registry entry for timing/blanking rate Possible generic LCD

[quadrophoeniX]

Hi there,

I am just goofing around with the 3d Vision set and wonder whether somebody has managed to fumble with the shutter on/off rate via the registry.

The reason I am asking and why this may become essential is that I found out that the limitation to native 120 Hz LCDs is actually completely virtual and could easily be circumvented by hacking this value.

Here are my findings:
Choosing generic CRT as display enables stereo AND shutter glasses, however.... although the sync is actually ok there is a ghosting of about 70% which means it is merely caused by the hold type nature of the display. If we could decrease open time (preferrably at the end of a frame) we also could increase ghosting down to the point of best compromise between ghosting and flicker/brightness reduction with native LCD support (through VGA) up to 75 Hz...

The other thing I found out is, that this could also work over DVI - we would just need a sync separation signal extracted from the DVI port...

Would be a shame if that couldn't be worked out somehow....

[iondrive]

You can read the "Post registry entries info here" topic and find the part about Parallax inversion. It mentions a GlassesSwitchDelay entry. It's for the old 3d glasses and drivers but maybe you can get a clue from there about what to look for in the new system. Also sometimes when you say LCD, I'm not sure if you mean LCD monitor or LCD glasses. You should try to help that guy out by posting a list of registry entries from the 3d-vision system.

[quadrophoeniX]

Hi, I've read your's and other posts before posting here...unfortunately the delay enable and delays for "minus" and "plus" (whatever that is) didn't seem to change anything...
Anyway omehow I feel - and hope - that's just the start....

oh, and by "LCD" i mean display... that shutter glasses mandatory are build with LCDs i tak as given and known....

[1140]

Hi there,
Choosing generic CRT as display enables stereo AND shutter glasses, however.... although the sync is actually ok there is a ghosting of about 70% which means it is merely caused by the hold type nature of the display.

I have tried cheap-ass shutter glasses on a 100Hz CRT giving no ghosting.
Tried it on an LCD but it gives about 100% ghosting on my DELL
Maybe you could force a black screen in between, but this would degrade color intensity.
So this would be outputted:
- frame0: Leftglass transparant / Rightglass blocked: Left view
- frame1: Leftglass transparant / Rightglass blocked: black
- frame2: Rightglass transparant / Leftglass blocked: Right view
- frame3: Rightglass transparant / Leftglass blocked: black

If you are able to leave the left shutter open for two frames, it possibly gives some acceptable results

[quadrophoeniX]

1140: the idea of inserting blackframes is not feasable since it would cut refresh rate in half giving 60:2:2 = 15 Hz per eye, you see? My thoughts when starting this thread actually were these:
fact 1: 3D Vision works with no ghosting in page flip on LCDs (!)... but only native 120 Hz ones - nvidia decided to block other LCDs
fact 2: they work with generic CRTs with any refresh rate with low ghosting in alternative shutter mode
fact 3: above can be forced on generic LCDs but shows massive ghosting. This has to be related to duty cicle of shutters on/off since...
fact 4: you can always go slower ( ) so if the system can handle 120 Hz rates with no ghosting there is no reason in the world it wouldn't be able at 60 or 75 Hz

Conclusion: we only need to find the parameter that controls duty cicle (or delay maybe) and adjust that to match timing for 60 Hz with 120 Hz
Solution: analize registry entries of a system with 3D vision and a native 120Hz LCD and compare these to the generic CRT entries.... so anybody with such a setting please post and I'll try to reproduce that, and for iondrive I'll post my registry tomorrow...
Alternatively I could think of deriving the Vesa signal from DVI and feed it into the transmitter while using the checkerboard mode.

I absolutely aware that this will be also affiliated with highly increased flicker, still I want to get this until full HD LCDS with 26' at least are available...

[iondrive]

Hi,

This is what I know about "GlassesSwitchDelay" and "GlassesDelayPlus" and "GlassesDelayMinus".
"GlassesDelayPlus"=dword:000004bb refers to the Alt-equals key combo.
"GlassesDelayMinus"=dword:000004bd refers to the Alt-minus key-combo.
"GlassesSwitchDelay" is a bit of a mystery but if you want to try something, do this. Set "GlassesSwitchDelay" to one, then start a 3d game, activate 3d mode, then hit Alt-equals like 10 times and the shutterglass sync should swap left and right timing. Hit Alt-minus some and it should switch back. The hotkeys did not work for me if I left "GlassesSwitchDelay" at 0. Tell us if this works for you if you decide to try it.

