[LAU] JACK MIDI latency -- was Re: Concerning schedtool

Robin Gareus robin at gareus.org
Sat Apr 19 17:47:56 UTC 2014


On 04/19/2014 06:18 PM, Ralf Mardorf wrote:
> On Sat, 2014-04-19 at 16:42 +0200, Robin Gareus wrote:
>> On 04/19/2014 04:13 PM, Jeremy Jongepier wrote:
>>> Why not use Jack2's built-in alsarawmidi driver?
>>
>> because it [excuse my language] sucks. very bad timing & jitter.
> 
> In the past I used jack2's -Xalsarawmidi option to get rid of bad timing
> and jitter. I can't speak for current jack1 and jack2, but some time ago
> -Xalsarawmidi improved those things a lot compared to jack1, which
> didn't provide this option and compared to jack2 without using this
> option.

Indeed they're pretty close these days. That was not true last time I
measured it. a2jmidid still has a slight edge, here. Same latency as
alsarawmidi but less jitter.


jackd2 from git (ab409a65df).
measured with https://github.com/x42/jack_midi_latency
using a PreSonus 1818VSL on a thinkpad x60s. linux 3.2.35 SMP PREEMPT RT

======================

## jackd -S -dalsa -dhw:2 -r48000 -p64 -n2 -Xraw

JACK settings: samplerate: 48000, samples/period: 64
               probable nominal jack latency: 128 [samples] = 2.67 [ms]
TOTAL: 10000 MIDI events sent+received.
 min=   183 max=   209 range=    26 avg= 193.6 dev=  4.69 [samples]
 min=  3.81 max=  4.35 range=  0.54 avg=   4.0 dev=  0.10 [ms]

 0.00 ..  3.78 [ms]:      0
 3.78 ..  3.80 [ms]:      2 #
 3.80 ..  3.81 [ms]:      0
 3.81 ..  3.83 [ms]:     13 #
 3.83 ..  3.85 [ms]:     44 #
 3.85 ..  3.86 [ms]:    127 #
 3.86 ..  3.88 [ms]:      0
 3.88 ..  3.89 [ms]:    142 #
 3.89 ..  3.91 [ms]:    270 ##
 3.91 ..  3.93 [ms]:    222 ##
 3.93 ..  3.94 [ms]:    491 ####
 3.94 ..  3.96 [ms]:      0
 3.96 ..  3.98 [ms]:   4930 ###########################################
 3.98 ..  3.99 [ms]:    263 ##
 3.99 ..  4.01 [ms]:    174 #
 4.01 ..  4.03 [ms]:      0
 4.03 ..  4.04 [ms]:    159 #
 4.04 ..  4.06 [ms]:    152 #
 4.06 ..  4.07 [ms]:    612 ######
 4.07 ..  4.09 [ms]:    200 ##
 4.09 ..  4.11 [ms]:      0
 4.11 ..  4.12 [ms]:    298 ###
 4.12 ..  4.14 [ms]:    170 #
 4.14 ..  4.16 [ms]:    188 #
 4.16 ..  4.17 [ms]:    132 #
 4.17 ..  4.19 [ms]:      0
 4.19 ..  4.20 [ms]:    733 #######
 4.20 ..  4.22 [ms]:    490 ####
 4.22 ..  4.24 [ms]:    153 #
 4.24 ..   inf [ms]:     34 #

======================

## jackd -S -dalsa -dhw:2 -r48000 -p64 -n2
## a2jmidid -e

JACK settings: samplerate: 48000, samples/period: 64
               probable nominal jack latency: 128 [samples] = 2.67 [ms]
TOTAL: 10000 MIDI events sent+received.
 min=   184 max=   209 range=    25 avg= 193.0 dev=  4.09 [samples]
 min=  3.83 max=  4.35 range=  0.52 avg=   4.0 dev=  0.09 [ms]

 0.00 ..  3.80 [ms]:      0
 3.80 ..  3.83 [ms]:     48 #
 3.83 ..  3.85 [ms]:     73 #
 3.85 ..  3.88 [ms]:     99 #
 3.88 ..  3.91 [ms]:    131 #
 3.91 ..  3.93 [ms]:    164 #
 3.93 ..  3.96 [ms]:   6318 ############################################
 3.96 ..  3.98 [ms]:    252 #
 3.98 ..  4.01 [ms]:    174 #
 4.01 ..  4.04 [ms]:    133 #
 4.04 ..  4.06 [ms]:    906 #######
 4.06 ..  4.09 [ms]:    223 #
 4.09 ..  4.11 [ms]:    151 #
 4.11 ..  4.14 [ms]:     96 #
 4.14 ..  4.17 [ms]:    220 #
 4.17 ..  4.19 [ms]:    623 ####
 4.19 ..   inf [ms]:    388 ###

======================

## jackd -S -dalsa -dhw:2 -r48000 -p64 -n2 -Xseq

interestingly has the lowest latency (one period less) but 3 times the
jitter.

JACK settings: samplerate: 48000, samples/period: 64
               probable nominal jack latency: 128 [samples] = 2.67 [ms]
TOTAL: 10000 MIDI events sent+received.
 min=   128 max=   179 range=    51 avg= 154.2 dev= 13.59 [samples]
 min=  2.67 max=  3.73 range=  1.06 avg=   3.2 dev=  0.28 [ms]

 0.00 ..  2.57 [ms]:     53 #
 2.57 ..  2.69 [ms]:    542 ##############
 2.69 ..  2.81 [ms]:   1834 ############################################
 2.81 ..  2.93 [ms]:   1078 #############################
 2.93 ..  3.05 [ms]:    764 ####################
 3.05 ..  3.18 [ms]:   1750 #########################################
 3.18 ..  3.30 [ms]:    887 ########################
 3.30 ..  3.42 [ms]:   1164 ###############################
 3.42 ..  3.54 [ms]:   1720 ########################################
 3.54 ..  3.66 [ms]:    207 #####
 3.66 ..  3.78 [ms]:      0
 3.78 ..   inf [ms]:      0



ciao,
robin


More information about the Linux-audio-user mailing list