14 May
2005
14 May
'05
8:21 a.m.
Florian's pratical experience confirms what I stated in my other message.
My take on the matter is to get a good soundcard with fast DAC/ADC (
2msec round trip AFAIK)
and use something like 64frames latency. 64 frames gives you still
good,reliable RT performance and does not
push the kernel to the limits nor does the setup part of DSP of dsp
algorithms get bigger than the
actual audio rendering.
eg in a sampler you don't have only to render the audio samples but
check for note events, ramping,
calculate filter coefficients etc which all take some time which could
be bigger than rendering the audio
of N samples for N a very small number (eg N < 4).
cheers,
Benno
http://www.linuxsampler.org
Florian Schmidt wrote:
On Fri, 13 May 2005 17:02:42 -0400
Lee Revell <rlrevell(a)joe-job.com> wrote:
I beg to differ. Especially because some AD/DA's introduce quite a bit
of additional latency. When i used my cs46xx based card i had to go down
to 32 frames to get an acceptable latency for playing my guitar through
jack-rack. I once measured the roundtrip latency of this card and found
out my AD/DA's added another 4-5ms of latency to the 1.333ms latency
which were caused by the periodsize of 32 frames. And it seems 6ms total
latency was fine for me :)
Flo
paul knows
that zyn is not RT-safe. he had several conversations with
me at LAC2005 about how to redesign the internals. it works fine at
larger latencies.
Judging from those messages, it seems like it should certainly work at
128 frames, maybe 64. But 32 is insanely low - too low for most sound
hardware. That setting is more appropriate for stress testing the
kernel, than for actual music making.