Investgating PAL 7800 picture issues (was 7800 RGB Musings)

[juansolo]

That's exactly the same problem I'm having. Well found! I'll give that a try in a couple of days. Cheers!

[emmanuelf]

Look like clock line impedance problem.

A resistor on the line close to maria chip is worth a try.

[juansolo]

Mini update... So I can't get over to where I can solder my own 7800 until Weds, however @marauder666 has quickly bodged in a buffer to his to test (we'll tidy this up later) and it's made a significant improvement. We'll experiment on mine shortly, but as I'm bypassing the 2600 side, there may be other options. As it is, it certainly appears to be the main source of the issue.

 

 

[juansolo]

So this is where we are currently via iMessage. Experiments have taken place.

 

First buffering the colourburst signal to the Maria to seperate it from the TIA one and stop them interfering. This seems to have a profound effect on the dot crawl (though doesn't remove it). This is the suggestion from the other thread and involves removing R74 and sticking it through a buffer (pink option).

 

Because I've not got the 2600 side of things working in mine, he's also tried pulling R41 and bypassing the TIA colourburst feed completely to send the signal direct to the Maria (blue option). Same result (so easier on my machine). 2600 will be in B&W on his wired this way, he'll have to go with the pink option if he wants to keep the 2600 side of things.

 

To try and eliminate what's left he's also tried pulling R37 and R35 to disconnect the chroma out from the TIA as that's had an effect on NTSC machines. However no real difference. That said, that signal travels all over the PCB on a PAL 7800, so if we've got a socket big enough, I'll pull the TIA and socket it so I can lift the pin.

 

Edited June 22, 2020 by juansolo

[emmanuelf]

Giving MARIA is own local color-burst oscillator is perhaps the only way to cure things.

With the buffer, what the signal look like on a scope on the maria pin ? Buffer + R on the maria side is perhaps sufficient.

[juansolo]

Yeah, that's probably what it'd take. I'm not able to do anything until Weds, I'll give @marauder666 a ping and see if he has chance to re-connect the buffer and stick it on the scope. We'll be having a proper fettle Wednesday regardless.

[juansolo]

Ok so back to mine and I've done the blue option mentioned above as my 7800 has most of the 2600 side isolated already so I've carried that on to it's natural conclusion. The result is that it changes it from uniform diagonal dot crawl to a much more random and weaker horizontal intereference. It's still there, but it's much less apparent. Indeed you'd barely notice it on a screen with a lot of overscan (I'll take a pic on my PVM when I get home this evening).

 

To rule everything out I also socketed the TIA and lifted the chroma pin. Something that apparently works well on a NTSC 7800 for reducing noise. On this it made no difference whatsoever. Again as I've disconnected everything else on mine, I'm going to leave it this way regardless.

 

At the moment I'm quite happy with things, It's certainly by far the best picture we've had from a PAL 7800. However we're not done, as you'll notice on the above schematic there's still a connection from the TIA to the colourburst circuit that as far as we can tell is to start it. making it something we can't disconnect and a remaining potential source of the interference. So we've ordered some PAL oscillators as @emmanuelf suggested above. It'll be interesting to see if that eliminates it completely.

 

The plan is to leave mine with the isolated 7800 and UAV and the other as 2600/7800 with the MK, and see how they compare when we've got this as clean as we can.

[juansolo]

I'll be the first to admit that I have no idea what I'm doing with a scope. We just tend to use it to make sure video signals are the correct voltages. But here's the output of the PAL colourburst that we're sending only to the Maria.

 

You can also have a laugh at our ancient Farnell scope

[juansolo]

UPDATE:

 

The blue option doesn't work. It's really, really good for a short period of time, then the oscillator stops and the picture goes B&W (looks like it needs to be connected to the TIA also). Looks like the other line is what keeps it doing it's thing and without the TIA getting it's colourburst, it just stops after a while. So the buffered route will likely be just fine. However we have the oscillators on the way, so I'll try that next before falling back on buffering.

 

EDIT: False alarm!

 

It's the colour pot (nearest the cart slot), needed a tweak. That's all.

 

Still not finished experimenting however Just waiting for a delivery.

 

Some before and after pics.

Edited June 24, 2020 by juansolo

[+-^CrossBow^-]

22 minutes ago, juansolo said:

UPDATE:

 

The blue option doesn't work. It's really, really good for a short period of time, then the oscillator stops and the picture goes B&W (looks like it needs to be connected to the TIA also). Looks like the other line is what keeps it doing it's thing and without the TIA getting it's colourburst, it just stops after a while. So the buffered route will likely be just fine. However we have the oscillators on the way, so I'll try that next before falling back on buffering.

