Graphic problems with AES AEC

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
Hi all, my first post, thanks for having me :)

I have an AES AEC here that has problems with the graphics, got it cheap on ebay.
Its the first original modeI I got, repaired about half a dozen japanese AES3-5/AES3-6 ("junk" from ebay and buyee.jp), one only needed the slot cleaned (they all do hey), the others had the classical rotten traces that needed to be fixed/bridged/jumped.

There was previously work done on that PCB, checked and re-soldered everything that was done.

I've switched the fast VRAM as I was told that this would cause the graphic problems, tested every single trace, all fine, there was some corrosion on the LSPC-A0, so I reflowed all the pins, no improvement, so I took it off to check for corroded traces, they are all perfect, put the LSPC-A0 back, works as before.

The LSPC-A0 has a discolouration that really looks like it burned up at some point.

Now, I have a strong suspicion that the LSPC-A0 is fried and causing this problem.
Ordered a MVH MV1 with an Z80 error off ebay, these LSPC-A0 are hard to find and not cheap, first gen NeoGeo ICs.

While I'm waiting for the (hopefully working) LSPC-A0 to arrive (takes a while as it comes from overseas), is there something else I should test/check for in the mean time?

I don't have an oscilloscope, just ordered a logic probe which also takes a few days to arrive. Multimeter, Soldering Iron, hot air station, flux, low melt solder (bismuth) and that stuff I have and I'm capable using it, not a pro though, just a hobbyist.
I burn my own Unibios and Diagnostic Bios, the latter says "all tests passed".

Also, I'm new here, is there really no search feature in the forum or am I blind?

I can show more pics/videos if needed.
 

Attachments

  • 270372186_10159956425748659_7158520077492370979_n.jpeg
    270372186_10159956425748659_7158520077492370979_n.jpeg
    455.1 KB · Views: 27
  • 270303141_10159956425798659_2684362895960125200_n.jpeg
    270303141_10159956425798659_2684362895960125200_n.jpeg
    633.2 KB · Views: 26
  • 270742502_10159958246383659_1952808871136835580_n.jpeg
    270742502_10159958246383659_1952808871136835580_n.jpeg
    541.3 KB · Views: 25
  • 270747991_10159958247703659_4198499872046637947_n.jpeg
    270747991_10159958247703659_4198499872046637947_n.jpeg
    747.5 KB · Views: 22
  • 270369952_10159958251078659_6779743162534477887_n.jpeg
    270369952_10159958251078659_6779743162534477887_n.jpeg
    845.7 KB · Views: 20
  • 270730459_10159958254658659_4332998875379288815_n.jpeg
    270730459_10159958254658659_4332998875379288815_n.jpeg
    905.2 KB · Views: 22

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
the MV1 arrived and I've transplanted the LSPC-A0, still the same problem, lots of effort for nothing and now I have a MV1 board to fix on top lol

so the DIAG bios says All Test Passed, which makes me think its the cart connector itself and the graphic glitches are actually caused by bad/missing connections to CROM, tested all the P0 - P24 to A0, B0, C0, all tested good

can't get a signal on 12M though, looks dead, no voltage at all, is known as root cause for graphics problems on Neo Geo Dev

my problem is now that I can't find it in any schematics/hints where 12M is being generated on this model of the PCB as there is no NEO-D0 and none of the A0, B0 or C0 have a pin for it, instead there is this daughter PCB has wires to the area thats supposed to be the main clock generator
I suspect that the 74LS368 is being used as clock divider to generate the 12Mhz, also the 68k is running and thats 12Mhz as well AFAIK

any hints would be appreciated :)
 
Last edited:

slugger_dan

Crazed MVS Addict
10 Year Member
Joined
Jul 15, 2008
Posts
134
can't get a signal on 12M though, looks dead, no voltage at all, is known as root cause for graphics problems on Neo Geo Dev
On a regular NEO-AES 12M comes from one of the 4 bit counters. I have it in my notes as pin 14 from one of the 74LS161s near the crystal. Don't have the daughter board in mine so you'd have to check for any changes in that section.
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
That isn't from the LSPC, that is from maybe PRO-B0.
Hello and thank you for al the sharing you do, huge fan :)

you're right, Its not the LSPC-A0, I had to swap it for another one to figure that out, in retrospect, if the DIAG bios is happy, high chance missing/interrupted traces to the cart port as its not tested by the diag bios.

