GBS 8200/8220 CFW Project

[rama]

If you find out what the problem was, please let me know.
Best of luck!

[Ryoandr]

Obviously when it happened what I said, I tried to connect to http://gbscontrol.com

Be careful as some browsers tend to autocomplete and add the .com if you just type gbscontrol
IIRC on my phone I have to fully write

Code: Select all

http://gbscontrol

to not trigger autocomplete

[NoAffinity]

Not sure if you guys are using your phones to access, or trying to do it from a desktop, but I've found direct-connecting to the web ui via phone is the easiest way. On my network, my wireless router in my work area issues IP addresses in a different range from the hard-wired network, so I can't get to a wireless device (aka GBS/ESP) from a hard-wired desktop.

But, whenever the GBS is powered, I simply turn off wifi on my phone and turn it back on, and the phone automatically connects to the gbscontrol wifi network. I have a shortcut setup on my home screen for the gbs web ui. So, the process is simply:

Power on GBS/ESP
Disable->enable wifi on phone
Tap on shortcut to CFW web ui

[Shiver_169]

Hello again.

Well I have already tried with what I have been told and I definitely think that the problem is in my home network or that there is a lot of electronic interference at home. And is that in the office after what I told them connect again the ESP8266 and I could see the network "gbscontrol" and connect to it, and as I understand if the data had already been stored in my home network this was no longer possible. I'm thinking maybe the network that I have at home is a problem and that is that I used a network through my cable company and I do not have a modem as such to connect the signal box generates wifi signal and then I connect my devices. I do not know if this is affecting the not being able to connect to ESP8266 although it is supposed that the network that this emits should be sufficient. I hope you can support me why I just do not know what to do anymore.

[AndehX]

Rama, could you maybe update your install guide with some of the newer hardware modifications on the last few pages?

[rama]

Shiver_169:
Do what NoAffinity suggested.
Wipe your WiFi settings once more, then simply don't connect the GBS to your home network.
The device will stay in AP mode, providing the 'gbscontrol' network.
You can use your mobile phone to configure settings.

AndehX:
When I have time for it.

[Shiver_169]

Shiver_169:
Do what NoAffinity suggested.
Wipe your WiFi settings once more, then simply don't connect the GBS to your home network.
The device will stay in AP mode, providing the 'gbscontrol' network.
You can use your mobile phone to configure settings

Ok I tried again this time with a little time at home what I did was install everything in my home computer and re-install the ESP8266 and did what they told me I did not connect it to my network I just tried to find the gbscontrol network and it worked very well all. I really appreciate your help.

Now, since this is excellent, I would like to ask you if it is possible to connect a VGA to HDMI converter to the GBS output or if you have already tested the CFW on the board that converts to HDMI and how does it work there?

[rama]

VGA to HDMI converters work perfectly, yes (both, the HDMI GBS8200 version and a regular board with external converter)
Since this question is asked a lot: Yes, the HDMI delivery means that many TV now display at 16:9 by default.
Unless the TV has an override to 4:3 mode (most do), the image will be 16:9 and there's nothing gbs-control can do about it.

[NoAffinity]

Hey rama,

Just a quick circle-back on testing component in -> component out. I'm sure I was doing something wrong when testing it before, as it's working perfectly now. I fed some PS2 480i into the GBS, and some pretty good 480p and 720p came out. Pass through didn't work, and the 1280 profiles didn't work, but the 1280's not working is likely the shortcoming of the one HD TV with component input that I have access to and was testing with.

Just wanted to close the loop on that, as the issue was clearly on my end.

The deinterlacing doesn't look too bad either. Playing some Hyper Street Fighter 2 and Outrun 2 Coast to Coast, it was very good. Playing Capcom Classics Collection Vol. 2, the quality suffered, noticeably heavy combing on drop shadows...but this collection does a pretty terrible job of providing good aesthetics from its ports. Even the Xbox version, set to 480p, going through OSSC or direct to HDTV, doesn't look great.

