From: Jeff W. <jww...@ya...> - 2005-03-07 17:41:03
|
--- Joe Emenaker <jo...@em...> wrote: > Jeff, how were you testing for errors? We you using > the loopback test, > or something else? If the loopback test can catch > these signaling > errors, then it could be integrated into an > "auto-tuning" wizard which > would keep increasing the buffer sizes and delay > times until the > loopback test succeeds. I didn't try the loopback test because I'm lazy and didn't want to reroute cables. I just noticed the frequency of signal errors increased dramatically when I started sending FCB1010 patches a couple weeks ago. Since that device has a 6010 byte patch, it led me to suspect there might be a correlation between the size of the patch and the frequency of signal errors. That's when I tweaked the buffer size and added the timeout on my local copy of MidiUtil.java. Since then, I've been running with a 256 byte buffer and a 100 msec timeout and while it hasn't eliminated the signal errors entirely, it has dramatically cut down on the frequency. I just downloaded the most recent change (with the buffer and timeout settings in the preferences) and I'm going to try a 128 bytes/100 msec for a while and see how that goes. Using the loopback test for detecting signal errors might be worth a try. But I think you would probably want to have it send a longer message. Right now it only sends 11 bytes. Even sending multiple short messages might skew the results because woudn't be much different than setting the buffer size to 11 bytes and sending a long message. __________________________________ Celebrate Yahoo!'s 10th Birthday! Yahoo! Netrospective: 100 Moments of the Web http://birthday.yahoo.com/netrospective/ |