OSSC (DIY video digitizer & scandoubler)

[FBX]

From what I read, they are down to 'bytes' left on the memory available to add new features. So this would be difficult to implement. However, I have a possible idea:

I assume firmware releases assign default settings, correct? Wouldn't it then be possible to make custom firmwares that assign custom defaults for each of the save slots? So for example:

I could do up profiles for Genesis, SNES, Saturn, Neo Geo, Playstation, etc, and have them as firmware default settings in slot 0, 1, 2, 3 ,etc. All people would need to do is download a firmware of their choice (it would contain a combination of console defaults on the save slots), and then flash it to the OSSC like we currently do with new firmeware.

Would need marqs to comment on the logistics and possibilities of this.

[EnragedWhale]

Doesn’t running the OSSC into the framemeister introduce the usual 2ish frames of lag that the FM puts into analogue sources?

Yes it does but for many people that delay is unnoticeable, especially with a low latency TV/monitor.

I second this. My W series Sony that is regarded as great for gaming simply doesn't have the compatibility with the OSSC's higher modes. With this method, I can have the best of both worlds! If FBX can make it as easy to use as he did for the Framemeister, that's a game changer for me.

I used to have a W6 Sony and used a DVDO edge to increase compatability with the OSSC as I didn’t think I could do better for largish, reasonably priced retro gaming display. After running into lag problems with certain systems I bit the bullet and bought a modern 4k Samsung and haven’t looked back. It has at least x3 compatability with all systems and whilst maybe not as sharp as my old Sony using FBX’s integer scaled FM profiles in screenshots, in motion it looks far better than the smeary mess the W6 used to turn into playing fast scrolling games like sonic etc. Now I just use the ossc direct to the tv for everything except the PS2 (I use the FM as it’s 99% 576i pal games) and enjoy the snappy response and detailed motion.

Long story short, the W series Sony’s aren’t the be all and end all for upscaled retro gaming I used to think they were.

This is all in my (idiot) opinion btw

[Xyga]

Long story short, the W series Sony’s aren’t the be all and end all for upscaled retro gaming I used to think they were.

They were, the 2013~2014 models ending with 'B', for a long time the only good low lag large displays available and they worked very well with an XRGB using a transcoder or a XRGB+DVDO combo still flat-out beating the Mini lagwise in both scenarios (assuming not the EDGE since it's the one that unlike the VP series produces lag with several systems)

note: In fact the Sony W of that time are still quite great today, why wouldn't they be? there aren't many flat panel displays doing so many things right all at once, believe me. And I'd take the scaling and PWM-free backlighting of the Sony's over anything Samsung anytime. Some Samsungs are quite good in some departments ok, but still not quite at the level of Sony in terms of combined accuracy and performance.

Then, the OSSC descended from heaven, and it happened that the W series don't like the odd signals it produces. Damn. But who would have anticipated that retrogamers would come to crave a multiplier device feeding mostly non-standard signals to a flat panel? x3, x4, and x5 were not exactly expected to work on any display just like that, actually we were happy that an accessible 'DISPL-like' device was coming like a savior.
I mean before the era of the OSSC, everything I've thrown at the W6 using a DISPL - that doesn't do any signal conversion, only doubling the source resolution kinda like the OSSC - worked like a charm and even better than with the XRGB2 and 3 that are supposed to work closer to the standards.

That 4K displays today are generally more compatible with the OSSC's odd resulting signals is good, but how many modes will be supported by a set exactly is always unknown at the time of the purchase and until the actual experiment. Flat panels displays - TVs and monitors - are still black boxes in that department.
I say it all the time but again never fall for brand fidelity, you can like a manufacturer all you want for some particular aspects, but nothing guarantees that they won't let you down with your next purchase. Samsung in particular are known for changing and redesigning a great deal of their set's built-in electronics and firmwares quite often, and compatibility can also change when that happens (I think it was 2015 to 16 that saw radical changes in compatibility areas for at least two major series)

Which reminds me; Sony who more rarely change the basics of their hardware/software (besides bringing Android to several series) have brought back their old Linux engine, maybe the same they were using with the good W's, to a number of their low-end Full-HD and even 4K it seems. Are they really the same though? impossible to tell since nobody will review the sets, especially not the FHD ones. If it's the same OS as before then the lag and scaling are probably just as good, but the compatibility with the OSSC might suffer the same restrictions.

