Hacking Nvidia 3DVision

[Petrus]

I use 3DVision glasses and a 85Hz DLP projector as my 3D system, it works great but I have to wear the glasses upside down to play games and it hurts my nose.

To solve this issue, I've programmed a micro-controller that synchronise on the IR signal from the 3D vision emitter and re-emit it reversed with variable delay.
It also outputs a DDC signal to drive my other shutterglasses.

If you want to build your own, here are the timings of the Infra-red pulses send by the 3D vision emitter :



Next step : Add remote control operation.

[edit] I updated the graph with more precise measurements.

[cybereality]

I've got no idea what to do with this information, but nice job!

[wireheadrick]

I felt I was misled when I bought the Nvideo 3d Vision (plus an extra pair of glasses) system thinking I could use them with any content driving my Mitsubishi DLP display. The 3 pin Vesa cable was the primary reason I thought the Nvidia system would work (USB connector to the emitter was for power only - simply buy a cellphone charger or the like to power the emitter - I thought). With the proper timing from the Vesa connector and 5v power from USB the emitter system has all the data it needs to be able to do it's thing in standalone mode. I guess Nvidia thinks otherwise and purposely left this capability out.

Well now I am ticked off. I could return it all for a refund but how long will that take and I just want to watch my own personally created 3d videos using stereoscopic player. I am not interested in any PC gaming in 3d (I'll stick to my PS3 and I suspect lots of other PS3 users will be interested in our hack when the PS3 starts supporting 3d). No other 3d glasses are in stock right now (anything with DLP link is sold out).

I am a seasoned and experienced electronics engineer so I think I could make this work in stand alone mode despite the Nvideo anti-consumer approach. Your diagram of the waveforms is exactly the info I need to move forward. I am prepared to hack my Nvidia emitter (with a microprocessor if necessary). I know IR drive signals are easy to create with the correct timing parameters.

So the question I have to ask you about your diagram is how did you measure these voltages (currents)? Any digrams/schematics on how you got these signals out of your emitter is greatly appreciated.

I have my emitter open and can easily see the large array of IR emitter LED's. I cannot get it to work because I don't have Windows Vista or Windows 7 or a proper 8XXX series Nvidia GPU (and I'm not buying either if I can help it). I am completely dependant on someone else tech savvy enough to connect up a 'scope and see the IR signal driving the IR LED's. All I have to do is copy that (LR polarity don't matter - my TV has the polarity switch in the menu) and sync it with the VESA 3 pin connector signal using a microcontroller (probably a PIC).

Any help or further clarification is greatly appreciated.

Rick

[wireheadrick]

Oh and can you get the timings for 60 Hz mode?

[wireheadrick]

Nevermind. I think I figured out your diagram. The 500uS was throwing me off.

That is until I studied the chart and realized the 500uS is the time that _both_ glasses are off. I suspect this value of time could well be adjusted to "tune" the glasses to minimize ghosting. Obviously the time between the two waveforms in your chart is dependent on the screen frequency. For example I would imagine the time between the left off/right to the right off/left on would be 1/60 or 16.7ms. It doesn't really matter since I will pick up the screen timing from the DLP 3d Vesa 3 pin connector.

I'm going to use an FPGA to do my first try since I can be more precise with the timing. It will let me experiment more until I'm ready to go to SW only with a processor.

What I already know:

The Vesa 3 pin connector has a gnd, a power (+5v), and a square wave signal (+3.3 v and gnd) with a frequency of 60Hz. I am assuming this square wave indicates left on/~right on (or visa versa), meaning when the square wave is high (8.333ms) one shutter is on and then when it is low (8.333ms) the other shutter is on (for a 120Hz DLP 3d display).

Well at least that's my starting point.

I'll report back on my results.

[Petrus]

I measured the timings with an oscilloscope. First, I used a photodiode with a pull-up resistor to get voltage change according to IR pulses.
For better précision, I measured the signal directly on the board.
I updated the graph with these more precise timings.

The shutters doesnt shut off instantly after the command are sent, they take 2.3ms from 100% opaque to 10% as I measured with my test bench :
http://www.mtbs3d.com/phpBB/viewtopic.php?f=26&t=8489" onclick="window.open(this.href);return false;
That's why they put 500µs between OFF and ON commands, we can change this timing and send ON command before OFF command to minimize ghosting on CRT/plasma monitor.
I personnally do this to have the same color on left and right image on my 85Hz DLP projector. It also eradicate a light red ghosting I had with standard timings.

[wireheadrick]

I successfully got the Nvida glasses to work with my Mitsubishi DLP.