[rama]

NoAffinity:
Maybe it was one of the later fixes that made it work

I fed some PS2 480i into the GBS, and some pretty good 480p and 720p came out. Pass through didn't work, and the 1280 profiles didn't work

Ah yeah, I almost expected that.
The 1280x1024 preset is very close to "1080p", having 1066 vlines, compared to 1125 for the HD preset.
So all I do right now is take the x1024 preset and simply bump up the vlines to 1090 or so.
Any more than that will require work on the scaling and such. I will get to that eventually, but for now it's the 1090 lines hack :p
It coerces my test TV (even at 1082 vlines) into 1080p mode just fine, but I suppose yours is a little more strict.

Pass through for Component is hard, unfortunately. It doesn't work the fast and easy route that I have for VGA output.
It's possible though, just needs a lot of effort on my side.
For the time being, it's probably best to stick to 1280x720. That mode should work on all TVs, no matter if VGA or Component.

The deinterlacer is probably maxed out right now. I also think it looks pretty good with most content, but yeah, drop shadows are fail :p
I don't have any of the other "popular" deinterlacers here, but I'd love to see one of those in action.
I think that just by seeing how they perform, I may be able to guess how they work.

AndehX:
I've got the SOG coupling caps done
https://github.com/ramapcsx2/gbs-contro ... placements

[NoAffinity]

I don't have any of the other "popular" deinterlacers here, but I'd love to see one of those in action.

Would some captures from a VP30 be helpful? If so, let me know what content would be most useful for you.

[AndehX]

AndehX:
I've got the SOG coupling caps done
https://github.com/ramapcsx2/gbs-contro ... placements

Cool good stuff. Now I understand what's going on. That being said, is it still worth changing these caps if you only use RGBS and not Component? or is this something that only affects SOG?

[Ryoandr]

C35 is for sync in RGBS, it allows to work with Composite Video or Luma as Sync without the need for an external sync stripper.

[rama]

Yep, only relevant for inputs. All source formats except RGBHV (VGA from a PC) benefit from the capacitors swap.
One of the caps is for the 3 RCA input, the other is for RGBS.

NoAffinity:
Not sure, it's worth an attempt!
I would like to see a 240p test suite drop shadow run (with the Motoko background).
Best would be a linear up <> down motion of the flickering shadow at various speeds, then some diagonals.
This test should show how deep the deinterlacer's buffers are, and whether or not it shows artefacting (one backbuffer).
Best (for us) would be if it shows artefacting ;p

Also, a short clip of the main menu would be nice, just selecting different items.
This produces a very slight jumpiness of the selected item text with the current GBS deinterlacer.

[NoAffinity]

Okay, will see what I can pull together. I've got the fast deinterlacer card in the VP30 currently, which impairs its ability to handle 240p (it treats it as 480i). I will capture with the deinterlacer card isntalled, and then remove it and get some additional footage without it.

This it the Mokoto background?

[Pilot001]

Hi all !

maybe this is kind a debugging.. maybe not.


So...
I have a machine with Coputer controll (CNC) from 1989.
Using a CRT , EGA conector.
I checked the V and H Sync s . 60Hz and 14,5 Khz (or 15KHz)
Had to add some ICs to get even a scren from the GBS 8200.
Now it seems kind a OK but aorund 1/10 of the screen is missing.
Tried in all the resulutions (640 * 480..... 1024*768) - still missing it.
If i move the pic on the GBS (H position) i can see the missing part .. but either the ledt side or the right side is missing.
So the frame GBS is sending is smaller then the actual one (can not get it on the screen even with H SIZE)

Here are some Pic

[rama]

Plug and Play deinterlacing cards? Times have changed :p

That's the background, yep.

[NoAffinity]

Here's the various VP30 w/ ABT-102 deinterlacing modes with SNES 240P and 480i source. I tried without the deinterlacer card installed, which at first was a hassle and required a hard reset to get the VP30 working properly, and then didn't capture right once it was working properly.

I'll revisit this effort without the ABT-102 installed, tomorrow, but for now here is the footage with it installed:

https://youtu.be/MbdoMn3QG6w

Once 1440p mode is available, this is the best way to view it. Youtube significantly reduces the butchery of 1440p.

[rama]

Great, this already gives some nice insights

This deinterlacer seems to use motion adaptive for the "film" and "game" modes.
The technique switches between bob and weave deinterlacing based on detected motion in time (fancy! :p).
It causes the artefacts in these specialized tests, but looks quite nice in 95% of cases.