On a regular NEO-AES 12M comes from one of the 4 bit counters. I have it in my notes as pin 14 from one of the 74LS161s near the crystal. Don't have the daughter board in mine so you'd have to check for any changes in that section.
Thank you!
Great input, my daughterboard has 3 74LS161s, it seems that this daughterboard actually should be creating the 12M as its the frequency divider, will look closer on that board. Please share your notes if you happen to find them, there isn't much detailed info about that model of the board, resorting to following traces at the moment.
I have two wires going to the daughter board (apart form the wires going into the plug of the daughterboard), one is ground, and I'm pretty sure the other one is CSYNC, but will need to check that. Of course the RGB pic is very dim/dark at the moment (already planing to replace the caps and Rs for the RGB output as @Xian Xi documented it once this problem is fixed).

Its seems that this schematics shows the PLL part:
However, this seems to be a "mixed schematics", showing how later revisions did that, better than nothing.

In the mean time I did the naughty and used the 68KCLK signal instead of the 12M (which I can't find atm, but both the 68KCLK and the 12M signals are 12Mhz clocks) and linked it to C4 pin A5 which is 12M and was previously dead, now graphics are fine with the 68CCLK (screenshot is composite), so its really the missing 12M, uploading some pics.

I have no intentions of using 68KCLK, really want to fix the missing 12M :) (desperation might change my decision though lol)

edit:
the flux on the last pic will come off ;)

PXL_20220115_093428397.RESTORED.jpgPXL_20220115_093613017.jpgPXL_20220115_093627197.jpg
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
I've socketed and replaced the 161 as I noticed it was a previous hacky job with two legs being only half length..

turns out that pin 14 of the 161 on this model is not 12M, it would not boot with it connected to the cart port C4 A5

so this is the clock divider circuit
clock divider.jpg

the middle one, LS74 (HD74LF74AP), is an IC with just two flip flops, one of them used as clock divider, pin 5 is 68KCLK, pins 2 and 6 are the 12M signal (presumably), this works! and it also fits the schematics of the LS74 when used as a clock divider :)

on to the next issue, the composite video is fine but RGB has sync issues, lets see if swapping the caps and resistors takes care of that or if there is another issue
 

slugger_dan

Crazed MVS Addict
10 Year Member
Joined
Jul 15, 2008
Posts
134
Please share your notes if you happen to find them, there isn't much detailed info about that model of the board, resorting to following traces at the moment.
I popped open the NEO-AES to fill in some blanks in the notes and confirm what I had, here's what I found. One note though: I just noticed my board is slightly different from the one you posted, mine matches the photo on the dev wiki (ignoring the daughterboard here). Is yours also "NEO-AES" on the right hand side or is it something else? Mine is probably older since the "LS04" on silkscreen is AS04 in yours, that's the upgraded part they went with later on. They look the same at a glance, components are in the same spot but the traces and silkscreen are a bit different.

Starting from 24MHz crystal:

X1 bottom - LS368(3) /1Y1
X1 top - LS368(4) 1A2
LS368(2) 1A1 - 100pF (left leg)
LS368(6) 1A3 - LS368(5) /1Y2 - 100pF (right leg)

Few more inversions before being divided down:
LS368 (7) /1Y3 - LS368(10) 1A4 -- /24MHz
LS368 (9) /1Y4 - LS04(3) 2A -- 24MHz

Then this 24MHz input clocks both the 161 + 74 to get 12MHz outputs:
LS04 (4) 2Y - LS161(2) CLK - LS74(3) 1CLK --- /24MHz

Then they go out to the cart slot. These should look the same on a scope but they are separate probably to avoid fanout:
12M - LS161 (14) QA - LS74(11) 2CLK
68CLKB - LS74(6) /1Q - LS74(2) (saw you posted this as 12M, is that definitely how it is on your board?)

