Matlab/Simulink and Audio/Midi

phoenix

Hello my friends,

Is anyone into analyzing or synthesizing MIDI/Audio signals using Matlab/Simulink realtime workshop (rtw)?

Or did anyone do it WITHOUT rtw? (As a simulation?)

How did you do it? Is it possible to access ASIO in realtime, has anyone built a Simulink block or written a matlab driver function which does this?

Greetings phoenix

jksuperstar · Posted: Mon Oct 30, 2006 7:42 pm Post subject:

Are you an RTW user?!!!

I think the cost of the system is just generally prohibitive for your typical audio user Smile

My last (cancelled) project was designing & developing an PCI/FPGA card for RTW, which I had working, and still have all the code & drivers I wrote Smile

phoenix · Posted: Tue Oct 31, 2006 8:53 am Post subject:

I use rtai- it is opensource Very Happy

https://www.rtai.org/
but less widely distributed.

jksuperstar · Posted: Tue Oct 31, 2006 12:38 pm Post subject:

I haven't used rtai before, it seems similar to Simulink, but I wonder how common the drivers are? I have no idea, but I've heard RTW is based on something other than windows, dos, or linux (I'm speaking of the target, not the host).

Anyway, on an RTW target, using a very fast system, I was able to get down to 33usec timing resolution, running little code. My aim was to run 40kHz servo loops, but obviously couldn't get there. Still, it was very useable. I wonder how close RTAI can get to that. I suppose it depends on how much of a kernel & linux system you are running.

jksuperstar · Posted: Tue Oct 31, 2006 12:43 pm Post subject:

Maybe this is a good idea for this other project:
http://electro-music.com/forum/viewtopic.php?highlight=electronix&t=12799

I haven't heard of any hard-real time linux audio systems out there.

phoenix · Posted: Wed Nov 01, 2006 10:03 am Post subject:

phoenix · Posted: Wed Nov 01, 2006 10:14 am Post subject:

jksuperstar · Posted: Wed Nov 01, 2006 5:06 pm Post subject:

I was speaking of compiling RTW in "stand alone" mode, where you boot up a PC with the special RTW floppy/CD, then a seperate PC running the simulink/rtw host connects & controls it. The other methods of running under windows, linux, etc, don't come close to the performance of stand-alone mode (as would be expected).

I did a fair amount of research, and the preemption and low-latency patches (now almost entirely part of the kernel source from 2.6.18 on) offer basically the same performance as RTAI does. Rather than bury linux proper on a layer "beneath" RTAI, the PREEMPT_RT patches modify the kernel to allow all the necessary spinlocks and preemption to hit a very low latency, real-time target. This allows one to use the same Linux API, rather than a seperate API like RTAI has.

I am already running this kernel, and can get below 2ms latency on a system that couldn't get below 12ms on Windows or standard Linux kernels. I think this is very good, and on a newer system (or PPC) would have even better IRQ response.

So, really, this type of effort may be a serious waste of time. Sad