The 2:2 modes, I'm not sure what they do there.
Maybe it's just the same deinterlacing settings, but with motion adaptive disabled (and favoring weave)?

In any case, it looks like these 2 basic setups are what we can choose from.
The motion adaptive line delay settings are better optimized on the VP30. I think I can copy that

[NoAffinity]

Cool, glad it helped. The results from non-deinterlacer-card scenario, which didnt record well, nonetheless looked identical. There is only video and film bias mode when the deinterlacer card is not present and the results look the same as other modes with the card present. That said, I'm not sure theres any value seeing what it looks like without the card, but if you feel it might add some insight, let me know and I will get some more footage.

Sent from my SM-G955U using Tapatalk

[rama]

The VP30 2:2 mode looks pretty solid in this test, but I already know how bad this'll look in motion (scrolling, etc).
So the only viable options are motion adaptive deinterlacing, or "none" deinterlacing, which is just OSSC style bob.

For the motion adaptive mode, it looks like the VP30 doesn't have (use) what the TVia guys called the "feedback bit".
If you want, could you record a round of an interlaced fighting game?
It doesn't matter if it's "game mode 2" or "video / film bias mode", any of those (but game mode 2 should be best).
I know the looks of Dead or Alive 2 or Tekken 3 best

[NoAffinity]

okay, will do. I do have Tekken 3 on PS1, so will get that. Would you prefer in-game play, attract mode, or doesn't matter?

[rama]

Any ingame mode is fine, as long as it's 2 fighters with lots of motion. Thanks

[NoAffinity]

Here you go, rama. I didn't do the 2:2 modes, but got all the others. There seems to be a good amount of action going on in attract mode. Hopefully it works for you.

https://youtu.be/oHbLGctd2jA

[rama]

Well, now I'm confused :p
Their adaptive mode clearly produces artefacts in the 240p drop shadow test, but it works flawlessly on Tekken 3.

With my tests, I always have to balance weave artefacts getting through vs "jumpy" static material.
Even when focusing on reduced weave artefacts, I probably can't reach the VP30 Tekken 3 performance.
So it must be more than just a small tweak. The VP30 probably does some more tricks.
I'm still pretty sure that the 5725 has all the required hardware though. It's just missing that perfect config ;p

[NoAffinity]

Did you also notice it when the test suite was switched to 480i output? If you saw it only when the snes was in 240p mode, and not when snes was in 480i mode and not at all with tekken (480i from ps1), then I would spspeculte that its more an issue with how the vp30 mishandles 240p once the deinterlacer card is installed (known shortcoming).

Sent from my SM-G955U using Tapatalk

[Ryoandr]

Are there any games that actually use flickering shadows in 480i ? Not counting ports on 480i-only consoles.

[rama]

Did you also notice it when the test suite was switched to 480i output?

Oh yea, the first part of the video was in downgraded video. I noticed that
With gbscontrol, I simply look at the filtered line count per frame, as decoded by the chip.
If it's an odd number of lines, it indicates regular 480i. If the line count is even, it indicates that the source messed with the odd/even indicators (ie: 240p hack).
It's a simple idea. I tried it and it instantly worked. Never had to change it
https://github.com/ramapcsx2/gbs-contro ... .ino#L3906

@Ryoandr
Oh, I'm sure there are some games that do it. Developers back then didn't care, probably because such content would still look fine on a CRT.

[NoAffinity]

Did you also notice it when the test suite was switched to 480i output?

Oh yea, the first part of the video was in downgraded video. I noticed that
With gbscontrol, I simply look at the filtered line count per frame, as decoded by the chip.
If it's an odd number of lines, it indicates regular 480i. If the line count is even, it indicates that the source messed with the odd/even indicators (ie: 240p hack).
It's a simple idea. I tried it and it instantly worked. Never had to change it
https://github.com/ramapcsx2/gbs-contro ... .ino#L3906

Would you says this is an improvement to how gbs cfw was previously handling deinterlaced sources?

I'll update mine, and give it a test this weekend.

[rama]

I haven't changed anything yet.
This was just showing the code that figures out whether 240p or 480i handling should be used