[EnragedWhale]

Long story short, the W series Sony’s aren’t the be all and end all for upscaled retro gaming I used to think they were.

note: In fact the Sony W of that time are still quite great today, why wouldn't they be? there aren't many flat panel displays doing so many things right all at once, believe me. And I'd take the scaling and PWM-free backlighting of the Sony's over anything Samsung anytime. Some Samsungs are quite good in some departments ok, but still not quite at the level of Sony in terms of combined accuracy and performance.

In my opinion motion clarity just wasn’t that great on my 2013 W6 Sony. My new Samsung is much better but not a patch on my old VT65 panny plasma.

Oh and don’t get me wrong, I’m no Samsung (or any brand) fan boy. I buy whatever suits my needs/budget at the time.

[Xyga]

Oh sure it wasn't the best you could find already at the time, but it was just fine and well-balanced for a VA, since then global response times have improved a bit on both major types of lcd panels (VA and IPS) though it mostly depends on RTC/overdrive acceleration and we still find many with crappy response despite having been manufactured recently.

Consider also that the higher resolution your source material is, the better the perceived motion clarity will be. basically a linedoubled source will look blurrier in motion compared to a linetripled etc until you reach an optimal 1:1 and assuming there's no heavy filter to bring the clarity down again.
But that's not all.

There's also the fact that Samsung use PWM for dimming which is active at all brightness levels but 100% (though sometimes even at 100%).
For some people this creates an illusion of faster response, "illusion" because while the eye will catch 'frozen frames' which the brain will think aren't blurred, these aren't synced to the displayed signal and therefore only solid at the very inconsitent and random times when the eye sees them because the backlight flashed that instant and the brain registered. A fixed or peripheral gaze however will more or less consciously experience afterimage motion artifacts adding to the general motion blur, and possibly induced fatigue/headaches due to the flickering.
Non-PWM displays don't do that and therefore show you the real 60Hz sample-and-hold motion with the performance of the pixel response of the panel without waiting/soliciting your visual center for some gymnastics on top of it. Technically motion is better on non-PWM displays, but since they don't generate the illusion I've described, some viewers see them as performing worse in that area.
Personally I can see what's wrong with PWM displays when the flicker frequency is low, and Samsung sets use some of the lowest there can be, because it's cheap to produce and also beause it makes a greater number of people sensitive to the fake-fast-motion trick.
In a competent testing comparison however, it's not even certain a display with faster pixel response times would rate higher than a slower but PWM-free.
Ideally there shouldn't be any PWM nor slow pixels, OLEDs eliminate both issues, but unlike plasmas they're still sample-and hold so one of the causes for motion blur remains.

Plasmas are completely different of course, they emit their own light and flash the whole frame in succession of subframes at high frequency, plus they don't have the issue of slow flipping crystals.
And blah and blah, we've been through in many discussions so I shall stop my rant here.

[Thomago]

OLEDs eliminate both issues

Regarding PWM - is that the case? I'm asking cause I was looking into buying a new smartphone lately and the ones with OLED screens seem to flicker at lower brightness levels.
(See here: https://www.notebookcheck.com/Test-Goog ... 540.0.html and https://www.youtube.com/watch?v=dhqYyWuXdHw)

[Xyga]

PWM can be used on OLED too yes I've read that, but like with LCD backlight there's no obligation, only designer's laziness.
To say something more accurate: OLED eliminates the issues related to backlighting, but never underestimate the ability of R&D tied to marketing and cost constraints to fuck things up. I guess even the best technology can be ruined by stupid ideas. ^^

[EnragedWhale]

Huh, just been reading up on PWM and you’re right my TV uses it. I honestly, hand on heart can’t see it (and I struggle to look at a crt these days) but if you can I can see why it would be a problem.

