GBS 8200/8220 CFW Project

The place for all discussion on gaming hardware
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Rodrik wrote: Back on interferences, I will try to fix this V3 GBS board following your inputs whatever time/skill it requires, it's more about the journey than about the result and/or cost :D I like modding things.
Nice, I like this attitude best!
This is how I got into my modding projects :)

When I got my first gen GBS-8220 (had to look hard for it!), I initially improved the Vcc stability as usual.
Removing C11 turned out to be very important.
The other tweaks (10uF and 22uF SMD caps) had the usual effect of smoothing out Vcc on the scope but not really making a difference in picture quality.
There was always that mild interference pattern, no matter which source / preset was active.
Suspecting the video driver chip, I turned to the ISL59833 datasheet: https://drive.google.com/drive/folders/ ... GFOMUFVdlE
Sure enough, it is an odd IC that generates noise as part of its operation. (These 2 things usually bite each other.)
If your remaining noise fits that description, then the fix is working on R58.
A small power bead slows down the current spikes the charge pump can generate. A reservoir capacitor should be used to deliver the current instead.

As for a replacement GBS, most designs are the same at the core.
The ones with yellow buttons, often with a "V3" or "V3.5" mark, appear to be the newest design.
They do require the 10 / 22uF caps and C11 removal but then deliver a great picture. They're also the cheapest versions currently available :)

The HDMI version is also very interesting, if your display accepts HDMI over DVI (I found the ratio of displays that support it vs no support is 50/50).
If it is supported, the HDMI version is more consistent when tuning for "best" picture quality / sharpest pixels.
Rodrik
Posts: 18
Joined: Tue Jul 03, 2018 12:28 pm

Re: GBS 8200/8220 CFW Project

Post by Rodrik »

rama wrote:Nice, I like this attitude best!
This is how I got into my modding projects :)
Thanks, it's also how I built an Internet Radio some years ago around a PI-II with a tactile interface :)
rama wrote:When I got my first gen GBS-8220 (had to look hard for it!), I initially improved the Vcc stability as usual.
Removing C11 turned out to be very important.
The other tweaks (10uF and 22uF SMD caps) had the usual effect of smoothing out Vcc on the scope but not really making a difference in picture quality.
There was always that mild interference pattern, no matter which source / preset was active.
Suspecting the video driver chip, I turned to the ISL59833 datasheet: https://drive.google.com/drive/folders/ ... GFOMUFVdlE
Sure enough, it is an odd IC that generates noise as part of its operation. (These 2 things usually bite each other.)
If your remaining noise fits that description, then the fix is working on R58.
A small power bead slows down the current spikes the charge pump can generate. A reservoir capacitor should be used to deliver the current instead.
I will do or at least try to do it. Parts are ordered (22uF SMD caps, 120 OHM Chip EMI Ferrite bead, some 0.22uF SMD caps and some Tantalum Capacitor 16V 220 uF)
And I am trying to desolder the trimpots without luck for the moment, but I miss a manual pump.
rama wrote:As for a replacement GBS, most designs are the same at the core.
The ones with yellow buttons, often with a "V3" or "V3.5" mark, appear to be the newest design.
They do require the 10 / 22uF caps and C11 removal but then deliver a great picture. They're also the cheapest versions currently available :)
Found a "v3.5" for a bit more of 18€ and ordered it.
rama wrote:The HDMI version is also very interesting, if your display accepts HDMI over DVI (I found the ratio of displays that support it vs no support is 50/50).
If it is supported, the HDMI version is more consistent when tuning for "best" picture quality / sharpest pixels.
Found one for a bit less than 30€, but will test my screens first.
I like those Dell 1908FP, I have 2 of them. If I test with a PI and a DVI/HDMI cable and it works, will this mean it will also work with the HDMI GBS?