Your timing chart worked perfectly!!

Thank you for posting it.

I will now move to a uController (PIC 12f675). The FPGA is impractical for a real solution but is great at being flexible in getting something working quickly.

I also will experiment with timing. Right now I do a simple edge detect of the DLP sync signal and then start the timing sequence from your diagram. Next I want to do an internal DLL which will allow me to slide the timing back and forth around the sync signal.

I also want to experiment with the 500us both off timing.

I'm kinda new to 3d. Any recommendations for a 3d test source for evaluating ghosting? I have some ideas but maybe the experts already have something?

Thanks again for the info!

[Petrus]

There are already some test patterns on the forum to measure ghosting :
http://www.mtbs3d.com/gallery/displayim ... ?pos=-3137" onclick="window.open(this.href);return false;
http://www.mtbs3d.com/gallery/displayim ... ?pos=-3136" onclick="window.open(this.href);return false;
http://www.mtbs3d.com/gallery/displayim ... ?pos=-3135" onclick="window.open(this.href);return false;
http://www.mtbs3d.com/gallery/displayim ... ?pos=-3134" onclick="window.open(this.href);return false;

Are you programming your µcontroller in assembler or C ?
I programmed my Atmega8 in C, it runs currently on the internal 8Mhz oscillator, but i will upgrade my circuit with a 16MHz crystal to get more accurate timings.
For the delay/advance setting, I use a compare interrupt of timer1
Basically, it works like this :
On each Vsync pulse (I use Vsync on the VGA plug as time base) I load the compare register with the current timer value + delay. once the timer reach the compare value, an interrupt is pulled and send the IR pulses. the IR pulses are handled with the other compare interrupt of the timer.

[wireheadrick]

Thanks for the links.

I program in Assembler for the PIC. I'm too cheap to buy the $400 C compiler.

I will probably do a simple fsm based on a timer count. I think I can get 3 uS resolution with this technique. I plan on implementing a simple low baud (300) rs232 CL/monitor port as well so I can adjust the timings on the fly without re-programming the chip. Eventually when the timings are solidified I will permanently code them in.

It won't be perfect but I love the 8pin SOIC form factor of the 12f675. It has an internal osc so the hw is very simple.

I plan on putting it all into a FOB like infrared emitter with a 3.5mm stereo connector to connect to the VESA 3 pin connector on the DLP. Power comes from the DLP so this is a pretty easy project I figure.

The hard part is done thanks to your diagram.

[Slah]

Great work! Could you please post the circuit? I also want to sync my 85hz DLP.

[Petrus]

I dont made a schematic for the moment, all the circuit is on a breadboard :

[Petrus]

As you can see, for the moment its just a bunch of wires that connects to different pins of the micro controller, no need of schematic for that.
The only interesting part would be the Comparator that converts the IR pulses from Nvidia's emitter to TTL voltage, here is the schematic :


And the circuit board in the bottom is a TTL to RS232 level shifter i use it for debugging purpose.

[Slah]

You cou please explain how I can reproduce this? I'm not very common with advanced stuff like micro controllers. I'm trying to sync my 85hz DLP projector.

Thank you!

[Petrus]

If you cant program a microcontroller, I'm afraid you have to keep wearing your glasses upside down.
It would be way too complicated to do this without a microcontroller.
If you have some electronics and programming skills, I strongly encourage you to learn how to use them, you can make so much stuff with microcontrollers !
You can begin with an Arduino board, there is a huge community of developers on this system.
http://www.arduino.cc/" onclick="window.open(this.href);return false;

[Petr]

I trying to create my 3DVision emitter. Now my problem is the wavelength of the IR-led. I have tried to use one led with 850 nm wavelength. I know that ir receiver needs a small level of signal to react. But the glasses reacts only on 20 cm. I assume that wavelength of native 3DVision leds is different. If anybody knows it, please share this info.

[Likay]

It's possible that the wavelength isn't that critical. It might be the drivercircuit for the ir-led that needs to speed up, especially when turning off the led (i have own experience of making ir-transceivers). One way to try is (if possible) removing the ir-filter in front of the nvidiaglasses sensor, dim the room and see if they work better (works better = wavelength issue, no change = timing issue). The ir-sensors themselves use to accept a quite wide range of ir-light and often even visible light. I'm a little reserved to really assume it's a fact though.

[Petrus]

I had also a range problem on first prototypes, I had to put 6 led and drives them with 1Amp pulses for each led to made it work with my setup (front projection with 3m between the screen and the emitter).
I don't know the wavelength of the diodes I use, maybe they are the wrong one.
I still need to modify my circuit to get more efficiency.