 

EDIT: False alarm!

 

It's the colour pot (nearst the cart slot), needed a tweak. That's all.

 

Still not finished experimenting however Just waiting for a delivery.

 

Some before and after pics.

 

So if I'm understanding everything here thus far. You have essentially turned this specific 7800 into a 7800 only unit correct? You stated taking the buffer circuit out of the mix for the 2600 side, so and even lifting the chroma pin out of circuit on the TIA so again, pretty much only a 7800 games console at this point?

 

Now, if some eagle eyed people might have noticed from some of the last few pics I've attached of recent installs I've done for people, they might have noticed that with the exception of 2 video tap locations, I'm no longer tapping my signals directly off the bottom of the resistor ladder. I found by accident that there are in fact unused vias that are closer to the signal sources that aren't populated on all the NTSC units I've worked on so far. It makes for a better looking install, makes sure the wires don't overlap any noisy components in the system for the most part, and slightly reduces the length of the wires needed and this has improved picture quality to my eyes on the 7800s I've done this way. The only reason I can't do it for the Sync and Lum 1 taps is because those unused vias are under passive components on the board and not able to be reached as easily in this case. Although nothing would stop a person from simply scraping solder mask off the actual traces on the board and just soldering to those instead.

 

[juansolo]

50 minutes ago, -^CrossBow^- said:

 

So if I'm understanding everything here thus far. You have essentially turned this specific 7800 into a 7800 only unit correct? You stated taking the buffer circuit out of the mix for the 2600 side, so and even lifting the chroma pin out of circuit on the TIA so again, pretty much only a 7800 games console at this point?

 

That 7800 is totally a best case scenario so we could rule things out by completely disconnecting the 2600 side of things as far as is possible (there are still 2 connections to the TIA that I can't bypass until we try an external oscillator). The other 7800 is my old one that Cleggy (Marauder) now has, that one still has the 2600 side connected and is using the MK board. The buffer on the colourburst line on the other 7800 (his machine is the one above the post with the schematic in) has equally as profound effect on the picture and keeps the 2600 backwards compatibility.

 

As I say, we're just experimenting currently and Cleggy did ask if I was gonna put the 2600 side back... At this point, probably not as I have a RGB 6 switch that I play 2600 carts on. There's not a massive difference currently between with and without. We'll do a proper back to back when we're done.

Edited June 24, 2020 by juansolo

[juansolo]

Cleggy's been continuing the experimentation while we wait for the oscillators to arrive. Quote: "Super Bodged! 74LS08 quad AND gate as a buffer like the bloke used in that ancient thread. Buffered 4.43Mhz to each video chip through it.  Noticed on schematic that LUM0 on the MARIA isnt buffered, so use another gate on the 08 for that.  Tweaked all 3 pots, have to increase brightness on monitor for VCS but its not bad"

 

Apparently another noticable improvement. He's now seeing the aliasing on the Ballblazer score that I can see on my isolated 7800.

 

So current plans are: Try the stand alone oscillator on mine running the Maria direct and see if there's any improvement. If there is, run a pair of them on Cleggy's machine replacing the buffers.

 

Edited June 25, 2020 by juansolo

[juansolo]

I blame the hot weather (it's currently uncommonly hot in the UK, we're so not made for this weather), so he's still sending me IMs... Seems he wasn't happy with the VCS side of things being dimmer and a bit shittier, so decided to have a go at fixing it. Pulled the 8k2 resistor on the VCS chroma line on the MK and stuck a 10k trimmer in there. That's done the trick! I've asked him to pull it and measure it to find out what it ended up set at. Says he's going to stick one on the 7800 side also to have a bit of a play and see if there is a more optimal setting for that also. Now bear in mind that this isn't currently how you'd install one of these standard as he's got the buffers in the circuit that wouldn't normally be there, but I think the pics speak for themselves.

[juansolo]

Update this morning is that doing the above has buggered the 7800 side. Buffering the chroma might be the solution there. More experiments required.

 

EDIT: ...and that made it worse. Backing those changes out, VCS side looks like it's just going to have to take the hit for now.