[quadrophoeniX]

Ok, on my task to get this thing to work with an ordinary LCD monitor/projector I played around with it an here is what i found:

external Vesa signal is only generated in CRT mode, 3Dvision trnasmitter gives ghosting as stated. So I put second video out in clone mode and hooked up my old revelator emitter to it, same ghosting - but at least it worked, which tells us the driver uses parts of the same mechanisms as the old one.
I further tapped sync signals by either the DDS signal of the primary DVI and by the horizontal sync of the second video out over VGA in clone mode and chased the signal through 2 monoflops and a PLL. Thus I could delay the sync signal up to 180° and more and reshape it to 50% duty cycle what both glasses need to sync up properly. (yeah, I know I could have used an Atmel, but my coding sucks while I am pretty good at breadboarding ) Unfortunately ghosting stayed as before, no matter which glasses I tried.

On a side note: revelator /H3D glasses do not work with 3Dvision emitter and vice versa. With the old glasses being synced just by page flip pulses this makes me believe 3D vision uses FSK, FM or some carrier modulation sync mode which would explain how they actually can achieve shorter aperture time than one frame by mantaining the above mentioned 50% duty cycle which seems to be the key for hold type displays.

I guess that this modulation pulse width is determined by the mode the driver is set to:

- The Generic CRT gives a Vesa signal for traditional shuuters but aparently can only work with pulse type displays.

- External sync of the 3D vision emitter only works without hickups in DLP mode but because the checkerboard cannot be easily decoded and the DDS signal does not carry syncsignal this is a dead end.

-Only the native 120Hz mode seems to give a open time with less than a frame cycle, but this mode is key locked by the correct DDS info and will not be available without such a monitor.

So if it is possible to get global support at all there are 2 options:
1. nvidia releases an universal driver
2. backengineering the emitter keying and adopting it to lower refresh rates

At this point I have no clue if one of both will eventually happen at all and if so who's gonna be first..... If ther should be progress I will start a new thread in the DIY section....

[RAGEdemon]

I further tapped sync signals by either the DDS signal of the primary DVI and by the horizontal sync of the second video out over VGA in clone mode

I am trying to get a vsync signal for stereo purposes over native DVI... have you managed this?

I'm not sure that I understand what DDS signal means or why you might be using the HSync signal... it's very interesting. Can you elaborate?

By any chance did you mean DDC and VSync?

I think I might be able to explain better the hurdles you have come across and perhaps help you overcome them... and I think your input would be invaluable for what I'm trying to do

Talk to me

-- Shahzad.

[RAGEdemon]

Hi RAGEdemon,

sorry for not checking my PMs on MTBS... I did not expect any response since you hinted you wouldnt check the Pms there too often either...anyway, I was indeed referring to our intiial converstaion about your post using the DDC data line on a DVI, and you can guess my surprise when I found the 3dvision drivers (with the USB emitter installed and connected) put out exactly the old signal when you choose generic CRT output. There is however no output signal on DLP (checkerboard) or anaglyph output.

What I have tried so far and all my resutls (also concerning crosscompatibility check with old (H3D/ MKetabyte + revelator) and 3Dvision glasses and emitters can be found here on MTBS.

As said I was able to get shutters flickering at 60Hz - 75Hz apparently in sync with the monitor but if I had to put that in nubers around 75% ghosting. Which means the glasses were transparent for 100% for the correct frame and 75% for the wrong side too which completetely eliminated the 3D effect. Allthough they were definately working at 60 Hz. And from the shutter speed I have the feeling 72 or 75 Hz on a standard TFT should be sufficient for §D - not as good as native 120Hz of course, but it's a "U-decide" thing, right? (pardon the pun).

I set up a little breadboard circuitry with a PLL and monovibrator that allowed me to derive a 50% duty cycle signal synced by the hor sync and shift phase (so delay the glasses sync and even alter the duty cycle and fed the signal into the minijack. The 3dVision glasses certainly accepted the signal and especially the alterations to the duty cycle gave weird symptoms - but again nothing that could be used for 3D.
I then decided to give it a shot on another monitor at home (my beloved 3007 Dell) but as meintioned this time the glasses wouldn't kick in. Fooling around with my breadboard again had some effects too.

