[LAD] MIDI jitter

[Ralf Mardorf]

On Tue, 2010-07-06 at 10:57 +0100, Daniel James wrote:
[snip]
> I notice
> http://www.rme-audio.de/en_support_techinfo.php?page=content/support/en_support_techinfo_steadyclock
> doesn't mention USB or MIDI at all.
> 
> Cheers!
> 
> Daniel

Hi all :) hi Daniel :)

I'll read this link tomorrow, I just did a short test, right after the
postman did give me the ordered equipment. Please take a look at all the
tests I did here.

The Terratec's MIDI might be ok, but ...

spinymouse11.2 at suse11-2:~> cat .alias
alias cpu-o="su -c\"cpufreq-set -gondemand\""
alias cpu-p="su -c\"cpufreq-set -gperformance\""
spinymouse11.2 at suse11-2:~> cpu-p
Password: 
spinymouse11.2 at suse11-2:~> uname -a
Linux suse11-2 2.6.31.6-rt19 #1 SMP PREEMPT RT Wed Nov 18 16:59:26 CET
2009 x86_64 x86_64 x86_64 GNU/Linux
spinymouse11.2 at suse11-2:~> alsa-midi-latency-test -l
 Port    Client name                      Port name
 14:0    Midi Through                     Midi Through Port-0
 16:0    TerraTec EWX24/96                TerraTec EWX24/96 MIDI
 20:0    USB Device 0x170b:0x11           USB Device 0x170b:0x11 MIDI 1
spinymouse11.2 at suse11-2:~> alsa-midi-latency-test -i16:0 -o16:0
> SUCCESS

 best latency was 0.98 ms
 worst latency was 1.42 ms, which is great.

spinymouse11.2 at suse11-2:~> alsa-midi-latency-test -Rrw20 -i16:0 -o16:0
> SUCCESS

 best latency was 0.99 ms
 worst latency was 1.11 ms, which is great.

Then I run glxgears and Firefox with windows always on top and moved the
Firefox window while running the tests.

spinymouse11.2 at suse11-2:~> alsa-midi-latency-test -i16:0 -o16:0
> SUCCESS

 best latency was 0.98 ms
 worst latency was 4.15 ms, which is great.

spinymouse11.2 at suse11-2:~> alsa-midi-latency-test -Rrw20 -i16:0 -o16:0
> SUCCESS

 best latency was 0.99 ms
 worst latency was 1.11 ms, which is great.

Then I tested if the hrtimer might change something, I dunno if the test
will use it automatically.

spinymouse11.2 at suse11-2:~> su
Password: 
suse11-2:/home/spinymouse11.2 # chgrp audio /dev/hpet
suse11-2:/home/spinymouse11.2 # sysctl -w dev.hpet.max-user-freq=64
dev.hpet.max-user-freq = 64
suse11-2:/home/spinymouse11.2 # modprobe snd-hrtimer
suse11-2:/home/spinymouse11.2 # cat /proc/sys/dev/hpet/max-user-freq
64
suse11-2:/home/spinymouse11.2 # exit

Firefox and glxgears still on top of the windows and I moved the Firefox
windows again during the test.

Note that I now used the -R switch for both tests.

pinymouse11.2 at suse11-2:~> alsa-midi-latency-test -Ri16:0 -o16:0
> SUCCESS

 best latency was 0.99 ms
 worst latency was 1.08 ms, which is great.

spinymouse11.2 at suse11-2:~> alsa-midi-latency-test -Rrw20 -i16:0 -o16:0
> SUCCESS

 best latency was 0.99 ms
 worst latency was 1.12 ms, which is great.

Just for comparison one test for the Swissonic USB MIDI device, without
running glxgears or Firefox or moving any window. HPET still enabled.

spinymouse11.2 at suse11-2:~> alsa-midi-latency-test -Rrw20 -i20:0 -o20:0
> SUCCESS

 best latency was 1.17 ms
 worst latency was 2.23 ms, which is great.

Now a final comment to those tests.

When I used the USB MIDI device + HPET the audible result wasn't usable
for music, but I had the impression that half of the jitter would solve
this issue.
If the results of the test are correct and if nothing would change when
running JACK and doing hard disk recording too, then I guess the PCI
MIDI could be ok.

I don't have much time today, perhaps tonight or tomorrow I'll mount the
new HDD and restore my 64 Studio's. When it's done I'll record some
music and additionally I'll ask Achim, http://achimjaroschek.com/ , to
stress the computer by playing the Roland drums and some hardcore
Classic or hardcore Jazz on the keyboards.
It's not only that he plays with all those music giants like Jasper
van't Hof, Peter Brötzmann etc., but he once throw his Apple through the
window and he always advise me not to make music using the computer
anymore.

I've got a good feeling, that around 1ms (when the -R switch is set)
would be good enough to make music, but again, even if the test says
2.23 ms for the USB device should be great, the USB device is unusable
for serious musicians, it results in music that might be done by am
idiot without any sense for music.

I couldn't use the USB device + HPET even for the simple Pop-Rock I
sometimes make.

Cheers!

Ralf