Other graphics clocks:
LS04 (10) 5Y - PCK1B
LS04 (12) 6Y - PCK2B

IMG_0546.jpg
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
I popped open the NEO-AES to fill in some blanks in the notes and confirm what I had, here's what I found.
Thank you so much, its really appreciated, I can't state enough how much this helped me :)

One note though: I just noticed my board is slightly different from the one you posted, mine matches the photo on the dev wiki (ignoring the daughterboard here). Is yours also "NEO-AES" on the right hand side or is it something else? Mine is probably older since the "LS04" on silkscreen is AS04 in yours, that's the upgraded part they went with later on. They look the same at a glance, components are in the same spot but the traces and silkscreen are a bit different.
Mine says NEO-AES in the right side of the PCB, its a US model, 17k'ish serial.
There seems to be a a gap of documentation in the net for these first models that you Sir just filled :)

68CLKB - LS74(6) /1Q - LS74(2) (saw you posted this as 12M, is that definitely how it is on your board?)
I don't have a scope and the pin on the port goes nowhere, so I can't really find it unless I find a rotten trace to tst against, didn';t want to scrape the solder mask off all traces in the area.
68KCLKB and 12Mt would look like the same on a scope as you said because 68KCLKB is inverted to 68KCLK, and so is 12M.
So in retrospect I just picked a 12MHz signal I could find, turns out the wrong one, I've re-tried with the new LS161 and pin 14 yields a 12MHz Signal/working Neo Geo! I might have picked the wrong pin previously, mixing it up while flipping the board etc.

Thanks to you I can see the actual 12M trace now and will follow it :)

I'll double check all the signals you gave me and will report back later, again this is very valuable for me, thank you again for your efforts!
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
so the LS161 was worked on before, noticed it but ignored it..

was able to trace out the 12M from the LS161 down the slot visually (lifted some Rs to have a better look), scratched away a little bit soldermask to test continuity in different spots, could only get continuity to the slot, not the LS161 pin 14

PXL_20220116_095430098.jpg

so I took out the socket that I've just put in again ti check underneath..

View attachment IMG_0022.JPG

so the via from pin 14 (the one with the flux around it, has no continuity to the trace at all!

I've fixed that (so no need to route a 12MHz clock signal with a flimsy jumper wire)


View attachment IMG_0024.JPG

:)
 

slugger_dan

Crazed MVS Addict
10 Year Member
Joined
Jul 15, 2008
Posts
134
Good to know there's more than 1 type of NEO-AES. Would be great if SNK gave the different board a different name but can't do much about that. At least you can tell them apart quickly by looking at the LS04/AS04 difference that I called out before. The photo I posted says LS04 on the board but I see a factory soldered AS04. Using the slower LS04 sounds like a bad idea considering how it's used so hopefully no consoles were actually made with those. I think SNK was just trying to be stingy again but maybe they had problems in testing and did a last minute parts change.

The one I posted has a 3k serial lots of 1989 parts in it. Probably the oldest revision of NEO-AES.
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
3k serial number, that was really one of the first run hey :)

LS04.. not the first time a manufacturer used the "wrong" components, Sega did that with the japanese Mega Drive VA0, using a daughterboard and with cheap components used outside of their spec (also as clock divider lol), I've replaced these with 54s instead of 74s (got this tip on reddit by someone called Rage95), hard to find the 54AS04, but the 74AS04 will do, thank you again :)

Interestingly, on the NEO-AES there is no ASIC except for the LSPC-A0, PRO-B0 and PRO-C0, I've heard that the first generation chipset is a bit faster /more performant than the subsequent ones.
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
So when I got this AES, I only tested composite because the cable was right on my desk.

Composite was fine the whole time, even with the garbled graphics.

