solving the CD32 joypad "mystery"

[jotd]

Toni, in a previous message you answered to

what could explain that some games work OK with the joystick, but when you plug in a CD32 joypad, red button is not recognized? (ex: Moktar, and another one I don't remember)

by

I guess they have 3rd button line active forcing pad in CD32 pad mode.

I can't imagine they're doing that on purpose (this isn't even a 2-button compatible game). And how to reset that? (I have another issue in another slave, I figured that I could reuse Asman/others ReadJoyPad code bits to do:

	moveq	#CIAB_GAMEPORT1,d3	; red button ( port 1 )
	bset	d3,ciaddra(a1)	;set bit to out at ciapra
	bclr	d3,ciapra(a1)	;clr bit to in at ciapra

before reading the button and

	bclr	d3,ciaddra(a1)		;set bit to in at ciapra		
	or.b	#$C0,ciapra(a1)	;reset port direction

afterwards. Would that be OK? and the bugs reported by many users about this damn red/fire button not registering on CD32 don't happen in WinUAE. Is that normal? I'd very much like this "feature" implemented so I could debug a lot of those issues...

[robinsonb5]

Originally Posted by jotd

I can't imagine they're doing that on purpose (this isn't even a 2-button compatible game). And how to reset that?

Look for writes to POTGO ($dff034).
Normally to have the middle and right mouse buttons (or two extra joystick buttons on pins 5 and 9) readable on each port, you'd write $ff00 to POTGO - if bits 8, 9 (port 1), or 12 or 13 (port 2) are low, then pin 5 won't be pulled up to 5v so the CD32 pad will be in shift-register rather than direct mode. (Most games probably won't write anything to POTGO, assuming that the system's left it in a sane state. The troublesome ones are probably writing an ill-considered value.)

Incidentally, in your CD32 pad reading code, after reading the buttons I'd restore POTGO by writing $ff00 to it, not $ffff - the latter is asking the chipset to put the ports in digital i/o mode, set pins 5 and 9 high, and then dump the capacitors and start the analogue counters - which makes no sense with the pins in digital mode.

[Amiga1992]

Sorry to butt in, don't have much to he;p with :/ but I remember someone here mentioning that if you would rewrite the pad reading routines, you could natively support a 3 button Megadrive pad without modifications. Is that true?

If so, while you guys are looking at joypad reading code, it would be a great addition to have. 3+1 buttons would be excellent, and those are much cheaper and easier to find than CD32 pads

[robinsonb5]

Originally Posted by Akira

Sorry to butt in, don't have much to he;p with :/ but I remember someone here mentioning that if you would rewrite the pad reading routines, you could natively support a 3 button Megadrive pad without modifications. Is that true?

If so, while you guys are looking at joypad reading code, it would be a great addition to have. 3+1 buttons would be excellent, and those are much cheaper and easier to find than CD32 pads

You can read two buttons of a megadrive pad without modifications. Technically it will be drawing a small amount of power from a signal pin rather than the power pin to run its multiplexing electronics, but in practice that's harmless. You do need to swap pins 5 and 7 in order to have control over the button multiplexing.

[Daedalus]

I remember someone (possibly Toni) saying before that different versions of kickstart have different default settings for the POTGO registers, so if a game is written for one and assumes the POTGO contents are set the way it needs them, it may not work with other kickstart versions where this is not true. Such a difference won't show up when using a single button controller, so you might only see it if you try to use a CD32 pad (or Megadrive pad for that matter) with such game/kickstart combinations.

Edit: Yep, here it is: http://eab.abime.net/showpost.php?p=989662&postcount=26
So if games for a 1.3 machine don't explicitly write to POTGO, they might work with multi-button controllers under 1.3 but not 2.0+.

[robinsonb5]

Originally Posted by Daedalus

I remember someone (possibly Toni) saying before that different versions of kickstart have different default settings for the POTGO registers, so if a game is written for one and assumes the POTGO contents are set the way it needs them, it may not work with other kickstart versions where this is not true.

