3D projector using Infitec method? - Page 4

yunti - "Will this work with frame sequential (pc source)"
defiancecp "As a note, frame sequential is losing importance"

You missed the point, maybe not what yunti meant, but what I want, it to be able to game at 1080p60. 3DTV Play won't do that.

"The box may include 2 HDMI inputs to support 2 full HD streams @ 60 Hz for gaming and presentations."

This is the vital thing. HDMI can't handle 1080p120 frame sequential, so for gaming you need dual hdmi.
Of course, it would be preferable to include a dual-link DVI input so nVidia 3D Vision would be supported (not using their active glasses obviously). Then this box could be used instead of a tiny 23" 1080p 3d monitor. At least with dual input we can use iZ3d or Tripledef drivers set to dual projector and get the color correction. I think this is a huge feature that you shouldn't leave out and will really sell this system.

I really don't think the frame interpolation is necessary. The projector can do that itself. They have been working on it quite a while and some have really good implementations.
Warren.

You are right. Expensive ones...
The idea was to get low cost projectors along with this box. That will appeal.
(More, FI has to be done only once at the box input, not twice in the projectors. Cost, cost... ).

Jack

That's a different thing then I was under the impression this was completely HDMI, so I was thinkiing about it from that perspective. If we're talking DL-DVI, that would be a REALLY nice addition.

The pannasonic ae-4000 LCD projectors are only 2k and have Frame Interpolation. But yeah, on second thought it would be nice to pair this up to some of the nice 1k projectors out now. The 2k projectors are nice, but not twice as nice.

DL-DVI would be the ultimate. I would pay 300 extra just for that option. But dual HDMI (seperate inputs for left and right eye) would let us game at 1080p with iZ3D drivers.
Not everyone is going to want to game on it so I would keep it a seperate gamer version. Actually it would be very wise from a marketing standpoint to hook up with nVidia and make a specific 3DVision version.

I think if you could price the regular version at $799 and the gamer version $999 you would do very well. Especially when you consider a passive 3D screen alone would be more than that.

It seems like you could strip down a couple BenQ W1000 projectors (less than $1000) and throw in a cabinet shooting out a single lens (like the lg cf3d) with this magic box built in and sell the whole thing for $3500 or so. That would be a pretty amazing 3D projector.

How about $299 ??
Jack

Actually, if we're still talking infitec, there aren't any unusual requirements for the screen.

I'd been planning on the 3d-theater, but from what I'm reading about it the color correction is a planned upgrade, not yet included... Plus nobody knows whether it'll handle the kind of correction needed for infitec...

So, if you release a box that will split hdmi 1.4 signals and definitely handle infitec correction for ~$500 or less I'm in. A little more than that and I'll consider... significantly more and I might have to pass.

299 would be awesome, I am just thinking 799 is the cost of two 3D-XL's.

"Actually, if we're still talking infitec, there aren't any unusual requirements for the screen. "
That was my point, if you buy the infitec adapter box you don't need a silver screen, so you come out ahead on money.

Yes, frame sequential, side-by-side, line interlaced, frame packed HDMI 1.4a, all usual formats, except Nvidia 3dvision which we are still investigating. Splitting and adjustment of all formats for dual projectors and color correction will be supported in all models from start; only frame interpolation is a postponed feature. Color correction will support both Infitec and Dolby, as well as anything else that could benefit from correction (or not - it can be switched off for polarized projection).
Input connectors, in addition to HDMI 1.4a, will support 2x DL-DVI, meaning up to 2x 1920x1080x32@60. Separate high end board will support 4K resolution.

About a cabinet/magic box - we are considering this and more; I cannot give more details yet.

I cannot confirm any price, but we are working with one chip manufacturer to make it low as possible.

That's very good news. Keep us updated.
Jack

Until frame packed supports 1080p at 60 fps then frame sequential definitely won't go away. Also nvidia doesn't support 3dtv play for 3 screens whereas it does with frame sequential. Those are 2 huge reasons for needing to use frame sequential. (note films aren't the only source for 3d - these boxes could be hugely useful for gaming too)

Sorry wnielsenbb. Just seen you have pretty much already answered this.

Pterodactyl if that's really true then you have me very excited. Bit worried to hope too much but will be amazing if you manage to get this out. (if you could ask your engineers to bear in mind lag then it sounds perfect)

I'm currently on a 3d vision surround which is limited to 3x 720p active projectors the best currently available with projectors (for a 3x setup).

