MV1FZS graphic glitches and sound problems

miisalo · Oct 26, 2015

Hello

Got non working mv1fzs cheap and started to fix it. Seller said that it didn't always boot and if it booted (only with unibios), it had graphic problems and no sound or sound problems.

First I replaced the bios socket, it was so loose that chip did't have good connection. After replacing the socket, normal bios got me to the Z80 error. I quickly socketed my extra AoF cart and burned diagnostic bioses and ran the tests.

All I got was SM1 unresponsive and continuing after that - no errors. The sound in the games are sometimes missing and sometimes it sounds like its playing wrong samples. Could that be just address line problem?

Graphic problems look like this:

Looks like odd or even pixels are missing.

I quickly scoped fix0-8 lines and there was signals there, also used logic analyzer to see that ZMC2 gad and gbd lines had something happening. One thing did happen when I probed even, load, dota and dotb signal - I lost more pixels:

I wasn't sure that did I managed to break solder while installing probes or had the previous owner fucked up smds bending the board but I quickly used heat gun for ZMC2, B1 and LSPC2 and got back to the first picture's situation.

Any ideas what could be the cause of there problems? I only quickly tested that those signal lines are not stuck high or low or are not floating, not checked if they made sense

Apocalypse · Oct 26, 2015

On the first picture it seems 1 out of 4 pixels is not decoded. Can be a latch in ZMC2. Can you check C-ROM data lines?
Don't think it has something to do with H (used for horizontal mirroring) or EVEN (used to switch pixels A and B).
You can't see anything with a logic analyser on DOTA or DOTB. There are just 0 if the pixel is black, otherwise 1. In your case it should follow this pattern (12Mhz): 1, 1, 1, 0 (black pixel not decoded).

miisalo · Oct 26, 2015

Apocalypse said:
On the first picture it seems 1 out of 4 pixels is not decoded. Can be a latch in ZMC2. Can you check C-ROM data lines?

Its every other pixel that is missing, there is about 160 lines on the screen. I don't much know about how mvs forms the picture but I'm now reading the wiki and that ZMC2 multiplexing seemed to be good suspect. C-ROM traces seems to be fine. Hard to use logic analyzer when probes go under the cart.

I just removed ZMC2 from spare mv-1fs I have. I could just try to use that hoping that I can do replacement it nicely with my heat gun. But I would prefer to identify the problem before randomly replacing components.

Edit: I hate autodiscorrect.

Apocalypse · Oct 26, 2015

miisalo said:
Its every other pixel that is missing, there is about 160 lines on the screen. I don't much know about how mvs forms the picture but I'm now reading the wiki and that ZMC2 multiplexing seemed to be good suspect.

The fact complete vertical lines are missing is absolutely normal if 1 out of 4 pixels is not decoded. MVS decodes pixels 8 by 8 from C-ROM which is latched (32bits). So 4bits per pixel. What I meant is only 3 pixels on 4 are visible and as ZMC2 (PRO-CT0) uses dual mux for decoding pixels if one latch is faulty a pair of pixels A/B will be missing. For each new horizontal line the missing pixels just match the preceding line vertically as each line has the same number of pixels.

[EDIT] have a look at my AES to MVS topic which can help you understanding the decoding cycle (quite technical though).
http://www.neo-geo.com/forums/showthread.php?260590-AES-to-MVS-(plays-AES-games-on-MVS)

miisalo · Oct 26, 2015

Thanks for all the info, I'll look into it.

miisalo · Oct 27, 2015

So it was not ZMC2. B1 maybe then?

Apocalypse · Oct 27, 2015

miisalo said:
So it was not ZMC2. B1 maybe then?

MASK4 (SM1) or RAM11 (6116) could explain the sound issue.
Concerning graphics, are you sure C0~31 lines are all ok?

miisalo · Oct 28, 2015

Apocalypse said:
MASK4 (SM1) or RAM11 (6116) could explain the sound issue.
Concerning graphics, are you sure C0~31 lines are all ok?

