OSSC (DIY video digitizer & scandoubler)

[Fudoh]

It's a timing issue, since you cannot just output all the black lines and top and bottom at once, but you have to stick to the signal's line interval. In the meantime new lines are coming in from the source and they need to get buffered.

You don't need a full-frame buffer to do that, but you need enough memory to buffer a few ms of incoming data.

[Ed Oscuro]

That's what I thought. I didn't say you need a full-frame buffer with the current method, though.

[rtw]

Did you mean that the output would be fixed at 720p/1080p and the input frame would be shown in its original size (or 2x etc.) at the center?

Yes that is what I meant. Thank you for the explanation. I suspect this is what the Framemeister does but as you say it will introduce lag.

[panzeroceania]

960p is classic quadrupling. I tried if often from various processors. Hardly any (TV) display will show it.

I would still implement 1080p, even if it's just 1280x960 active pixels with borders all around.

I know several people who would kill for this

[Xyga]

What about 1120p leaving 40 lines out in a 'forced' overscan area ?
That would make 224p sources like the MD display fullscreen without borders at the top and bottom (and maybe only a few actual game lines out as well, but heh).

[marqs]

What about 1120p leaving 40 lines out in a 'forced' overscan area ?
That would make 224p sources like the MD display fullscreen without borders at the top and bottom (and maybe only a few actual game lines out as well, but heh).

Yes, it's basically same as RGB32 suggested earlier. Out of all the proposed "1080p" modes, that's definitely the most realistic to implement on a FPGA, since it is a plain 5x vertical line multiply. I can't say for sure if the current hardware is powerful enough to run that, but that's one feature for which support could be added in future revisions.

Status update: I managed to squeeze the code into an inexpensive low-end Cyclone IV FPGA, and I'm currently drawing schematics for the next version (almost done). My plan is to finish the pcb layout and send it to manufacturing before Bloodborne comes out.

[Xyga]

What about 1120p leaving 40 lines out in a 'forced' overscan area ?
That would make 224p sources like the MD display fullscreen without borders at the top and bottom (and maybe only a few actual game lines out as well, but heh).

Yes, it's basically same as RGB32 suggested earlier. Out of all the proposed "1080p" modes, that's definitely the most realistic to implement on a FPGA, since it is a plain 5x vertical line multiply. I can't say for sure if the current hardware is powerful enough to run that, but that's one feature for which support could be added in future revisions.

Oops missed his post. Vertical position adjustment is a good idea as well, though it always seemed to me that scalers have a horrible time trying to v_shift (v/h pan on the other hand is done smoothly).

Anyway keep up the good work ! *thumbsup*

[invzim]

Status update: I managed to squeeze the code into an inexpensive low-end Cyclone IV FPGA, and I'm currently drawing schematics for the next version (almost done). My plan is to finish the pcb layout and send it to manufacturing before Bloodborne comes out.

Exciting stuff!

You're quite a bit further along than me - did you consider the MAX10 part? For the PCB stuff I need a bit of CPU processing, so current plan is to stick an STM32 to it to handle CPU processing, fpga programming and firmware upgrades.

[Konsolkongen]

I just discovered this project yesterday. No idea how I could have missed it

I think it looks extremely promising already! The Gunstar Heroes shot already looks quite a bit better than what the Framemeister can produce, due to it's RGB444 > YUV422 subsampling.
The scanlines look great as well, but they are a bit too dark for my taste. Will the user be able to adjust the brightness of these?

I'm also very glad that you decided to add 720p as all TVs I've owned scales this better than 480p. Panasonic TVs are especially retarded when it comes to 480p as it forces overscan which can't be disabled.

As I understand there will be no framerate conversion with this device? I wonder if this won't cause trouble with a lot of TVs since most retro consoles are a bit off the standard 59.94Hz. The Neo Geo CD (and probably AES too) comes to mind, because it runs at 59.4Hz according to the Framemeister. And the MVS as well as other arcade systems running at 59.1Hz or even weirder refresh-rates will probably not work at all?

In my experience a Panasonic plasma (haven't owned any of their LCD models) is much more tolerable to obscure refresh-rates than other TVs.
Using the Neo Geo CD with my Framemeister and a locked sync output, the Panasonics would display no stuttering at all, where there was some on the other TVs I tested.

The tested TVs were Panasonic S20, Panasonic ST50, Samsung E6500 (plasma) and a Sony W905A.

If this mod can be modified to OC every system to proper standards then that's fine by me! I'd rather have a ~1% increase in speed (in most cases even less) than stutter. And I doubt anyone could ever tell the difference. It's not like the ~17% difference between 50Hz and 60Hz :p

This wasn't meant as a criticism, but merely a heads up that some systems might not work as well on some of the more strict TVs

[marqs]

Status update: I managed to squeeze the code into an inexpensive low-end Cyclone IV FPGA, and I'm currently drawing schematics for the next version (almost done). My plan is to finish the pcb layout and send it to manufacturing before Bloodborne comes out.

Exciting stuff!

