GBS 8200/8220 CFW Project

The place for all discussion on gaming hardware
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

rama wrote:What may be the issue:
We see much more combing in those "not really 480i" titles that are actually 240p games that have been forced to 480i.
If I crank up the deinterlacer filters so they recognize motion in these titles, all the "real 480i" titles are almost purely bob.

So I don't know, I think I'll focus a little more on actual 480i content.
For the converted titles, it may be better to offer a pure bob mode (which is what gbscontrol had before the deinterlacer got first implemented).
Bob looks pretty acceptable on the converted titles, as it seems to flicker less than on real 480i stuff.
Of course, bob has no combing artefacts, since no weave is done at all.

Oh, and pure bob can do scanlines, although it's those interlaced ones. Still looks okay'ish ;p
Ah yes, 240p converted to 480i...I didn't consider that. But, I was thinking after posting, even when in 480p mode, SFAA doesn't look that great. And that explains it.

Great ideas for solutions! I'm still going to try to track down some better 480i 2d content, if it exists.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

One cool game I had in stock is Sega Ages Phantasy Star Generation 1, the modern PS1 remake.
It is all 2D 480 line artwork, displayed in 480i. This allows tuning the thresholds so it doesn't waver in still scenes.

Pretty much any 3D title should work well with the current deinterlacer. Except maybe if it uses yet another interlace method.
Coming from PCSX2, I learned that PS2 games have like a dozen different methods to mess up an image ><
Ryoandr
Posts: 269
Joined: Mon Jun 12, 2017 4:12 am

Re: GBS 8200/8220 CFW Project

Post by Ryoandr »

PS2 has a lot of different interlaced games

low res (240 ?) interlaced into 480i, lot of early titles, CVS2, VF4
real 480i games, many later titles, VF4Evo
480i with a 30fps cap, I can't remember a specific title, but it might be a thing with multiplat FPS, as many came adapted from og Xbox, and this console was 30fps realm.

I can imagine your pain...
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

I've found some titles, including Wii and Xbox original 2D, that I think will give a good additional sample.

It's interesting to see how the process is handling something like Capcom Vs. SNK 2, with 2d sprites and 3d backgrounds. The backgrounds look great, the sprites have noticeable combing. I don't think it's at all bad, just an interesting contrast when you're looking for it.

The Sega Genesis 240p test suite, and possibly the SNES test suite, is/are switchable to 480i mode. Would the drop shadow tests or anything else from the test suite be of any value?

side note: I ordered myself a magewell pro, taking AndehX's advice on ditching the elgato. :p I just checked tracking, and looks like it was delivered today. :D
Ryoandr
Posts: 269
Joined: Mon Jun 12, 2017 4:12 am

Re: GBS 8200/8220 CFW Project

Post by Ryoandr »

CVS2 is different on platform other than PS2. While the sprites are lowres, the display as a whole is proper full res 480i. Only PS2 version displays that weird lowres interlaced mode.
Also for some reason, the K groove Just Defend effect is massive on PS2 only.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

NoAffinity wrote: Would the drop shadow tests or anything else from the test suite be of any value?
Yep, the drop shadow test in 480i shows the full range of the 3D deinterlacer.
Stable lines with no artefact at all means it's 100% weave / full combing if a game would draw such material.
The opposite is full transparent, meaning the deinterlacer is doing 100% bob and the 3D part is effectively off.
In games this will look like gbscontrol did earlier, a constant bob flicker but no artefacts.

These 240p in 480i games would require 100% bob to eliminate all combing, which isn't the case for regular 480i content.
Regular 480i stuff can work with a mix of the 2 extremes, and this is what the deinterlacer is supposed to deliver.
The deinterlacer has special detection for low pixel difference in 2D and 3D, and will choose weave if it detects it.
If the material moves more than the threshold, bob will be selected.
Weave and bob are mixed, even in the same frame by the way. It works on areas around the detected movement. Pretty clever!
Anyway, this doesn't seem to work at all in those special games :/
NoAffinity wrote:I ordered myself a magewell pro, taking AndehX's advice on ditching the elgato.
What are these cards on the 2nd hand market these days? :p
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

rama wrote:
NoAffinity wrote:I ordered myself a magewell pro, taking AndehX's advice on ditching the elgato.
What are these cards on the 2nd hand market these days? :p
Well, I feel like I got a pretty decent deal on "a customer return. Item is LIKE NEW, and has little to no signs of use. All original packaging included. Tested in shop to ensure proper working order. Warranty Included!" with all the accessories, at $201 shipped. Most importantly for my purposes - the break-out cable and full height bracket are included. A little more than I was wanting to spend, but again, getting a good deal justifies a little extra spending. 8)
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

