[hatari-devel] Chaos AD problem : $fffffc02

[Laurent Sallafranque]

I've checked the DSP part. Nothing special at first glance.

I'll check it more.

I think many sound problems may come from the fact that the real Falcon 
DMA sound has 2
32 bytes buffers.

I've already thought about coding them in the crossbar, but I need some 
infos

- when are these buffers filled by the Falcon ? (at VBL, every xxx 
cycles, ...)
- Is there a "burst mode" like in the videl DMA ?

I may try to read a few bytes every HBL, as Nicolas did for the STe and 
see if this increase the sound quality.
(after 1.5)

Regards

Laurent


Le 13/06/2011 22:45, Nicolas Pomarède a écrit :
> Le 13/06/2011 22:32, Eero Tamminen a écrit :
>> Hi,
>>
>> On maanantai 13 kesäkuu 2011, Laurent Sallafranque wrote:
>>> Thanks a lot Nicolas.
>>>
>>> Another demo that runs now ;)
>>>
>>> For me, it has freezed while showing the tridi DNT crew logo (near the
>>> end of the demo I guess).
>>
>> For me it works to the end (there are still several screens after the
>> blue-white tridi DNT logo), when using the old UAE core.
>
> In my case, it was also using the old UAE core (although I get one 
> freeze once during the kind of "fractal color cycling" tiled screen)
>
>>
>>> I agree, the sound is awfully noisy (one of the most noisy sounds I've
>>> heard under hatari).
>>
>> Credits at the end say it's DSP sound.  :-)
>
> Yes, and there're a lot of dsp traces if your turn them on
>
>>>>
>>>> Hopefully, the demo doesn't uses the 6301's code that was supposed to
>>>> decode the byte. The 6301 code is using some kinds of 'sync decoding',
>>>> by relying on the 6301's timer counter and the instructions cycle
>>>> length.
>>
>> At what frequency that timer is running, are you saying that it could
>> be using IKBD for synching the sound?
>
> No, the ikbd is used at the start to "decrypt" the demo (during the 
> floppy disk picture) ; it's a kind of protection for preventing other 
> people to rip your code.
> Once the demo starts, the ikbd is not used anymore (it would be 
> impossible to use it for any real time effect, as it takes approx 8000 
> 68000 cpu cycles to send 1 byte)
>
> I was refering to some "sync code" in the 6301 decoder, because it was 
> supposed to use the content of the 6301's free clock run counter to 
> decrypt the bytes, which means you can't run the code "by hand" at a 
> slower speed, you need to really run it on the ikbd if you want to get 
> proper decoding, which makes things harder to debug if you want to 
> crack the protection (this is similar to the protections that use the 
> content of $ff8209 to decode some part of a program).
>
>
>
> Nicolas
> _______________________________________________
> hatari-devel mailing list
> hatari-devel at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/hatari-devel
>