You're quite a bit further along than me - did you consider the MAX10 part? For the PCB stuff I need a bit of CPU processing, so current plan is to stick an STM32 to it to handle CPU processing, fpga programming and firmware upgrades.

Actually I haven't looked at MAX10 FPGAs until now - I thought they would just be CPLDs like earlier MAX series. Internal flash & oscillator are really nice features which would reduce the number of required extra components on PCB. Also, the internal flash apparently could be used also for Nios2 instruction memory, which would be great as the current chip I'm targeting at (EP4CE15) leaves me only ~32KB for software. Unfortunately, the availability of 10M16 chips seems to be quite weak currently so I'm still probably using Cyclone IV for now, but will definitely look at MAX10 series in the future.

[marqs]

I just discovered this project yesterday. No idea how I could have missed it

Well, I've deliberately avoided creating too much hype until everything is implemented on a single board and working well enough.

I think it looks extremely promising already! The Gunstar Heroes shot already looks quite a bit better than what the Framemeister can produce, due to it's RGB444 > YUV422 subsampling.
The scanlines look great as well, but they are a bit too dark for my taste. Will the user be able to adjust the brightness of these?

Yes, you can freely set the strength as shown in the controls of the prototype.

I'm also very glad that you decided to add 720p as all TVs I've owned scales this better than 480p. Panasonic TVs are especially retarded when it comes to 480p as it forces overscan which can't be disabled.
As I understand there will be no framerate conversion with this device? I wonder if this won't cause trouble with a lot of TVs since most retro consoles are a bit off the standard 59.94Hz. The Neo Geo CD (and probably AES too) comes to mind, because it runs at 59.4Hz according to the Framemeister. And the MVS as well as other arcade systems running at 59.1Hz or even weirder refresh-rates will probably not work at all?

In my experience a Panasonic plasma (haven't owned any of their LCD models) is much more tolerable to obscure refresh-rates than other TVs.
Using the Neo Geo CD with my Framemeister and a locked sync output, the Panasonics would display no stuttering at all, where there was some on the other TVs I tested.

You've understood correctly, there is no framerate conversion so this is not a system for everyone. As you said, Panasonics display slightly non-standard refresh rates without issues, but unfortunately they are not as tolerant with horizontal/vertical area. For example, my VT30 doesn't accept the 720p generated by the linetripler as it has more blanking lines than the CEA 720p mode, so 480p is the only option with that TV.

The tested TVs were Panasonic S20, Panasonic ST50, Samsung E6500 (plasma) and a Sony W905A.

If this mod can be modified to OC every system to proper standards then that's fine by me! I'd rather have a ~1% increase in speed (in most cases even less) than stutter. And I doubt anyone could ever tell the difference. It's not like the ~17% difference between 50Hz and 60Hz :p

This wasn't meant as a criticism, but merely a heads up that some systems might not work as well on some of the more strict TVs

Yeah, the reality is that some console/TV -combinations just won't give good results by default, unless you apply mods like that.

[Konsolkongen]

Yes, you can freely set the strength as shown in the controls of the prototype.

Thank you, I missed that as well

You've understood correctly, there is no framerate conversion so this is not a system for everyone. As you said, Panasonics display slightly non-standard refresh rates without issues, but unfortunately they are not as tolerant with horizontal/vertical area. For example, my VT30 doesn't accept the 720p generated by the linetripler as it has more blanking lines than the CEA 720p mode, so 480p is the only option with that TV.

Will it be fixable so it does follow 720p standards?

[bobrocks95]

Will it be fixable so it does follow 720p standards?

From what I understand, that would require building the 720p signal from scratch, which may be impossible for the FPGA and would create lag since it's not a simple line-tripling, so it's kinda against the point of this project.

[marqs]

Will it be fixable so it does follow 720p standards?

From what I understand, that would require building the 720p signal from scratch, which may be impossible for the FPGA and would create lag since it's not a simple line-tripling, so it's kinda against the point of this project.

Exactly. Framemeister already does a good job of general scaling to standardized modes by using a separate ASIC, so no point in trying to replicate that

[Konsolkongen]

I'm considering getting a Panasonic Plasma as well, so it would be a shame if overscanned 480p was the only option

I can see that the prototype has VGA out, is this doable on the final version too? If that worked, I could just route 480p VGA through my Gefen VGA to DVI scaler plus, which does a pretty good job upscaling to 1080p

[bobrocks95]

I tried the custom 1280x720 mode you posted a while back, as well as some other weird modes people proposed, and nothing seemed to work on the new Sony TV I just bought

I am just using a laptop connected to it and the Intel graphics manager, so maybe something's wrong there (I know it asked for a back porch as well, and I just threw in a value that made the total pixel count match up).

I'm guessing not though. So does this confirm that the only option for me using this would be the basic 480p linedoubling mode?

[marqs]

I'm considering getting a Panasonic Plasma as well, so it would be a shame if overscanned 480p was the only option

I can see that the prototype has VGA out, is this doable on the final version too? If that worked, I could just route 480p VGA through my Gefen VGA to DVI scaler plus, which does a pretty good job upscaling to 1080p

It'd be doable with some PCB area/cost penalty, but a better option would probably be making an alternative board with VGA output only.

