How to create SFX for Moonsound's FM registers ?

My PSG sfx loop is just this: (Artrag's psg sfx replayer)

;out buffered sfx data 
  xor   a				          ; --- FIXES BITS 6 AND 7 OF MIXER ---
  ld  	hl,AYREGS+Mixer
  set	  7,(hl)
  res	  6,(hl)
  ld    c,$a0
	ld    hl,AYREGS
.sfxloop:
  out   (c),a             ;first write to psg register ??
  inc   c
  outi                    ;then write volume value ? not sure
  dec   c
  inc   a
  cp    13
  jr    nz,.sfxloop
;/out buffered sfx data 

So we write to psg register $a0 after which we write the value to $a1.
That's all.
Any idea how to re-route this to FM registers ?

More on topic, I suppose you could just VGM log a "song" playing real fast (60 or 50 steps per seconds). That way you can even have multi channel sound effect (detune & delays etc.) which I imagine can give that extra kick you are looking for... especially as it will be played in contrast to the epic 24 channel wave music

Or even with 18 fm channels at you disposal, someone like ARTRTAG could figure out how to make some cool ISR samples by adding up a bunch of sine waves, I am sure of it

Nope, 640kb Moonsound-RAM and 2op ought to be enough for everyone.

Ok, the reason is this code when it starts playing the music:

    ld  a,(speed)   ; 3A speed
    sub 2           ; D6 02
    ld  (spdcnt),a  ; 32 spdcnt

That makes it overflow and set a wrong value in spdcnt when tempo is 24. Speed is calculated as 25 - tempo and thus speed is 1 when you set tempo 24. This makes spdcnt 255 instead of 0 which makes it not work. Change the code to sub 1 and it should work.

So, use a hex editor on the MoonBlaster 1.4 disk image and do the following replacements to make it support tempo 24 (1 step / interrupt):

3E 18 32    ->    3E 19 32
D6 02 32    ->    D6 01 32

Now you should be able to edit FM sound effects in MoonBlaster 1.4 .

For playing them back I wouldn’t use the MoonBlaster 1.4 replayer though (with I/O ports modified to play on OPL4), it seems much more CPU efficient to record the sound effects as VGM and then process that down into a short register data stream to include in your code with a small playback loop.

From what I understood from Bengalack’s post about that in another thread, I think that was more of a problem of the conversion script initially having trouble dealing with timing variation of the I/O (some of which is bound to be there). That’s something I’ve had to deal with in different VGM conversion contexts as well.

It should not be a problem unless MB’s processing for the playback takes longer than 1/60th of a second, which is hopefully not the case. So I think it’s probably good enough. Anyway it’s something that can be dealt with by massaging the data after the effect is created.