OSSC (DIY video digitizer & scandoubler)

[Fudoh]

The OSSC provides output 720p output for 240p sources, but compatibility will completely depend on your display. PC monitors are more likely to work than TVs.

Remember that the XRGB-3 only provided line doubling in B1. For scaling you had to use B0 which uses a full frame buffer and introduces lag accordingly.

[BuckoA51]

Have you had more time to mess with games that have 240p to 480i transitions?

Yes, connecting the OSSC directly to the TV it's a second or so, but every active splitter/switch you put in between can slow it down.

So 720p is not possible for this device then?

720p is supported in line triple but doesn't work with every TV.

Isn't Sonic 2 splitscreen just using standard interlaced output? I tried that some months ago while testing a MD, and it worked as expected.

Sadly I can confirm it doesn't work here using OSSC to Sony KDL-40Z4500, the Megadrives interlace mode is kinda weird. 524i, 15.69khz VMod 480i LO 525 VSM 0. I'll test it on some other sets too.

[Edward_Tz]

the OSSC itself does this really fast, but on a digital display will most likely still trigger a resync on the display's side. That's still MUCH faster than the FM though.

I have an XRGB-3 currently. The screen just flashes black for a second if that. I assume it's comparable to that?

I know it's a non issue for some but long 240p to 480i transition times are a deal breaker for me.

On the videogameperfection site it mentions that adding a DVI to HDMI audio integrator may be a problem? I never used any devices like that since I have no need.

[BuckoA51]

See my reply above

[RocketKnight]

I suppose a line quadruplication or even quintuplication mode would only be possible with a framebuffer and a faster IC, right?

Something like (240-2*8-2*4)*5=1080 for 224p or 240*4+60+60=1080 for 240p internal resolutions would be really nice.

[Guspaz]

Don't need a framebuffer for that (don't want a framebuffer, that means lag), just needs a faster chip and more work RAM.

[Smashbro29]

Let him build some experience with it and hopefully make some profit to reinvest in a future, maybe more complete revision, if he feels it can be done.
Just repeating what's been said many times but the additional features we're asking for certainly would push the price up and indeed propulse the thing to actual consumer-level, market-player category.
Such step up does imply a lot more things to think about and be prepared for I believe, unless you already have a network and ready structure I mean (otherwise people who turned a hobby into a business from scratch have loooooong stories to tell lol)

That's fair, I can only speak as a user/consumer. I know it's not intended to replace a framemeister but it's about 2 features out from not only replacing it but beating it outright.

Borti, can we get the make/model of your TV for the compatibility list?

You're a few features shy of kicking the mini to the curb and then some.

This isn't supposed to be kicking the Mini to the curb, it's a replacement for the older XRGB3/2 series line-doublers.

Even just the audio would make this an instant buy for me.

Give us $7k for the license fee then and you can have audio

If he made some kind of funding campaign for it I would back it as much as I could and yell it from the rooftops anyway I could. I'm sick of waiting for Micomsoft to make things, the community is capable of making better easier to use equipment that's fully open source.

But I guess I'm just too hype, I'll probably just wait for future revisions. I respect that he wants to do things right or not at all.

[BuckoA51]

You can't do nice deinterlacing without a frame-buffer, this project is about making a lag-free line doubler. For the audio I believe the fee is yearly. Frankly I find that outrageous and I'm glad it's not paid out of principle. It has to account for a significant amount of the Framemeisters cost.

[paulb_nl]

I don't mind that there is no audio from the OSSC. Just connect the audio to an external speakerset. It will sound better than your tv anyway.

About the royalties, maybe DisplayPort can be used? Its free and there are many $10 DisplayPort to HDMI cables for sale.

[lettuce]

The OSSC provides output 720p output for 240p sources, but compatibility will completely depend on your display. PC monitors are more likely to work than TVs.

Remember that the XRGB-3 only provided line doubling in B1. For scaling you had to use B0 which uses a full frame buffer and introduces lag accordingly.

Whats the url for the TV tested at the moment?

[BuckoA51]

Have a peek here - https://docs.google.com/spreadsheets/d/ ... li=1#gid=0

I'll add more info to the wiki soon too.

[Asbrandt]

You can't do nice deinterlacing without a frame-buffer, this project is about making a lag-free line doubler.

Well. You can do Weave deinterlacing without having to delay for a buffer to fill. (You do need to -store- the incoming field for recall in the -next- field output, but that can be done without latency.)
But Weave only really works out on CRT monitors, on a modern LCD it needs a little something extra, such as darkening the previous field's lines a bit to roughly imitate phosphor falloff.

