Graphic problems with AES AEC

[maki]

Okay, disconnected C0 from REF (pin 112 on the PRO-C0 IIRC), REF is generated by the LSPC-A0

This is REF measured at pin 117 of the LSPC-A0:

https://imgur.com/9ddCwYi

thats a lot of noise.. bummer

sadtrombone.mp3

I'll check if the trace is connected to some other components that could cause this, but right now it looks like I bought a MVS that has a shot A0 as replacement..

Edit:
REF is also going to the PRO-B0 pin 108, haven't found connections to other components yet but there might be some

Edit 2:
so I lifted the pin 1-8 on PRO-B0 to check if its the source of REF
its not, so far everything points to LSPC-A0 as source..

 

[maki]

Went as far as replacing the LSPC-A0 again with the one that was originally in there, REF is still doing that funky stuff, not as bad I think.
So, pretty sure the PLL is not working as it should.

I'm thinking now about replacing the PLL daughterboard with a simple clock divider first to get the RGB sync locked, if that works, maybe even rebuild a PLL with modern ICs to regain the composite video.

Anyone ever done this or has links?

EDIT:
I was over complicating things.. I'll try to solder the crystal directly to the mainboard, if the problem persists, I will then trace the divided frequencies.
I did this few weeks back and sync was still off, but didn't have a scope back then to check. If this is really the case. the REF/SYNC problem is not related to the PLL.

 

[slugger_dan]

I was over complicating things.. I'll try to solder the crystal directly to the mainboard, if the problem persists, I will then trace the divided frequencies.
I did this few weeks back and sync was still off, but didn't have a scope back then to check. If this is really the case. the REF/SYNC problem is not related to the PLL.

If anything removing the daughterboard is better for troubleshooting, removes a lot of variables, the 24M clock isn't being adjusted by some analog circuit, less points of failure to think about and the boards were added for composite quality rather than fixing any sync issue. Composite video on my NEO-AES without daughterboard is hot garbage but it does seem to have more stable sync than my 3-6, which has the daughterboard hack built into it, which means the sync rate is prone to slight changes. On one TV, the 3-6 sync rate moves around enough that every few minutes, the TV OSD popup with a “hey the sync rate changed” message. Never happens on the NEO-AES. Maybe coincidence, but modern TVs are sensitive to this sort of thing.

Getting a working OSSC for the HDTV + a CRT would also help, just to have different displays setups to try on. Plugging it directly into a HDTV is probably the least forgiving setup.

 

[maki]

If anything removing the daughterboard is better for troubleshooting, removes a lot of variables, the 24M clock isn't being adjusted by some analog circuit, less points of failure to think about and the boards were added for composite quality rather than fixing any sync issue.

Yeah absolutely, my expectation would be that without the daughterboard and the crystal on the main PCB, sync with RGB should be stable, and Composite should be really bad. Last time I tried this I had not scope, but sync was off.

Composite video on my NEO-AES without daughterboard is hot garbage but it does seem to have more stable sync than my 3-6, which has the daughterboard hack built into it, which means the sync rate is prone to slight changes. On one TV, the 3-6 sync rate moves around enough that every few minutes, the TV OSD popup with a “hey the sync rate changed” message. Never happens on the NEO-AES. Maybe coincidence, but modern TVs are sensitive to this sort of thing.

The way I understand this, is that the PLL will change the master clock and the subcarrier is the fixed point to which the master clock will be adjusted, REF is just a feedback used to speed up/slow down the master clock, since CSYNC is derived from the master clock, its going to swing as well.

I've found out that there is programmable oscillators, used as dual frequency oscillators:
https://nfggames.com/forum2/index.php?topic=5744.0

Now with that and another sort of mod, it could be adjusted to either have great Composite or perfect RGB, I'd prefer to keep Composite, but my focus is on getting the RGB sync right. Thats something for another round of soldering/testing though.

I'm a bit short on time these days, real world has demands, started a new job.

