[Milkymist-devel] [Democompo] Altparty

[Bengt Sjölén]

mod player with some fixes is now committed in both the old dev attic  
subdirectory and in the demo src.

fixes are for example:
* that arpeggio is now independent modulation of frequecy removing  
some previous problems with arpeggio changing that actual note that  
would result in a continously climbin arpeggio
* set sample position is now jumping to the right offset which fixes  
the broken rythm of the demo mod.
* linear interpolation had the wrong order of the samples resulting in  
a reverse and broken sawtooth-like interpolation
* a bunch of new interpolations added, should perhaps be software  
selectable or at least compile-time selectable by defines but is now  
only a bunch of #if 1 to enable and #if 0 to disable. they are:
	floor - no interpolation - crispy sound with lots of aliasing
	linear - that just makes a linear ramp between sample values - fast
	cosine - which is not ready for use since it actually calls cos but  
should do this with fixed point arithmetics
	 	and perhaps a lookup table - sounds quite good though and will be  
fast
	cubic interpolation - which is broken
	cubic 2 interpolation - which is a working 4-point cubic  
interpolation - the currently enabled one - uses
		64-bit ints now just because I haven't bothered yet about fitting  
the fixed points to smallest ints.
	hermite - which is not ready for use since it uses float, also it has  
bias and tension configuration values
		that are just at some basic defaults which it any research into  
would give the best result

about declicking there is declicking for volume changes but not for  
ending samples in the peak of a waveform or so. when a new sound is  
started with an attack it's normally not audible if there is a click  
of breaking the old playing sample on that channel since it is in the  
attack of the new sound and in rythm as well. ending a sample on a  
peak is not at all forgiving though and one way to remove most of this  
is to make sure that the waveform ends at 0 instead of anywhere else  
in the waveform.
we should probably add some form of declicking in these situations too  
but it is not all obvious which is the best option because you don't  
want to start a new sound with a ramping up however short, since you  
might kill the attack of that sound, but you might want to ramp down  
when stopping a sample...

/b



On 29 sep 2009, at 21.55, Sébastien Bourdeauducq wrote:

> Hi,
>
> I doubt we will be able to make it to MAIN#4, so I suggest we try  
> our chance
> for Altparty (www.altparty.org) at the end of October in Helsinki. I  
> will be
> able to get there (MAIN is way too much transportation/schedule  
> hassle for me)
> and the program looks amazing. Of course, if we still have something  
> to show
> for MAIN, that would still be very cool :)
>
> Bengt's ML401 board is now operational, and he will be joining us  
> for the
> graphics programming effort. The outline of what should be done is  
> available
> at:
> http://www.tmplab.org/wiki/index.php/DemoParty_Compo_with_MilkyMist
>
> If you have any problem compiling the Mico32 toolchain for Mac OS,  
> just let us
> know on this list and we will help you.
>
> Also, please commit your fixes to the MOD player so that we can test  
> with the
> latest song from Alban and get rid of the last bugs fast.
>
> Thanks!
>
> Sébastien