Good point - so the troublesome games may in fact be those that aren't writing to POTGO at all.

[Amiga1992]

Originally Posted by robinsonb5

You can read two buttons of a megadrive pad without modifications. Technically it will be drawing a small amount of power from a signal pin rather than the power pin to run its multiplexing electronics, but in practice that's harmless. You do need to swap pins 5 and 7 in order to have control over the button multiplexing.

So you still need to swap pins 5 and 7. If you do, can you access all buttons on a 3 button Sega pad?
You would need to write support for them, though, right?

A "dongle" swapping those pins should be easy to build.

[robinsonb5]

Originally Posted by Akira

So you still need to swap pins 5 and 7. If you do, can you access all buttons on a 3 button Sega pad?
You would need to write support for them, though, right?

A "dongle" swapping those pins should be easy to build.

Yup, the 3-button pads are really easy to support: pin 7 high = buttons B & C on pins 6 and 9. Pin 7 low = buttons A and Start on pins 6 & 9. So yeah, it still needs software support. (The six buttons pads can also be read by pulsing pin 7 with the correct timing - they multiplex the direction lines as well as the fire button lines.)

[Amiga1992]

Originally Posted by robinsonb5

Yup, the 3-button pads are really easy to support: pin 7 high = buttons B & C on pins 6 and 9. Pin 7 low = buttons A and Start on pins 6 & 9. So yeah, it still needs software support. (The six buttons pads can also be read by pulsing pin 7 with the correct timing - they multiplex the direction lines as well as the fire button lines.)

Interesting, was just gonna ask about 6 button pads too.
This would be great to have. I just don't understand why this has not been requested more and realized years ago. Making adapters would be really stupid easy.
Anyway sorry for derailing! But I hope this could be added to CD32Load/WHDLoad too, as I said, CD32 pads are expensive, harder to find, and to be honest, not really good layout-wise.

[jotd]

Still issues with this damn pad, and it appears that the revision of the A1200 mobo could be the culprit. Check this mantis whdload bug: http://mantis.whdload.de/view.php?id=3141

And that explains a lot: my A1200 is a rev B (probably 2B, don't have it here), and I always had issues with the joypads. However, lowlevel library seemed okay. Is this confirmed, and is there a workaround/fix once and for all? This button thing has been driving me crazy for 20 years now.

Note: it seems that there's a problem reading the buttons on my TF232 (CD32 expansion). Any other ppl with this issue?

[Toni Wilen]

Does it only happen with official CD32 pad or also clones? IMHO They are not 100% identical. Also does the mantis bug happen without Blizzard installed? (This needs to be narrowed down more) Check/ask also CIA revision, is it 391078-01 or -02?

[jotd]

Toni I only have official CD32 pads (yes, there are even more issues with Honeybee and such). I haven't tested CIA revision on my A1200. I could use your CIAtest tool, I know.

I don't know about the mantis bug, but at the time I seem to remember that I thought that CD32 pad routine was failing because of a CPU dependent issue (because of TST.B $BFE001) but I was told this wasn't the case.
So I tried turning the expansion off and was puzzled to see that even A1200/020 no fast the issue was the same.
(I'm talking about custom CD32 pad read, I never had any issue with lowlevel.library, though)

[Toni Wilen]

I have 2B A1200 (and -01 CIAs) but I don't have official CD32 pad (I really need to find one..)

Could you include source of button reading code that is confirmed not working with 2B+CD32 combination?

EDIT: -01 would explain the problem if CIA is in output mode because -01 bug makes CIA to always return output mode state, not pin's actual state.

EDIT: Does same happen if you tick recent WinUAE's -01 CIA option in advanced chipset panel?

[jotd]

ok Toni here are the results of my tests (sorry for the delay):

- motherboard revision: it's 1A (oooooold!!)

