Re: High latency in PtII javasound library
Hi JS, Edward created ptolemy/actor/lib/javasound/ClipPlayer.java, which plays a sound file. This actor is in the CVS tree, it is not yet released. The DE demo RealTimeComposite uses ClipPlayer. See http://chess.eecs.berkeley.edu/ptexternal for CVS access. However, it sounds like you want to play audio data directly, not files. Ptolemy II does includes an interface to the Java Media Framework (JMF), see ptolemy.actor.lib.jmf. There is an AudioPlayer actor in the jmf file. The C codegen interface also has an interface to SDL, so that can be used to generate sound. You could also use the EmbeddedC actor to call the C based SDL from Java. One thing to look at with AudioPlayer is how large chunks of audio data are that you are passing around. Try changing the transferSize of AudioPlayer and see how that works for you. _Christopher Hi, I'm trying to build an application with PtII that can: 1- read audio files 2- extract raw information (ie. amplitude) to control something else (like, say, the intensity of a light or the saturation of a video) 3- play the file I can do (1) and (2) pretty easily with the AudioReader actor BUT if I plug it into the AudioPlayer the sound output is really poor. Even with this very simple app (AudioReader - AudioPlayer) I can hear clicks indicating a poor latency (and unacceptable for a media application). Christopher pointed out a long time ago that the C codegen facility had an interface to Simple DirectMedia Layer. Could that be a solution? What about JMF? -- J. S. Senécal http://drone.ws --- - Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED] Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED]
Re: High latency in PtII javasound library
Hi Christopher, Thanks for the quick answer. Yes, as you say, I want to be able to do at least simple sound filtering (ie. have the sound pass through a set of filters and getting a good output). I have tried changing the transfer size but it doesn't seem to help at all. From earlier tests I had found out that even for a very straightforward script using Javasound, the required buffer size is pretty high (2048 on a Mac iBook G4) (see https://lethe.koumbit.net/trac/drone/wiki/DevNotes/SwitchingToJava#Lowlatencyissues). We were able to achieve a much more acceptable performance (buffer size = 64) using the JASS library (see http://www.cs.ubc.ca/~kvdoel/jass/). I think this is the key, so I'll try to reimplement using JASS instead. ++ js Christopher Brooks wrote: Hi JS, Edward created ptolemy/actor/lib/javasound/ClipPlayer.java, which plays a sound file. This actor is in the CVS tree, it is not yet released. The DE demo RealTimeComposite uses ClipPlayer. See http://chess.eecs.berkeley.edu/ptexternal for CVS access. However, it sounds like you want to play audio data directly, not files. Ptolemy II does includes an interface to the Java Media Framework (JMF), see ptolemy.actor.lib.jmf. There is an AudioPlayer actor in the jmf file. The C codegen interface also has an interface to SDL, so that can be used to generate sound. You could also use the EmbeddedC actor to call the C based SDL from Java. One thing to look at with AudioPlayer is how large chunks of audio data are that you are passing around. Try changing the transferSize of AudioPlayer and see how that works for you. _Christopher Hi, I'm trying to build an application with PtII that can: 1- read audio files 2- extract raw information (ie. amplitude) to control something else (like, say, the intensity of a light or the saturation of a video) 3- play the file I can do (1) and (2) pretty easily with the AudioReader actor BUT if I plug it into the AudioPlayer the sound output is really poor. Even with this very simple app (AudioReader - AudioPlayer) I can hear clicks indicating a poor latency (and unacceptable for a media application). Christopher pointed out a long time ago that the C codegen facility had an interface to Simple DirectMedia Layer. Could that be a solution? What about JMF? -- J. S. Senécal http://drone.ws --- - Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED] Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED] -- J. S. Senécal http://drone.ws Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED]
Re: High latency in PtII javasound library
Hi, I just ran through my emails and there are actually a few issues with JASS. First, it does not any longer uses the original native library that they developed... it uses Javasound. So it just won't help. Also, they have a strange license, not FreeBSD, not GPL, so, doesn't look that good after all. I'm not sure what you mean by C code generation. Are you referring to building a native library? Thanks, ++ js Edward A. Lee wrote: Yes, I've been frustrated by this as well... Unfortunately, Christopher's suggestion of changing the transferSize of AudioPlayer hasn't ever solved the problem for me... JASS looks promising... Please let us know if you make progress here. However, I suspect that to get really reliable performance, C code generation is probably the only really effective way to do this effectively... Edward At 09:51 AM 2/2/2008, JS wrote: Hi, I'm trying to build an application with PtII that can: 1- read audio files 2- extract raw information (ie. amplitude) to control something else (like, say, the intensity of a light or the saturation of a video) 3- play the file I can do (1) and (2) pretty easily with the AudioReader actor BUT if I plug it into the AudioPlayer the sound output is really poor. Even with this very simple app (AudioReader - AudioPlayer) I can hear clicks indicating a poor latency (and unacceptable for a media application). Christopher pointed out a long time ago that the C codegen facility had an interface to Simple DirectMedia Layer. Could that be a solution? What about JMF? -- J. S. Senécal http://drone.ws Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED] Edward A. Lee Chair of EECS and Robert S. Pepper Distinguished Professor 231 Cory Hall, UC Berkeley, Berkeley, CA 94720-1770 phone: 510-642-0253, fax: 510-642-2845 [EMAIL PROTECTED], http://www.eecs.berkeley.edu/Faculty/Homepages/lee.html -- J. S. Senécal http://drone.ws Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED]
Re: High latency in PtII javasound library
Hi Christopher, I am pretty confident that the basic issue at hand here is latency in the sound playback. I have had the same problem in the past with captured audio in fact. Java Sound clearly sucks when it comes to audio playback/latency. I have read that this was fixed in Java 5, but apparently not on Mac OS X from what I have read. I have been successfully using the Portaudio library [1] in the past and I have just found out there is a java binding project on Google Code called JPAB [2]. I am pretty confident that this approach would work on any platform. As for Metronome, this is another set of issues that go beyond audio only, it's the issue of real-time applications. And the main problem for now is that there are no open-source implementations of the RTSJ available yet. [1] http://portaudio.com [2] http://code.google.com/p/jpab/ Christopher Brooks wrote: Hi JS, Ptolemy can generate C code in various ways: 1) Take a regular Java model, generate C code for the entire model. This is what we usually do with code generation. 2) Take a regular Java model, generate C code for only certain actors. This is uses Gang Zhou's EmbeddedC actor. 3) Take a regular Java model, generate C code for the model (like #1), but call some Java code from the C code. We do this for the Java Plotter. We could try calling the SDL C code directly. I think the problem with #3 is that we will face the same issues that javasound faces. It might be interesting to get the Java source and see if we can see how the javasound interface works. Perhaps there is something in the javasound implementation that is optimized for large files. Another idea is to profile the model and see where the time is being spent. Looking at various real-time java implementations might be of interest. You might also look at David Bacon's work at IBM on Harmonicon: http://domino.research.ibm.com/comm/research_projects.nsf/pages/metronome.harmonicon.html Harmonicon uses Metronome, which is a real time GC: http://domino.research.ibm.com/comm/research_projects.nsf/pages/metronome.index.html We intend to do some work in the area of audio, as it is very easy for a human to tell when something is wrong. _Christopher Hi, I just ran through my emails and there are actually a few issues with JASS. First, it does not any longer uses the original native library that they developed... it uses Javasound. So it just won't help. Also, they have a strange license, not FreeBSD, not GPL, so, doesn't look that good after all. I'm not sure what you mean by C code generation. Are you referring to building a native library? Thanks, ++ js Edward A. Lee wrote: Yes, I've been frustrated by this as well... Unfortunately, Christopher's suggestion of changing the transferSize of AudioPlayer hasn't ever solved the problem for me... JASS looks promising... Please let us know if you make progress here. However, I suspect that to get really reliable performance, C code generation is probably the only really effective way to do this effectively... Edward At 09:51 AM 2/2/2008, JS wrote: Hi, I'm trying to build an application with PtII that can: 1- read audio files 2- extract raw information (ie. amplitude) to control something else (li ke, say, the intensity of a light or the saturation of a video) 3- play the file I can do (1) and (2) pretty easily with the AudioReader actor BUT if I p lug it into the AudioPlayer the sound output is really poor. Even with this very simple app (AudioReader - AudioPlayer) I can hear clicks indicating a poor latency (and unacceptable for a media application). Christopher pointed out a long time ago that the C codegen facility had an interface to Simple DirectMedia Layer. Could that be a solution? What abo ut JMF? -- J. S. Senécal http://drone.ws Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED] Edward A. Lee Chair of EECS and Robert S. Pepper Distinguished Professor 231 Cory Hall, UC Berkeley, Berkeley, CA 94720-1770 phone: 510-642-0253, fax: 510-642-2845 [EMAIL PROTECTED], http://www.eecs.berkeley.edu/Faculty/Homepages/lee .html -- J. S. Senécal http://drone.ws --- - Posted to the ptolemy-hackers mailing list. Please send administrative mail for this list to: [EMAIL PROTECTED] -- J. S. Senécal http://drone.ws