My latest assumtions are these:
The driver may either need to detect a monitor which is capable of refresh rates above 60 HZ to kick in or the non working glasses at home are due to my 30''er is using both TDMS on dual link.
The strange massive ghosting could be related to overdrive that cannot be disabled on my monitor or it's so with every TFT due to the hold type display.

If the latter is the case this could be overcome by reducing the on time of the shutters. Old type shutters don't allow for this because the high/low signal switches both eyes alternating ther is no option for turning both sides on or of simultaneously.
If so, the 3Dvision glasses can do it and they indeed seem to use some keyed transmision which will be somewhat hard to re-analyse.

So yes, as long we are stuck with arbitrary limitations your idea sounds really good. I do not see, however, how you want to offer a blue line decoder with Vesa sync output without the need to pull out a soldering iron. Are you talking about a small interface dongle with DVI I/O and Vesa? I would be glad to assist if you can use any help. Also I would suggest to perhaps move further discusion to the open community on MTBS since I don't expect Andrew to be too happy if we start using the "IZ3D" term too often here - but then again, we wouldn't if we hadn't to, right?

And maybe, this time, they do get the message and start moving just a little bit? It's a easy "3 steps to heaven" program:
1. drop the 120 Hz limitation (eeeaaasy - and if it was a hidden expert option or registry tweak)
2. XP support (shouldn't be too hard with Quaddro drivers supporting it allready - and, while you're at it, add native Open GL support for Vista too )
3. customizable color tables for anaglyphs (also not too hard - was ther in the old control panel and 3D vision color correction shader proves adaptive filters)

[edit] As I understand, you are going to pick up the projected image and detect the frames with a Schmitt trigger? Funny - a plan I had a while ago and discarded because I had no idea on how to properly read out the parallax information, apparently you have one... It may be worth noting that the nvidia emiiter also needs a 5V present on the Vesa connector (see my post here) in order to accept the signal. [/edit]

Hi quadrophoeniX,

Please tell me if I have the wrong end of the proverbial stick here but are your experiments all on a TFT monitor?

You probably know all this but if a TFT is being used and i'm not misunderstanding you, that would explain why you are getting so much ghosting.

or it's so with every TFT due to the hold type display.

Exactly

TFT displays do not refresh the entire display one frame at a time which the glasses need to work. The pixels just change colour/brightness from the previous frame to the next one. Another problem is that TFT displays, even the best 120Hz gaming models have too high a retention rate to be useful if even the entire display were to be refreshed - I think eD tried this once with their LCD driver - they inserted a black frame every other frame to force a complete refresh from what I remember.

Reducing the off time should decrease the ghosting, if the whole screen refreshed at the same time, I agree... to what extent, I can't even begin to speculate. I guess it would be monitor dependent. It would be disastrous if the hold time of the pixel was more than 2 frames... which is the case with a lot of the older TFT displays... everything would just be a blur... I think anaglyph might be the only method of 3D on TFT based displays

You would ideally need a CRT or a DLP projector.

I find it fascinating that it worked at 60Hz at your work but not at home... I wonder what changed and if indeed a TMDS connection is the culprit as you suggest.

I came across a universal monitor driver which makes the graphics card think your monitor can support any frequency:

http://www.videoi.com/~pietro/monitor/" onclick="window.open(this.href);return false;

It should get past the 60Hz block of the monitor and it shouldn't damage anything. You shouldnt have to apply a higher refresh - just as long as the 3d vision driver thinks your monitor can handle it then it might work? It's installed from updating the drivers from system > hardware in control panel.

Can you get access to a CRT?

And I wonder what signal would be present on the input the the IR LEDs in the emitter. At some point ill try to put an oscilloscope on it... if nothing else it would be interesting to see what kind of encoding they are using. Perhaps we might be able to reduce the open time of the shutters.

If they took the points you suggested to heart and made a proper driver then they would get better business. I just don't know what their problem is.. they seem to be blind to what is obvious to most people: Give more features, give more compatibility, and above all, give customers what they want! If you don't, they will simply go elsewhere - as oh so many have.