After fixing the 12M signal, I've tried out RGB, and that one has of sync issues/rolling pic and the frequency seems to be changing, I can change the amount of rolling/Sync problem a bit by turning the variable capacitor (TC2, next to the 24.167829 crystal, https://wiki.neogeodev.org/images/f/f4/Neogeo_aes_schematics_pal_2-page-011.jpg), but I can't get it to stop rolling/syncing wrong.
TP1 gives me 2.2V
TP2 0.45V (should be 2.7V)

no oscilloscope yet, seems like I really need one for this stuff, wish I didn't sell mine years ago..

After swapping a 24.00 Mhz into the mainboard (disconnecting the daughterboard), I get a stable rolling/sync issue, same when I use the original 24.167829 crystal from the daughterboard.

The MV1-1 MVS PCB I have here, where I swapped the LSPC-A0 from this AES into, seems to have an issue with SYNC (pin 126) that its generating, I can hear noise/disharmonics on it with my logic probe that matches the flicker on the screen, not with the AES though.

I'm afraid now that there is an issue with both LSPC-A0, however what baffles me is that Composite is syncing just fine, only RGB doesn't play ball here. Does this sound familiar to anyone?
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
Update:
replaced the two C1815 on the daughter PCB and socketed all three 161 and the single 74HC04P as I'm getting replacement for these soon

TP2 now gives me 2.7V, which is the expected 2.7, so I'm assuming at least one of these transistors wasn't working properly

now the picture is already way more stable, almost no jitter/sync issues, just a minor "jump" every few seconds, thats progress :)

Edit:
scrap that, back to rolling.. at least the voltage is right
 
Last edited:

Neo Alec

Ned's Ninja Academy Dropout
20 Year Member
Joined
Dec 7, 2000
Posts
11,923
I've had good luck just grabbing attenuated TTL sync instead of using the sync output by the Sony encoder.
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
I've had good luck just grabbing attenuated TTL sync instead of using the sync output by the Sony encoder.
good idea!
might just use the SYNC signal from the LSPC-A0 Pin 126/CXA1145P Pin 10 directly for a quick test

weird thing is that I don't see any sync problems when using Composite Video
RGB with Composite sync has the sync problem, same with using the CSYNC signal

that daughter PCB has only 3 connections apart from GND and VCC:
the output for the main Oscillator X1, the subcarrier (?) for Composite Video, and a blue wire that goes to the PRO-C0 Pin 112, ULN2003, a darlington transistor, I'll check that one out

sourcing that varactor diode D3 is hard, 105p@1V?

seems foolish to try to find the problem here without an oscilloscope.. lets hope its just the CXA1145P
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
There is no ULN2003, that signal goes to the PRO-C0 and the LSPC-A0, possibly somewhere else.

So I've desoldered the CXA, socketed it, bent the pins 10 & 11 upwards (CSYNC in and out), tried to use the TTL level SYNC from the LSPC-A0 pin 126.

Turns out that the CXA1145p has two legs that are shorter (same as with the LS161), so this one was already replaced, really need to pay more attention to previous work, I've reflowed almost all pins on the PCB, so after that I can't really tell easily if someone was working on that part before, need to do that different next time.
On a PCB like these, if you remove a component like this, please socket it :)
This makes it obvious were work has been done and very easy to replace these components again, also the traces/pads often come off when soldering, so its best avoided/reduced to a minimum.

Tried different resistors with a resistor decade, would either get the full TTL level, or 0V, nothing in between, having my doubts that the LSPC-A0 can drive this.

I've ordered a new CXA1145P.
I have a few THS7374 here, and a few LM1881, will try the sync stripper on the Composite video to see if that yields a clean CSYNC, will do the same for the actual CSYNC output, without scope I can only make assumptions.. current working theory is that the CXA is shot and leaking high frequency noise into CSYNC.

I'm not optimistic that I can analyse the PLL circuit without a scope, let alone fix it, so my hope is that its really just the CXA

Edit:
so using the LM1881 SyncStripper on Composite Video yields the same RGB jittery video, same when using the SyncStripper on CSYNC.. this both should work if the signal is fine, out of ideas at the moment..
 