Edited June 26, 2020 by juansolo

[juansolo]

Well it was a day of utter frustration (well not completely as I finished my MiSTer Joystick upgrade and fixed the battery on my Dreamcast). The 7800 however was being a complete sod. This nonsense is where we're at and it's the best picture we've had out of the one where we're trying to maintain 2600 backward compatibility. The 7800 side is like mine in that it still has some shimmering (still awaiting parts to potentially address that), but the VCS is near perfect.

 

When we've got some sort of conclusion at the end of all this I will do a summary of what we ended up with...

Edited June 28, 2020 by juansolo

[juansolo]

This is where the above machine is at the moment, it's the best he's had out of it to date. He's had to pull back the 2600 side a bit as it has a knock on effect to the 7800. The problem/flaw is how it gets the luma levels... Ideally it'd need a pair of 4050s to keep everything separate until the luma pin of the svideo socket. He's also noticed a problem with the sync on panels for the 7800 side. It works fine on the PVMs but they're pretty tolerant of oddness. I'll be back over mid-week sometime and take mine back over to see if mine is the same.

 

Edited June 28, 2020 by juansolo

[juansolo]

Latest update: My machine is still awaiting the oscillator that's coming from France and is taking forever to get here... Cleggy's (@marauder666) machine I saw on Wednesday, he's still poking around with that and mine with the scope. Turned out mine had a really nice sine wave with the colourburst and the 74LS08 on his (the recommendation from the other thread) made it square. So he's now removed that and replaced with a pair of transistor buffers on the colourburst signals (removing the one he added to LUM3 also as it doesn't really seem to make much difference) to get them back nice and siney.

 

This still has the buffers on the two chroma signals at the other end with the trimmers to balance them. Now I got to see this prior to pulling of the 74LS08 and it was bloody good. I mean really, really close to mine (the one with the 2600 disabled to give the best possible picture with the on board oscillator) on the 7800 and with a surprisingly brilliant 2600 picture. He was having issues with noise on the 7800 chroma which the transistor buffers have sorted.

 

Another change made to the Magic Knight board was replacing the 1N4001 diode with a BAT85 Schottky diode. This helps with balancing the brightness between the 7800 and 2600.

 

 

So should the external oscillators not work for whatever reason, this might be as good as it gets and if that's the case I'll write up the two options: Mine that is the easy one to do, but loses the 2600 back compatibility, and his that keeps it and is really actually rather brilliant, but is a bit fiddly to set up.

 

The external oscillators might be the holy grail of PAL picture goodness, if they work... But even if they do, they're difficult to get hold of so might not be a plausable solution for everyone.

Edited July 3, 2020 by juansolo

[juansolo]

Yay, the clocks arrived!

 

So what we have now in the control 7800 is the clock sat in the vacant 74LS32 socket as the power and ground pins line up, and it has a decoupling cap on the power, which is nice. All we then need to do is remove the resistor that corresponds to the clock pin and replace that with a wire from that pad to the clock in of the MARIA. To further rule out the on-board clock causing interference, we pulled the transistor to stop that.

 

 

The results however are not as big as we'd hoped. There's a marginal improvement, but there's not enough to warrant not using the original clock. It has a nicer waveform on the scope, indeed all the remaining interference is random, there's no more diagonal interference which is much more noticable and distracting. There's not really anything else we can do with this one, it looks like this might be as good as it gets.

 

The other machine, we have plans and more parts on the way (CD4066 if you want a guess at where we're going next). There are potential improvments there and a potential simplification of the changes to that one (it should remove the need to blend/balance the chroma). We're not done yet.

 

Edited July 6, 2020 by juansolo

[juansolo]

Ooops, accidental post.

 

I'll bung a picture or two here on the big PVM when I get home.

[juansolo]

Not sure what happened there, couple of spurious posts while trying to edit the first, sorry about that...

 

Anyhow, back home and on the big PVM. Holy crap! Errm, that's made a hell of a difference. The vast majority of the interference is gone and what's left you'd only notice being super analytical. All the uniform interference was coming from the clock it seems! Picture is bloody brilliant. Might need to tweak the colour a little to get it a little more vibrant, but that's just a tweak of the colour pot. Very pleased with this and I consider my 7800 done now.

 

 

 

 

 

Solid blocks of colour before were borderline intolerable, now they're pretty much noise free.

[juansolo]

Bit of tweaking to get the colour a bit more there, and I think we're done with this one. Cleggy is now grafting in a pair of clocks into the other 7800.

