On Sun, 14 Feb 2016 14:43:01 +0000
Harry van Haaren <harryhaaren(a)gmail.com> wrote:
Hope that clears up exactly whats going on :)
A good definition is looming...
Using the linux generic kernel as is, the following measurements were
taken by connecting a single RCA cable from playback 3 to capture 6 and
using qjackctl (which must be restarted before each run since jackd is
launched from the command line).
The goal now is to be able to have a delay that cannot be noticed
without having xruns, crackling noise, i.e, while keeping processing
running smoothly. There is no choice than to drop the delay below 5ms
to get best response when playing. So this means 128. While 64 could
be great, I am not so sure about overdoing it.
If I understood correctly what will help in getting this are the
low latency kernel, and a re-assignment of real time processing of IRQs.
/usr/bin/jackd -T -ndefault -dalsa -dhw:M1010LT -r44100 -p512 -n2
configuring for 44100Hz, period = 512 frames (11.6 ms), buffer = 2
periods
1597.528 frames 36.225 ms total roundtrip latency
extra loopback latency: 61 frames
use 30 for the backend arguments -I and -O
/usr/bin/jackd -T -ndefault -dalsa -dhw:M1010LT -r44100 -p256 -n2
configuring for 44100Hz, period = 256 frames (5.8 ms), buffer = 2
periods
829.529 frames 18.810 ms total roundtrip latency
/usr/bin/jackd -T -ndefault -dalsa -dhw:M1010LT -r44100 -p128 -n2
configuring for 44100Hz, period = 128 frames (2.9 ms), buffer = 2
periods
445.529 frames 10.103 ms total roundtrip latency
/usr/bin/jackd -T -ndefault -dalsa -dhw:M1010LT -r44100 -p64 -n2
configuring for 44100Hz, period = 64 frames (1.5 ms), buffer = 2 periods
253.529 frames 5.749 ms total roundtrip latency
/usr/bin/jackd -T -ndefault -dalsa -dhw:M1010LT -r44100 -p32 -n2
configuring for 44100Hz, period = 32 frames (0.7 ms), buffer = 2 periods
157.529 frames 3.572 ms total roundtrip latency