform sampling.

#Knowing the ideal pulse width it is usually quantized to the nearest multiple of a high frequency bit clock resulting in a realized pulse width, which is different from the ideal pulse width. As a consequence of this process a quantization error will arise. The quantization error depends upon the deviation of the ideal pulse width as calculated by signal processing from the realized pulse width. The quantization error will increase according to the deviation between ideal and realized pulse width. In prior art like US 2002/0180547 A1 tabs from a fine buffer delay chain are used to obtain fine timing resolutions.

#Other objects and many of the attendant advantages of the present invention will be readily appreciated and become better understood by reference to the following description, taken in conjunction with the accompanying drawings, in the several figures of which like reference numerals identify like elements, and in which:

#FIG. 1 shows a timing diagram of some of the signals related to pulse placement in the case of natural sampling,

#FIG. 2 shows a block diagram of prior art pulse width modulation (PWM),

#1/TR: pulse rate TR: pulse rate period R: value controlling the ramp counter TW: pulse width TWi: pulse width of individual pulses (TW1 for pulse 1, TW2 for pulse 2 and so on) W: value to control the width counter 1/TB: bit clock frequency TB: bit clock period TBr: bit clock period for individual pulses (TB1 for pulse 1, TB2 for pulse 2 and so on) D: value to control the D counter 335 in FIG. 5 q: quantization error qi: quantization error for individual pulses (q1 for pulse 1, q2 for pulse 2 and so on) N, M: parameters used in the N/M PLL synthesizer in FIG. 5 TC: period of a reference clock in FIG. 5 TM, TV: periods of signal 330 and signal 332 in FIG. 5

#FIG. 1 shows a timing diagram of signals related to the pulse placement in the special case of natural sampling. Time is laid off as abscissa 6, usually called X-axis whereas the signal level is laid off as ordinate 7, usually called Y-axis. As unit of the X-axis the bit clock period (TB) 3 is shown, whereas the signal level at the Y-axis is shown only qualitatively without a unit. Four signals 1, 2, 4 and 5 are shown in the timing diagram. A ramp 5, increasing from zero to its maximum value within one pulse rate period (TR). Ideal pulses 2, each individual pulse characterized by having a pulse width TW1, TW2, TW3, TW4. The leading edge of each individual pulse occurs at the same time as the level of the ramp 5 is zero. The trailing edge of each individual pulse occurs after the duration of the individual pulse width TWi, i.e. TW1 for the first pulse, TW2 for the second pulse and so on as shown in FIG. 1. The analog signal 1 which represents an example of the waveforms generated by the method disclosed in this invention. The quantized pulses 4 with leading and trailing edges spaced by bit clock periods (TB).

#The pulse widths TWi of the ideal pulses are determined by the points in time at which the ramp 5 intersects the analog signal 1. These ideal pulses include no quantization error. As mentioned earlier the pulse rate period TR which is the same as the ramp rate for the ramp 5 shown in FIG. 1 has to be exactly constant over time in accordance with the present invention. The ramp rate is exactly constant, if TR=R×TB. The pulse widths TWi of ideal pulses 2 must be rounded to the quantized pulses 4 in a real system, like the one disclosed by this invention. The timing error thus introduced is +/−0.5×TB, and the average absolute error to be expected is 0.25×TB. This is a consequence of the fact that the edges of the quantized pulses 4 in contrast to the edges of the ideal pulses 2 have to be positioned spaced at multiples of the bit clock period TB.