[PD-ot] how low (latency) can you go?
Hans-Christoph Steiner
hans at eds.org
Tue Dec 19 04:52:45 CET 2006
On Dec 18, 2006, at 10:52 AM, marius schebella wrote:
> Hans-Christoph Steiner wrote:
>> Honestly, I doubt you could tell the difference between 4ms and
>> 14ms in a blind test. There have been many studies on this. If
>> you can detect such tiny differences in onset time, you would be
>> superhuman.
>
> please test me! are these studies made with instrumentalists and
> performers? (I did not read them yet, but know about perception
> tests in general) It is not a big deal to play scales with notes of
> about 40-50 ms. so 10 percent delay makes you believe you play
> slower. playing fast musical events makes the listener think it is
> one event, but actually the performer has to trigger two ore
> several events.
> it is not just pressing a key and asking did I percept the delay..
> marius.
Yes, let's do some tests! I think they would be great patches to
have. This stuff makes much more sense when you actually experience
it, rather then when reading obtuse academic papers.
One key distinction here is which system you are talking about. The
sense of touch/feel and muscle control are very fine grained. Good
drummers can control events on scale of 5ms or less. The brain's
subconscious audio processing can detect very small temporal changes
and events. But when talking about the perception of audio, things
are much slower. All this data is processed, packaged up, then
delivered, and that takes a lot longer. So when talking about
playing, listening, then adjusting what you are playing, that is a
much slower loop.
That's only part of the question of handling latency. I read a study
a while back that directly testing the affects of latency on highly
skilled musicians. It showed that the musicians could compensate for
things around 50ms, then there was a window around 300-700ms was the
most difficult, then longer was a bit easier, but still hard.
Perhaps there is a study somewhere that tests the latency of an
instrument versus the level of virtuosity (e.g. a violin), that would
cover this issue well.
I wrote up a simple test of the latency of audio and visual
perception. I get pretty consistently between 200 and 250ms for
sound and 300ms for visual.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test_your_audio_latency.pd
Type: application/octet-stream
Size: 1436 bytes
Desc: not available
Url : http://lists.puredata.info/pipermail/pd-ot/attachments/20061218/8652a549/test_your_audio_latency.obj
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test_your_visual_latency.pd
Type: application/octet-stream
Size: 763 bytes
Desc: not available
Url : http://lists.puredata.info/pipermail/pd-ot/attachments/20061218/8652a549/test_your_visual_latency.obj
-------------- next part --------------
One thing I wonder is what is the latency of musicians who are
improvising together? I am guessing it is much closer to the 250ms
reaction time than the 10ms low-latency mark. But it would be hard
to measure since there is a lot of prediction happening, rather than
purely listening/watching and responding.
And another references, but more related to a previous topic:
(performers are much more sensitive to jitter than latency)
http://openmuse.org/transport/fidelity.html
.hc
------------------------------------------------------------------------
Mistrust authority - promote decentralization. - the hacker ethic
More information about the PD-ot
mailing list