[PD-ot] how low (latency) can you go?

Bryan Jurish moocow at ling.uni-potsdam.de
Sat Dec 16 22:52:54 CET 2006


On 2006-12-16 20:44:22, Hans-Christoph Steiner <hans at eds.org> appears to 
have written:
> 
> On Dec 16, 2006, at 1:14 PM, Bryan Jurish wrote:
> 
>>
>> On 2006-12-16 19:02:08, Hans-Christoph Steiner <hans at eds.org> appears 
>> to have written:
>>> And always, when talking about latency, I feel the need to point out 
>>> the speed of sound: 340 m/s or .34 m/ms.  So if your speakers are 2m 
>>> away from you, that's 6ms of latency.  You could spend days tweaking 
>>> your machine to get 3ms less latency, or you could move 1m closer to 
>>> your speakers.  Puts things into perspective...
>>
>> indeed 'tis true.  but 5ms vs. 10ms makes a major 
>> subjective/perceptual difference if we're talking about playing an 
>> electro-acoustic instrument in and getting a munged signal out...
> 
> Do you have any references on that?  I think it could be perseptable to 
> a very experienced musician, but I don't think there is a major effect.  

well, i wouldn't call myself a *very* experienced musician, but plugging 
a guitar into the computer and just running the signal through a simple 
[adc~]->[dac~] in pd with 10ms reported latency (latency.pd gives 
10-11ms) is easily detectable, it being a pain in the wazoo to play 
(probably has to do with the fact that i am at best only a mediocre 
guitarist, but that's not the point :-/).  also, 10ms is roughly the 
minimal "willed action -> muscular implementation" latency on your 
average homo sapiens architecture (get a stopwatch, try it out ;-)

 > From the studies that I have read, jitter has a much larger affect
 > than  latency in terms of affecting performance.   (I need to find
 > some references myself... ;)

i think you're right there.  i know Reinhold Kliegl and Ralf-Thomas 
Krampe have done a lot of rhythm perception studies in Potsdam, with 
professional musicians, amateur musicians, and non-musicians as 
subjects, and the pros pretty much blew everyone else away precision-wise.

 > Performers are actually quite adept at adjusting to latency. Just
 > think  of an orchestra or choir:  there could be 20ms or more of
 > latency from  one side of the stage to the other, yet its not hard 
for > a moderately  trained group to handle it.

as a former orchestra inmate myself, i seem to recall concentrating 
mostly on keeping a constant time myself -- the "big picture" ("big 
sound?") just isn't available to a 3rd tier 2nd violinist -- i went by 
the conductor's tempo and the momentum i got from those around me: the 
only real trick was keeping it constant, but that's "just" training: 
hence perhaps the jitter-related problems even for the pros...

> Another key factor in performance is 
> physical feedback, which has a much faster feedback loop that the audio 
> system.  Trained musicians can feel a mistake and correct it long before 
> the sounds are consciously perceived.  (I first saw that in print on one 
> of Miller's papers http://crca.ucsd.edu/~msp/Publications/icmc93.ps)
> 
> One realm were I could see that latency would be important would be when 
> measuring response time to a stimulus, like in a psychological 
> experiment.  But even still, as long as the latency is steady (i.e. low 
> jitter), the latency is not a big deal.

it can certainly be compensated for if it is known, but for me as a 
guitarist, it amounts essentially to learning an entirely new style, 
since i'm just plain not good enough to compensate before the glitches 
become painfully audible...

marmosets,
   Bryan

-- 
Bryan Jurish                           "There is *always* one more bug."
jurish at ling.uni-potsdam.de      -Lubarsky's Law of Cybernetic Entomology



More information about the PD-ot mailing list