Yes, I have triple tested C0-31 lines. And now when I actually started to think, sprite data is not missing, to problem is when rendering the screen. If I move character one pixel, it will show the pixels that where missing and hide the shown pixels with those lines on screen. So its getting the data, isn't it? Non of the FIX0~7 lines seems to be stuck (that would explain the lines?) and I even replaced the 74ls245 that looks like it's used to select between sfix or cart srom just in case.

This really starts to look like conflict between fix and sprite data.

About sound hardware, SM1 and RAM signals are not stuck, but not sure if chip enable in the SM1 goes as it should. I'm about to socket it and test with another chip. But wouldn't the diagnostic bios find the ram problem?

E: And suddenly I realized that player 2 A and C buttons are on. C1 pins seems to give like 2V for those lines. No bad traces or anything. I think I would need spare part board to fix this.

Apocalypse · Oct 28, 2015

miisalo said:
Yes, I have triple tested C0-31 lines. And now when I actually started to think, sprite data is not missing, to problem is when rendering the screen. If I move character one pixel, it will show the pixels that where missing and hide the shown pixels with those lines on screen. So its getting the data, isn't it?

I'm not sure of it at all. When making tests with my converter I could move character but there were always the same vertical lines missing on screen.

miisalo said:
About sound hardware, SM1 and RAM signals are not stuck, but not sure if chip enable in the SM1 goes as it should. I'm about to socket it and test with another chip. But wouldn't the diagnostic bios find the ram problem?

What kind of BIOS is installed on your board?

miisalo said:
E: And suddenly I realized that player 2 A and C buttons are on. C1 pins seems to give like 2V for those lines. No bad traces or anything. I think I would need spare part board to fix this.

Humm... So your board seems to have a general issue (plugged to a faulty PSU by previous owner?).

miisalo · Oct 28, 2015

Originally it had standard bios and now I have unibios on it, but I tested it with the diagnostic bios set (also made the diagnostic cart so it did Z80, sound RAM etc tests ).

I think so too that board has been killed with faulty psu or shorting something.

What comes to the picture, isn't zmc2 just getting two words (2x16bit) from carts C ROM from address that lpcp2-a2 has generated. Which means that sprite's pixels always come in same C0~31 line to zmc2 regardless the sprite position in the screen. And if I can see every pixel by moving the sprite, it means that lines are good. Its the effect of having jailbars over the picture that is normally just stuck FIX line.

Apocalypse · Oct 28, 2015

miisalo said:
What comes to the picture, isn't zmc2 just getting two words (2x16bit) from carts C ROM from address that lpcp2-a2 has generated. Which means that sprite's pixels always come in same C0~31 line to zmc2 regardless the sprite position in the screen. And if I can see every pixel by moving the sprite, it means that lines are good. Its the effect of having jailbars over the picture that is normally just stuck FIX line.

You're right concerning access to C0~31. It could have been one of the latches in ZMC2 but as you replaced it... Really strange, I don't get what's wrong. The fact that missing lines have equal interval should mean it's related to the sprite data decoding. You can see that S-ROM access is good and that texts are not affected by jailbars.

miisalo · Oct 29, 2015

Driving me bonkers. But so far has been really good learning experience, i have been reading a lot and studying how the mv-1fzs board works using multimeter, scope and logic analyzer so time well spent. When the problem is not ZMC2 itself, could be B1 or some of the signals controlling either chips. Some clock lines have ringing on them but I'm not sure if that is abnormal level.

I have mv-1fs board that had working graphics but sound side has total destruction after berrytogard had bit of an accident with the polarity. i harvested the zmc2 from there but when I tried to remove C0, that damn chip seems to be glued (or flux has glued it down). I gave it heat and even legs were free, I could not lift the chip. To be honest, I probably gave it so much heat before realizing that its not solder holding it down that i probably damaged it. I'll give the chips flux cleaner bath and see if I can then remove'em. In this point I consider both boards more like learning material than something I would treat like (as we Finns would say) "sore cunt". Probably going solder test wires to chips so I can use logic analyzer while playing with it.

Razoola · Oct 29, 2015

Try this Miika, I'm interested in what the result is and it may help you too. Run the system with a unibios and bring up the ingame menu when you are at an area where the issue happens (like your screenshots). Then in the patch engine do the following patch.