An other thing, I noticed the pulses timings have to be very precise. After a timing adjustment I got a significant range improvement.

[Petr]

Another question: the receiver(or shuter-glasses logic) reacts on signal level (for example switching on 0.5 using the max and min level as 1 and 0) or it reacts on speed of level changing?
If the second one, then the time of full disabling/enabling of the led is not critical, otherwise every type of led need to have own shifts for disable-time (pulse device has to provide specific timings for every type of led panel).

[Petrus]

The voltage of the LEDs on the emitter is sharp so it works with signal level (_|¨|_) ,not signal change ( _/\_)

[awmayhall]

hey petrus im new to the forums here but i have been working on this little project of my for quite some time, and it seems that you have fond an answer to the problem ive been having, currently im using an array of 555 to generate the signals for the nvidia 3d vision it works in the sense that we can get the 36khz required but the pulses for the left eye on and off along with the right eye on and off are not present, i can build a micro controller for this thing but im running into problems... wondering if you could post your source code for your micro controller. I plan on attempting to build something similar if you had a schematic it would be a lot easier for me. But my main concern is i would like it if at all possible to get the code so it would be a little less time consuming. I myself am a rather experienced programmer however this project has been a big headache. It would be such a big help but i understand if you don't want to post your code. Well Congratulations on your design something that Ive been attempting for several weeks and still continuing to fail... thanks for the valuable information its helped a lot!

[Petrus]

The schematic and code are already on the forum here : http://www.mtbs3d.com/phpBB/viewtopic.php?f=26&t=10183" onclick="window.open(this.href);return false;
I updated it today.
Keep us informed about your work

And for my IR led problems I resolved it by soldering The IR emitting circuitry on a separate board to avoid the high resistance contacts of the bread board.

[awmayhall]

haha sorry i didn't see that, and thank you so much ill keep you updated defiantly

[baobrien]

I'm also working w/ awmayhall on the project. I've used the timing diagrams to design a board using 555 timers to generate a signal. 12 in total. I'm stilling calculating the R's and C's, but can post the schematic and board soon.

[baobrien]

Schematic for 555 based signal generator
Boardfor 555 based signal generator - uses 'fat' traces to allow for easy toner-transfer

All done in KiCad

[Petrus]

Hi baobrien
So you made a chain with 555s wired as monostables, each circuit triggering the next one, right ?

But why some 555s have common trigger input ?
Did you made a Pspice simulation ?
Pspice student version is free and can simulate 555s, you probably cant simulate all your circuit at once without exceeding the limitation but it worth a try and it will help me and others to understand how your circuit works.
http://www.electronics-lab.com/downloads/schematic/013/" onclick="window.open(this.href);return false;

You can also make a new topic about your project, It will certainly interest other members like Slah, he asked me how to reproduce my project without microcontroller. I tell him that it will be too complex but I didn't though about chaining 555s.

It will be great if we can do an old-school analogue nvidia emitter that anyone with basic electronics skills can build.

By the way, what type of display will be synchronized with this emitter ?

[baobrien]

I'll draw up a timing diagram of what i've got going on. There's probably an easier way to do it with 555s, but their dirt cheap, along with the 74' chips.

[baobrien]

Timing diagram for left-to-right transition

Also, I finished the schematic from earlier.




This is designed to allow the glasses to operate on any monitor, regardless of weather it is '3d enabled' or not.

Description of microcontroller interface:

sup - Suppress - when pulled high, the signal is not allowed to the IR LEDs
es - eye set -when pulsed high, the eye flip-flop is loaded with whatever is in ed
ed - eye data - selects eye to be used during this and the next cycle

es and ed are used to sync with whatever software is running.

All custom software must be used with this.

[Petrus]

I just figured out that a 555 cannot be triggered by the output of an other 555.
A low state on the trigger input will force the output to 1.

I think a 74221 monostable is more suitable for this application. :
"Once fired, the outputs are independent of further transitions of the A and B inputs and are a function of the timing components" from 74221 datasheet

[baobrien]

I was afraid of that. I'll just have to use a microcontroller.

[baobrien]

It might work if I threw an or-gate onto each timer so that the output of the timer gated the trigger input


Of course, at this point, the design is just conceptual, and won't be produced.

[Petrus]

I made a circuit to use my system with DVI connexion.
The circuit is composed of two sub-circuit : A EDID (Extended display identification data) EEprom and the Sync extractor :


If I connect my optoma projector directly on DVI, the 3D vision driver switch immediately in discover mode.
I had to wire an I2C EEprom on the DDC data and clock pins, the EDID on the EEprom replace the EDID of the projector and fool the 3d vision driver that believe there is a compatible display connected.
This part might also interest those who have problems with not available resolution/refresh rate.