Last edited:

Neo Alec

Ned's Ninja Academy Dropout
20 Year Member
Joined
Dec 7, 2000
Posts
11,923
Yeah, you're probably right the solution will be somewhere else.

FYI, I pulled the signal from pin 10 of the CXA1145P. Should have a 470 ohm resistor and a 220uF cap.
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
Thank you, I've just tried that, I do get the same (broken) sync with it (not bending up pins/disconnecting the CXA1145P) with these values.

So I've ordered another CXA1145P, but also a new D3 (Varactor) and the CX23065A, the last two are hard to find and did cost me.. lets hope these will fix the PLL, after that, no scope, no way IMO, might have to bite the bullet and get one (there is other uses for it here as well), but this might take a while.

edit:
so TP2 broke down to 0.45ish V again (and swinging at that), that supposed to be 2.7V and very stable, so one of the C1815 transistor got fried again, so it seems like ordering a new CX23065A was the thing to do here.
 
Last edited:

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
Just a quick update:
I've replaced both transistors (C1815), the varicap, the crystal ("borrowed" that from my Japanese AES) , and even the CX23065A..
that was after I've replaced the logic ICs on the daughterboard (a single 74HC04P and three 161, for the latter I've used the SNJ54LS161AJ as I got some in bulk) and the two electrolytic caps

The problem remains: RGB has sync issues, composite Video is fine

even got an old analogue scope with a bandwidth of 60MHz, with that I can measure properly up to 12MHz, the pixel clock and everything below it, but not the main clock) compared REF and DIVO, both are looking okay
measuring the main clock accurately would require at least 120MHz bandwidth, thats a bit over my budget at the moment

learned some things on the way about PLLs, realised why my experiments with omitting the daughter board wouldn't work/fix the problem, the PLL is adjusting the main oscillator frequency to match the subcarrier frequency divided/DIVO

every now and then TP2 breaks down to 0.4V, should be 2.7V, sync issues in both cases

there is a hand soldered ground wire on the back of the daughter PCB, goes from ground to ground and the wire is routed between the pins, looks siily but without it, the frequencies start swinging, HF is weird and very new to me, but its clear that the daughter boards design and layout was very quick and dirty

next thing is that I'm going to replace the passive components, starting with the ceramic capacitors with very small capacitance, then the coaxial cables for the subcarrier frequency input and the main clock output

so far this has been a rabbit hole, lots of work, plenty of replacement parts that are hard to find and expensive, but still not working
good thing is I'm doing this as a hobby and not for a living
 

slugger_dan

Crazed MVS Addict
10 Year Member
Joined
Jul 15, 2008
Posts
134
next thing is that I'm going to replace the passive components, starting with the ceramic capacitors with very small capacitance, then the coaxial cables for the subcarrier frequency input and the main clock output

The encoder uses the same sync input for RGBS/composite too so that’s the weird part here. If something in the digital section was bad then both RGB/composite should be bad but here 24M is at least good enough to get stable composite.

Was everything unique to RGB ruled out already? Random things I'm thinking of: the resistor/cap for SYNC->AV port, the AV port sync pin itself, RGB cable and any extra circuit in there (got spares?), RGB sync settings on your monitors, different monitors if you have any especially if you have a CRT incase your other display is fussy about sync, any VHOLD sync settings or any other sync setting on the monitor. Some monitors can show the current sync rates and you can see it change in realtime, could help. Another thing is checking the sync output on a scope since something’s up with the vertical blanking pulse if the picture is rolling.
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
Thank you for the suggestions @slugger_dan :)

The encoder uses the same sync input for RGBS/composite too so that’s the weird part here. If something in the digital section was bad then both RGB/composite should be bad but here 24M is at least good enough to get stable composite.
I was puzzled as well (and still am tbh), but I the realised that the purpose of the PLL to sync between the subcarrier and the main clock, and doing so by adjusting the frequency of the main clock, so the composite subcarrier is "always correct", the main clock is the flexible part.
However, the main clock needs to be matched to REF and I believe this is the source of the problem (this is the reason btw why omitting the daughter board didn't yield stable sync).

