Hello, It's getting late here, so my wits aren't all passing muster, but here are a few thoughts. 1) I've never used nama without jack, have you? 2) Jack offers built-in extra latency compensation via the -I and -O options, and I think it should be the preferred way of handling this particular issue. 3) Based on this, I don't think we need worry about -z:multitrack. 4) It makes sense to set the buffersize (-b) ourselves, so we always know for certain what it is and can compensate for it. 5) We probably want to disable all buffering when recording. (rtlowlatency) 6) We can allow some buffering in monitor mode. (rt) 7) Theoretically, we might want very large buffering when performing certain operations such as caching and mixdown to avoid any possible xrun. *However,* it is worth noting that a project with inserts using other jack clients (and even more so the soundcard) would still need to run real-time during mixdown. I think that's about it in the pearls of wisdom department. The e-mail you quote regarding sensible buffering parameters does not really apply to nama because 1) it focuses about the single operation of recording a lot of audio to disk reliably and 2) it's 7 years old. Running a stable RT kernel (2.6.33.7.2-rt30) with jackd using a period size of 128 frames, I can run ecasound with -b:128 all night and, so long as the system doesn't come under stress, I don't experience any xruns. On a modern system, it really seems like overkill to use a buffersize over 1024 samples. Cheers, S.M. --