On Tue, Sep 6, 2011 at 11:29 PM, Mauro Carvalho Chehab <mchehab@xxxxxxxxxx> wrote: >> Basically the above starts at the *maximum* capture resolution and >> works its way down. One might argue that this heuristic makes more >> sense anyway - why *wouldn't* you want the highest quality audio >> possible by default (rather than the lowest)? > > That change makes sense to me. Yet, you should try to disable pulseaudio > and see what's the _real_ speed that the audio announces. On Fedora, > just removing pulsaudio-oss-plugin (or something like that) is enough. > > It seems doubtful that my em2820 WinTV USB2 is different than yours. > I suspect that pulseaudio is passing the "extra" range, offering to > interpolate the data. I disabled pulseaudio and the capture device is advertising the exact same range (8-48 KHz). Seems to be behaving the same way as well. So while I'm usually willing to blame things on Pulse, this doesn't look like the case here. >> Even with that patch though, I hit severe underrun/overrun conditions >> at 30ms of latency (to the point where the audio is interrupted dozens >> of times per second). > > Yes, it is the same here: 30ms on my notebook is not enough for WinTV > USB2 (it is OK with HVR-950). > >> Turned it up to 50ms and it's much better. >> That said, of course such a change would impact lipsync, so perhaps we >> need to be adjusting the periods instead. > > We've added a parameter for that on xawtv3 (--alsa-latency). We've parametrized > it at the alsa stream function call. So, all it needs is to add a new parameter > at tvtime config file. Ugh. We really need some sort of heuristic to do this. It's unreasonable to expect users to know about some magic parameter buried in a config file which causes it to start working. Perhaps a counter that increments whenever an underrun is hit, and after a certain number it automatically restarts the stream with a higher latency. Or perhaps we're just making some poor choice in terms of the buffers/periods for a given rate. Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
