BitBlitz: 15k RGB -> VGA

[Crafty+Mech]

BitBlitz

The purpose of this project is to create a converter board that will accept 15k progressive RGB input, and output VGA. I started the project in November 2012 because of my disappointment with cheapie solutions that suffer from a variety of display quality issues. Those boards are cheap, but they are jack-of-all-trades designs (supporting interlaced & progressive) and employ a full-frame buffer for conversion, which produces a lag of at least 1 frame. My hope is to be able to provide an optimized line-doubling design, virtually lag free (63 microseconds), and built for use with legacy 240p hardware in mind.

Status: 75%

I was waiting to post a project thread until I had a nice HD demo running on a Shmup like Raiden, but figured I needed to announce this project to light a fire under me to finish the design and start pcb production. The demo video below is so-so quality wise, but shows a 60-in-1 board running in CGA mode, and line doubled to VGA.


YouTube link: 15k RGB -> VGA Converter Demo


The breadboarded prototype as of 5/13:

[fagin]

Looks interesting from a technical standpoint.

Look forward in seeing your final results.

[jdubs]

Awesome man!!! Great project!

What's the target price?

-Jim

[dieKatze88]

My stupid question of the year, does it support classic games that switch between 240p and 480i on the fly?

[Crafty+Mech]

What's the target price?

I was originally hoping to keep the cost under $50 USD, but in the ballpark of $75 is looking more realistic. Hard to say though until I finalize the design, and will ultimately also depend on the size of the first production run.

My stupid question of the year, does it support classic games that switch between 240p and 480i on the fly?

Right now the design only supports 240p conversion, but I've talked with Fudoh about a simple way to convert 480i to 480p, so 480i support is still a possibility. As for on the fly switching, if I do add 480i support it will be a dip switch configured mode, rather than the auto-detect needed for the situation you describe.

Forgot to mention that scanline support will be built into the board, so no need for a separate SLG.

[dieKatze88]

What's the target price?

I was originally hoping to keep the cost under $50 USD, but in the ballpark of $75 is looking more realistic. Hard to say though until I finalize the design, and will ultimately also depend on the size of the first production run.

My stupid question of the year, does it support classic games that switch between 240p and 480i on the fly?

Right now the design only supports 240p conversion, but I've talked with Fudoh about a simple way to convert 480i to 480p, so 480i support is still a possibility. As for on the fly switching, if I do add 480i support it will be a dip switch configured mode, rather than the auto-detect needed for the situation you describe.

Forgot to mention that scanline support will be built into the board, so no need for a separate SLG.

Hmm. OK. I'm just trying to find something a bit more suited for the sillyness that is the Arcade DDR games. The old ones use 480i for the menus but 240p for the gameplay, but don't appear to change the sync signal because the monitor just takes it in stride. The problem is of course, with music games, lag is forbidden. So I'm trying to find the least stupid option possible.

[matrigs]

RetroPix
line-doubling design

upscaled

So, which one is it?

[Crafty+Mech]

So, which one is it?

I'm not sure I understand your question, the 320x240p progressive source in my demo has been upscaled to VGA (640x480). The algorithm employed is line doubling, which is to say that every line of source video is buffered and then written twice to the output. If I've used terminology in the wrong context it is unintentional, the core functionality of the RetroPix is to convert 15k progressive RGB to VGA.

[RGB32E]

Looks great! Any chance for supporting higher resolution output for HDTVs (1080p/1280x960/ect)? Scaling 640x480 to 1080p works ok with the XRGB units, but you'd be able to likely do a better job! What is your current ETA for getting boards fabbed?

[Crafty+Mech]

Looks great! Any chance for supporting higher resolution output for HDTVs (1080p/1280x960/ect)? Scaling 640x480 to 1080p works ok with the XRGB units, but you'd be able to likely do a better job! What is your current ETA for getting boards fabbed?

No plans for HD resolution support at the current time, but depending on the success of the initial release, the project could evolve in new directions.

My plan is to finalize the design next month, with a July/August release. If the demand is strong enough, I can have the boards made locally, which will speed up production.

[Fudoh]

If quality and price turn out fine, I'm absolutely certain that this will outsell basically any other home-electronic project of the past years.