Onto my side of things - As you know, I'm trying to get a sync signal on native DVI when it is interfacing through TMDS. You are quite correct that ideally a soldering iron would be needed. But alternatively, one can just twist wire around the pins of the optosensor and push bare wire into the input jack of the DLP connection (unless it's male?) assuming they already got the 2.5mm jack with the bundle - I did not. But it's no problem to make my own thanks to nobsi's schematic

I think you are correct - soldering will be needed eventually for others. I have a few soldering irons lying around so not a problem personally.

About picking up on parallax: the blue line code on iz3D's driver already does this... they draw a blue line at the bottom of the screen which only appears for frames intended for one eye and not the other. If I put the sensor here, it should give me a solid signal.

Sensor arrives today... I'll mess around and hopefully get something to work over the coming days

-- Shahzad

[quadrophoeniX]

Ah, I start to see where you coming from... At least I think so: you want to detect the blue line in the projected image with an opto to do the sync output, right? Ok, that would intuitively ship around lag/delay issues for sure. Still you need some sort of filter to separate the blue line code, electronically or optically. I suppose you are heading for the latter. First i though you would yuse an optoisolator for decoupling the sync, which would call for a PLL / oscillator stage like I played around with.

You understood correctly I was using generic TFTs and knowing the basics of stereoscopy I never expected the effect to be overwhelming - I just do that out of curiosity and to be prepared once bigger screens come out (I just do not have the space for projection). Still, the same problematic of circumventing the hold time goes for the 120Hz displays, so the 3Dvision drivers should have a compensation in form of adaptive on times built in.

Until we can access this anaglyph probably will be the route to go with TFTs (Allthough, I actually was using my shutters on Win98/GF6800/24''TFT). This why I also started a thread on customizable color tables for anaglyphs on nvidia's and IZ3D's forums - IZ3D gave a "it's on the list" already, the are much moe responsive. It's as you said: these limitations are really a shame and completely counterproductive. The more open the support the more people will jump on the mission, probabbly start with a pair of shutters and later move over to a 120Hz TFT (or DLP back-pro or projector). The people at nvidia however don't seem to follow this somehow...

I probably will have access to a CRT later this month and see what I get out of it. Also my DVI cable witht the DDC tap got dodgy - it didn't like the strees of the probe and running the long way from my monitor to the scope. I need to make a new one and this time I will make a Y-loom like the Revelator splitter, that should last longer. In the meantime I will give that refresh override tool a go an keep you posted here.

[quadrophoeniX]

By any chance did you mean DDC and VSync?

Yes, sorry for mixing that up. So to sum it up correctly, I was using your "pin7" solution on the DVI and by putting a DVI/VGA adaptor on the second output I could use an adaptor with 75ohm loads on the RGB lines, grounding the monitor ID pins 4&11 (emulating a native 1024x768 display) and tapping V-sync on pin 14. I put the second virtual monitor in clone mode and compared the V-sync of the secondary output with the shutter sync of pin 7 on the primary by triggering on the latter - both were absolutely synchronous. So I knew I dindn't need to hack further into the DVI cable. I reshaped that signal with a PLL to with variable pulse width and triggering a monovibrator with adustable duty cycle with the falling slope -> adjustable delay and shutter useable sync independantly from the stereo mode the drivers were set to. So I was able to verify working shutter on a TFT - at least on one eye - but hey, I'm "one eyed, hopeful"

Next up: a dual link DVI-DVI breadboard with with a header to tap every relevant signal plus 3pin Vesa socket over pin7 - calls for a proper multilyer pcb design in EAGLE, however.

[RAGEdemon]

Wow... that's some undertaking

I'd love to see what kind of signals are on those lines.

Someone else suggested using a an active DVI splitter and messing with the signals on the other port so they are isolate from the main line running from the card to the monitor. - instead of using a y-cable or splicing. Seems like a nice idea

Sorry or the late reply - have had family staying over for a few days.

Built the sensor which responds well... too well. It is picking up the individual revs of the colour wheel of the DLP I think, giving a LOT of noise on the square wave. Will try to clean it up using a couple of caps.