The second part is the Sync extraction circuit, it extract a single sync pulse out of the DVI blue signal.


It works like this :

DVI blue signal :

Signal A : Blue-
Signal B : Blue+


The sync pulse is extracted from the Blue signal with a transistor based differential amplifier (signal "RB")
The signal is then shifted by C3, R11 and R4 just above the 555 trigger level (VCC/3).
During a sync pulse, the signal goes below VCC/3 and trig the 555 monostable.
The result is a positive pulse (signal "B").

The circuit works fine, but I have an other problem. There is some flashing dots on the black area of an image even if I plug the projector directly on the video card, without my circuit in between.
This problem doesn't appear on my LCD monitor.



Is someone already resolved this problem ?

[Petrus]

After some measurements, I found that the differential impedance of the DVI input of the projector is only 60ohms instead of 100ohms on the LCD monitor.
The DVI specification says the differential impedance must be 100ohms +/- 15%.
Apparently, the projector doesn't follow the DVI specification I think this impedance mismatch is the cause of the flashing pixels.

[tritosine5G]

-yeah there might be reflections if not terminated properly.

[atariguy]

I thought you could use 555s to trigger each other, like this

http://www.play-hookey.com/digital/time ... encer.html" onclick="window.open(this.href);return false;

or am I (most likely) misunderstanding?

[Petrus]

I thought you could use 555s to trigger each other, like this

http://www.play-hookey.com/digital/time ... encer.html" onclick="window.open(this.href);return false;

or am I (most likely) misunderstanding?

Yep, with the additional Capacitor/resistors on the inputs (like on your link) it will work.

[Chris1976]

I also have IR distance problem while I was happy with working this working IR timing. 1m is the maximum even I drive three emitters with 200mA each. Finally I bought 850nm and 940nm for testing and got the result. 940nm is the better one which gives me 5m distance wide angle with 3 emitters of 200mA each. You guys can try 940nm emitter for 3D Vision. Thanks for everybody's discussion because it became the best glasses.

[atariguy]

Yes, I found 940nm LEDs to be by FAR the best solution.

Thanks so much for the DVI sync circuit, it works a treat, and is so easy to couple with $5 of parts to drive just about any glasses.
Cheers!

[kuquet]

I understand it is possible to make a synchro extractor from a DVI signal, or HDMI.
I was wondering if this principle could work with a PS3 or a bluray player 3D, a video projector, 3D glasses operating with an infrared transmitter equipped with a VESA input pin mini-DIN-3?

[icesterftl]

I made a circuit to use my system with DVI connexion.
The circuit is composed of two sub-circuit : A EDID (Extended display identification data) EEprom and the Sync extractor :


If I connect my optoma projector directly on DVI, the 3D vision driver switch immediately in discover mode.
I had to wire an I2C EEprom on the DDC data and clock pins, the EDID on the EEprom replace the EDID of the projector and fool the 3d vision driver that believe there is a compatible display connected.
This part might also interest those who have problems with not available resolution/refresh rate.


The second part is the Sync extraction circuit, it extract a single sync pulse out of the DVI blue signal.


It works like this :

DVI blue signal :

Signal A : Blue-
Signal B : Blue+


The sync pulse is extracted from the Blue signal with a transistor based differential amplifier (signal "RB")
The signal is then shifted by C3, R11 and R4 just above the 555 trigger level (VCC/3).
During a sync pulse, the signal goes below VCC/3 and trig the 555 monostable.
The result is a positive pulse (signal "B").

The circuit works fine, but I have an other problem. There is some flashing dots on the black area of an image even if I plug the projector directly on the video card, without my circuit in between.
This problem doesn't appear on my LCD monitor.



Is someone already resolved this problem ?

Do you mean to say that nVidia 3D Vision outputs blu-line sync when configured for DLP-Link projector?
Or is it your own software Stereoscopic Player that inserts the blu-line sync pulse?

The experts are claiming that there is no way to get correct eye phase for 120Hz DLP-Link projectors to work with DLP-Link LC shutter glasses.

Please advice.

icesterftl

[Petrus]

The sync extractor I designed do not extract blu-line L/R sync, it extract the vertical sync of every frame to synchronize my IR emitter. I synchronize the L/R with the 3D vision emitter.

The vertical sync is sent to the monitor via the blue channel. I think that's why you misunderstood the function of the circuit.