Simple sample player (no DMA)

[obiwanken]

Hi
I want to play samples using no DMA.
is it possible ?

For example (not working):

;---------------------------------------------------------------------
lea Sound(pc),a0
move.l #(SoundEnd-Sound)/2,d0
move.w #$0040,$dff0a8 ; Aud0vol
move.w #$0080,$dff0a6 ; Aud0Per
Loop
move.w (a0)+,$dff0aa ; Aud0dat

move.w #$0100,d1 ; Pause
Wait
dbra d1,Wait

sub.l #$0001,d0
bne Loop
rts

Sound
incbin "dh1:work/samples/crom_16khz.snd"
SoundEnd
;--------------------------------------------------------------------
What's wrong, any suggestions ?
Thanks.

[Toni Wilen]

You need to clear channel's interrupt request bit. Writing to AUDxDAT does nothing if interrupt request is set.

btw, much better method is to wait for channel's interrupt bit (poll INTREQR), clear it immediately, write next word to AUDxDAT and so on.

[Asman]

Originally Posted by obiwanken

What's wrong, any suggestions ?

Sound is in CHIP memory I guess. If not then use SECTION sfx,DATA_C.

[obiwanken]

Many thanks , now working !

Just one line added and result is awesome.

[Thorham]

Originally Posted by Asman

Sound is in CHIP memory I guess. If not then use SECTION sfx,DATA_C.

Audio data only has to be in chipmem if you're going to use audio DMA. For CPU playback, data can be anywhere

[obiwanken]

Yes audio data is in FastMem.
Thanks.

[leonard]

Originally Posted by obiwanken

Many thanks , now working !

Just one line added and result is awesome.

Hi!

I wonder why you need to play sample by polling and using CPU only. Is it only to "demonstrate" something, or is there a special interest? I'm curious about it.

[CFou!]

Originally Posted by leonard

Hi!

I wonder why you need to play sample by polling and using CPU only. Is it only to "demonstrate" something, or is there a special interest? I'm curious about it.

I'am interested too by the answer

[pandy71]

Originally Posted by leonard

Hi!

I wonder why you need to play sample by polling and using CPU only. Is it only to "demonstrate" something, or is there a special interest? I'm curious about it.

Higher sampling rates, override CHIP limitations, increase number of the independent audio channels etc.

[leonard]

Originally Posted by pandy71

Higher sampling rates, override CHIP limitations, increase number of the independent audio channels etc.

Oh I see, but you just write two 8 bits samples with a 16bit write, right? why don't you write this to a small buffer, played by PAULA at the highest freq? So you keep have your own CPU mixing routine.

I just realized you can use the same trick as with ATARI-ST digisound: you can use several PAULA voices to play a "more than 8bits" sample.

[CFou!]

Originally Posted by pandy71

Higher sampling rates, override CHIP limitations, increase number of the independent audio channels etc.

i had not thougth to that!!! very interesting

[pandy71]

Originally Posted by leonard

Oh I see, but you just write two 8 bits samples with a 16bit write, right? why don't you write this to a small buffer, played by PAULA at the highest freq? So you keep have your own CPU mixing routine.

I just realized you can use the same trick as with ATARI-ST digisound: you can use several PAULA voices to play a "more than 8bits" sample.

As Paula have limited amount DMA time slots there is no way to reproduce more than 4 voices in hardware unless you accept lower sampling speed (worse quality) - with CPU however time interleaving should be possible thus more "independent" channels.

Higher sampling rate means or ECS/AGA or tricks in OCS with horizontal line shortening or CPU (Copper) software driven AUDxDAT.

Higher resolution than 8 bit imply use 2 DMA channels + calibration (never perform any measurements what kind of accuracy can be achieved in this way - i assume something between 10 and 12 bits should be possible - with modern PC sound card used as a audio analyzer perhaps this calibration can be performed even better than previously and 12 - 13 bits sounds reasonable).

Still i believe that 8 bit with dithering and noiseshaping can provide higher quality than combined 2 Paula channels with pseudo 14 bit mode.

If CPU can be replaced by synchronous and cycle accurate Copper then most of the Paula (Agnus) limitations can be override.

[8bitbubsy]

Originally Posted by leonard

[...]
I just realized you can use the same trick as with ATARI-ST digisound: you can use several PAULA voices to play a "more than 8bits" sample.

How?

[obiwanken]

Originally Posted by leonard

Hi!

I wonder why you need to play sample by polling and using CPU only. Is it only to "demonstrate" something, or is there a special interest? I'm curious about it.

Yes, for "demostrate", I want to play "Africa by Toto" song (4650 kb lenght).

[pandy71]

Sorry for using this topic but can be closely related - i have strange idea how to play in non-DMA mode with DMA active.

General principle how Paula works is quite clear:

If period value is to low - last data stored in AUDxDAT register are repeated.

Is there any chance to set 1 word DMA in AUDxLEN, data for example $0000 word (or similar) in CHIP RAM location addressed by ADUxLCH/L, set period higher than maximum supported by normal DMA, then manually (CPU - Copper) write to ADUxDAT data words in regular time slots to increase sampling frequency?