21 Feb
2023
21 Feb
'23
12:36 a.m.
Hey hey,
I'm experimenting with c++ trying to program something nice for MIDI. I'm now
experimenting with clocks, using both the standard c++ chrono library and the
boost chrono library. My example program sets a desired delta between ticks of
250ms. I see that there is a difference, since Boost chrono can also use
pthread thread parameters foor realtime scheduling and priorities.
With the standard chrono and thread library my ticks are usually out by
anything between 150.000 to 290.000 nanoseconds. Using boost a tick is out
anything between 110.000 to 124.000 nanoseconds. Yes, much better. But,
assuming that I correct the drift, does it make a difference for MIDI
instruments?
I know one of my MIDI instruments with a clock sync'able delay can be rather
touchy with MIDI clocks, but are there good estimated guidelines experience
values how precise ticks should be?
Best wishes and thanks for any help,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
Top down, on the strip
Lookin' in the mirror
I'm checkin' out my lipstick <3
(Britney Spears)
21 Feb
21 Feb
9:57 a.m.
If you are worried, I would try to use alsa sequencer queues, where you
specify when you want the event to be sent and ALSA makes sure its as
accurate as possible.
https://github.com/alsa-project/alsa-lib/blob/master/test/playmidi1.c
This example sets the queue to use tempo but its possible to use clock time
events, relative or absolute.
Anyway, a back of the napkin calculation shows that the old MIDI protocol
at 32650 kbit/sec needs around 245 022 us to send a byte. Missing a byte
should not make a huge difference. So you can not really get more precise
than that.. on old DIN5 MIDI. USB MIDI is faster but many times limited by
the same speed for compatibility.
Regards,
David
On Tue, 21 Feb 2023 at 00:35, Jeanette C. <julien(a)mail.upb.de> wrote:
Hey hey,
I'm experimenting with c++ trying to program something nice for MIDI. I'm
now
experimenting with clocks, using both the standard c++ chrono library and
the
boost chrono library. My example program sets a desired delta between
ticks of
250ms. I see that there is a difference, since Boost chrono can also use
pthread thread parameters foor realtime scheduling and priorities.
With the standard chrono and thread library my ticks are usually out by
anything between 150.000 to 290.000 nanoseconds. Using boost a tick is out
anything between 110.000 to 124.000 nanoseconds. Yes, much better. But,
assuming that I correct the drift, does it make a difference for MIDI
instruments?
I know one of my MIDI instruments with a clock sync'able delay can be
rather
touchy with MIDI clocks, but are there good estimated guidelines
experience
values how precise ticks should be?
Best wishes and thanks for any help,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
Top down, on the strip
Lookin' in the mirror
I'm checkin' out my lipstick <3
(Britney Spears)
_______________________________________________
Linux-audio-dev mailing list -- linux-audio-dev(a)lists.linuxaudio.org
To unsubscribe send an email to linux-audio-dev-leave(a)lists.linuxaudio.org
--
David Moreno Montero
dmoreno(a)coralbits.com
+34 658 18 77 17
[image: Coralbits.com] <http://www.coralbits.com/>
http://www.coralbits.com
10:06 a.m.
Hi David!
Feb 21 2023, David Moreno Montero has written:
If you are worried, I would try to use alsa sequencer
queues,
I'm working with RtMidi. I don't think I have access to that low
level.
...
Anyway, a back of the napkin calculation shows that
the old MIDI protocol
at 32650 kbit/sec needs around 245 022 us to send a byte. Missing a byte
should not make a huge difference. So you can not really get more precise
than that.. on old DIN5 MIDI. USB MIDI is faster but many times limited by
the same speed for compatibility.
Thank you. This is helpful. Maybe I should just
give it a go, make up an
actual clock rate and see what the delay on my outboard synth does. I'm
definitely below that value.
...
Best wishes,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
Skip on the drinks Head to the floor
Makin' my way Past the show
My body's taken over And I want some more <3
(Britney Spears)
10:31 a.m.
On 21/02/2023 10:06, Jeanette C. wrote:
Yeah, RtMidi API is awfully limited when compared to what ALSA
sequenceer can do. Especially for timing and routing. But it is simple
and widely available on multiple platforms, so it is often still a good
choice.
Jacek
If you are
worried, I would try to use alsa sequencer queues,
I'm working with RtMidi. I
don't think I have access to that low level. 
22 Feb
22 Feb
9:28 a.m.
Hi Jeanette,
On 2/21/2023 12:36 AM, Jeanette C. wrote:
[...]
With the standard chrono and thread library my ticks
are usually out by
anything between 150.000 to 290.000 nanoseconds. Using boost a tick is
out anything between 110.000 to 124.000 nanoseconds. Yes, much better.
But, assuming that I correct the drift, does it make a difference for
MIDI instruments?
I know one of my MIDI instruments with a clock sync'able delay can be
rather touchy with MIDI clocks, but are there good estimated guidelines
experience values how precise ticks should be?
the kind of precision required for timers depends very much on what
precisely you are trying to do. Are you trying to implement MIDI Time
Code (MTC) or MIDI Clock (MC)? Are you trying to implement changing the
timing of incoming MIDI events? Or something else completely?