Thanks again for your support and your work on GBS control, I love it :mrgreen:

Rodrik
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

I'm not sure what the requirement for HDMI over DVI is that some monitors have and others lack.
Chances are good that when one device works with it, the GBS will as well.
The HDMI transcoding looks to be pretty minimalistic (which is fine).

That 3.5 model is good. Just remember to get rid of C11 and add a few more of those 22uF caps :)

Removing the 3 trimmers can be annoying. Sometimes I just go medieval on them, with a screwdriver.
Just have to be very careful not to damage the PCB and especially the traces.
If the traces do get damaged, a repair isn't too hard. It'll just look ugly ;p

Newest development on the research front:
It's possible to enable bypass mode for all inputs (not just RGBHV). In this mode, the chip performs sync separation and produces H + V sync.
The video is sampled on the ADCs, goes to the "decimators" (data reduction units that enable oversampling support) and then directly out to the DAC.
In effect, this allows a real "Multisync" monitor to display the content without any format conversion.
By chance, my NEC EA190M supports this mode (as 712x242 / 15.7kHz).

It's quite possible that many more displays actually support these modes, but no one ever tests them!
It's hard to test because of the need to separate csync and build adaptors for the cables ;p
User avatar
donluca
Posts: 852
Joined: Sat Feb 28, 2015 8:51 pm
Location: Italy
Contact:

Re: GBS 8200/8220 CFW Project

Post by donluca »

rama wrote:Removing the 3 trimmers can be annoying. Sometimes I just go medieval on them, with a screwdriver.
ROFL

Pls don't *ever* do that, get a nipper and cut the legs, then proceed to remove what remains of the leg with the soldering iron and a plier.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

The pot legs are near impossible to reach with a clipper.
Desoldering them is tricky because of the thick PCB with lots of thermal mass.

I haven't found an easy method to do it yet.
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

I got them off pretty reliably with my hot air station
Rodrik
Posts: 18
Joined: Tue Jul 03, 2018 12:28 pm

Re: GBS 8200/8220 CFW Project

Post by Rodrik »

Trimpots removed!

Was not easy and as you can see I did damage the trace for the Blue, fixed it later by a small wire to L3/R17.

Image

I am now waiting for the SMD components to progress more because I still have very visible interferences.
Rodrik
Posts: 18
Joined: Tue Jul 03, 2018 12:28 pm

Re: GBS 8200/8220 CFW Project

Post by Rodrik »

Also, would it be possible to make a 1280x1024 output to match 5/4 screens?
I think this would allow to keep the 4/3 aspect ratio by putting a small black border on top and bottom.
Is that makes sens or I miss something?
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Honestly, the trimpot area looks the same on most of my boards. Always one or the other trace is ripped.
It's an easy fix, so congrats :p

The x1024 resolution is important for even scanlines and vertical scrolling.
I'll surely try and cram it in there some day.
It may allow for some aspect ratio tricks as well, I don't know.
User avatar
Syntax
Posts: 1774
Joined: Wed Aug 09, 2017 12:10 am
Location: Australia

Re: GBS 8200/8220 CFW Project

Post by Syntax »

Don't see how you guys manage to damage the board removing the pots.
I just cut the output leg and pull the pot forward to cut the input and ground leg. Then desoldering is a breeze.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Got a pic of your cutter tool, Syntax?
User avatar
Syntax
Posts: 1774
Joined: Wed Aug 09, 2017 12:10 am
Location: Australia

Re: GBS 8200/8220 CFW Project

Post by Syntax »

I really should of taken some pics when i tackled this today.
I just used regular flat/side cutters.
I cut the plastic top off the pot first, then did a vertical snip on the output leg.
This allows the top of the pot to peel off and you can grab what's left easily with the snips, pull it forward which gives you room to cut the other two legs.
I was never worried about damaging traces as the legs are left encapsulated this way, and heating with an iron from the other side allows them to simply fall out.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

I'll try this on my next board.
I don't have a flat cutter and good quality tools are expensive. Guess some day I'll have to invest.
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

User avatar
Syntax
Posts: 1774
Joined: Wed Aug 09, 2017 12:10 am
Location: Australia

Re: GBS 8200/8220 CFW Project

Post by Syntax »

Yep those at that price are perfect.
When they get dull sharpen them on a stone.
Buy 2, one for cutting wire insulation and one for cutting everything else. The later will end up pretty chipped after awhile.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Passthrough mode used to be ADC to DAC directly, so no fancy processing was possible.