The sloping towards 0 is due to an improvised oscilloscope someone made using a soundcard on a computer. When looking on a proper oscilloscope, a pure square wave will be observed. Still, you can see all the noise.

Still, we can clearly see a square wave sync'd directly to the display over DVI. With a little cleaning and perhaps amplification, i think this should be a success.

Will investigate waveforms on a proper oscilloscope and make a stabiliser in the coming days.

-- Shahzad.

[quadrophoeniX]

Clever this, no idea haven't been thinking of using a soundcard based scope. Now I can proceed at home and don't need to run BNC wires across my lab .
The falling slope (understanding Fourrier transformations - what I sume both of us do) is natural with these kind of scopes since they are limited to the frequency bandwith of the soundcard input. It's a neglible effect for our purposes anyway. Still, I do prefer them good ol' CRT scopes everytime over a digital one...

Not sure which problems an active DVI splitter could introduce... and to get one just to find out it's not passing through all signals or delaying beyond reasonability seems a little risky....

[edit] is that signal with or without optical filtering? What I mean does that include the blue line parallax inforation allready or is it "just" the flickering of the projection?) [/edit]

[edit again] can't believe it, I really managed to put a typo in "Fourier" [/edit again]

[RAGEdemon]

SUCCESS!

Everything works great - Sensor mounted to a stick standing on the floor picking up the blue line directly from the bottom of the screen. This gives a clean double pulse - the colour wheel must be working at 120Hz to minimise rainbow artefacts. A 555 timer can divide the signal by 2 and another one can then make it into a 50% duty cycle Signal to output... luckily I had exactly this lying around from years ago in the form of my ghosting removing circuit. This processed the signal nicely, and finally inverted for correct parallax.

Beautiful Active Stereo3D on what used to be a non-syncing projector, using the iZ3D driver. No reason why the nvidia glasses shouldn't work on the DLP port either.

thanks for the explanation about the slopes... I had no idea it was due to the fourier transformations

-- a VERY happy Shahzad

[cybereality]

SUCCESS!

Congratulations man! I knew you could do it.

[quadrophoeniX]

Ah great.... can you tell me sensor type and source and if you applied optical filters etc? Does the sensor go straight into the first timer IC or do you use an op-amp buffer? I would like to give that a go as well.

I am just a little confused by "No reason why the nvidia glasses shouldn't work on the DLP port either." - Do you refer to using nvidia's emmiter/shutters which I thought you have, or to the "DLP port"? I mean, did that solve your 60Hz prblem with the nvidia shutters or is that still persistent?

It would good to see the 3Dvision shutters compatible with the IZ3D drivers because that would get rid of the stupid linitation to Vista and above and bring back XP support.

Regarding my workaround to make nvidia's drivers support different stereo modes through registry settings you can know find my progress on the official nvidia forums at: http://forums.nvidia.com/index.php?show ... ntry573476" onclick="window.open(this.href);return false; which, for a start, allows for customized settings of the anaglyph colors (links found in the same thread) and maybe more.

[RAGEdemon]

That is a remarkable find Klaus. Just amazing. thank You. we all appreciate yur work on this... hopefully people can experiemnt with the setting and we can finally get a more compatible driver.

From the blue line, the signal, although double per frame, was clean and did not need capacitors or buffers to clean it up.

I fed the signal straight into the first timer and out came the perfect square wave. the most difficult art is positioning the sensor to pick up the blue line, and then shielding it with black tape so other light does not refract through the crysteline casing and contaminate the microlens.

I am planning on writing a DIY tutorial but I wanted to try it with the 3Dvision glasses first too for completeness... finding a 2.5mm jack is taking a few days.

As you say, the problem of 60Hz incompatibility is very probably still an issue but above 62Hz, i don't think it will be a problem - a lot of users have bought 75Hz+ Stereo compatible projectors such as the BenQ 720p

If enough people are interested in making this thing supported by iZ3D then we can ask them to make a blue square instead of one thin line... should be a relatively simple for them as the line already exists.

thank You for your help and insight, Klaus

-- Shahzad.

[neol]

@quadrophoeniX, @RAGEdemon

am not entirely sure what exactly this thread is about. but did you guys succeed getting real 3D for 60Hz CRT/LCDs ? instead of 120Hz LCDs.