Quote:
Originally Posted by twobob
Terrifying latency. Something must be up. There's a big LOUD surge, then it remembers to turn down, that's a bit odd also...
Something must be up. Even at it's maximum period size of 4096 it's struggling to initialise Full duplex in one swallow. I suspect this may be the kernel glitch I was spouting about. would love to pull some decent errors from somewhere.
All stracing gives me is:
ioctl(5, SNDRV_PCM_IOCTL_READI_FRAMES, 0xbe9559b0) = -1 EAGAIN (Resource temporarily unavailable)
repeated a lot... Then:
octl(5, SNDRV_PCM_IOCTL_READI_FRAMES, 0xbe9559b0) = -1 EPIPE (Broken pipe)
Which is pretty much what I observed before when trying to full duplex the card.
Something must be up. Did I say that already?
Meh. Ill abandon full duplex for the day in favour of tea and twiddling freqs toward the GM++ Viddy viewer. much more fun.
|
I guess you will need to stick with half-duplex audio, and use a PTT (Push To Talk) button for your kindle VOIP calling.
And I am leaning to a choice of CELT or Speex for gmvid (to be chosen when encoding). Speex would be great for the MIT lecture series. Perhaps we could dynamically switch between them so the intro music before and after lectures would not sound horrible. And the code is public domain. And Speex can encode down to 2kbps bitrate.
Or to be low overhead we could use raw ADPCM interleaved with dithered video, and let an outer compression loop (or external compressor like gzip) handle it.
Ideas? Building libav just for a little player proggie like gmplay seems like overkill. I think better to use that only in the ENCODER (i.e. continue to use ffmpeg|raw2gmv|gzip for encoding on a host PC).