help puredata + rosegarden

[adrjork]

Hi, I've built a patch in puredata that outputs many MIDI-notes-out per seconds (128 MIDI-notes-out at 25 times per seconds). These MIDI-notes are on channels 17-24. So I linked PD to ALSA (bus 2, obviously because MIDI channels are 17-24), and I set Rosegarden as PureData MIDI receiver.
Then I receive with Rosegarden the MIDI-outs from PD, but here I've a problem: Rosegarden records MIDI-notes but not all! Few notes are recorded in Rosegarden.
In Mac (with PD + IAC driver + Logic), Logic receives all MIDI notes making "dense" graphics of MIDIs, while here in Linux (with PD + ALSA + Rosegarden), Rosegarden seems to loose many many signals!
Can you help me?
Thanks!

[adrjork]

If you want to give me an hand, please download my patch "v2MSy" at http://adrjork.altervista.org/puredatatutorials/v2MSy-tutorial_091031.pdf
or at
http://adrjork.altervista.org/puredatatutorials.html
Please, help me guys.
Thanks

[D. Michael McIntyre]

On Wednesday 04 November 2009, adrjork wrote:

> If you want to give me an hand, please download my patch "v2MSy" at
> http://adrjork.altervista.org/puredatatutorials/v2MSy-tutorial_091031.pdf

Well, I burned a couple of hours on this, to no effect.  I can't get it to do
anything with any video file I tried.  I click >select, navigate through a
file dialog, and nothing happens, no matter what file I choose.

I never got it to transmit any MIDI messages at all.

Anyway, it would be interesting to see if the problem here really is with
Rosegarden itself, or ALSA more generally.  Asking anything to swallow 192,000
messages per minute is a pretty tall order.  Can MusE handle the message
stream without dropping anything?  How about KMidiMon?

I don't think there's any MIDI reproduction engine on the planet that could
actually play that stream of data at full speed anyway.
--
D. Michael McIntyre

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Rosegarden-user mailing list
Rosegarden-user@... - use the link below to unsubscribe
https://lists.sourceforge.net/lists/listinfo/rosegarden-user

[GARY SEVIOUR]

Im not a puredata user but I will try to help as best I can

> channels are 17-24
I thought midi only allowed 0-15 channels?

Have you armed the respective tracks/channes and made the correct jack connections? You will be recording to 7 tracks at once. maybe you are only recording one channel of it at the moment?

>128 MIDI-notes-out at 25 times per seconds

A lot of notes in musical terms but I would  guess a decent PC could handle it.
Have you tried editing your patch to produce less notes? If so did RG record all of them.
I suspect that the issue is not much to do with RG rather the issue is with jack , ALSA  or the speed of your linux machine compared with a MAC(this is just a guess).

I've not noticed any problem with polyphony myself. 128 notes is a lot, is it just to stress test the machine?
I guess concerning the number of notes ber second you will need a latency of less than 40ms (1000/25 = 40) (just guessing again, could be nonsense).
On my computer so far I have been able to get the latency down to around .4ms (but choose >=1ms for safety). On the other hand higher latency tends to give higher reliability, so what I said might actually be contrary to the reality.

My experience of having very high polyphony on zynaddsubfx and a latency of around 21ms on quite an old machine (1Ghz, 384mb ram) was that you would start getting loads of lag when the processor was pushed to it limits (some one going mad on the keys with loads of sustain) while all the notes would be played eventually, no notes seemed to be dropped at all.

Hope some of my ideas (and thats all they are until proven) have helped.

--Gary

--- On Wed, 4/11/09, adrjork <adrjork@...> wrote:

> From: adrjork <adrjork@...>
> Subject: [Rosegarden-user]  help puredata + rosegarden
> To: rosegarden-user@...
> Date: Wednesday, 4 November, 2009, 4:04 PM
>
> Hi, I've built a patch in puredata that outputs many
> MIDI-notes-out per
> seconds (128 MIDI-notes-out at 25 times per seconds). These
> MIDI-notes are
> on channels 17-24. So I linked PD to ALSA (bus 2, obviously
> because MIDI
> channels are 17-24), and I set Rosegarden as PureData MIDI
> receiver.
> Then I receive with Rosegarden the MIDI-outs from PD, but
> here I've a
> problem: Rosegarden records MIDI-notes but not all! Few
> notes are recorded
> in Rosegarden.
> In Mac (with PD + IAC driver + Logic), Logic receives all
> MIDI notes making
> "dense" graphics of MIDIs, while here in Linux (with PD +
> ALSA +
> Rosegarden), Rosegarden seems to loose many many signals!
> Can you help me?
> Thanks!
> --
> View this message in context: http://old.nabble.com/help-puredata-%2B-rosegarden-tp26199159p26199159.html
> Sent from the Rosegarden - User mailing list archive at
> Nabble.com.
>
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal
> Reports 2008 30-Day
> trial. Simplify your report design, integration and
> deployment - and focus on
> what you do best, core application coding. Discover what's
> new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Rosegarden-user mailing list
> Rosegarden-user@...
> - use the link below to unsubscribe
> https://lists.sourceforge.net/lists/listinfo/rosegarden-user
>

