MIDI jitter - Was: outfitting a computer for songwriting in linux?
Jeremy, we shouldn't discuss this in the original thread.
There are latency test measurements in the mailing list archive. PCI and
PCIe cards show better results, IOW less MIDI jitter. There are also
explanations why USB suffers from more jitter.
The point of difference is, if the MIDI jitter is audible. It is! The
measurements might not provide correct results. To measure a system by
itself is a bad test and to measure a MIDI loop in addition is tricky.
Jakob (living in Augsburg) from a Linux mailing list (perhaps LAU or
LAD) years ago send me a board to loop through MIDI and that provides an
audio output, to record the MIDI signal to an audio track. I never had
time and interest to continued my MIDI jitter tests, so I still didn't
use it. I guess this would be a better test, then the latency test.
Howsoever, it's better to use PCI or PCIe, IIRC some USB interfaces
even fail(ed) the latency test that measures a MIDI loop. The USB
interface I own doesn't fail the test, but I'm anyway not satisfied and
use PCI and PCIe MIDI only.
That's it, I can't provide more information and I'm not willing to
search the mailing lists archives.
Re: MIDI jitter - Was: outfitting a computer for songwriting in linux?
On 07/26/2015 12:17 PM, Ralf Mardorf wrote:
> Jeremy, we shouldn't discuss this in the original thread.
> There are latency test measurements in the mailing list archive. PCI and
> PCIe cards show better results, IOW less MIDI jitter. There are also
> explanations why USB suffers from more jitter.
> The point of difference is, if the MIDI jitter is audible. It is! The
> measurements might not provide correct results. To measure a system by
> itself is a bad test and to measure a MIDI loop in addition is tricky.
> Jakob (living in Augsburg) from a Linux mailing list (perhaps LAU or
> LAD) years ago send me a board to loop through MIDI and that provides an
> audio output, to record the MIDI signal to an audio track. I never had
> time and interest to continued my MIDI jitter tests, so I still didn't
> use it. I guess this would be a better test, then the latency test.
> Howsoever, it's better to use PCI or PCIe, IIRC some USB interfaces
> even fail(ed) the latency test that measures a MIDI loop. The USB
> interface I own doesn't fail the test, but I'm anyway not satisfied and
> use PCI and PCIe MIDI only.
> That's it, I can't provide more information and I'm not willing to
> search the mailing lists archives.
Thanks, I'll dig through the archives to see if I can find related messages.
On 07/26/2015 11:42 AM, Ralf Mardorf wrote:> On Sun, 26 Jul 2015
11:01:46 +0200, Jeremy Jongepier wrote:
>> In the meanwhile the whole planet is using USB for MIDI.
> That's an assertion without substance.
I don't think it is. There are virtually no manufacturers that produce
great numbers of PCI(e) interfaces with MIDI and besides, the number of
new computers with PCI(e) is not that great too. That leaves basically
one choice, USB MIDI.
>> Please explain what is wrong with it, if possible with test results or
>> something else that confirms your assumption. I've never experienced
>> any jitter issues, not even with external MIDI gear.
> Search the LAU and LAD mailing list archives or edit a
> four-on-the-floor kick MIDI track. Play an external synth and make an
> audio recording of that kick, record the kick again and you'll notice
> that the kicks are never in sync, you'll hear an early reflection like
> shift or a moving phasing.
I'll give this a try, good suggestion, thanks.
If you make syncopated or measure free music
> this is a serious issue, but it already is an issue if you want to
> sync four-on-the-floor kicks or similar. A lot of people claim that
> MIDI is not good enough to record "hand made" music, it always suffers
> from quantisation. To record really "hand made" music there's the need
> to do audio recordings. That's not completely wrong, but MIDI isn't that
> bad as many people nowadays think that it is. If you use a C64 or Atari
> ST and sync by klick or by SMPTE to a tape recorder, you get perfect
> I noticed that shift of MIDI events is related to the used audio
> latency. The longer the latency is, the less precise is the timing of
> the recorded MIDI events on audio tracks. I never was able to get less
> audio latency than 5.3 ms (-r48000 -p256 -n2).