[emphatic]

If quality and price turn out fine, I'm absolutely certain that this will outsell basically any other home-electronic project of the past years.

QFT.

[matrigs]

So, which one is it?

I'm not sure I understand your question, the 320x240p progressive source in my demo has been upscaled to VGA (640x480). The algorithm employed is line doubling, which is to say that every line of source video is buffered and then written twice to the output. If I've used terminology in the wrong context it is unintentional, the core functionality of the RetroPix is to convert 15k progressive RGB to VGA.

I'm thinking about non-240p resolutions. think of ketsui with it's 224 vertical lines. In case of just "line doubling" you would end up with 448 lines and the rest filled black, in case of "scaling" you would have it stretched to 480 lines.

I was curious how you approach this.

Actually, out of sheer curiousity: Would someone care to explain to me why line-doubling is so difficult to do "right"? I was convinced that it is as easy as it sounds, but looking at the Extron Andorra devices which do it pretty bad, i got pretty confused.

[undamned]

I was waiting to post a project thread until I had a nice HD demo running on a Shmup like Raiden, but figured I needed to announce this project to light a fire under me to finish the design and start pcb production.

Ha! Boy do I know about that one! I waited till I was about a year into my current side-project before making any sort of announcement
-ud

[Crafty+Mech]

So, which one is it?

I'm thinking about non-240p resolutions. think of ketsui with it's 224 vertical lines. In case of just "line doubling" you would end up with 448 lines and the rest filled black, in case of "scaling" you would have it stretched to 480 lines.

I was curious how you approach this.

Actually, out of sheer curiousity: Would someone care to explain to me why line-doubling is so difficult to do "right"? I was convinced that it is as easy as it sounds, but looking at the Extron Andorra devices which do it pretty bad, i got pretty confused.

Ah ok, now I understand what you meant. My design would be a straight-up line doubler then, as 224 vertical lines would end up as 448 lines for the output.

Speaking from my own experience with this project, line doubling done right demands a very precise, and stable timing circuit. The timing circuit must stay locked to the source HSync, in order to generate the buffer clocks, and doubled VGA Hsync pulse. During the Vsync pulse, most sources drop the Hsync signal, which means the timing circuit must also be very quick to lock again once the Hsync pulses resume (or else the result is image distortion on the output). In addition, the ADC module must be very stable, and well protected from other portions of the design. Otherwise, ripple/pixel artifacts will appear on the output, in particular with moving sprites & scrolling backgrounds.

So I think the difficulty comes from taking the time to do it right, and testing against a variety of sources. With this kind of approach, the source & output are effectively coupled via a line buffer. This allows the design to deliver virtually lag free performance, versus a frame buffer having a delay of at least one frame.

[matrigs]

Sound really exciting to do!

I would be really interested in one of these for my vga sega naomi cab. I can adjust the geometry manually so i don't care about black borders at all.

Hmm. OK. I'm just trying to find something a bit more suited for the sillyness that is the Arcade DDR games.

Is it really necessary to have 480i in the menus? You could juss put an Extron RGB in between and get rid of it. Or is the text unreadable in this case?

[dieKatze88]

Sound really exciting to do!

I would be really interested in one of these for my vga sega naomi cab. I can adjust the geometry manually so i don't care about black borders at all.

Hmm. OK. I'm just trying to find something a bit more suited for the sillyness that is the Arcade DDR games.

Is it really necessary to have 480i in the menus? You could juss put an Extron RGB in between and get rid of it. Or is the text unreadable in this case?

The problem with doing that is the Extron RGBs will 240p it for you, and it looks good, but they introduce like 4 or 5 frames of lag when they're in the only mode the 573 output and various inputs I have agree with.

[Crafty+Mech]

If quality and price turn out fine, I'm absolutely certain that this will outsell basically any other home-electronic project of the past years.

The pressure is on! I'm really picky about image quality, which is why I started this project. I have a stack of Shmup pcbs that I am itching to play, but just can't bring myself to hook up with the Gonbes board I have. I know there are a lot of enthusiasts out there who are picky about their pixels like I am, so I'm looking forward to sharing this product with everyone. My first retro gaming project, the "mini SLG", just reached 200 units sold. That project has been fun and exciting, and has helped fund the development of the RetroPix board.

