[timidity-talk] extended GUS patches

[Greg L. <gr...@li...>]

I'm sorry for hijacking Mark Constable's thread on Best Quality
settings.  The subject header line I gave earlier was a mistake.
Also, I find that I failed to direct a reply to TAMUKI Shoichi
to the list.  So, in case others are interested, here it is:

TAMUKI Shoichi wrote:

 > Please let me know why the velocity and stereo layer information
 > are stored in patch header.  In my opinion, they are reasonable
 > to store in sample header.
 >
 > For example, there is a certain extended GUS/patch structure:
 >
 > [Patch header information (18 samples)]
 > [Sample for low freq. range, low vel. range, left]
 > [Sample for low freq. range, low vel. range, right]

...

 >
 > MIDI players that know only about ordinary GUS/patches will choose
 > either among the samples of each frequency range.


Hi.

I don't exactly follow what you're suggesting.  If a midi player
knows only about ordinary GUS patches, how can it read extra
information in the sample headers in order to figure out which
samples it needs to load?  It won't know about that extra
information, because it knows only about ordinary GUS patches.

I've arranged things so that the first velocity layer following
the patch header will look just like the single set of keyrange
samples that a naive midi player expects to see and knows how
to deal with.  I also made sure that the utility unsf, which
breaks up soundfonts into extended GUS patches, orders the
the velocity layers with the widest velocity range first.  That
way, a naive midi player which can read only one layer will
get the best one available.

Greg