"Academic studies have indicated that jitter becomes noticeable between
1 and 5ms, with a weighting around 2-3ms (Van Noorden 1975, Michon
1964, and Lunney 1974). Note that these tests only measured the
perception of jitter while listening, not while performing. Intuitively
I expect performing musicians to be even less tolerant to it."
"As sample buffer sizes are reduced, the jitter reduces"
When we measured the MIDI loops a few years ago people didn't believe
that I'm able to notice the jitter and they posted links about "good"
drummers who are more out of time than MIDI jitter is. Even while it's
true that a good musician is out of time too, it's done with human
touch, the human "jitter" provides the good groove. Machine's jitter
breaks the groove, it's arbitrary jitter, The musicians jitter is
aligned to the emotion the music should communicate.
However, the link above also claims that _it is noticeable_ and we
should keep in mind that a MIDI loop measurement doesn't care about
MIDI jitter of the receiving synth and jitter caused when the audio
signal is synced to the MIDI events. When recording very short click
sounds with a short attack, it's still hard to find the position were
the audio signal starts, when taking a look at the audio files, so
measurements a hard to do.
Another nice test is to record the MIDI events parallel to the audio
signal. With a high resolution for the MIDI events (quasi no
quantisation) playing the audio and MIDI track parallel should be
identical, but it isn't.
Taking a look at audio recordings of MIDI tracks, I noticed that jitter
peaks are periodically. Even while the graphics doesn't share the IRQ
with the audio gear, the periodic jitter peaks differ when using
Re: MIDI jitter - Was: outfitting a computer for songwriting in linux?
PS: We also care about the tone of the notes, not only about the timing
of the notes. Tone of a note is not only related to the kind of touch,
but also of the timing, regarding phases to the other instruments that
are played. I suspect that human jitter is less randomly as machine's
jitter. Neuroscience can't explain a lot of human abilities.
I didn't read that tests again, but I guess we later made tests with
better results. Were the 2010 tests already made with jackd2
IIRC there were (later) tests made with jackd2 from svn and the
-Xalsarawmidi switch while HR TIMER - ALSA MIDI playback queue timer
res 1000000000 Hz was enabled. 2011 or later. My PCI cards and/or my
PCIe card had 0.0x ms jitter for the MIDI out to MIDI in loop
measurement. ^^^ Or perhaps even for the recorded MIDI instruments on
the audio tracks.
I don't remember if recordings to audio tracks were really that
synchrone too. IIRC there are several issues to consider. I guess the
0.0x ms were ok that time when recording to audio tracks, but IIRC there
always was around 1 ms for USB and it was audible.
Without jackd2 -Xalsarawmidi jitter increased rapidly.
jackd2's -Xalsarawmidi was new or fixed, I don't remember the year.
You always need to keep in mind that multi-track recordings could be
done one track after the other and the jitter of the MIDI events
recorded to the audio tracks has got positive and negative delay, so
1ms could become 2ms and 5ms could become 10ms.
An early reflection might be >= 5ms, but a similar effect already is
noticeable if the delay is shorter. If you record a synth or hihat
sample 2 times from a MIDI track to audio tracks to get a fixed phase
sound and there's just minimal jitter, everybody will hear the phasing
Better than measurements is simply making music. If there are no
audible issues, when recording MIDI to audio, then there are no issues.
If there are audible issues, then there are issues, what ever a
measurment would show.
In my experiences there are less often issues when using PCI/PCIe MIDI.
If it's wanted to use two external percussive sounds to become one
sound, then I wouldn't record them one after the other, as I did with
old sequencers, nowadays I try to record them at the same time or to
make a sound sample or ... I at least never ever would consider to use
PS: I won't to discuss this, resp. I have no time to continue this
discussion. Resume: If a musician doesn't notice an issue it's ok, if a
musician notice an issue, than it doesn't matter if most other people
are unable to notice it too. It's more likely that somebody is able to
notice jitter when using USB, than when using PCI/PCIe.
Linux-audio-user mailing list
[hidden email] http://lists.linuxaudio.org/listinfo/linux-audio-user