So the only thing working right is the Composite Video because the PLL is not adjusting that, its fix, and it happens to be a whole multiple of REF and DIVO

Forgot to mention it, but I even replaced the CXA1145P.. yeah I'm getting desperate.

Was everything unique to RGB ruled out already? Random things I'm thinking of: the resistor/cap for SYNC->AV port, the AV port sync pin itself, RGB cable and any extra circuit in there (got spares?),
I tried using an LM1881 Sync Stripper (see above), no matter if I used CSYNC or Composite Video as source to get a striped sync, both had sync issues with RGB, that fits the thesis that REF (and thus the sync of anything RGB related) is off.

I have a Japanese AES that I'm using as reference, its an AES3-5 model so I can't directly compare everything but good enough to double check.

Currently I'm using an old flat screen TV that has an SCART RGB input, as my OSSC needs fixing (new THS7353 is on the way), the TV works perfectly fine with my Japanese AES, MVS, Saturn, SNES and DreamCast.
Without the OSSC I can't really see the refresh rate directly, hoping for the spares to arrive next week, it could be that the AES refresh is just a little but too low for the TV to properly work with it.

On the scope RGB and CSYNC look fine, same for composite video, REF looks a bit dirty but not too bad, will post pics later today.

I also try swapping the subcarrier crystal, since its used as fixed reference, it may be the cause of this.

Also, I'm really close to buying a universal frequency counter lol, I'm getting a Loewe Concept 55 CRT TV in 2 weeks as well

EDIT:
after explaining how this is supposed to work ("Rubberducking" FTW), it could be that the subcarrier frequency is off a bit too much, and the PLL can't match it. After all, everything is related to that frequency, it is the "fixed point" here so to speak, if that is off, everything else will be as well.
So far all my adjustments have been around the PLL daughterboard, there is a varcap for the subcarrier itself that I haven't touched yet.
However, to measure frequency precisely, I will need a frequency counter. Might have to do this wit trial & error.
 
Last edited:

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
Here are some oscillographs from the working AES3-5, this one works fine for Composite Video and RGB.

I can't measure the CSYNC output on the AV port, because this model has Composite Video on both pins, there is tutorials in the net how to restore CSYNC, but I don't want to mod this console, want to sell it on as is, future buyers can have some fun modding it, also happy to do it for them if requested.

Anyways, I measured the CSYNC on the CXA1145P, its attached.

TP2 on this one is exactly 2.6V

AES3-5 NTSC subcarrier:

AES3-5 NTSC subcarrier.jpg

AES3-5 DIVO (upper, CH1) vs REF (lower, CH2)
AES3-5 REF vs DIVO.jpg

AES3-5 CSYNC

AES3-5 CSYNC.jpg

These signals are much cleaner than what the NEOAES had, will post pics of that soon.
 
Last edited:

Neo Alec

Ned's Ninja Academy Dropout
20 Year Member
Joined
Dec 7, 2000
Posts
11,923
Sorry if I missed this...
Have you tried more than one TV and cable? The Neo Geo's sync is notoriously incompatible with modern TV's.
 

maki

Edo Express Delivery Guy
Joined
Jan 1, 2022
Posts
332
another quick update, REF and thus SYNC are totally off on the NEO-AES

here is a short video of the NEO-AES, the red channel and CSYNC, synced to CSYNC


As you can see its very "dirty", the displayed frequency breaks down/swings, that ghosting effect/noise is due to the sync "swinging", its not stable, only VDP and the Subcarrier are in sync, since they use the same "broken" signal for syncing.

REF is the cause of that, have to check where it comes from, its either generated in the LSPCA-A0 or PRO-C1, documentation is lacking but since the REF trace was broken (rotten via due to flux), I can disconnect PRO-C1 easily by removing the jumper again, so I can see where it comes from, lets hope its the C0, have a replacement for that, not for the A0...
 
Top