MiSTer FPGA Forum

Good Afternoon,
I've come across an issue were the audio is not playing correctly when configured to be a A1200. WHDLoad games the audio works fine. Its when I boot a game from a floppy (ADF image) the issue occurs. The best way to describe the issue is it sounds like one or two channels are not being out putted to the speaker. But it does play in stereo which is odd. I did think it was my DC-10 nano. I have a spare so I hooked that up and did the same. If I use the A500 configuration then the audio is fine. Audio plays fine on real A1200 hardware. Games I have had issues with are "ARCHER MACLEAN'S POOL", "Jimmy White Whirlwing Snooker" "Lotus 2". Those are the only ones I have played from floppy as I don't have a lot of ADF files yet(Stuck on a faulty Windows PC). I could use the A500 config but I have found that the A1200 config runs the snooker and pool games more smoothly.

If this is a known issue then forgive this post. I did do a search before posting but could not find any thing.

DCACHE on/off doesn't matter?

Hi there,

That's odd - I just posted something similar here (I unfortunately didn't see this post first):
https://misterfpga.org/viewtopic.php?f=4&t=2194

I noticed in 68020 (A1200) mode, some games on floppy have illegible sound. I will try the games you mentioned.

My post was mainly about slightly glitching on WHDload games though (not many). I know you say WHDload games work fine, and by and large they do, but if you look at my post, if you are running the MegaAGS image (or even just loading WHDload games from Workbench), as long as your config is 68020/AGA/KS3.1 (basically an A1200), if you try and run Hostages and/or Shufflepuck Cafe for example, you'll notice the title screen music glitches and the channels momentarily drop sound with regularity. Most games are fine, but for some reason the sound is a bit messed up even when running the WHDload version of the game, if the config is set to 68020/AGA.

Of course, if I change to an A500 configuration and run the floppy ADF, things are fine.

I did a search too and really didn't come up with anything. I wondered whether there is something related to the 68020 implementation, but wanted the thoughts of people on the forum before considering creating an issue for it. It might just be the way that it works or a limitation of sorts.

Good Afternoon Moviesman,

I've done a quick test of Hostages using the "MegaAGS" image and my own WHDLoad install and also booting from floppy with A500 options and I cant find a difference between any of them. I've never played Hostages but it sounds fine to me.

rcade wrote: ↑Sat Mar 06, 2021 4:56 pm DCACHE on/off doesn't matter?

DCACHE on causes the issue if I turn it off then the audio works fine. Thanks, I guess this post can be closed.

Hi,

The issue is there, but if not familiar with the game(s) it is potentially hard to notice. I did some tests on WinUAE and the issue is there too when using 68020 as the processor - however if you enable 'cycle exact' it goes away. As far as I know 68020 implementation in MiSTer is not cycle exact, which explains why the issue exists.

In the other thread I linked above, there is another explanation from robinsond5 about the CPU running too fast for those particular games, with an explanation for how the routines work, which makes sense.

Also, I saw a comment on discord today from Optiroc, stating "AGA/020 and HDD loading introduces a lot of non-accurate behavior..... Most things work but far from everything works without glitches here and there."

I think unless the Minimig 68020 implementation is made cycle-exact, there will always be oddities, glitches and incompatibilities here and there because of the CPU speed being much higher than a regular A1200.

Of course, 68000/OCS/512k chip+512k slow/kickstart v1.3/PAL configuration should have no problems at all on non-AGA releases, but you need to use floppies.

movisman wrote: ↑Sun Mar 07, 2021 3:53 pm Hi,

The issue is there, but if not familiar with the game(s) it is potentially hard to notice. I did some tests on WinUAE and the issue is there too when using 68020 as the processor - however if you enable 'cycle exact' it goes away. As far as I know 68020 implementation in MiSTer is not cycle exact, which explains why the issue exists.

[...]

As far as I know, 68020 on WinUAE is not cycle exact either. Toni Willem explained it some time ago and basically it can't be achieved because of the lack of information. The Motorola 68020 has not been documented at the same level that the 68000, unfortunately, so until that happens we will have to wait.

Maybe the software implementation on WinUAE is more advance that the one we find on Minimig, just a thought.

It shouldn't need to be completely cycle exact - but I think at the moment for artificially tiny loops which run completely from cache, like the timing loop used in some music players, it's much faster than a real '020.