[xwax-devel] timecoder woes

[Albert S. <asa...@uw...>]

Hi all,

I've been trying to use timecoder.c in another open source application,
and I'm having a really tough time getting it to work.

Basically, I get audio samples from the soundcard as signed shorts
(paInt16 to be exact), and pass them off to timecoder. The buffer sizes
I use are fairly large (in the thousands), but I've tried small buffer
sizes with no luck either.

The basic breakdown of how I use timecoder is the following:


timecoder_init(&timecoder);       //Init the timecoder
char* timecode = "serato_2a";     //Select a timecode source
timecoder_build_lookup(timecode); //Build LFSR table

After receiving a buffer's worth of samples, the following is called:

timecoder_submit(&timecoder, this->m_samples, this->m_SamplesSize);

... where
m_samples is an array of signed shorts
m_SamplesSize is the number of samples

Then I immediately do the following:

int when = 0;
timecoder_get_pitch(&timecoder, &dVinylPitch);
dVinylPosition = timecoder_get_position(&timecoder, &when);

and every time except the odd fluke time, dVinylPosition ends up being
-1 (error)

The pitch and position as readback from my program are:
At 33 RPM:
Debug: dVinylPosition: -1.000000, dVinylPitch: -12.500000, when: 0
Debug: dVinylPosition: -1.000000, dVinylPitch: -6.367286, when: 0
Debug: dVinylPosition: -1.000000, dVinylPitch: -12.500000, when: 0
Debug: dVinylPosition: -1.000000, dVinylPitch: -6.367286, when: 0

(Although tonight dVinylPitch seems to go to around 0.50 at 33 RPM and
0.00 when stopped - progress?)

In order to try to see what was going on in timecoder, I enabled the
debugging information and tried running both xwax and my code.

When running xwax, the debugging output looks like:
 +2017 +/14723 -/6233 (12930)   = 0 (0)   386   [crossings: 2 44]
  +754 +/14723 -/11124 (12937)  = 1 (1)   387   [crossings: 1 23]
 -2285 +/11124 -/2057 (12915)   = 0 (0)   388   [crossings: 2 44]
   +93 +/14731 -/11862 (12915)  = 0 (0)   388   [crossings: 1 21]
 +1822 +/14731 -/11006 (12922)  = 1 (1)   389   [crossings: 1 23]
  -357 +/11006 -/10510 (12877)  = 0 (0)   390   [crossings: 2 44]


When running my code, the debugging output looks like
(m_SamplesSize: 1412)
 +1016 +/3758 -/3424 (12226)    = 1 (1)   2     [crossings: 32 1412]
Debug: dVinylPosition: -1.000000, dVinylPitch: 0.499717, when: 139205596
 +1041 +/3855 -/3498 (11993)    = 1 (1)   1     [crossings: 32 1412]
Debug: dVinylPosition: -1.000000, dVinylPitch: 0.499717, when: 139205596
 +1041 +/3855 -/3498 (12044)    = 1 (1)   1     [crossings: 32 1412]
Debug: dVinylPosition: -1.000000, dVinylPitch: 0.499717, when: 139205596
 +1081 +/4037 -/3635 (11668)    = 1 (0) x 0     [crossings: 32 1412]
Debug: dVinylPosition: -1.000000, dVinylPitch: 0.499717, when: 139205596
 +1081 +/4037 -/3635 (11639)    = 1 (0) x 0     [crossings: 32 1412]
Debug: dVinylPosition: -1.000000, dVinylPitch: 0.499717, when: 139205596
 +1046 +/3892 -/3525 (11551)    = 1 (1)   1     [crossings: 32 1412]
Debug: dVinylPosition: -1.000000, dVinylPitch: 0.499717, when: 139205596

If you look through the lines there, you'll see that 380-ish column,
which represents the "number of successful error checks" looks off. (The
high number of crossings makes sense since I'm using large buffers.) I
haven't the slightest clue what the meaning of all of this is. However,
I'd like to mention once again that I've tried it with small buffers (16
and 32), with similar results.

Looks like the pitch info is more or less right, but something else is
terribly wrong. I've spent hours trying to figure this out and I'm still
scratching my head.

If anyone has any ideas, I'd appreciate if you could toss me a
lifeline. :)

Thanks,
Albert