Hello, with a 8192 samples for analysing audio in 48000hz you should get 170ms latency, unless I've forgotten something in the equation. With increasing the amount of the window size in [fiddle] you will be able to detect lower notes, for exemple with 2048 samples the lower note detected is a A3 flat ( 207.65hz), your latency will be about ~42ms + audiobuffers. I've found some mathematical references, you might be interested:

http://aubio.piem.org/articles/brossier04realtimesegmentation.pdf

With a linux box the audio buffers can be reduced to ridiculous amount in comparission with what you can get with windows, following what driver you are using (OSS or ALSA) before the sound glitches. For some reason I ignore, the driver that still work under an amount of audiobuffers of 5ms is OSS.

Julian Villegas julovi@yahoo.com a écrit :Tks for your answer, I know I can have lower latencies with Linux, and Mac Os, (Probably in that order too)... but in fact I have a satisfactory low latency in my patch in XP... maybe I didn't explain correctly my problem (or you didn't read well), but what I'm trying to know (and it's independent from the OS) is what delay I'm having by using a 8192 samples window in fiddle~ (85 ms), or the delay value specified in the audio properties (8 ms), since none of those seems to be the actual one.

Faites de Yahoo! votre page d'accueil sur le web pour retrouver directement vos services préférés : vérifiez vos nouveaux mails, lancez vos recherches et suivez l'actualité en temps réel. Cliquez ici.