I'll be honest, I've never even heard of Magewell. Needless to say you can't get them over here in the UK :mrgreen:
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

AndehX wrote:I'll be honest, I've never even heard of Magewell. Needless to say you can't get them over here in the UK :mrgreen:
The Magewell's get mentioned a lot in the same conversation as Datapath's, Micomsoft's and Startech's. Confirmed compatibilty with OSSC and 240p input. Seems to be a pretty robust card:

https://www.bhphotovideo.com/c/product/ ... pture.html

Hoping to get some better color balancing and brightness results from a higher end product. 8) 8)
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

i'd love a datapath E1S but those are not available over here either :/
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

12 bit ADCs, wow.
The GBS uses 3 x 8 bit ADCs on the input and has 4 x 10 bit for output.
This is enough to produce high quality conversions, such that I can't tell an RGBHV source is outputting directly, or through the ADC > DAC chain of the GBS.
Not sure what the extra bits are used for. Maybe for some kind of oversampling or so?

The prices are harsh, but well, niche products and all.
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

Here we go, some purely 480i testing, focusing on 2D:

https://youtu.be/9j-mMVm2nPM

I was hoping to tack on some Dreamcast 480p at the end, to demonstrate the CFW doing it, but I get no signal with VGA input. I tried all resolution profiles and pass thru, with no joy. I think we may have talked about this a while back, and the sync->ground resistor might have to be removed for VGA to function? Or maybe that was something to do with component input, can't fully remember, but if there's any ideas why VGA input isn't working, please give some suggestions. To confirm, I took the same VGA output from the DC to the OSSC, and it worked fine.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

I think I've got to implement 3 methods after all, as the best choice really depends on the material a lot.
Anyway, there is a bug in the implementation (already fixed locally).
I'm not sure what the effects of it are exactly, but for one thing, it uses more memory bandwidth than necessary.

The video shows that there's some color space conversion (or manual adjustment) that eats the lowest blacks.
You can see it in the 240p test suite SMPTE color bars test. The 3 vertical black bars on the lower right should show.
I just checked with my SNES, and I definitely send out full range. It should look similar to this:
Image

It's a little harder to see, but also the 2 large white areas in the test picture should display as a different level.
As long as this color space compression is applied, the image won't look correct.

I don't have an OSSC so I can't verify, but possibly the bad conversion happens at the HDMI output.
Maybe the OSSC has color space conversion options? Maybe the capture card has.
Try to get the blacker than black bar showing :)

Edit:
If you can fix it, please tell us what the problem was. This surely happens elsewhere and I'd like to see if I can do anything about it.
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

The issue is likely indeed due to color conversion. The OSSC outputs 4:4:4. The Magewell Pro HDMI confirms 4:4:4 is incoming. However, youtube compresses to 4:2:0. So, I set the magewell to spit out NV12, which is also what youtube likes if what I've read from multiple sources is accurate.

My thought process is: better to handle the process myself than to leave it up to youtube.

But, I will give a test with a 4:4:4 video format and get a screenshot. For youtube upload, I think what I've provided is the best we're going to get, unless somebody knows better or has some advice for overcoming that limitation.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Hm, I don't think the color subsampling (or lack thereof) should affect these luma levels.
Do you have the full range in your raw footage?

Edit:
Still using OBS? Could you check your ini file and see if the range is set to full? (Also try to swap it, maybe).
From their forum:
Go to %appdata%\obs-studio\basic\basic.ini (paste it in to start/run or windows key + R).

Then for example, in the video section:

[Video]
ColorSpace=709
ColorRange=Full
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

rama wrote:Hm, I don't think the color subsampling (or lack thereof) should affect these luma levels.
Do you have the full range in your raw footage?

Edit:
Still using OBS? Could you check your ini file and see if the range is set to full? (Also try to swap it, maybe).
From their forum:
Go to %appdata%\obs-studio\basic\basic.ini (paste it in to start/run or windows key + R).

Then for example, in the video section:

[Video]
ColorSpace=709
ColorRange=Full
Yes, using OBS. What you see on youtube is pretty spot-on to what the raw footage looks like. I did set color range to 'partial' within OBS...another tip I found a few places, for encoding for youtube upload. There is an option to limit color range on the output from the Magewell. I'm wondering if it will do internal correction so that the full range is represented across a limited range (is that even possible?). I will play around with that tonight also.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

NoAffinity wrote:Yes, using OBS. What you see on youtube is pretty spot-on to what the raw footage looks like.
Okay, this means the luma problem is before youtube or any further conversions.
It's likely happening at the HDMI stage. Either the OSSC or the capture card (or both) need to be set to full range at the HDMI stage.
(Output for OSSC, input for capture card :p)