     

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Rosegarden-user mailing list
Rosegarden-user@... - use the link below to unsubscribe
https://lists.sourceforge.net/lists/listinfo/rosegarden-user

[Julie S]

Hello Adrjork,

I took a look at the tutorial you have created...very neat.

I don't do much with video myself, but interfacing a midi device to alter video on the fly is pretty cool.

I wanted to mention that you are sending 3200 events a second (128 * 25)  and some of these events are multi byte so that potentially 9600 individual bytes sent each second.

A traditional MIDI connection would choke on this, but not a high speed x100 connection.  As you discovered, on the mac they were able to handle this.

So the first step is to see if ALSA can handle this...which I believe was mentioned.

The second thing to consider is RG internal resolution.  Maybe Chris C or Michael knows the details better.  Does RG record 3200 events a second.  I thought our resolution was 9600 events a second or something like that.

If it is at 9600 then there may be some slight jaggedness in the streams if the two are not perfectly in sync.  Hmmm...

Some further input from others that are closer to the code may be able to answer this better for you.

I hope this helps.

Sincerely,
Julie S.



     

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Rosegarden-user mailing list
Rosegarden-user@... - use the link below to unsubscribe
https://lists.sourceforge.net/lists/listinfo/rosegarden-user

[adrjork]

>I don't think there's any MIDI reproduction engine on the planet that could
>actually play that stream of data at full speed anyway.

Well, Michael, I can only say that Logic can do it, without problems. I must say also that, instead, Ableton Live can't do it without crashing! So it seems to me that only few softwares can handle this MIDI flow! Rosegarden doesn't crash during my test. So I thought that the problem was an incorrect setting of some parameter... I don't know.
I use a Linux audio-dedicated patform for my test: Apodio (Ubuntu-Hardy i386) on Macbook-pro/Santarosa. I can try also with Muse (I don't know KMidiMon...) But...I "feel" that the problem could be just a correct setting in ALSA...

[adrjork]

Hi Julie, thanks a lot.
1st question: I'm a beginner, as I said, so...never heard about x100 connection! Could you explain me something?
2nd question: what is RG resolution?
3rd: how can I sync the 2 streams?
Thanks.

[Julie S]

Dear Ardjork:

Try this first in RG before reading the rest.  If this doesn't work for you, then read the rest to see if anything helps.

Try First:
==========
An issue you are facing in RG is that your MIDI data is being forced onto a single channel if you are trying to record it on a single track.  RG will allow this, but playback is only through the device selected for the track, it does not re-split the data to the other channels.  This is a RG fundamental design constraint.

You must arm and record to 16 tracks in your scenerio, and make certain each track is only listening and record its channel data.

Another thing to consider is that a NOTE ON message of 0 velocity is a note OFF to most midi devices.  So this could cause some trouble for music recording applications.

Rethought on process:
=====================
I wrote the section below first.  Since then I reread your .pdf about the project and see that you only transmit pixel changes and do not try to resend every note each second.  So any references below would only apply to maximum throughput.

Answer the Questions:
=====================
You wrote:
> 1st question: I'm a beginner, as I said, so...never heard about x100
> connection! Could you explain me something?

Here is my long answer for you.

OLD MIDI
--------
Well here is a old school MIDI device using a DIN connector.
http://www.tigoe.net/pcomp/code/serial-communication/midi

In the Midi communications section of the text, a few paragraphs in it states:
> MIDI is a serial communications protocol, operating at 31,250 bits per
> second. Each byte has 8 bits, plus a start bit and a stop bit.
> It operates at 5 volts DC.

There is only one way communication so two connections are needed for two way communication.  THere are not hadnshaking or acknowledments.

With a little math we can see that there are 10 bits to each MIDI code 8 of which are the actual message
so we get 3125 messgages each second, and a note on event will have three bytes a status byte, an note number byte and a velocity byte
3125 / 3 approx. 1041 note on events each second.  There are optimizations for streams of note on events on the same channel that are 2 byes each, but this is not the case for you.