0x00 - 0x3A000B

This will cause the system to use its onboard FIX tiles and not the game cart stored FIX tiles (audio will crap out also). Of course the correct game tiles will no longer be showing but if the vertical lines are also gone or change in some way it will could rule out a few things or help pinpoint an area to check over.

Raz

miisalo · Oct 29, 2015

Tried and nothing happened to vertical lines.

Razoola · Oct 29, 2015

That's a shame, I guess you did see the onscreen tiles corrupt though? It was wortha go at least given how simple it was to try.

miisalo · Oct 29, 2015

Razoola said:
That's a shame, I guess you did see the onscreen tiles corrupt though? It was wortha go at least given how simple it was to try.

Yes, tiles went to mess. But this is really good tip for the future, I have spent some time analyzing the fix lines and even replaced the ic controlling the bus. This is good way to test it.

GadgetUK · Oct 29, 2015

EDIT: I've removed the stuff I suggested here as you seem to be getting 1 line missing out of a quad, so it's not a toggle type issue somewhere, it has to be data transition somewhere. One bit constantly low, and since moving brings that 'invisible' row back into display I would assume internal fault on the LSPC or B1. Worth checking all data lines are connected on both the LSPC and B1 but I guess you've done that already.

EDIT2: I am not even sure if a data bus issue would cause this due to the format of the 4 bit pixels. It has to be internal to the LSPC I think. Have you tried test mode to output a full colour screen, eg. just red background, green background etc, just to rule out internals? It might not prove anything doing that but if you did see the same lines in the test mode solid colours it would indicate an internal fault to me.

mikejmoffitt · Oct 29, 2015

Do the lines scroll with the sprites, or are they static?

miisalo · Oct 29, 2015

Static lines.

GadgetUK · Oct 29, 2015

In test mode what happens to the cross hatch - or solid filled backgrounds in test mode?

Xian Xi · Oct 30, 2015

If it's affecting both the C rom Data and S rom Data then I'd rule out the ZMC2 since the fix data bypasses it. When you play an MVS cart, on the Neo Geo boot screen is both the Neo Geo spinning logo and the text under it have lines? I think I might have had this problem before and it was a TTL that the FIXD0-D7 were going to.

miisalo · Oct 30, 2015

GadgetUK, cross hatch and backgrounds are ok.

Xian Xi, Only spinning logo has the missing lines, so its C ROM data only? ZMC2 I replaced with known working one, it was faster than setting the logic analyzer for all the lines (=soldering test wires).

Because the bios socket was fucked up and the bios pins were also bent, the guy who owned the board has probably used bit too much force when swapping to unibios and back to normal. That means that board was probably bent and so I checked chip legs with magnifier and trying to move the legs. I did some rework on the chips but now I'm not sure if I did it on B1. Magnifier is not so good tool, I wish I had proper microscope.

Xian Xi · Oct 30, 2015

Have you tried another riser card? The one you have might be corroded in the board interconnects.

miisalo · Nov 17, 2015

Xian Xi said:
Have you tried another riser card? The one you have might be corroded in the board interconnects.

Nope, I don't have similar riser cards. Multimeter and scope tells me that is should be ok.

I replaced the B0 just for fun and still the same... I will continue with this once I have done some fixing on my six slots (seems to need recapping the audio).

My_Anal_Fissure · Nov 22, 2015

Using CMVS made by Xian Xi. Seeing a glitch on some of my older beat em ups like Burning Fight and Robo Army (see pics).

Issue is intermittent. Anyone seen this before?

MV1FZS graphic glitches and sound problems

Astra Superstar

Edo Express Delivery Guy

Astra Superstar

Edo Express Delivery Guy

Astra Superstar

Astra Superstar

Edo Express Delivery Guy

Astra Superstar

Edo Express Delivery Guy

Astra Superstar

Edo Express Delivery Guy

Astra Superstar

Divine Hand of the UniBIOS,

Astra Superstar

Divine Hand of the UniBIOS,

Astra Superstar

Ace Ghost Pilot

Mickey's Coach

Astra Superstar

Ace Ghost Pilot

JammaNationX,

Astra Superstar

JammaNationX,

Astra Superstar

n00b

Attachments