Hello Andy,
Thanks for the info . I will check it with some other (cheaper)
soundcards, but the E-MU 0202 is the only one I have which really
supports 192 kS/sec for the analog input.
you wrote:
But I'm impressed you can get that good averaging from a sampling rate
much lower than that
It was a suprise for me as well, thus the scepticism. The principle is
as follows:
- Interpolate the samples from the soundcard (192 kS/s) by a factor of four
- Run a few dozen of those interpolated samples (around the rising edge)
through a 2000-point FIR filter, windowed-sinc lowpass with ~96 kHz
cutoff = nyquist frequency of the original signal. Going larger than
2000 points doesn't improve the result any further, going below 1000
points makes it worse.
- Use the four samples from that signal around the zero-crossing (or
steepest slope) for a cubic interpolation
- Find the zero-crossing in that interpolated function, to calculate the
fractional part for 'number of samples between two rising GPS sync pulses'.
All further calculations (statistics, momentary sampling rate, averaged
sampling rate) are based on that result.
If anyone is interested, I will publish the sourcecode (plain old "C")
after some clean-up.
73,
Wolf .
|