[Guspaz]

Even when you do the faster approach to weave deinterlacing, where you buffer the first field and then draw the output frame while the second field is coming in, you're halving your framerate, and getting the chicken-teeth artifacts on top of that.

You can do bob deinterlacing without any latency or framerate reductions, basically just linedouble each field, and you preserve the framerate, but then you're going to get flickering. This is what the GCVideo does for linedoubling mode on 480i, IIRC.

Another approach that you can do is pretty much the same as weave, only you double and blend the fields when you're outputting. You avoid the flicker, and you avoid the chicken-teeth artifacting, but you halve the framerate and get ghosting on motion.

All three approaches have problems. None is perfect. All can be done without adding any latency. Some require a framebuffer.

[Asbrandt]

Even when you do the faster approach to weave deinterlacing, where you buffer the first field and then draw the output frame while the second field is coming in, you're halving your framerate, and getting the chicken-teeth artifacts on top of that.

You can do bob deinterlacing without any latency or framerate reductions, basically just linedouble each field, and you preserve the framerate, but then you're going to get flickering. This is what the GCVideo does for linedoubling mode on 480i, IIRC.

Another approach that you can do is pretty much the same as weave, only you double and blend the fields when you're outputting. You avoid the flicker, and you avoid the chicken-teeth artifacting, but you halve the framerate and get ghosting on motion.

All three approaches have problems. None is perfect. All can be done without adding any latency. Some require a framebuffer.

Maybe I got my terms wrong, I thought Weave was where you treat each field as a full frame at 60hz and insert the lines from the previous field in the gaps.
For example if you're on an "even" field it goes (N,1)(N-1,1)(N2)(N-1,2)...(N,240)(N-1,240), so on so forth. If the N-1 lines were darkened a bit, it would approximately be what 480i does on a normal tube TV.

And yes, all of the approaches have visual problems, some alleviated on CRT monitors, but that's the cost of prioritizing speed over visuals, plenty of options already if you want pretty over speed.
Games that render at 480p60 but output 480i are especially cruel, there is essentially nothing you can do to make these look even remotely decent, but to my knowledge games that do this -and- don't have true 480p support are rare.

And it should be noted that even Bob has it's uses. For example on the PS2 there are quite alot of arcade ports that output 240p frames as 480i fields, Bob is perfect to correct this.

[Guspaz]

I could be mixing them up too, and yeah, as you've pointed out, you could keep 60Hz with weave and just fill the gaps with the previous frame instead of buffering every other frame. Those chicken-teeth, though...

[Paianni]

Wouldn't a 720p output be perfect for UHD TVs? Cos 720 is exactly a third of 2160. The TV would line-triple it.

[telemetry]

it needs a little something extra, such as darkening the previous field's lines a bit to roughly imitate phosphor falloff.

It's unclear (to me) how an LCD would have the bandwidth to do this -- you're already spending every field drawing every other line, so there's not really any time (at 60Hz) for a "falloff" on each line, since it has to get drawn again in ~30ms anyway.

The best you have time for is *one* color degradation in the intervening field, which isn't really enough to create the "falloff" illusion (which otherwise sounds pretty reasonable) -- it just seems like it would have a horrible "screen door" bobbing effect with 50% scanlines that shift back and forth.

Now, if you had an actual 240/480Hz refresh rate and the ability to actually apply meaningful fadeoff between every half-frame, that'd be something.

[Xyga]

Wouldn't a 720p output be perfect for UHD TVs? Cos 720 is exactly a third of 2160. The TV would line-triple it.

Only in an ideal world.
People keep spamming the same thing about 4K TVs making upscaling easy, but products segmentation reality is there to destroy logic and turn that statement into a fallacy.
An unhealthy number of TVs screw up even the simplest upscaling job, because manufacturers tune their built-in scalers essentially for video and broadcast contents to look nice when upscaled, which is not the same approach as getting input PC resolutions and contents to look right.
Name brands have their good and bad habits in that field (most bad) but to be safe just buy a Sony.

PS: and it's not just about PC resolutions anyway, when you look at it many 4K destroy the integers of 480p, 720p and even 1080p.

[viletim]

You can't do nice deinterlacing without a frame-buffer, this project is about making a lag-free line doubler. For the audio I believe the fee is yearly. Frankly I find that outrageous and I'm glad it's not paid out of principle. It has to account for a significant amount of the Framemeisters cost.

The HDMI license you mean? Micomsoft didn't bother with it. Just look at the unit and the box it comes in. The trademark HDMI logo isn't used anywhere. The main reason a company becomes a member of this club is to use the trademarks in advertising. Otherwise, what's the point?

The emulation consoles that have been popping up recently are all unlicensed. My local 'tech gadget' shop sells lots of unlicensed HDMI products too. If you ever see a product with a HDMI connection that doesn't used the trademark logos/graphics then there's a reason for that!

