Re: [timidity-talk] Awful Crackling Noises at ends of sounds

[Robert M. J. <bob...@be...>]

Eric A. Welsh wrote:

>>Crackling shouldn't be related to processor speed.  When timidity runs
>>out of CPU, at least under Windows, it stutters instead of crackles.  I
>>used to run a Pentium 133, and it wasn't difficult to use up all the CPU
>>if reverb and/or chorus was enabled.  The result was always stuttering,
>>as the sound paused until the audio buffer was full again.
>>    
>>
>
>What I said before wasn't entirely correct.  While running _out_ of CPU
>only stutters, running _low_ on CPU can produce some crackling noises.
>This generally only happens when the user is using an interpolation mode
>that is too slow for that computer.  If you run timidity in the ncurses
>display mode, it will display the current % "audio queue".  If this number
>drops below around 85% or so, timidity will start to stutter, because the
>audio buffer isn't full by the time it is needed to play it.  To avoid
>stuttering, timidity does some things to reduce CPU usage once the buffer
>drops below 100%.  Below 100%, it will switch from whatever the current
>interpolation method is to using linear interpolation, then switch back to
>the original interpolation method once the buffer is at 100% again.  On
>modern processors, this usually looks like the buffer stays at a constant
>100%, since the interpolation method is constantly switching back and
>forth between linear (too fast) and the other mode (too slow).  Because of
>difference in the output from the interpolation routines, the sample
>amplitudes before and after the switchover may not be smoothly continuous,
>and can cause a crackling noise.  I've heard this myself, so I know it
>happens.  If the CPU is _really_ getting hammered, and switching over to
>linear interpolation isn't helping, then timidity starts killing voices.
>Once the buffer gets to around 85% (the exact criteria is a bit different,
>but it usually winds up being around when the 85% mark is reached),
>timidity starts killing notes.  First it kills notes that are OFF, but
>haven't finished releasing yet, starting with the most quiet volumes.
>If it still hasn't reached the estimated "safe" number of voices
>(calculated based on prior observations on how many voices are "good" or
>"bad" for buffer depletion), it starts killing of voices that are still
>ON.  Abruptly killing off notes without decaying them first can also
>result in a popping noise, since the sample amplitude abruptly drops from
>something large to zero.
>
>So, if the % audio queue is hovering at 100%, crackle may be generated due
>to choosing too expensive an interpolation method.  Try using a smaller -N
>value (the default Gauss-like interpolation defaults to -N 25, which is
>probably too slow for some older computers).  If -N 5 still isn't getting
>the % audio queue above 100%, use a different resampling mode all
>together (-EFresamp=L for 4 fast point interpolation).  If that STILL
>isn't helping, use -EFresamp=l for linear interpolation.  If you're still
>in trouble, which could happen on Pentium speed machines, lower the
>sampling rate with -r, change the chorus mode to -EFchorus=1 or disable it
>entirely with -EFchorus=0, set -EFreverb=g or =G (global versus per
>channel), or disable reverb entirely.  By turning off enough special
>effects, and lowering the sound quality options, most machines can play
>any midi just fine.  Even my old Pentium 133.
>
>-Eric
>
>
>
>-------------------------------------------------------
>This SF.Net email is sponsored by: InterSystems CACHE
>FREE OODBMS DOWNLOAD - A multidimensional database that combines
>robust object and relational technologies, making it a perfect match
>for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
>_______________________________________________
>Timidity-talk mailing list
>Tim...@li...
>https://lists.sourceforge.net/lists/listinfo/timidity-talk
>
>  
>
That's what figured. I always make my waves without extras stuff.
Even on this one.
my batch file looks like this:

timidity -int --modules=32 -Ow *.mid *.MID
I'm in a case sensitive enviroment.
Especially after I put Windows Services for Unix and Cygwin.
Bobby