Today I got it working through the IF and VDS units.
This means, besides other cool things, that the GBS will be an SD + HD YPbPr to RGBHV converter.
A quality one :)

Edit:
So basically, the new mode just bypasses the framebuffer. No memory used, no memory controller noise necessary (clock can be turned off).
The source's timings are used directly, so this is more useful for HD (640x480 and up).
My test display supports SD as well, and the new mode works with it.
In HD modes, it's a color space converter + image positioning tweaker + anything to anything adapter ;p
Ryoandr
Posts: 269
Joined: Mon Jun 12, 2017 4:12 am

Re: GBS 8200/8220 CFW Project

Post by Ryoandr »

Wait, so it could do SD passthrough ? Sync cleaning for those troublesome upper skew troubles ? That would be amazing.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

I suppose so :)

Video timings are in sync lock mode then. What goes in, comes out (but it's possible to do a little processing, like fixing sync issues).
The video is being digitized and can be converted to anything else. Component to RGB or back, color changes, green / purple tint fixes, etc.
The ADC conversion is of pretty good quality, if the parameters are known or can be assumed reliably.
The DAC is excellent as well. It doesn't even require an extra buffer and can drive good quality RGB directly.

With the price of these boards and this new flexibility, this allows for some crazy stuff.
For example, when RGB modding a console, it may be more economically viable to install a full GBS instead of an RGB bypass board! :p
Just remove the RGB encoder, wire the SNES PPU output to the GBS, tap the dot clock while we're at it for some extra quality, out comes pristine RGB over VGA.
No expensive SNES cable required even ;)
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Hmm, the devil is in the details again. There are more issues with mode switches between interlaced and progressive.
Guess I don't know yet how good it'll be.
I'll push the preliminary code soon.
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

So I still have this weird shimmering/speckle interference. I dunno if you can pinpoint what's causing it Rama? I've just added the 5 SMD caps and the cap across the buck converter, but I still have this weird issue.

https://www.dropbox.com/s/jsv7jwnabh1fp ... 1.mp4?dl=0

Do you think it could be the metal heatsink? I'd like to replace it with a ceramic one, but the metal one is stuck on there pretty good.

Edit: ok so I manage to get the heatsink off, and well... no expense spared on the thermal adhesive...

ImageImage

I switched the GBS on for a few seconds just to see if the interference was gone without the heatsink there, but no, it's still the same.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

It looks like sync jitter, which I'm working on all the time. It might be a recent change to the default values I did.
Are you using the dev branch?

Connect a terminal and send a few 'v' commands while watching the image. Does it improve?
The command cycles through the 32 Sync Processor phase adjust steps.

There also seems to be a memory bandwith issue in the background.
Try sending 's4s2cs10' and vary the '10' a bit. The default is '1e' and you might get better results anywhere between '0b' and '2e'.
This is the memory playback FIFO delay, the most important factor for bandwith.
If this helps somewhat, but you just can't find a perfect value, use the hammer: Switch memory clock frequency :p

Oh, and there seem to be a lot of solder blobs on your board that need to be removed!
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

rama wrote:It looks like sync jitter, which I'm working on all the time. It might be a recent change to the default values I did.
Are you using the dev branch?
No, im using a build from a month or 2 ago.
rama wrote:Connect a terminal and send a few 'v' commands while watching the image. Does it improve?
The command cycles through the 32 Sync Processor phase adjust steps.

