Further to the previous post on the linux.conf.au 2005 audio, a gentleman named Matt kindly wrote in with these thoughts:
Hi Michael, Prompted by some of the other Fluendo guys, I took a look at the sample ogg file you uploaded of the LCA streams. We think we've figured out roughly what the problem was, so I thought I'd let you know... We think the machine doing the DV reading/encoding was falling behind (possibly due to too high CPU usage, but possibly because of an issue in the dv grabber gstreamer component - we're discussing what we might do to solve some of these issues in the future. Unfortunately, the result of this is that the raw DV data is simply lost - the remaining data was encoded fine (there are no errors in the vorbis data according to my validation tools), but there's no possible way to recover the bits that went missing. There are a couple of minor muxing problems in the actual ogg files, which can be fixed easily enough (I've written some tools to help with this, and Conrad Parker has some too, let me know if you want to know how to fix these up - they confuse some players a bit), but the talk in that sample file is still pretty much incomprehensible due to the missing data after repairing the ogg-level errors. Mike
So there you go.