In a perfect world if I had the money I’d just have two OLEDS

[Xyga]

In a perfect world we would all have stocked on good CRTs and plasmas before they went extinct.

[marqs]

I assume firmware releases assign default settings, correct? Wouldn't it then be possible to make custom firmwares that assign custom defaults for each of the save slots? So for example:

I could do up profiles for Genesis, SNES, Saturn, Neo Geo, Playstation, etc, and have them as firmware default settings in slot 0, 1, 2, 3 ,etc. All people would need to do is download a firmware of their choice (it would contain a combination of console defaults on the save slots), and then flash it to the OSSC like we currently do with new firmeware.

Would need marqs to comment on the logistics and possibilities of this.

A firmware includes default settings that can be overridden by profiles. The idea is that with a PC app, one could configure settings via GUI and import/export profiles (e.g. in XML format). As a final step it'd generate a binary consisting of the slots (optionally including a firmware as well) that can be written to a SD card and loaded on OSSC. That way users could share their settings in readable/editable format and there would be no need to share customized binary/firmware images.

In my opinion, having FBX profiles on the OSSC would be a game changer. Marqs, is there anything we could do to help? Would you need scripts written? Any specific OS required?

The app is still in planning stages, so at this points all ideas / suggestions are welcome (please use this VGP thread for it). As for your OS question, the implementation should preferably use cross-platform frameworks so that it's not tied to any specific OS.

[FBX]

The idea is that with a PC app, one could configure settings via GUI and import/export profiles (e.g. in XML format). As a final step it'd generate a binary consisting of the slots (optionally including a firmware as well) that can be written to a SD card and loaded on OSSC. That way users could share their settings in readable/editable format and there would be no need to share customized binary/firmware images.

[GigaBoots]

So I thought I'd post my OSSC->FrameMeister test photos because it encapsulates a lot of why I ultimately decided against using the FM as a framebuffer of sorts for the OSSC so I could record SNES.

Both Photos have had contrast increased to exemplify important details

OSSC Direct

Spoiler

OSSC Framemeister

Spoiler

Video: https://www.youtube.com/watch?v=n3BSjlF ... e=youtu.be


I'd rather have the image exhibit small-grain noise than have a machine smooth over all the noise and harsh edges. Perhaps I don't have the FM set in the right mode (1280x960, DVI Mode, Picture) but I feel that won't change how it fundamentally alters the image. I liken this difference to a camera that uses a lossy compressed codec and smooths over details and filiming in raw. Most videographers consider the fine, smaller grain noise to be more idea and "filmic" and I think that stays the same here. The way the "smoothing" changes every frame anyway, so it tends to look worse in motion.

I'm curious as to why my Genesis has brighter output at the top of the colored squares and dimmer at the bottom. I should probably just tolerate it though, god knows how much of a headache that might become to fix.

[FBX]

I don't understand how your color bars are looking like that. It should be showing all possible shades of the Genesis RGB in that test pattern. I'll show you my results with OSSC and OSSC + Fm in here in an hour after I get screenshots.

[GigaBoots]

I don't understand how your color bars are looking like that. It should be showing all possible shades of the Genesis RGB in that test pattern.

Are you talking about how everything near 1 is black? I noted in the original post, it's been contrasted up to exemplify details in the video. I've added more markup to that text to make it more outstanding.

[FBX]

I don't understand how your color bars are looking like that. It should be showing all possible shades of the Genesis RGB in that test pattern.

Are you talking about how everything near 1 is black? I noted in the original post, it's been contrasted up to exemplify details in the video. I've added more markup to that text to make it more outstanding.

Okay I will do the same on mine to see if I get similar results.