Thanks to you and the information you shared about the crystal and components in your AES I should be able to find the culprit, otherwise information is very rare on these old models on the net. Its probably going to be a counter used as frequency divider that has hiccups/counts wrong thus causing this..

EDIT:
replacement parts for my broken OSSC (need to replace the THS7353) got stuck, was hoping that Mouser would be faster but since its being sent from the US to Germany, things can get stuck. I'm getting a CRT TV next week (Loewe Concept 55, multi standard Loewe CRTs aren't hard to find around here)

 

[maki]

Got the multi norm CRT TV (Loewe Concept 55, 1991) and RGB works just fine.. so it was my LCD TV all along, I feel foolish for fiddling so much with the AES and then it turns out that it worked all along, even though its mentioned here and in many other threads there is even a thread here just about the Neo Geo compatibility with TV sets..

I go the replacement THS7353 for my OSSC, turns out the LDO if fried, waiting for a new one, can't say until then if it will work with the OSSC but I'm optimistic.

Thanks everyone for the help

 

[slugger_dan]

Well that's a good outcome in the end. The sync issue on newer TVs is partly why I’m holding onto my PVM-9042. Great little CRT with all the inputs and useful when diagnosing/repairing old hardware.

Topic is partly about the video daughterboards so I’ll add this: if I buy a NEO-AES that happens to have one of these, I’m removing it at the first sign of trouble and will just restore the console to its original state. Long wires carrying a relatively fast clock signal, possibly dodgy soldering of those wires from some overworked SNK technician, possible jittery clocks causing erratic sync and other problems, 30 year old caps used to control a sensitive part of the console plus the grossness of it. All to fix a problem that doesn’t exist if you use RGB. I’d save the board itself since the same parts show up on AES 3-X incase one needs repair but that’s about it.

 

[maki]

If the daughterboard acts up its getting evicted immediately lol.. might throw it out in the future, your reasoning is sound and tbh at first I treated it as something "worth preserving" but it does trade certain things to improve Composite Video, I'm only using it with RGB so its not a good trade..

The sync is not just very dirty on both first gen chipset Neo Geos I have (this AES and a MVS MV1-1), its also very "slow" and tends to break down in frequency more often and lower (really to to 15KHz and possibly below it fo a brief moment) than the Generation 2 chipsets, I had a few AES and MVS of latter generations and none had these issues.

The "dirtyness" is a feature of the LSPC-A0, which does generates all the syncs from the master clock:
SYNC - 126 - Composite Sync
BNKB - 127 - Vertical blank
CHBL - 128 - Horizontal Blank
REF - 117 - reference used by the PLL

Interestingly, on the 2nd gen models there is ceramic filter caps for CHBL (220pF), BNKB (330pF) that are missing on my first gen models, C61 and C57 on the far left of the diagram:

https://wiki.neogeodev.org/images/b/be/Neogeo_aes_schematics_pal_2-page-003.jpg

Adding these caps on the first gen models did calm down the noise on BNKB and CHBL, but I can't tell of the composite sync is generated from the horizontal and vertical blanks or vice versa, composite sync is still very dirty, hard to tell if it made a difference. I might try adding a 560p cap to TTL SYNC.

I've already tried using a LM1881 to fix the dirtyness, didn't help. Might try an EL1883 in the future (I have a some here), its very similar but better as it can even deal with a chopped sync by MacroVision. Then there is the ISL59885 which can do a full sync restoration, expensive but looks very interesting for this use case. I've also read that people had success with using Extron conditioners that have ADSP, I have one of those here but need to wire it up to the CRT (which doesn't like TTL CSYNC).


Apart from the sync being dirty, its also very 'slow', sometimes even below the specs IMO, this is a function of the used master clock crystal, MVS comes with 24.000 stock, and the primitive frequency counter in my scope indicates that with it the hor. sync breaks down below 15KHz briefly.
The AES uses a faster crystal for that reason (24.168..) and current DFO solutions seem to aim at 24.3ish MHz.