[matrigs]

thought of some kickstarter, maybe?

[Crafty+Mech]

thought of some kickstarter, maybe?

I have thought about a Kickstarter, or international equivalent, to raise funds for the first production run. It will depend on what the fabrication cost is per board. If I use the company down the street from me the turn around would be a week, so I could pre-sell a batch and then ship them out two weeks later, avoiding the need for fundraising.

[deltronik]

I registered to say I can't wait to see your progress on this project. I would jump on a presale of these when you're ready for your first production run.

I am currently a happy mini slg owner and can attest to Crafty's customer service. I'm glad to see you tackling this project.

-del

[Crafty+Mech]

I registered to say I can't wait to see your progress on this project. I would jump on a presale of these when you're ready for your first production run.

I am currently a happy mini slg owner and can attest to Crafty's customer service. I'm glad to see you tackling this project.

-del

Thanks Del! I'm working through a list of minor quirks at the moment, and building a prototype board. A new video will be posted soon!

[robbo43]

Looks great!

Will the board support 288p 50hz (PAL) without framerate conversion?

Cheers

Rob

[yojo!]

you never thought of adding support for EGA to VGA conversion too?

[Crafty+Mech]

Looks great!
Will the board support 288p 50hz (PAL) without framerate conversion?

The Vsync signal is passed through the converter, so whatever refresh rate is put in, is what will be on the output. This preserves the refresh rate of signals that are not quite 60hz. So inputing a 50hz PAL signal will not result in 60hz on the output, which is I think what you are asking. I'm not aware of a method to convert 50hz to 60hz without frame buffering.

you never thought of adding support for EGA to VGA conversion too?

To keep this project simple in scope, my focus is 15k modes with a maximum horizontal resolution (including overscan) of 512 pixels. Slower signal speeds equals less expensive chips, and easier board layout. I don't have any plans at the moment to support EGA/Medium res in the future.

[robbo43]

Looks great!
Will the board support 288p 50hz (PAL) without framerate conversion?

The Vsync signal is passed through the converter, so whatever refresh rate is put in, is what will be on the output. This preserves the refresh rate of signals that are not quite 60hz. So inputing a 50hz PAL signal will not result in 60hz on the output, which is I think what you are asking. I'm not aware of a method to convert 50hz to 60hz without frame buffering.

you never thought of adding support for EGA to VGA conversion too?

To keep this project simple in scope, my focus is 15k modes with a maximum horizontal resolution (including overscan) of 512 pixels. Slower signal speeds equals less expensive chips, and easier board layout. I don't have any plans at the moment to support EGA/Medium res in the future.

Great stuff! That's exactly what I was asking - 50Hz In = 50Hz out. Looks like this could be a great solution for PAL owners.

[Fudoh]

Good luck finding a VGA input that accepts 576p50.

[Crafty+Mech]

Looks great!
Will the board support 288p 50hz (PAL) without framerate conversion?

Great stuff! That's exactly what I was asking - 50Hz In = 50Hz out. Looks like this could be a great solution for PAL owners.

You'll definitely get 50hz out, but as Fudoh mentioned the challenge will be finding a display that will sync to 576p @ 50hz through a VGA input. I'm not familiar at all with PAL equipment, so I will have to defer to our European members for a definitive answer.

[Fudoh]

some PC monitors will sync to it, but hardly any with smooth scrolling. No TV I know will do it.

For PAL usage I would recommend adding a VGA to HDMI or VGA to component converter after the linedoubler.

[robbo43]

some PC monitors will sync to it, but hardly any with smooth scrolling. No TV I know will do it.

For PAL usage I would recommend adding a VGA to HDMI or VGA to component converter after the linedoubler.

Absolutely, the use of a VGA to HDMI etc converter would be required. I would probably try and pick up a used VP30. If the scaler could have RGBS or RGsB selectable output the option could be there to use the cheap Vehda VP20 clone to scale to TV.

This brings me to the next question, have you decided on input/output connection options yet?

Cheers

Rob