You are trying to send 128*125 = 12000 messages down the pipe -- well over the maximum for MIDI 1 spec of 1041 note on messages a second.

USB & Firewire MIDI
-------------------
I remember MIDI Manufcaturing Association specifying a x10 and x100 communication spec. for USB, but I'm not finding it.

http://www.midi.org/aboutmidi/tut_midicables.php

the site about mentions that the communication as device dependent, but they did propose a standard for firewire.  Here is a link to a PDF--look at page 6 and 7:
http://www.midi.org/techspecs/rp27v10spec(1394).pdf

My take on it this that MIDI X1 (standrd) sends 8 messages at maximum speed of every 320 microseconds this translate to 3125 messages a second amazing just like the old MIDI USB spec above.

There is a X2 speed and a X3 speed defined.  But the bit loophole is this phrase for each:
  or faster if a higher Data Channel rate is determined using the negotiation procedure
 
So the communication bay be much much faster depending on device negotiations.  But looks like they are implying multiples of the base rate of 3125 messages a second.


You wrote:
> 2nd question: what is RG resolution?

RG Resolution
-------------
I was hoping Chris C or another more knowledgeable person would provided the definitive answer about our timer / event recording resolution.  I thought it was set to 9600 events a second (a note on event is one event I believe).

You wrote:
> 3rd: how can I sync the 2 streams?

I see you would like to syn your motorized fader on one bus with the data from the other bus.

RG will do that, but you must arm 16 tracks (each recording one channel) dedicated to the faders channels.

Then arm each channel video MIDI channel (upto 16 tracks) each accepting input from one channel only (like for the faders).

In the RG paradigm and using your current setup, you would record everything as piece of music (make certain you set a tempo and select a good time signature.  I would suggest a tempo of 60 bmp (1 beat per second) this would capture 25 frames a second, and a time signature of 1/4. Therefore each measure will be 1 second of data.

>From there you can record all at one time, or record on separate takes.  These will be segments on the main canvas.  You can shift these around on the canvas to you timing likings and split them easily in one second intervals.  RG will easily allow breaks in 1 / 2^n intervals and some others as well.  Getting exact 1/10 or 1/100 splits might require more computations and a little luck.

Other Thoughts:
You might get more bang for you buck if you used system exclusive messages.  You could pack each system exclusive message with the with 2 bytes one for the note and one for the velocity.  That way you can save a lot of overhead than using the note ON message method.  The trick is that some systems have a byte limit per bulk transaction.

Y0u could beak the data up into 2 messages in various ways, maybe the first is the pixels affected, then the next are the values for the pixels.  That way you still keep your minimized data flow scheme in place and avoid transfer length limits.

Basically you just start with the system exclusive header pack the velocity and note number as bytes in sequence then end with the end of exclusive message byte.

You can send these out 25 times a second.  Most applications will record these, and as extra timing security you could preface this with a Note ON message of at say Middle C (60 decimal), at velocity 127, and then send the system Exclusive message, then a NOTE ON for middle C at velocity 0.

This will allow easy editing in the matrix editor since you can select the quarter notes (with some catches) and allow some shifting of things as well.

This is still much more compact than your method, and should not exceed system exclusive message bulk limits and still work with many firewire MIDI x2 and over transmissions.

This I would assume will work with RG if indeed the resolution its event resolution is 9600.

If you choose to transmit over multiple channels say a 4x4 grid of video (16 channels) you may start exceeding RG and possible MIDI (but this is device dependent).

I hope this helps.

Sincerely,
Julie S.



     

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Rosegarden-user mailing list
Rosegarden-user@... - use the link below to unsubscribe
https://lists.sourceforge.net/lists/listinfo/rosegarden-user

[Julie S]

Hello Ardjork:

I wrote a solution concerning system exclusive messages.

This will not work, since it can not be assigned to specific channel.  System exclusive messages are read on all channels.

Also after some more thought it is not any more compact, since your initial grid of 128 x 128 only used one channel.  There is no savings using the system exclusive messages.

Sorry about that.

Sincerely,
Julie S.



     

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Rosegarden-user mailing list
Rosegarden-user@... - use the link below to unsubscribe
https://lists.sourceforge.net/lists/listinfo/rosegarden-user

[adrjork]

Julie, your answers are really detailed, and you are really kind to help me!
Thanks a lot, really.
I made a test with rosegarden, but next days (when I'll feel better - I'm ill...) simply I'll try with some other misi receiver software (like Muse).
I'm very curious to try with frinika because it's free and it's cross platform (while, all other midi software I've found out are all dedicated to a single platform)
With frinika I could make a test in Mac and in Linux...