 

No appologies for the 7800 pron attached All taken on the PVM (which is a nightmare to photograph).

 

 

[juansolo]

In Conclusion:

 

The PAL 7800 combined video is a mess. One of the signals is going to suffer at the expense of the other, and if that’s going to be the case, getting the absolute best picture possible out of the 7800 means sacrificing 2600 backwards compatibility. We’re still working on the other one, more to follow on that. For what it's worth, this is by far the easiest mod of the two (though we're working on simplifying the other) and there will always be a compromise trying to keep 2600 backward compatibility given the way the two signals are blended.

 

The Disclaimer:

 

1) This is going to break the 2600 side of the circuit (though it is reversible), only 7800 games will work after this upgrade.
2) If you’re not confident in your soldering abilities, don’t do this.
3) I’m using the UAV output board just because it makes life really easy for this particular upgrade.
4) Part numbers are per my Rev C PAL 7800 board, which differs from just about all the schematics out there that I can find… I've also attached a clipping of the video side of things from the schematic for reference. As I say, don't trust the numbers to match up, use the photo and schematic as reference and a multimeter to double check everything.

 

 

 

Step one: Part Pullage

 

1) Unsolder the RF modulator - We’re going for the best output possible here so all this is surplus.
2) Unsolder the U13 - The 14pin IC (73LS32N) above the MARIA.
3) Remove R25 - The top resistor out of the pair that are between the MARIA and the 73LS32N.
4) Remove C54 - The capacitor at the top of the right hand side bank of components above the 73LS32N.
5) Remove R35 - The 2nd resistor down from the same area.
6) Remove R74 - The lower of the two resistors to the right of the MARIA.
7) Remove R41- The 3rd resistor down in the colourburst circuit (the white outlined section on the right of the PCB).
Remove Q9 - The transistor at the top of the colourburst circuit.
9) Remove R66 - The 3rd resistor up in the set between the RF modulator and the cartridge port.
10) Optional - You can pull the inductor L6 and transistor Q8 also if you wish. They were used by the modulator and are no longer required.

 

Step two: Get some bits

 

Get yourself a UAV from The Brewing Academy, a 14pin DIL socket and a CMOS oscillator that runs at 4.433619Mhz (which aren’t as easy to get hold of… FWIW, we got ours from here: https://www.ebay.co.uk/itm/Oscillator-C-Mos-Ttl-100PPM-5V-Value-of-choice-Batch-4-Pre-order-5-7-Days/183606390301?hash=item2abfcb2e1d:g:saAAAOSwe~lcG-T4).

 

Step three: Stick it all in there

 

1) Solder in a socket to U13, the place formerly occupied by the 73LS32N and put the oscillator in there.
2) Run a wire from the left hand pad of R66 to the top pad of R74.
3) Wire in your UAV directly to the pins of the MARIA as follows:

 

LUM0 - Pin 42
LUM1 - Pin 47
LUM2 - Pin 45
LUM3 - Pin 44
CSync -  Pin 48
Chroma - Pin 43

 

Don’t connect T-Col to the TIA.
Wire up the audio as recommended in the UAV instructions.

 

You might want to adjust the 7800 colour using R42, the trim pot nearest the cartridge port. If it’s got goop on it, de-goop it first.

Edited July 7, 2020 by juansolo

[juansolo]

Cleggy has made a point, the title of this thread has become misleading... Is there a way to change it to something like "investigating PAL 7800 picture issues (was 7800 RGB Musings)" or something like that?

[+DrVenkman]

12 hours ago, juansolo said:

Cleggy has made a point, the title of this thread has become misleading... Is there a way to change it to something like "investigating PAL 7800 picture issues (was 7800 RGB Musings)" or something like that?

Tag a mod and ask.

[juansolo]

The other mod is now probably in it’s final form. The biggest remaining problem we had was having the chroma signals from both the TIA and MARIA blending together. After a session probing around we found a way to trigger a CD4066 when it was running either a 2600 game or 7800, switching in only one buffered chroma output (the trim pot on there is no longer required), therefore removing any need for blending. The other thing this mod has done has disconnected the MARIA from the onboard colour burst clock (still being used for the TIA), so it’s now using it’s own isolated oscillator.

 

This is likely as good as it’s ever gonna get. Documenting this might take a while…

 

 

 

 

 

EDIT: Of course we've just found an easier way of doing it by looking at the Longhorn mod...

Edited July 9, 2020 by juansolo