Re: [hatari-devel] Hatari plays all voices in phase

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 28/12/2015 15:56, Troed Sångberg a écrit :
> I think you should only compare with target hardware, not SC68 :) I only
> pointed out that I know Ben tries to emulate the in/out-phase thing,
> which Hatari doesn't. In my tests the conclusion was:
>
> 1) Target hardware will "randomly" be in and out of phase when playing
> the voices
> 2) SC68 will sometimes not be in phase
> 3) Hatari will always be in phase
>
> For the, I think, separate issue with the hi-hats, SC68 is indeed closer
> to the target volume of the hi-hats than Hatari. I've right now listened
> through the recording by evil at
> https://www.youtube.com/watch?v=sBufLphwVlg and compared to the original
> sndh in VLC (with the sc68 plugin) and my own .wav recording from hardware.
>
> 7an, who is a musician, doesn't consider sc68 to be _correct_ either -
> but I as a non-musician will say that at least the perceived volume of
> the hi-hats seems close to my target recording :)

I have some ideas about the cause for the phase difference ; it's 
something that I tried to find documentation from yamaha but was never 
able to find.

Basically, it comes to :

 - how does the ym2149 handle the "period" counter to do a square wave 
: does it increments a counter until it reaches "period" value and then 
toggles between 0 and 1 to give the square wave, or does it decrements 
the "period" counter until it reaches "0", then toggle between 0 and 1 
(most old chips will often decrement and compare with 0 as it requires 
less logic gates)

 - what happens when a new value is written in "period" ? Does it 
restart the current wave immediately ? Does it becomes the new value to 
reach in case we use an increment, or is it latched until the end of 
current phase ?

This could be measured with a logic analyser, but unfortunately I don't 
have one :(

I saw many projects emulating the ym2149 (soft or fpga) and claiming 
"perfect emulation", but in the end, no one has a clear answer to this, 
it's often based on how other YM chip worked, but experience proved that 
not all YM chips had the same inner work.

I will try to change the way Hatari latches/restarts the new value in 
"period" to use a different method, hopefully it will give a better 
sound and an empirical proof of the correct working.

Nicolas