There also seems to be a memory bandwith issue in the background.
Try sending 's4s2cs10' and vary the '10' a bit. The default is '1e' and you might get better results anywhere between '0b' and '2e'.
This is the memory playback FIFO delay, the most important factor for bandwith.
If this helps somewhat, but you just can't find a perfect value, use the hammer: Switch memory clock frequency :p
I'll give all that a try, although I have no idea how to change memory frequency
rama wrote:Oh, and there seem to be a lot of solder blobs on your board that need to be removed!
it's silicon from the heatsink :P

I have a ceramic heatsink on order which should in theory perform better, even if it doesn't have an effect on interference.
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

Ok, I tried the various commands you mentioned, and none of them made any noticeable difference
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Hmm, that's odd that nothing affected the issue..
The memory timings are available on the web ui, or by sending 's0s40s2c' where your options are '0c' to '7c'.

It doesn't look like an interference pattern, definitely has a jitter sync look to it.
Are you using an LM1881? If so, is it being powered by 3.3V or 5V?
3.3V is a much preferred construction, since this is available on the GBS.
The drawback is that the chip is specified for 5V or more. At just 3.3V, the sync flanks can be slow, especially the falling flank.
If you manage to build a 5V variant, the CSYNC output will be too strong for direct use, requiring level translation.
The simple fix is to use a series resistor, but this slows down the flanks again.
Quite tricky, all this :p

I want to build some kind of MCU sync stripper that fixes these issues.
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

I fixed it. It was being caused by my BVM lol. I had the GBS connected to my BVM's output's. I have it connected directly to my gscartsw now and the speckle interference is gone. Stupid Sony.
User avatar
Syntax
Posts: 1774
Joined: Wed Aug 09, 2017 12:10 am
Location: Australia

Re: GBS 8200/8220 CFW Project

Post by Syntax »

The Gscart regenerates/cleans sync
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

It was going through the gscartsw before, it was just going through the BVM too. Now it's only going through the gscartsw.

Any ideas when/if you'll be adding 1080p support rama? I can't help but feel like the only way to get a razor sharp pixel image is with native resolution output
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

1080p input?
That has a good chance of working with the new sync lock mode.
I already have 1080i working, and this mode doesn't require scaling or memory to work.

Regarding sync, every converter or processor changes the timing relations.
It might be that the BVM adds a digital step. Then it can sample the source at bad times and cause jitter.
This is the same issue that I'm trying to solve on the GBS as well.
Thankfully, there seems to be just one really bad spot for most sources, at around 150 degrees (level 28 of 32).
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

I meant 1080p output lol. We have 640x480 and 1280x960, but an output native to the display it's connected to would be preferred as i'm sure it would result in sharper pixels.

But yeah, the BVM just introduces more variables to the quality, so I'm just joining the 2 VGA cables together with one of those VGA joiner things, instead of "joining" them together via the BVM.
jbilander
Posts: 19
Joined: Sat Aug 11, 2018 8:12 am
Location: Stockholm

Re: GBS 8200/8220 CFW Project

Post by jbilander »

rama wrote:
So basically, the new mode just bypasses the framebuffer. No memory used, no memory controller noise necessary (clock can be turned off).
The source's timings are used directly, so this is more useful for HD (640x480 and up).
My test display supports SD as well, and the new mode works with it.
In HD modes, it's a color space converter + image positioning tweaker + anything to anything adapter ;p
Hi,

I'm new to this forum but have been following this thread on a distance. Thanks for all the great work you put in. The news about the new passthrough mode caught my interest...

Would this allow running the Amigas non-laced 15Khz signals untouched as passthrough to the monitor (effectively being like a simple RGB->Vga adapter) whilst for laced modes let GBS handle deinterlacing and pass that deinterlaced signal on to the monitor? That would be great. My monitor Benq BL912 (1280x1024) syncs great with Amigas 15KHz 50Hz PAL non-laced signals through a simple RGB->Vga adapter but flickers a lot in PAL High res laced (640x512) which is the mode I would prefer to use when running classic workbench.
Post Reply