L/R color balance corrector for 85Hz DLP porjector

[Petrus]

If you use a 85Hz DLP projector for S3D with LCD shutter glasses, yo have probably observed a color difference between left and right eye.

I've made a test pattern to see this issue without glasses :


Once projected with my Optoma EP739H I get this:


Looks pretty bad.

So, to correct this issue, i build an electronic circuit :


This circuit dims the video signal according to RV1, RV2 and RV3 potentiometers synchronously with the DDC signal that drives the glasses.
SW1 and SW2 are a single double switch that invert the side (L/R) of the correction.
there is also a switch to invert le DDC signal for the glasses(SW3).


A : DDC signal (pin 12 on VGA)
B : red video signal

with the circuit activated and adjusted :


It works great at the price of brightness loss, we can compensate a part of this loss by pushing the RGB gain at 150% on the projector menu.

I want to buy the Nvidia 3D vision glasses, but is the DDC signal still sent on the VGA port ?

This is a (poor ?) translation of the same topic on the french section :
http://www.mtbs3d.com/phpBB/viewtopic.php?f=41&t=6052" onclick="window.open(this.href);return false;

[Okta]

Very nice work. Unfortunately Nvidia glasses sync from the usb and im guessing there is no ddc output.

[iondrive]

interesting,

dimming the signal for only one eye (or differently for each eye?). I would have used the other approach of changing the timing of the glasses but that would also result in dimming. Maybe I will use this someday but I'll probably go with the timing solution instead. Nice work and thanks for posting. Some people need this.

--- iondrive ---

[Petrus]

I will try the timing approach by introducing a "dark time" when the left and right
shutter are closed at the same time during 1/3 of the frame according to the diagram found in this document :
http://cmst.curtin.edu.au/local/docs/pubs/2007-05.pdf" onclick="window.open(this.href);return false;

But I suspect this method to increase the flicker, we will see.
It would be easy as I already hacked the nvidia 3D vision glasses by reading the IR signal from the emiter an re-emit it as I want.
But with modified timings I could not be able to use my other glasses as a second pair.

And Okta is right, there is no DDC signal with The 3DVision. I put a flip-flop on the Vsync pin to re-create a DDC signal and synchronise the circuit, it is more stable than with the DDC signal, I just have to set the switch the first time I go in 3D if it's not good and it stay synchronized until I shut off the projector.

[iondrive]

If you do a project like that, you can make it have variable times and then if you want to try something, you can try them with a fast LCD monitor. If you adjust the glasses to have long dark times and short clear times, you might be able to get partial good results with a fast LCD screen in 3d. I wanted to try this eventually but have been constantly distracted with other things. Good luck.

--- iondrive ---

[Petrus]

I upgraded my circuit with variable dark time, it's a bit tricky to adjust, but once done it works perfectly.
The flicker is hardly noticeable just like before.

An other thing I have noticed is that the 3DVision emitter is not very accurate when the CPU is loaded, that result in ghosting on vidéos, particularly on "Summer in Heidelberg".

I have resolved this by synchronising my circuit with the Vsync pin on the VGA plug. I only use the IR signal from the 3DVision emitter to know which image (left or right) is displayed.

[tritosine5G]

HM. Now I think about this , how this interacts with the fact that DLP micromirrors need time to accumulate light?

This is why 120hz DLP is so bad ,contrast is 1/3, brightness 2/3 gone.

[Petrus]

HM. Now I think about this , how this interacts with the fact that DLP micromirrors need time to accumulate light?

This is why 120hz DLP is so bad ,contrast is 1/3, brightness 2/3 gone.

I don't see what you mean by "DLP micromirrors need time to accumulate light"

Shutterglasses absorb 75% of the light ideally (50% by polarisation and 50% by shuttering)
It the real world it is worse. The best results I measured is 16,5 % of the light that pass trough the glasses with the 3D vision @120Hz (I don't measure it @85Hz, it would be a bit more).

With the color correction this brightness will be reduced by an other 1/3 theoretically, so the eyes sees only about 12% of the projector brightness.

The 2000 lumens of my projector drop to only 240 lumens in 3D, which is quite dark.
For example, Tomb raider underworld is almost unplayable when Lara Croft is in dark rooms, but in clear environments it is pure pleasure

DLP projectors is the best display method with shutterglasses in my opinion, it is ghost free and you can display a big picture.

[tritosine5G]

Well say you want to use 2 projectors, and use an external shutter in front of projection lenses. It wouldn't be good , because single chip DLP needs its time to reach the brightness level, all colors are multiplexed by the chip. Looks like at 120hz we are starting to arrive at a speed constraint.

http://web.uni-weimar.de/cms/fileadmin/ ... 004ipt.pdf

o well. Looks like you can't drive DLP shuttered. Also multiplexing the dlp chip wont let enough time for the micromirrors accumulate contrast. Ahh.... : (((

Acer pj is 3200 contrast, 2600 lumens, these are measured to be 2600 contrast and 2100 lumens @ 60hz.
120hz / 3d mode, 1050 contrast , 750 lumens. Uhh! * glasses 0.37 transmittance *0.5 duty cycle.

I think 120hz is too big price to pay with DLP, also shutterglasses is a deal im not going to accept. External shutter would work good with 2 pj's and mechanical shutter maybe.

I would try your 85hz solution. Im sure its better than 120hz?

[tritosine5G]

A reference is very good in that paper, where the author says we would want pulsed projectors for shutter use. Sounds like its true, I would prefer a polariser color wheel for single chip DLP. But 85hz then sounds slow.

[Petrus]

Can you explain what you want to build please ?
If I understand well, you want passive glasses with two DLP projectors but what are the shutters for ?

I really don't see why a 120Hz input would be darker than 60Hz. With a 4x color wheel speed, the projector will produce 240 full images per second (60*4) so the DLP chip is fast enough for 120Hz.

Looks like the reason why 120Hz is darker than 60Hz on the acer projector is just because of a poor design of the electronics before the DLP chip.

You increased my curiosity of how the DLP creates color shades and how fast are the micromirrors. I just need to make a super fast photodiode amplifier and see how it works, I will post the results here.

[tritosine5G]

DLP can do about 24.000 hz , thats not the problem here , but packing the light into timeslices is.

There are screens like the Kodak ektalite, which would provide even up to 16x gain over matte white. Then you would have problems with high gain, but thats easily resolved with a static iris on the pj, giving you improvement over the 1000 to over 4000. Colors are destroyed anyway, but maybe blacks survive?

All 120hz pj's exhibit this as far I know. Don know about 3 chips. LED dlp can be used I can send you a doc how to mod a standard dlp onto LED lightsource. You can throw the colorwheel out entirely. PM me your email adress pls. Contrast can be easily bettered with better light routing. Darkchip3 is very capable.

[tritosine5G]

Im trying to get local simulator cockpit builder fans interested in this.

http://books.google.hu/books?ei=Gnf6S4y ... N&start=40
very interesting. Ultimately , shutter should be used down in those tunnels if you get 10x gain on axis, with a 160 usd paint , and ektalite surface...

[tritosine5G]

BTW, how is your comp, and how smooth 85hz feels? Compared to 60fps / 60hz / 2d?

[Petrus]

I made some measurements on my DLP projector to see how it works.
I measured the light output with an oscilloscope while projecting shades of red :

100% : Somewhere between 100 and 75% :

75% : 50% : 25% :

As you can see the light is modulated to create shades of color.

As for the speed :

from 0 to 1 : from 1 to 0 :