Kind regards,
FPS
--
https://dfdx.eu
10:52 a.m.
Hi Florian,
thanks for your considered response and questions. Here's my goal and
progress so far...
Feb 22 2023, Florian Paul Schmidt has written:
...
Are you trying to implement MIDI Time
Code (MTC) or MIDI Clock (MC)? Are you trying to implement changing the
timing of incoming MIDI events? Or something else completely?
...
I'm trying to create a MIDI clock, here's the example code I use, based
on the midiclock.cpp example that comes with rtmidi:
https://www.dropbox.com/s/b0bh9pxsqzmar5j/midiclock.cpp
I could use the boost thread library, which allows to set thread
attributes of the underlying thread library (pthread), which appears to
be a little more precies.
The example shows that, most of the time, the clock is quite good. But
ever so often the clock stumbles quite audibly.
I'm working towards a kind of sequencer application with quantised and
unquantised events. I suppose unquantised is relative. Let's say adapted
microtiming to an internal clock at a higher rate (five or ten times
faster). That may include filtering or otherwise modifying events.
So many things yet to work out. Currently it's a "study project" to
understand "all things MIDI" and working out/finding solutions to many
of the involved challenges.
Best wishes and thanks agian,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
I saw your smile
Stay with me a while <3
(Britney Spears)
12:03 p.m.
On Wed, Feb 22, 2023 at 10:52:49AM +0100, Jeanette C. wrote:
This is very strange:
miss_time = now - wakeup;
wakeup = now + (delta - miss_time);
Which just means:
wakeup += delta;
So it's not wrong, just confusing by suggesting this
somehow compensates for the previous tick being late.
It doesn't, nor should that be done.
--
FA
1:28 p.m.
Hi Fons!
Feb 22 2023, Fons Adriaensen has written:
...
This is very strange:
miss_time = now - wakeup;
wakeup = now + (delta - miss_time);
Which just means:
wakeup += delta;
So it's not wrong, just confusing by suggesting this
somehow compensates for the previous tick being late.
It doesn't, nor should that be done.
...
Thanks Fons! I was really confused about that myself. Mixing up the
mechanism of sleeping for a specified duration and sleeping until a
specified point in time.
Best wishes,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
With you I get so high
Lost in the crystal sky
You are this melody
That's where you take me <3
(Britney Spears)
3:42 p.m.
On Wed, Feb 22, 2023 at 01:28:21PM +0100, Jeanette C. wrote:
Thanks Fons! I was really confused about that myself.
Mixing up the
mechanism of sleeping for a specified duration and sleeping until a
specified point in time.
Only the latter is correct.
BTW, just using
sem_timedwait ()
in a RT thread I get
Min ~ 60 us
Max ~ 135 us
Average ~ 74 us.
Ciao,
--
FA
23 Feb
23 Feb
1:51 p.m.
On Wed, Feb 22, 2023 at 03:42:00PM +0100, Fons Adriaensen wrote:
BTW, just using
sem_timedwait ()
in a RT thread I get
Min ~ 60 us
Max ~ 135 us
Average ~ 74 us.
And using
clock_nanosleep (CLOCK_MONOTONIC, TIMER_ABSTIME, &T, 0);
in a RT thread I get
Min ~ 5 us
Max ~ 105 us
Average ~ 23 us
No need for any special libraries...
Ciao,
--
FA
2:36 p.m.
Hi Fons,
thanks for this tip, The functions look good and helpful. But perhaps
I'm not getting something. I made a very simple test program, no RT
thread, no special thread at all.
https://www.dropbox.com/s/ox9owa6fdd459dt/nanosleep.cpp
I tried including both time.h and ctime. The result is that the program
just returns and the wakeup time and now drift apart ever more. Does it
need a specific thread environment?
Best wishes,
Jeanette
Feb 23 2023, Fons Adriaensen has written:
...
And using
clock_nanosleep (CLOCK_MONOTONIC, TIMER_ABSTIME, &T, 0);
in a RT thread I get
Min ~ 5 us
Max ~ 105 us
Average ~ 23 us
No need for any special libraries...
Ciao,
--
FA
_______________________________________________
Linux-audio-dev mailing list -- linux-audio-dev(a)lists.linuxaudio.org
To unsubscribe send an email to linux-audio-dev-leave(a)lists.linuxaudio.org
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
Do you even know what I like
Just what I'm living for,
What I adore <3
(Britney Spears)
3:28 p.m.
New subject: [SOLVED] Re: Re: Admissable clock inaccuracy for MIDI
Feb 23 2023, Jeanette C. has written:
...
I made a very simple test program, no RT
thread, no special thread at all.
https://www.dropbox.com/s/ox9owa6fdd459dt/nanosleep.cpp
...
It just needed a function to do clever increasing of the timespec
nanoseconds.
Best wishes and thanks again,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
Do you even know what I like
Just what I'm living for,
What I adore <3
(Britney Spears)
634
days inactive
637
days old
linux-audio-dev@lists.linuxaudio.org
11 comments
5 participants
participants (5)
-
David Moreno Montero -
Florian Paul Schmidt -
Fons Adriaensen -
Jacek Konieczny -
Jeanette C.