Once the levels are clipped to limited, these levels are gone and can't be restored.
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

rama wrote:
NoAffinity wrote:Yes, using OBS. What you see on youtube is pretty spot-on to what the raw footage looks like.
Okay, this means the luma problem is before youtube or any further conversions.
It's likely happening at the HDMI stage. Either the OSSC or the capture card (or both) need to be set to full range at the HDMI stage.
(Output for OSSC, input for capture card :p)

Once the levels are clipped to limited, these levels are gone and can't be restored.
The OSSC only outputs full range, marqs has confirmed that multiple times. The Magewell confirms it is receiving full range. The output from the Magewell, for these particular clips, was also set to full range. Possibly setting OBS to 'partial' is affecting the input and not the output, causing a mismatch.

I think the best option will be letting the Magewell handle the conversion, but again, I'll do some testing tonight, with the SMPTE color bars displaying and see if I can't find the magical combination.
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

rama wrote: I just checked with my SNES, and I definitely send out full range. It should look similar to this:
Image
Here's my SNES color bars. Going through my PEXHDCAP (color range set to bypass on capture card and Full in OBS. Color space set to 709)
Image
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Okay, this is how it should look.
The ADC gain is a little low (weak max white), but that would be fixed with auto ADC gain.

NoAffinity: Good luck! :)
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

ah yes ADC gain. I forgot about that! I was using my capture card to adjust the brightness. Thanks for that.
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

rama wrote: NoAffinity: Good luck! :)
Thanks! It's also very possible I adjusted something manually that obliterated the "lighter shades of black". I was using standard color bars to fine tune, but will use the SMPTE bars moving forward.
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

rama - any thoughts on VGA->VGA and not getting anything from Dreamcast through the GBS?
User avatar
AndehX
Posts: 790
Joined: Sun Oct 18, 2015 11:37 pm

Re: GBS 8200/8220 CFW Project

Post by AndehX »

I'd imagine you just have to enable passthrough?
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

AndehX wrote:I'd imagine you just have to enable passthrough?
I tried all profiles and passthrough. OSSC never reported getting a singal from GBS.
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Oh, I forgot.
Yes, you need to remove the 100 Ohm resistor for (most) VGA sources (it depends on the source's drive strength).
They're TTL and meant for ~ 1kOhm termination. The GBS installed resistor is already enough.
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

rama wrote:Oh, I forgot.
Yes, you need to remove the 100 Ohm resistor for (most) VGA sources (it depends on the source's drive strength).
They're TTL and meant for ~ 1kOhm termination. The GBS installed resistor is already enough.
That's right...my memory is refreshed. I was going to install a simple on-off switch on the resistor for this purpose but never got around to it. Maybe now is the time.
Higgy
Posts: 54
Joined: Thu Mar 15, 2018 1:29 pm

Re: GBS 8200/8220 CFW Project

Post by Higgy »

Rama - what voltage rating for the 10/22uF caps and the 1nF please? We can get those values added to the Wiki

I was also going to ask about size. I used 0805 for the 1nF. Would 1210 be ok for linking parallel to C23 C41 C42 C48 ?

Thanks

(and am I right in thinking that a higher voltage rating does not matter? I know that higher the voltage rating meant they were physically bigger on electrolytic capacitors at least)
rama
Posts: 1373
Joined: Wed Mar 08, 2017 3:15 pm

Re: GBS 8200/8220 CFW Project

Post by rama »

Let's settle on 10uF for the large local storage caps, as more capacity seems to be at the upper end of current market offers.
We want as high a voltage rating as possible, but that increases size. We want to limit size to 0805, so they can still be soldered besides the 0605 stocks.
Ebay has some cheap listings for:
0805: 10V
0605: 6.3V
Actual voltages are 3.3V for the main supply (C23 C41 C48), and 1.8V for the digital supply (C42).
By the way, I just noticed that on some GBS board designs, C41 is more crowded. It can be replaced by C43.

The small SOG coupling caps of 1nF all come at voltage ratings much higher than required.
Here it's better to focus on size (0605 preferred, 0805 okay), and their material (NP0 / C0G preferred).
User avatar
NoAffinity
Posts: 1018
Joined: Mon May 07, 2018 5:27 pm
Location: Escondido, CA, USA

Re: GBS 8200/8220 CFW Project

Post by NoAffinity »

I ordered some of these, 10uf 0805 6.3v. A good option for those of us in the US. :)

https://vod.ebay.com/vod/FetchOrderDeta ... .S1.R1.TR1

rama - can you please clarify this. I don't quite understand.
rama wrote: By the way, I just noticed that on some GBS board designs, C41 is more crowded. It can be replaced by C43.
Post Reply