feeding binary frames to a DLP projector?

[thevesel]

How can you modify/make a DLP projector to have it display binary frames instead?

Technically they are displaying binary frames, but only to create a feel of grayscale frames and then with a color wheel (if a single chip DLP pj) to create a colored frame to our eyes.

If I could tell the projector to treat a 24 bit frame as 24 binary frames instead and take off the color wheel, I could achieve a 24 * 60 Hz/ 120 Hz refresh rate binary projection which is what I need for my project.

 

I've read somewhere on the internet of using an FPGA to convert the 24 bit frame data sent via HDMI to binary frames but that's all the info there is...

[Dmills]

A question for Texas Instruments application engineering more then a technical theater forum I cannot help but feel.

The TI applications guys are usually very good and well worth asking.

 

Now consider that in a standard projection setup the 8 bits making up an intensity are not mapped in a particularly useful way, it is not done as 8 identical time slices such as you would need for your idea, but is instead done as a duty cycle taking the total number as a binary value, not saying TIs driver chipset cannot do what you need, but no video projector will work it that way.

It is even possible that the chip is internally working at a LOT less then 8 bits and is playing sigma delta games with individual pixels...

 

Take your projector and experiment, produce a solid gray RGB frame having value #010101, #030303, #070707, #0f0f0f, #1f1f1f....#7f7f7f and look at the light output using a photodiode and a scope to see what waveforms the DLP oputputs in response, compare with 1,2,4,8,10,20,40,80.

 

IIRC TI has a DLP development platform available that may be a better starting place.

 

Regards, Dan.

[thevesel]

Texas Instruments support has not been really helpful.

Also the development platforms (dlp chips+leds+optics+board) they offer are very overpriced and beyond my budget, They cost more than two brand new 3000 lumen video projectors. They do not have cheaper options, nor licensed 3rd parties.

So the only option I see is modding the board of a off-the shelf projector. I hoped someone had done such experiments and had the knowledge here.

[Dmills]

Surprised by TIs apps people not being helpful, I usually find them to be rather good.

 

Dev boards are always expensive, being a kind of low volume thing, but I don't think a video projector is a sane starting place for this as the interpretation of the data is completely different and even if the projectors driver chip set can be talked into doing what you want (Which needs the datasheet for that part, TI again), it will be a fine pitch BGA and breaking into the I2C bus may be very tough.

Looks to me like the lightcrafter DLP eval module is only $600 and unlike the projection chipsets the thing being designed for binary pattern usage will do what you want out of the box with little programming (And, big win, the thing has sane documentation).

 

TI seem to sell two ranges of DLP chipsets, the ones for video and the stuff that supports 'structured light mode' (which is what you want), their DLP pages are really very good on this stuff.

You might find this http://www.ti.com/lit/ug/dlpu004b/dlpu004b.pdf document interesting, particularly the bit on structured light mode which seems to be what you wish to achieve.

 

If memory serves George from Chamsys had some dealings with DLP back in the day, might be worth dropping him a PM.

 

Regards, Dan.

[thevesel]

Yeah, I thought about using 3 of those "Lightcrafters", one for each channel and combining the result with a dichro cube, but the projection refresh rate of this product is misleading.

"Up to 4000 Hz" is when you preload the frames in to the mDDR memory. I think you can only store about 96 frames there...

And the 1440 Hz 1 bit streaming in "external pattern mode" is confusing. I'm not sure how binary frames at 1440 Hz can be streamed. With the mini-HDMI input it says the "framerate" is 60 Hz and "pattern rate" is 1440 Hz. The USB 1.1 port is too slow for that so that can't be it. 1 bit frames can also be preloaded into the builtin 128 MB RAM according to the manual, but that's not enough space either. I mentioned this in their "E2E" forums but haven't gottena response so far.

 

And I don't think 1440 Hz will be fast enough anyway.

All the projects done so far have ran DLP chips in the 4000-6000 Hz range.

For 180-198 2d slices for each volume (half-revolution) 4000 Hz allows 20 fps per 2d slice. With 1440 hz you could try something like 120 slices and 12 fps but don't know if that amount of slices and that flicker inducing fps would give any decent impression of a 3d volume.

[Dmills]

So it is volumetric projection?

 

1440 is 60 * 24 so the HDMI input is probably the limiting factor (one image plane per bit, set them to have equal duty cycle, mux 24 binary frames into one 24 bit HDMI frame, job done), for more I think you need to talk to the thing over the parallel LVDS bus at silly speeds, could be done but board layout is going to be **NASTY**.

 

To get the real speed you need the 64 bit LVDS DDR data path to the driver chip, and even then you have a problem with bandwidth going into the FPGA (10GigE wont be fast enough at maximum possible pattern rate), in any case this is well outside the domain of the video projector.

 

You do have scary PCB layout skills available I take it? DDR2 memory is a ballache and looking at the specs on the bus used, this thing is **ALL** like that (Length matched, impedance controlled diff pairs, dozens and dozens of them).

 

Regards, Dan.

[thevesel]

So it is volumetric projection?

I thought I had linked to it, here are the details of the project: https://electronics....metric-displays

 

1440 is 60 * 24 so the HDMI input is probably the limiting factor

Even then, the "frame rate" is said to be "60 Hz" and "pattern rate" 1440 Hz in 1 bit mode. I don't really understand how the term "pattern" they use and "frame" can be different in binary frame mode and one be 60 and the other 1440.

 

I also tried to email them this question instead but I cannot find an email for their DLP department.

[Dmills]

Even then, the "frame rate" is said to be "60 Hz" and "pattern rate" 1440 Hz in 1 bit mode. I don't really understand how the term "pattern" they use and "frame" can be different in binary frame mode and one be 60 and the other 1440.

Where is George when you need him, I am pretty sure he did work on DLP hardware back in the day?

 

In standard video mode the 24 bits per pixel are broken down into 3 * 8 bit slices (RGB) and the 'on time' fraction for each slice is determined by the 8 bit value divided by 255 (Or equivalently the light source brightness is stepped to achieve the same effect - only really possible with LED sources.

The effect of this is that each bit in the 8 bit byte has a different weighting (The low bit being 1/45900th of a second or thereabouts (1/255 * 1/60 * 1/3), the high bit being 1/360th of a second or thereabouts (There may be some overhead in the timing I am unsure and cannot be bothered to read up, they may even be doing some kind of dithering to reduce the required switching rate, maybe even delta sigma of some sort), this is video mode to which the the frame rate number applies.

 

In structured light mode they simply switch the data interpretation so that every bit in the 24 bit data has the SAME weighting (1/1440th of a second), now one 60Hz frame contains 24 independent binary patterns that will be displayed in sequence, 24 * 60 = 1440.

 

Honestly the datasheets are not that bad, and they even have an intro to the technology as an app note.

 

The odd things a bunch of stage techs pick up over the years....

 

Regards, Dan.

[thevesel]

What is George's username in this forum?

 

And any ideas what TI department I should contact for DLP questions? http://www.ti.com/ge...ocs/contact.tsp

[Dmills]

As I say, I am almost sure that Mr McDuff (MadMac on the forum) had dealings with DLP way back in the day, but I only met the man a few times and I just remember him showing off a knackered DLP chip with a static image on the damaged array that he used as a keyring back then (That would have been last Century).

 

No idea at all who at TI to contact, I just spent 10 minutes reading a few of their introductory app notes over lunch because I thought the tech was cool.

 

Regards, Dan.