On Oct 2, 2006, at 4:46 PM, Keith Moore wrote: >> >> Have you ensured that two successive bounces of the same type yield >> complete cancellation? There may be some randomness factor in >> something that may prevent identical results in any case. This is >> worth verifying first. >> > > The first test was a sinple (just a few audio and audio instruments) > bounce... one real time, one offline. > Phase cancellation was perfect.... silence. > > The next test was quite a complex track, around 30 tracks of audio & > audio instruments, including VSL, Post's "Old Lady", (a few tracks > frozen) plus lots of automation. > Phase cancellation revealed a marimba. some reverb and what sounded > like leakage from a delay line, I had used the Stereo Delay plug > (bussed) fairly extensively. My question still stands--there may be something about the content or processing that does not come out the same way twice which is different between the two tests. You should try doing two passes real-time and two passes offline and seeing if those compare successfully to themselves before comparing them with each other.