[fluxcore]

Any scope for audio through the HDMI output? I do realise this is both a physical and development overhead.

[marqs]

I tried the custom 1280x720 mode you posted a while back, as well as some other weird modes people proposed, and nothing seemed to work on the new Sony TV I just bought

I am just using a laptop connected to it and the Intel graphics manager, so maybe something's wrong there (I know it asked for a back porch as well, and I just threw in a value that made the total pixel count match up).

I'm guessing not though. So does this confirm that the only option for me using this would be the basic 480p linedoubling mode?

You can still try that mode with a PC monitor to verify that the problem is not with Intel GPU, although I wouldn't be surprised to hear that a Sony TV does not accept the mode. If that's the case, linedouble (480p) is the only option. That's the reason I was originally somewhat reluctant to implement linetriple, but after seeing the result (on displays which support it) I have to say it was well worth the effort .

[BuckoA51]

If HDMI has royalties and DVI does not, why not just go with DVI? DVI to HDMI adapters are super cheap anyway (and yes, contrary to what some people will tell you, DVI can carry sound too).

[marqs]

If HDMI has royalties and DVI does not, why not just go with DVI? DVI to HDMI adapters are super cheap anyway (and yes, contrary to what some people will tell you, DVI can carry sound too).

Yes, that's probably going to happen if the board is going to be sold complete at some point. However, I doubt DVI transmitter chips can embed audio to the output. I wonder if it'd be legal to just use the same HDMI transmitter chip but with a DVI connector...

[bobrocks95]

You can still try that mode with a PC monitor to verify that the problem is not with Intel GPU, although I wouldn't be surprised to hear that a Sony TV does not accept the mode. If that's the case, linedouble (480p) is the only option. That's the reason I was originally somewhat reluctant to implement linetriple, but after seeing the result (on displays which support it) I have to say it was well worth the effort .

Works fine on my monitor, which is disappointing, but not surprising. Well, I've got plenty of options open for displays anyway, I'm sure I can find plenty of use regardless. People say the Sony handles 480p well anyways.

[Xyga]

Id's say there must be something really off with the modeline then, not the TV, because the Sonys as well as most TV's will handle 1280x720p perfectly within the most common/standard EDID values.

Guys, if you start asking TV's to do things that aren't industry standards then this nice board will end up being unusable with most sets on the market, which would be too bad !

[dwards]

This is a really cool project and I am looking forward to your first release. I have a couple questions I could not find answers to.

1) Will the SCART input accept Composite Video + SYNC or will it require CSYNC? Trying to see if SYNC STRIKE is necessary.

2) Has the Cyclone IV FPGA also dropped the cost of the board? Any idea on estimated cost of final board?

Thank you,
Ed

[bobrocks95]

Id's say there must be something really off with the modeline then, not the TV, because the Sonys as well as most TV's will handle 1280x720p perfectly within the most common/standard EDID values.

Guys, if you start asking TV's to do things that aren't industry standards then this nice board will end up being unusable with most sets on the market, which would be too bad !

If you read a couple posts up, marqs says that's why he was initially reluctant to implement the linetripling, because it's far from CEA/EDID/whatever standards. The primary mode of linedoubling will work for everyone, the linetripling is an optional addition that works well on the few monitors/TVs that will display it.

Was there a realistic case for 1080p support somehow? It'd be quadrupled to 960p wouldn't it, since it's impossible to add borders with the FPGA, right?

[Fudoh]

since it's impossible to add borders with the FPGA, right?

the FPGA can do it, but the buffer-free design doesn't allow it.

[Konsolkongen]

Will it be able to do a 480i/p > 240p conversion before linedoubling to 480p/720p? That would be great for Gameboy Player on the Gamecube, as well as some of the lazier retro compilations on PS2 and Xbox.

I have very little understanding of how your project works, so it may be a dumb question. Still, a feature a lot of us have been wanting in the XRGB

[fluxcore]

Will it be able to do a 480i/p > 240p conversion before linedoubling to 480p/720p? That would be great for Gameboy Player on the Gamecube, as well as some of the lazier retro compilations on PS2 and Xbox.

I have very little understanding of how your project works, so it may be a dumb question. Still, a feature a lot of us have been wanting in the XRGB

From first post:

The features of the current system include:
-digitizes&scandoubles 240p/480i to 480p, 288p/576i to 576p, and digitizes 480p and higher
-quick 240p<->480i switching

Is that close enough?

[Konsolkongen]

That's not what I meant at all

I mean converting bad ports in either 480i or 480p to 240p first, so we can get nice scanlines there as well.

I suppose 480p with scanlines can work, but most of the times these ports in 480p are softened and look quite bad. Converting to 240p sharpens the image a lot. Or at least, that's the case with the 240p hack for the Gamecube's Gameboy player. It looks pretty damn good

[fluxcore]

Oh... you want to intentionally downscale it? Interesting.