CurtPalme.com Home Theater Forum

Melifluonze · Posted: Sat Apr 04, 2015 2:57 am Post subject: Open Platform CMS

After visiting Nash's theater and viewing the blend, I've been working my way towards one as well. I have two TVOne boxes and I'm working on the projectors. I just got my hands on a 136" 2.35:1 acoustically transparent screen as well. But the big thing that I've been looking into is developing a custom open source/open platform Color Management System (CMS) solution for blending CRTs, and for adding CMS to any processor you have by sticking this solution in line.

I don't have tons of time, and it's my fault, but I do rotate through all the stuff I do and I eventually get things done. Sort of a combination of real world responsibilities, pack-ratting (or hoarding, as my GF puts it, hehe!) and a ton of ADD. What I want to do is start a thread on this. My goal is to keep what I'm doing completely open to everyone, so when it's finally figured out, anyone that wants a CMS box can get one on the cheap and download the software/firmware for it without paying for it, and anyone that wants to can enhance it and share the enhancements with everyone.

So here goes...

I'm not an expert at color or image science, but like some (most?) of you, I mess with hardware (FPGAs, ASICs, embedded processors, software, etc.). Here are my initial thoughts... To do what is needed to blend two displays together you've got the blending function. Splitting the two displays out of the video frame and generating two outputs with a blend zone, that's the EASY part. To be honest, I'll probably put the blending function right into whatever this ends up being anyways, so you don't have to buy anything else (or someone else will put it in) but let's start with the CMS.

I'm not going to reverse-engineer anything, or copy someone else's product, so please don't tell me to make it more like XYZ... It's going to be a kit/experimental platform, and you do what you want. I'm not going to compete with anyone or try to make a living here... That being said, what I do want to do is have the hardware be powerful enough to be expandable by everyone that makes their own, so maybe we can end up with an open source video processor of sorts... but that's stepping WAY ahead.

Goals:

- Stand alone box (not a full blown HTPC)
- Cheap to start, maybe like $200-$300 for the base CMS box kit
- Start out by handling one HDMI in / one HDMI out, to tune one projector to the other
- Video timing in = video timing out, and maybe gets confined to 24 Hz 1080p processing to start...
- Anyone can buy the board off the shelf (maybe make a more custom and powerful board kit later)
- A common sense GUI of some sort

Here's what I've done so far... I researched FPGA platforms first, and I'm a Xilinx guy. Anything that has enough capability to do 10 bit 1080p HDMI at 150+ MHz is going to cost some money. I found a Digilent Atlys™ Spartan-6 FPGA Development Board that costs about $400, and has the capability of doing both HDMI in and out (two outputs, just right), but the Spartan-6 is small and is barely capable of doing what is needed. The problem is memory, because there's not enough in the part to hold an appropriate CMS table... And interface speed to reach the frame rates and resolution people will want. My ideas may be too wild so maybe we could still fit it into the part if only the blend zone was corrected, for example. I have one of these boards now, and it comes with a good Verilog design example to do HDMI in and out.

https://www.digilentinc.com/Products/Detail.cfm?NavPath=2,400,836&Prod=ATLYS

Melifluonze · Posted: Sat Apr 04, 2015 3:21 am Post subject:

I thought about Kintex 7 devices, and Zynq (with a processor on board), but they are expensive and would need some custom board work. They might still be on the table, but it'll be a future thing.

So, then I took a look at single board computers. We have cell phones now that do 1080p without blinking. Compulab has just come out with the SBC-QS600 Single Board Computer (SBC) that uses a Qualcomm quad core Snapdragon processor on a module, the CM-QS600 computer-on-module. This little beastie can run Linux, and can be purchased as a development kit for about $300, with HDMI in and out support. The cool thing is, it runs at 1.7 GHz and has lots of memory to hold tables and an on-board GPU, which I think we could use to do most of the arithmetic. It also has flash for storing configuration. And it also has ethernet, wireless networking, multiple USB 3.0 ports, flash card port, micro PCIe port, etc. and it's teeny!

http://www.compulab.co.il/products/sbcs/sbc-qs600/

Even though I'm an FPGA guy, I think the SBC is possibly the way to go, especially if this is an open-source sort of project. It'll fit in a box, runs on almost no power, with an open OS, drivers provided, and has enough guts to do more than just CMS in the future. I'm going to pick one up to play with it. I'm just not convinced that it'll do the trick yet though... And it really can't stutter and judder due to something the Linux OS wants to do... Linux isn't really an RTOS.

My initial thought on the whole CMS thing was to provide the capability of entering an IRE level and color (Red, Green, or Blue), and being able to tune that IRE level to be perfect across the raster and in the blend zone by adding an adjustment in a table for some number of zones across the raster. The zones would be in a grid that you could manipulate (compress/expand, squish into the blend zone, sort of like doing geometry adjustments with the grid up on your projector). You could then pick a grid square and increase or decrease the level in the square. This would do a soft boost to the values in that square with respect to the others around it.

So you could set the grid for each IRE level (how many???) and then adjust the gain correction in each grid square you needed to. Once done, the gain correction and x y position of the grid squares for each IRE would be saved to the flash for your master table.

Each pixel that came in over HDMI would have an intensity that mapped into an original IRE level for that position that you set, and based on the coordinates within the raster, the gain adjustment for that position would be applied from the table created in memory. VERY powerful. And yes, it could be automated later...

I'll see if I can draw some pictures and stuff tomorrow and get more down in here. Anyone that knows this stuff better, please update my thought process... I don't know that you really need more than I just laid out... do you need an offset and gain? How would the soft graduation from the value in the center of the grid to the edge be done?

- Melifluonze

carpfisher · Joined: 30 Nov 2006 Posts: 19

Ooo, that's a nice project. I've often thought of projects like this too, but have always been stumped when it comes to hdcp. Do you have something up your sleeve for a practical system?

Melifluonze · Posted: Mon Apr 06, 2015 1:13 pm Post subject:

Heh... Easter took over and I got dragged off to family...

I do have something up my sleeve for a practical system. My goal is to put together a hardware kit that will run the software/firmware we put together, so anyone can buy the hardware for basically cost and then either use it, or help develop it and make it better.

Next step for me is to get some diagrams/pictures done.

- M

koldby · Posted: Fri Apr 17, 2015 8:05 pm Post subject:

I cannot contribute to this, but your thoughts sounds very interesting indeed!
Koldby

**AnalogRocks** · Posted: Fri Apr 17, 2015 11:31 pm Post subject:

I'd love to see this happen.

Melifluonze · Posted: Tue Apr 21, 2015 1:59 pm Post subject:

Oh, it's happening. It just won't be light speed.

I just received the single board computer from Compulab. I think it's missing the HDMI in board, but I'll see if I can get some pictures together for you guys and try to bring it up.

So far, I'm not getting the best support from Compulab, but we'll see. This might be a dead-end.

- M

carpfisher · Joined: 30 Nov 2006 Posts: 19

Would it be correct to assume that it didn't come with the HDMI-in board because you need an HDCP license to be able to buy one? This is where I always come unstuck.

Melifluonze · Posted: Wed Apr 29, 2015 3:57 pm Post subject:

No... or at least they told me that they had not completed the development of the HDMI input module yet, even though it is advertised on their site as if it was available.

I'm currently in the process of returning the development kit. We'll see how that goes. I tried to get them to accept me as an early adopter/tester, but that didn't go.

- M