> On Tue, Dec 02, 2003 at 07:38:54PM -0800, Mark Knecht wrote: > > More data, but I'm not sure there's an answer there yet. The first file > > creates a very loud glitch sound, which the second one doesn't. The > > midline value seems better on the first, assuming this represents DC > > offset. > > Mark, > > I might suggesst taking this data andthe two Ogg files to the > voris list at xiph.com - the real vorbis experts hang out there. I'm > sure it'd be a matter of a few minutes for someone familiar with the > format to tell you what the difference is. > > > -- > Ross Vandegrift Ross, Thanks for the idea. I may do that. However, this morning I tried another experiment that (I think) brings this back squarely into the Alsa camp. If I start playing the offending ogg (and indeed most files) in xmms, using OSS emulation, I get the loud popping noise, and then the file plays fine to the end. However, if I pause the play back, and then restart it, I get the popping noise every time I restart it. While I can see that if the file has a DC value, it has a DC value, I cannot really see that it's likely that simply pausing and restarting in the middle should create the noise all by itself. I suppose on could argue that when xmms stops the DC offset has an opportunity to drop to zero, and then starts again, but if that was the case why wouldn't an alsa app playing the same file create the same noise? Overall, I still think that most likely culprit is created by a few things: 1) The content of the file (although 99% of my files do it, I do have 1 that does not.) 2) The fact that I'm using the AI-3 D/A as my playback path. The AI-3 wants to run at 48KHz if it doesn't receive a clock for ANY period of time. I generally notice in hdspconf both the clock frequency moving when this noise occurs, as well as hdspconf showing that the HDSP 9652 goes from lock to sync to lock, indicating to me that they two devices (HDSP and AI-3) are not handling their clocks correctly. Anyone in the San Jose area wanna let me borrow a different D/A? ;-) Thanks, Mark
