On Wed, 2003-06-25 at 12:12, Brian Edmonds wrote: > Alan Murrell <[EMAIL PROTECTED]> writes: > > The error appears usually after around 20mins of recording, but will > > sometimes appear after 6mins, 11mins, etc. > > This sounds almost identical to the problem I've had with my DC10+ since > day one (back in January 2002).
I occasionally get ring buffer overflows when recording from my Matrox Marvel. > I've never been able to record more > than 20min with it. For me it happens much rarer than that. I have some theories about it in my case however. > I've tried both ALSA and OSS drivers, multiple CVS > Zoran drivers and a few different 2.4 kernel releases. Can you try a 2.5 release? That would indicate that it's less likely to be simple kernel latency. I wonder if somebody who knows more could at least explain what the error means and give a brief technical summary of the process of capturing and how the audio ring buffer is involved in it all. I am really quite good at problem solving and a decent understanding about the general workings of the kernel but I have to understand the capture process first. How is this "audio ring buffer" used? How does it relate to the sound card? For the record, I believe the audio ring buffer overflows I get are due to latencies in the marvel linux driver. I suspect the latencies get really bad when the picture is noisy, like when it goes out of sync (i.e. from the source). I think the jpeg compressor gets tied up trying to compress that and does not give enough resources to the other parts of the capture process, overflowing the buffer of audio samples. As I write this I wonder if simply increasing the size of the audio ring buffer would solve some of these problems or would it just create other problems? b. -- Brian J. Murrell <[EMAIL PROTECTED]>
signature.asc
Description: This is a digitally signed message part
