aengus

I'm looking for some information and references on appropriate sampling rates for physical controllers, specifically control surfaces with knobs and faders and so on. Personally I find that around 10-25ms sampling period seems to work fine, but I wonder are there any good studies or definitive results? I want to record incoming control data in Max accurately, but efficiently. Really my question is: how fast or at what frequency can someone typically move a knob or fader? or is it usually limited by the hardware in generally available controllers?

sampling-rate-for-physical-controllers

> Really my question is: how fast or at what frequency can someone

> typically move a knob or fader? or is it usually limited by the

> hardware in generally available controllers?

Generally available controlers is not very specific. If its Midi, its

event driven, the highest rate is the speed Midi can send data (ca. 1 ms

per command) I prefer that approach over any "Sampling rate" sometimes

called polling. A sample rate of 10-25 ms might be fine, but is always

too slow for catching rythmic events. I need the event in the moment it

happens. 1 ms is fine for that. If you do it with polling/sample rate,

you'd need a sample rate of 1 kHz which produces most of the time 

unneeded data and locks your machine for no advantage...

aengus schrieb:
> Really my question is: how fast or at what frequency can someone
> typically move a knob or fader? or is it usually limited by the
> hardware in generally available controllers?

Generally available controlers is not very specific. If its Midi, its
event driven, the highest rate is the speed Midi can send data (ca. 1 ms
per command) I prefer that approach over any "Sampling rate" sometimes
called polling. A sample rate of 10-25 ms might be fine, but is always
too slow for catching rythmic events. I need the event in the moment it
happens. 1 ms is fine for that. If you do it with polling/sample rate,
you'd need a sample rate of 1 kHz which produces most of the time 
unneeded data and locks your machine for no advantage...

-- 
Stefan Tiedje------------x-------
--_____-----------|--------------
--(_|_ ----|-----|-----()-------
-- _|_)----|-----()--------------
----------()--------www.ccmix.com


I understand that rhythmic events require much higher sample rates, but I'm thinking of an automation-type application -recording and playback of some continuous parameter value such as a level or the centre frequency of a filter- I can only think of ways of doing that in Max which involve metro-driven polling of a value. Also in my case, I want to process the control data and my processing requires that the data are recorded at a fixed sample rate.

Is there a definitive minimum for that sample rate? Or even a generally accepted range?

On Mar 4, 2008, at 3:52 AM, aengus wrote:

> Is there a definitive minimum for that sample rate? Or even a 

For knobish things, I've found that a 25-50 ms polling rate, with 

(optionally) interpolation between points on playback, is a pretty 

good compromise between maintaining gestural accuracy and data 

The Gesture Recorder stuff I put up on my site does that:

> Is there a definitive minimum for that sample rate? Or even a  
> generally accepted range?

For knobish things, I've found that a 25-50 ms polling rate, with  
(optionally) interpolation between points on playback, is a pretty  
good compromise between maintaining gestural accuracy and data  
reduction.

The Gesture Recorder stuff I put up on my site does that:
http://www.xfade.com/max

There isn't really a fixed "standard" when it comes to control latency. 

If you talk to most musicians, they will want "as little as possible", 

but you are going to be limited by the speed of whatever hardware 

communication protocol you use. Pretty much any hardware interface is 

going to have some amount of latency, but it will vary with the 

particular implementation. You will have to find your own happy balance 

between data-rate, processing speed, hardware limitations, and 

perceptual effectiveness. Most people I talk to find that up to 50ms is 

musically tolerable, and sometimes more if temporal interpolation is 

used effectively. I would probably try to keep it below 50ms if possible.

There isn't really a fixed "standard" when it comes to control latency.  
If you talk to most musicians, they will want "as little as possible", 
but you are going to be limited by the speed of whatever hardware 
communication protocol you use.  Pretty much any hardware interface is 
going to have some amount of latency, but it will vary with the 
particular implementation.  You will have to find your own happy balance 
between data-rate, processing speed, hardware limitations, and 
perceptual effectiveness.  Most people I talk to find that up to 50ms is 
musically tolerable, and sometimes more if temporal interpolation is 
used effectively.  I would probably try to keep it below 50ms if possible.

> Most people I talk to find that up to 50ms is musically tolerable,

> and sometimes more if temporal interpolation is used effectively. I

> would probably try to keep it below 50ms if possible.

Yes, if its a constant latency, no if its jittery (f.e. varies between 0 

and 50 ms) - polling creates jitter. Midi is fine, fast and could be 

thinned without loosing that precision with speedlim for example...

Andrew Benson schrieb:
> Most people I talk to find that up to 50ms is musically tolerable,
> and sometimes more if temporal interpolation is used effectively.  I
> would probably try to keep it below 50ms if possible.

Yes, if its a constant latency, no if its jittery (f.e. varies between 0 
and 50 ms) - polling creates jitter. Midi is fine, fast and could be 
thinned without loosing that precision with speedlim for example...

-- 
Stefan Tiedje------------x-------
--_____-----------|--------------
--(_|_ ----|-----|-----()-------
-- _|_)----|-----()--------------
----------()--------www.ccmix.com

Sampling rate for physical controllers