MIDI Output Buffer Size and MIDI Output Delay preference (Re: [Jsynthlib-devel] Signal Errors on Mac OSX)

[Hiroo H. <hir...@co...>]

Hi Jeff,

Now you can specify MIDI output buffer size and MIDI output delay time
by MIDI preference dialog.  These have been on my TODO list and I've
finally implemented them.

Enjoy.

Jeff> Just a note Mac OSX signal errors. I tried Plumstone
Jeff> but was unable to get it working. However I did some
Jeff> experimentation with CAProvider yesterday and
Jeff> discovered a couple of modifications to the
Jeff> core.MidiUtil.java class that will at least cut down
Jeff> on the frequency of signal errors. First of all, I was
Jeff> having a lot of problems with the Behringer FCB1010.
Jeff> This device only has one type of patch with a length
Jeff> of 2352. Since it's a fairly large patch and since I
Jeff> wasn't getting as many signal errors as with the other
Jeff> devices, I suspected that the problem might be some
Jeff> kind of a timing issue.
Jeff> 
Jeff> First, I changed the value of BUFSIZE from 0 to 256
Jeff> which causes the MIDI data to be sent in multiple
Jeff> blocks of 256 bytes. Then I added a 100 millisecond
Jeff> delay at the end of the send(Receiver rcv, MidiMessage
Jeff> msg) method. The whole thing looks like this:
Jeff>     /**
Jeff>      * MIDI Output buffer size.  If set to '0', Whole
Jeff> Sysex data is
Jeff>      * sent in one packet.  Set '0' unless you have
Jeff> problem.
Jeff>      */
Jeff>     private static final int BUFSIZE = 256;
Jeff> 
Jeff>     /**
Jeff>      * Send a <code>MidiMessage</code>.  If BUFSIZE is
Jeff> non-zero, data
Jeff>      * size will be limited to the size.
Jeff>      */
Jeff>     public static void send(Receiver rcv, MidiMessage
Jeff> msg)
Jeff>         throws MidiUnavailableException,
Jeff> InvalidMidiDataException {
Jeff>             int size = msg.getLength();
Jeff>             
Jeff>             if (BUFSIZE == 0 || size <= BUFSIZE) {
Jeff>                 rcv.send(msg, -1);
Jeff>                 log("XMIT: ", msg);
Jeff>             } else {
Jeff>                 // divide large System Exclusive
Jeff> Message into multiple
Jeff>                 // small messages.
Jeff>                 byte[] sysex = msg.getMessage();
Jeff>                 byte[] tmpArray = new byte[BUFSIZE +
Jeff> 1];
Jeff>                 for (int i = 0; size > 0; i +=
Jeff> BUFSIZE, size -= BUFSIZE) {
Jeff>                     int s = Math.min(size, BUFSIZE);
Jeff>                     
Jeff>                     if (i == 0) {
Jeff>                         System.arraycopy(sysex, i,
Jeff> tmpArray, 0, s);
Jeff>                         ((SysexMessage)
Jeff> msg).setMessage(tmpArray, s);
Jeff>                     } else {
Jeff>                         tmpArray[0] = (byte)
Jeff> SysexMessage.SPECIAL_SYSTEM_EXCLUSIVE;
Jeff>                         System.arraycopy(sysex, i,
Jeff> tmpArray, 1, s);
Jeff>                         ((SysexMessage)
Jeff> msg).setMessage(tmpArray, ++s);
Jeff>                     }
Jeff>                     rcv.send(msg, -1);
Jeff>                     log("XMIT: ", msg);
Jeff>                     try {
Jeff>                         Thread.sleep (100);
Jeff>                     } catch (Exception e) {}
Jeff>                 }
Jeff>             }
Jeff>         }
Jeff> 
Jeff> I realize this is only a temporary workaround. While
Jeff> it doesn't totally eliminate signal errors I'm getting
Jeff> way fewer of them than I was getting.
Jeff> 
Jeff> I have only made these changes to my local copy and I
Jeff> don't plan on committing the changes, since they would
Jeff> only apply to Mac users and would only be a temporary
Jeff> workaround at best. But for any of you who are Mac
Jeff> users and don't want to use Plumstone, this might be
Jeff> another alternative.
-- 
Hiroo Hayashi