Post by thread:[OT]: VGA signal splitting...for a video walllikei

Jinx wrote: > > James Tu wrote: > > > Has anyone ever used hardware that allows you to split one > > > VGA output from a graphics card into multiple monitors...such > > > as for a video wall. > > > I want to be able to specify, for example, a rectangle defined by > > > pixels (1 - 100)x(1 - 200) goes to monitor 1 and the rectangle > > > defined by pixels (101- 200)x(1-200) goes to monitor 2, and > > > so forth...these numbers are just examples. > > I'm no video expert, but I reckon you could do this with micros. Uh huh. Something running around 100MHz, sampling RGB around 20MHz each channel. IOW 30us H time/640 pixels. Then you want to antialias the samples since it sounds like you'll still want VGA out. > Although my thinking is probably up the creek wrt to interlacing > and other technical vagaries. If picture-in-picture is doable by > a micro then so should video walling PIP is done with specialized chips that get initialized by the micro. Few micros can run at video rates. > It seems to me you don't need to monkey with the picture info > much, just direct it. Say you took the top left third of a display Oh, you do have to monkey with the picture data quite a bit. There is a lot of retiming to do if you want VGA in and out. > (for example splitting the display over 3x3 screens). It will be > made up of pixels and lines. You could count the pixel clock > with a micro and direct, via a video switch, those pixels to > Monitor1. The middle third would go to Monitor2, the last 3rd > to Monitor3. So then each monitor will show 1/3 of the image, with the remainly 8/9's blank. > Same for the next scan line, until you reach the > scan line that marks the end of the top third of the original > picture, then send pixels to Monitors 4 5 & 6, and 7 8 & 9 to > end > The tricky bit would be filling the monitors' displays. To do this Yep, thats the tricky part. > you would have to send 3 copies of the pixel and the line, so > expanding the 9th part of the original to a full screen width, in > the process reducing the resolution. But you can't display info > you don't have anyway Which is why you have to antialias the data. > Sound OK, or should I wait for the other shoe to drop ? There It just dropped. You have to think about your timing quite a bit more. What are the other monitors doing while you're sending out the first monitor's data? What is THAT monitor doing when the next monitors data is coming out. The CRT scan doesn't stop just because you don't have any video for it to display. > might be a "can't do" - maybe previously accessed monitors > need refreshing. I know people on the list have driven displays > with fast Scenix micros, perhaps it really is possible It's quite possible, but your approach is too simplistic to work. You have time dialate the incoming data so that each 1/3 line is a full line time on the outgoing monitor. You have to replicate that line 3 (or 4) times in order to fill the vertical space, and to maintain the aspect ratio. The cheapest way to do this is with 3 x Analog CCD chips (that don't seem to be made anymore). But you also have to have enough memory to capture the incoming lines while you're busy outputing the earlier ones x3. And then you have to do some filtering so that you don't see huge square pixels on the monitors. If you are willing to hack your VGA driver so that the data gets written to the screen in a special format, it IS possible to do 3x3 or 4x4 screen with nothing more than sample and holds (Really. I've already prototyped it, but didn't see a big market for it so never took it to production). Robert -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

I do not know what your resources are but fyi a commercial scan converter starts at $500 and there seems to be no end to it. That is the price of a cheap computer or a moderately expensive XVGA card with extras (like a proper video output that can be fed to a video wall splitter that is much cheaper than $500 probably). hope this helps, Peter PS: Hint: usually buying parts off the shelf to build something highly optimized like $SUBJECT will end up costing much more than the finished product, even if development time is not counted in. So the short answer is probably, look into high speed FPGAs with capability to interface to SRAM or VRAM (and do not try to save too much money imho). You can probably achieve 50MHz sampling using a 100MHz clock on the FPGAs and 10ns RAM. -- http://www.piclist.com#nomail Going offline? Don't AutoReply us! email spam_OUTlistservTakeThisOuTmitvma.mit.edu with SET PICList DIGEST in the body