[BuckoA51]

The point is not getting sued one would assume.

[viletim]

The point is not getting sued one would assume.

Sued for what? Not trademark infringement.

BTW. It's the same situation for USB. You can just go ahead and implement it. Being a fully paid member of the USB club is not legal requirement. Avoiding trademark infringement is.


EDIT: I should clarify that my original point was that no business would pay the high cost of a HDMI membership and not bother to use the HDMI logos/trademarks.

[bobrocks95]

Isn't the code for HDMI transmittal copyrighted in some way? Or can you freely reverse engineer it and buying the license just gives you pre-written code? Or what?

[BuckoA51]

Sued for what? Not trademark infringement.

Just using HDMI in a finished product requires an integrators license and the $5k/year fee. That's according to the HDMI.org website. Perhaps they aren't really enforcing it against manufacturers/distributors who have really low runs and we're worrying too much, but it's definitely not the case that it's a free for all unless you want the HDMI trademark.

Though, they maintain a list of every licensed "adopter" as they call them, and I couldn't see Micomsoft's name on the list.

[borti4938]

Borti, can we get the make/model of your TV for the compatibility list?

At the moment I mainly write on the adaptation of the firmware. At the moment I have the audio-ADC running at 96kHz I have written an option for downsampling this signal to 48kHz.
There was also something I wasn't satisfied with the menu: settings with two option are wrapping around; settings with three or more options does not. I changed it (except for horizontal and vertical masking)

Now I will go over to more intesive testing I will update compatibility list every time I observe for me and my TV sets something 'new'.

[cleeg]

Good news about Sonic 2 split screen, the XRGB Mini doesn't like it. Thanks for answering Marqs.

[marqs]

Isn't the code for HDMI transmittal copyrighted in some way? Or can you freely reverse engineer it and buying the license just gives you pre-written code? Or what?

You'll find example code (not copyrighted) for that HDMI TX with on the net. There apparently are also some manufacturers who publish datasheets and reference code for their HDMI TX chips. Still, HDMI is not very friendly to hobbyist developers since the actual HDMI specifications are not openly available. The licensing terms are also unclear as can be seen from the different opinions here and in other similar threads concerning the use of HDMI as part of some electronic device.

There was also something I wasn't satisfied with the menu: settings with two option are wrapping around; settings with three or more options does not. I changed it (except for horizontal and vertical masking)

The whole menu generation code needs to be rewritten (e.g. using structs for different options types, submenus etc.) as the current implementation is from the time there was only a couple options . Haven't really had enough time for improving the fw code since January, so it'd great if you can help.

Good news about Sonic 2 split screen, the XRGB Mini doesn't like it. Thanks for answering Marqs.

There's definitely something funny if TVP7002 detects it as 524i even though 525 lines are correctly drawn to the output (LO). Strange that it doesn't work on Bucko's TV - I'll need to scope MD interlace signal when I have a chance.

[bobrocks95]

Isn't the code for HDMI transmittal copyrighted in some way? Or can you freely reverse engineer it and buying the license just gives you pre-written code? Or what?

You'll find example code (not copyrighted) for that HDMI TX with on the net. There apparently are also some manufacturers who publish datasheets and reference code for their HDMI TX chips. Still, HDMI is not very friendly to hobbyist developers since the actual HDMI specifications are not openly available. The licensing terms are also unclear as can be seen from the different opinions here and in other similar threads concerning the use of HDMI as part of some electronic device.

I think I also remember you mentioning that you didn't have an HDMI Analyzer to see if you were matching the spec or not, so that was another hurdle as well, right?

[marqs]

I think I also remember you mentioning that you didn't have an HDMI Analyzer to see if you were matching the spec or not, so that was another hurdle as well, right?

I don't indeed own a HDMI Analyzer (or high-bandwidth oscilloscope), so I can't really debug signals coming out of the TX chip. However, DVI/DisplayPort are not easier in this regard.

[marqs]

The compatibility chart is now organized a bit with similar systems grouped and groups split into tabs. That should improve readability when more results are added.

[BuckoA51]

Good news about Sonic 2 split screen, the XRGB Mini doesn't like it. Thanks for answering Marqs.

I've yet to find a single display that works with Sonic 2 splitscreen and OSSC.

There's definitely something funny if TVP7002 detects it as 524i even though 525 lines are correctly drawn to the output (LO). Strange that it doesn't work on Bucko's TV - I'll need to scope MD interlace signal when I have a chance.

It's not just my TV, I've tested monitors too, so far nothing works with it except MD->OSSC->DVDO Edge.