[LAU] Play with latencies

Jostein Chr. Andersen jostein at vait.se
Fri Dec 12 07:12:07 EST 2008


Hi,

I have played a little bit with the connections of latencies, distances 
and notes, just to get things in perspectives.

We always try to keep the latencies as low as possible and I'm no 
exception: my every day standard is 2.9 ms, but decreases down to 0.72 
in recording situations and maybe up to 20-50 ms when mixing (working in 
Ardour).

So how long time (latency) does it tak from when you tap rhythm with the 
foot to you or other can hear it, and how is it between the members of a 
big orchestra? Big latencies are apparently very common in real life 
situations and it's fun to know about it.

So here is the result of a wasted hour (hope every number is right), enjoy.

  Time in  Length Note,
       ms    in m 120 bpm Comments
======== ======= ======= ===========================================
  1000.00  343.00     1/1 en.wikipedia.org/wiki/Speed_of_sound
   500.00  171.50     1/2 Common delay speed (echo)
   250.00   85.75     1/4            "
   125.00   42.88     1/8            "
   100.00   34.30
    62.50   21.44    1/16 Distance between people in a big orch.
*  46.40   15.92         Latency value
    31.25   10.72    1/32 Wow, playing fast!
*  23.20    7.96         Latency value
    15.63    5.36    1/64 Playing real music that fast? Get a life!
*  11.60    3.98         Latency value
    10.00    3.43
     7.81    2.68   1/128 Academic interest (musically)
*   5.80    1.99         Latency value
     5.00    1.72         Dist. from ear to foot
     3.91    1.34   1/256
*   2.90    0.99         Latency value (64 f/p, 44100, 2 p/b)
     1.96    0.67   1/512
*   1.45    0.50         Latency, common dist. to monitor
     1.00    0.34         Not an uncommon distance to mic or wife
     0.98    0.33  1/1024
*   0.73    0.25         Latency value
     0.49    0.17  1/2048 Still of academic interest (musically)

* The latency numbers is taken from qjackctl's setup dialog box when 
frames/period are from 16 to 1024 on a 44100 sample rate system with 2 
periods/buffer.

This shows that the quite common rule of keeping everything under 
approx 10 ms make sense, right? :-)

Jostein



More information about the Linux-audio-user mailing list