[FBX]

Okay before posting mine, I want to give a heads up on easy 1:1 scaling of OSSC custom signals into the Framemeister:

1. Turn ZOOM OFF.
2. Make sure Visual_Set dimensions are at defaults (64,32,32,32).
3. Not sure if this matters with ZOOM off, but set Visual_Set's Auto_Scaler to OFF and set H_Scaler to 5 and V_Scaler to 6 anyway.
4. Go into the HDMI_Set options and make sure the HDMI-1-Direct is turned OFF. (same thing for 2 if you decide to use port 2)
5. Go to Screen_Set and set it to SMART_1X. This is important because it will force the Framemeister to 1:1 scale the OSSC feed.

First OSSC under my best possible settings for my Genesis (with overblown constrast):

Spoiler

Now unfortunately since my card captures in RGB color space, it forced the Framemeister to output in RGB mode. Some of you know the FM screws up the color balance in RGB output, which can be seen in the image below with green and red channels out of alignment (and again contrast was intentionally overblown):

Spoiler

And for posterity, here is the raw image of the OSSC direct:

Spoiler

And the OSSC + FM combo raw:

Spoiler

Sharpness seem okay, but the real issue is the loss of color quality for the red and blue channels. It makes them kind of 'fuzzy' looking.

-FBX

[Xer Xian]

For what it's worth, I also observed quite an ugly noise in my captures from the 240p test suite for the GC using component cable and an OSSC+Crystalio 2: one, two (picture is quite soft - I've since learned that this particular VPS needs to be sharpened up a bit on digital inputs).

I attributed the noise to my lossy capture device, but probably at least a part of it was there to begin with after all..

[FBX]

By the way, I started experimenting with the gscartsw-lite added into the chain (due to it's digitization of sync signals), and found that the 320 mode H. samplerate of 427 indeed can work just as well if not a hair better than 428. I methodically tested all the video LPF settings and found the following gave the cleanest possible picture on 320 mode:

Video in proc:

Video LPF = 35Mhz (weird I know, but this minimized artifacts the most in solid colors for this setup)

Sampling opt:

Sampling phase = 123 deg

<adv timing 320x240:

H. samplerate = 427
H. synclen = 32
H. backporch = 52
H. active = 320
V. synclen = 3
V. backporch = 24
V. active = 224

Output opt:

240p/288p proc = Line4x
Line4x mode = 320x240 optim.

Example result:

[Triple Lei]

That was from Line 4x?

And I was about to suggest we drop 4x altogether in favor of new features. Personally I only use either the more compatible, integer-friendly Line 3x or the always-amazing Line 5x.


Speaking of 5x, I only recently found out my setup (OSSC > DVDO Edge > Monoprice Item 5369 + 2029 > HDfury (VGA out) > PC monitor) was compatible with Line 5x on NES and SNES after all. I didn't have any trouble with 5x on Genesis, PS1, or my supergun with an older HDMI cable, but switching to a cable I knew was higher speed (18 Gbps) did the trick for NES and SNES. And it wasn't like I was replacing every cable in my chain; just the cable from the OSSC to the DVDO Edge.

So I guess the lesson is, maybe SNES wasn't as troublesome as we thought with 5x? Just pay a little more for an 18 Gbps HDMI cable, or a lot more for 48 Gbps.

---

EDIT: Okay, I had some blanking out issues on 5x after switching displays and consoles. The blanking out happened even with my original setup with confirmed working cables. But eventually it went away after trying different settings. I think it was RaphM's cycling trick that did it!

EDIT 2: All right, I think NES isn't as "easy" to get to work in 5x as the SNES is. I cycled Line 5x format 1920x1080 to 1600x1200 and back to 1920x1080.

Output opt.:
Line5x mode 256x240 optim.