- ciabuttontest displays :

--*- --*-

when I press fire:

***- ***-
***- --*- (sometimes)

Joyporttest by Psygore (handly tool!):

- CD32 joypad: only red & directions are recognized, lowlevel lib or not!!
- wico joystick #1: same thing
- wico joystick #2: blue button is recognized as well.

Could you include source of button reading code that is confirmed not working with 2B+CD32 combination?

I can't include Joyporttest source because Psygore almost never releases his source code. But with lowlevel or without it, it doesn't work.

I tested with one of my latest releases, confirmed working by several EAB members here: http://www.whdload.de/games/Beast3.lha (SOTB 3, source of joypad read is in the install archive: ReadJoyPad.s).

I have added a joypad/joystick automatic detection and it works fine.

Guess what: the game doesn't recognize joypad buttons, not even the blue one, but when I plug my good Wico joystick, the second button works...

Is there a chance to be able to read joypads with a rev 1A A1200 mobo? I have a spare mobo (rev 1D). Maybe I could install that one instead.

Last Q: how can I check CIA revision?

[Toni Wilen]

Originally Posted by jotd

I tested with one of my latest releases, confirmed working by several EAB members here: http://www.whdload.de/games/Beast3.lha (SOTB 3, source of joypad read is in the install archive: ReadJoyPad.s).

Do you mean this routine also does not work with your A1200? If yes, could you make it quick and dirty tester program? (Just add interrupt and dma disable, wait for vblank, check buttons, for example flash red background when red is pressed and blue when blue is pressed, preferably also check buttons 3 and 4 because blue is also shared with normal second button, repeat loop forever.)

(I don't want to do it because I want program that is 100% guaranteed "not working". Even vertical position of read may affect results)

EDIT: Does it work if you double or triple number of dummy CIA waits?

Last Q: how can I check CIA revision?

Test program result matches -01 model behavior but only 100% sure is to open the cover and check chip labels, 391078-01 or 391078-02. (Check both CIA's, I guess it is also possible other CIA is -01 and another is -02...)

[jotd]

Toni I wanted to create a whdload joystick button tester program for a while. I'll do that. With variable delay on the CIA waits, all right.

The color stuff is the debug mode I added in JST because it just drove me CRAZY! (now I know why).

Rev 1A is probably model 01. This is the one where you can create software autofire right? I'll open the machine again tonight.

[Toni Wilen]

Originally Posted by jotd

Toni I wanted to create a whdload joystick button tester program for a while. I'll do that. With variable delay on the CIA waits, all right.

That would be good idea. "Standard" test that everyone can run easily and quickly.

I suggested changed CIA waits because this is most likely "analog problem", second and third buttons are connected to Paula POT lines which also have capacitors for analog/paddle position detection.

Capacitors will delay "digital" state changes and if something has drifted too much (old capacitors, increased internal resistance, possibly low +5v line, weak Paula output drivers etc..), it may make state change delays too long.

It may not have anything to do CIAs.

Rev 1A is probably model 01. This is the one where you can create software autofire right? I'll open the machine again tonight.

Yes, if software can change button state to released, even when physical button is pressed, it has "broken" (-01) CIA.

[jotd]

If I can create the ultimate joypad read routine that works with all joypads on all amigas I'll die a happy man.

(not that I'm planning that to happen anytime soon).

Do you think the timings would have an effect with WinUAE?

[malko]

Originally Posted by jotd

If I can create the ultimate joypad read routine that works with all joypads on all amigas I'll die a happy man.

(not that I'm planning that to happen anytime soon)[...]

So do I .
We are waiting updates of JST & CD32Load .

[Toni Wilen]

Originally Posted by jotd

If I can create the ultimate joypad read routine that works with all joypads on all amigas I'll die a happy man.

(not that I'm planning that to happen anytime soon).

I'd prefer to get non-working one first that works some hardware but not all. I want to find out the real reason. Don't fix it

Do you think the timings would have an effect with WinUAE?

No. It would be similar to keyboard hanshake pulse. Different keyboards have different requirements and no one really knows exact (or even close enough) values.