Sampling opt.:
Adv. timing 256x240
Allow TVP HPLL2x off

EDIT 3: Looks like switching from Line5x format 1920x1080 to 1600x1200 and back to 1920x1080 is a winner. I played some SNES in 5x which had some drops, then switched 5x formats (waiting 5 seconds or so between switches) and the picture remained stable from then on. Then I fully powered off the SNES and OSSC and switched HDMI inputs on the Edge, then switched HDMI inputs back to the OSSC and played on the Famicom in 5x. Switching between the Line 5x formats did the trick there too.

EDIT 4: I got that 48 Gbps HDMI cable and I haven't had any hiccups in Line 5x.

[Guspaz]

My projector (primary TV) works with 4x, but not 3x or 5x, so dropping support for 4x would basically limit the OSSC to 1x and 2x for me.

4x is the largest non-overscanned resolution for 1080p displays, so I think it's more important than 5x mode.

[Zappyraccoon]

For what it's worth, I also observed quite an ugly noise in my captures from the 240p test suite for the GC using component cable and an OSSC+Crystalio 2: one, two (picture is quite soft - I've since learned that this particular VPS needs to be sharpened up a bit on digital inputs).

I attributed the noise to my lossy capture device, but probably at least a part of it was there to begin with after all..

Are you just passing through the 240p signal into the Crystalio II? Or are you sending 4x or greater? I don't have any softening on my unit.

[GigaBoots]

Just wanted to post these unadjusted versions of the image.

OSSC->Framemeister

Spoiler

OSSC Direct

Spoiler

Even unadjusted, you can see in the FM-buffer image horizontal smearing across colors that shifts them more green or magenta, like digital sensor noise. It's very odd.

[Thomago]

4x is the largest non-overscanned resolution for 1080p displays, so I think it's more important than 5x mode.

Indeed.

[Xyga]

4x is the largest non-overscanned resolution for 1080p displays, so I think it's more important than 5x mode.

Indeed.

I don't see it that way, with the OSSC one must always think of the displays first and what happens in practice;

5x is the only way to disable any scaling done by the display (well, when it works), and the 'overscan' it does is also the only chance to eliminate borders in cases the display lacks working zoom and overscan of its own (most monitors but also quite a lot of TVs). And 5x also provides optimum clarity (again, when it works ^^)

Compared to 4x which is more compatible (though mainly with monitors rather than TVs) but quite difficult to upscale properly (all but one display I've tried 4x on made the scanlines uneven and generally also made the whole picture look coarse af) I'd say 5x is pretty damn more important!
I understand the love for 4x but compared to the importance of 3x (best on most displays) and 5x, I see 4x more as a bonus for people who want a sony broadcast-like look.

If any possibility ever exists to make 5x more compatible with common displays and that would mean dropping 4x, I'd vote to sacrifice 4x.
Anyway there could always be two separate firmwares, one with 4x, one with 5x. ^^

[Thomago]

Line5x 1080p mode sure is nice in terms of clarity, but the inevitable forced overscan sours the deal for me. I will readily sacrifice clarity over 'getting the whole picture'.

[Xyga]

It's not right fo all sources I agree, but It's great for the MD for instance, in like 90% of the games if not more the very few lines you lose top and bottom have absolutely nothing of importance graphics or gameplay-wise, and those ugly, sometimes flashing, colored borders are finally gone. Fantastic.

[Elrinth]

same with NES...

has anyone gotten the X68000 via rgb out -> scart working?
i don't have framemeister back yet so i can't check if it's the new cable i bought or is the OSSC which cannot sync.

[TestType]

It's not right fo all sources I agree, but It's great for the MD for instance, in like 90% of the games if not more the very few lines you lose top and bottom have absolutely nothing of importance graphics or gameplay-wise, and those ugly, sometimes flashing, colored borders are finally gone. Fantastic.

You know the OSSC has a masking feature that allows you to eliminate any anomalies at the edges of the image. No need to go 5x just for that.

[Xyga]

But it's ten times better to have the picture use the available real estate.

[Thomago]

I would argue it's 0.5 times better