Using 3 of your boxes will let me move to 3x 1080p screens with 6 projectors using 1080p DL-dvi input into each of the 3 boxes.


(I assume your comment on 3d vision just means 3d vision certification. At the very worst if it isn't certified we could use a hacked .inf to get it recognised so it works)

Let us know when you have the finalised spec sheet and pre orders open! Thanks.

Pterodactyl, I just found out about your project and I find it definitely interesting.
However since I'm doing passive polarised, I'm not that dependant on the Dolby3D colour correction than others.

What's more I currently drive my projectors from a PC, I'm doing it at FullHD 60Hz per projector, the software is quite buggy but it kind of works, so I don't really need your product right now.

The one thing I am worried about though is future-proofing my installation, and that's where your product interests me.

Over the next few years, as the industry will embrace 3D more and more, the 3D-ready software that will come will completely outperform my current software (because it's more hacks than actual functional software) but these future software won't support dual-projectors at all. I can already predict the only workflow they'll use :
Software -> GPU standard 3D API (eventually replaced with Windows standard API) -> single Hdmi1.4 or DisplayPort1.2 -> Display
So the feature that really interests me is splitting Hdmi1.4a or DisplayPort 1.2 to the projectors. But one thing that I absolutely want is the full FullHD @ 60Hz per projector.
So If I were to use Hdmi, the Hdmi spitting must support the optional 300MHz bandwidth of Hdmi 1.4a in order to support 1080p Frame Packing at 60Hz per Eye.

One more thing : there is currently no GPU that claims to support 300MHz Hdmi1.4a, so even if you released your product with it, I wouldn't buy it right now.
However AMD does claim to support DisplayPort 1.2 right now, so if you provided a DP1.2 input with splitting to dual-Hdmi for the projectors, that is something I could use right away, and it would already improve my buggy software nightmare.
I could even consider just simply a DisplayPort splitter and get separate Display Port to Hdmi converters.

I'll have to check the bandwidth figures. We are going to start with HDMI, but technically we can provide any kinds of video inputs and outputs; obviously it depends on actual market situation.
About polarized projection, we might add some assistance with reducing ghosting by subtracting adjustable amount of opposite channel's image. Would that improve the projection (we don't have a polarized setup yet in the office)?

Lag: there might be none, or there might be large Delay regarding color correction and splitting is practically non-existing, measured in nanoseconds. However, for some input formats where we are receiving left and right images sequentially, we must introduce additional delay of about half a frame, to output both images simultaneously to projectors.
Motion interpolation will affect things in a similar way: we must receive at least sufficient parts of surrounding frames to be able to start calculating intermediate ones. No help here. At least, if we support audio, we will keep it in sync.

Nvidia 3d vision: AFAIK it is not a simple 120 Hz stream of sequential frames, but additionally information about left and right frames is embedded in HDMI signalling in some proprietary way. That may require obtaining some licensing rights (as well as for just mentioning the name).
HDMI 1.4a is a different thing, "certification" here just means telling Nvidia "yes we are a HDMI 1.4a device, please add us to the list".

Preorder should be officially set up after several weeks (we've already completed the web, then realized we weren't happy with the design), but you can already send me a PM.

Ghost Busting would improve polarised crosstalk by a substantial amount.
My Silverscreen is one of the best currently available and it still bleeds enough to be noticeable in some situations.

I have never tried software ghost busting, I only read an article once about it. (not very detailed) So I'm supposing a lot of things here.

Non-destructive ghost busting would solve ghosting as long as the picture is not black (you can't substract light to the absence of light).
Since crosstalk on polarised projection is mostly proportional to light intensity, the only adjustments would be : max crosstalk (pure white on pure black) and a gamma curve compensation. Although fine tune settings would be a plus, just in case the projector's gamma curve isn't a regular curve.

Destructive ghost busting would raise the black levels in order to create enough margin to be able to lower the amount of light below the black colour, this is terrible for contrast but if you do ghost busting, it would be good to have the setting.

Thanks, that matches what I've had in mind. I'm 99% sure we'll include the feature.

Ghostbusting is a good idea. I have made experiments with AVS scripts and it works very well (except on blacks of course, as BlackShark said).
But projectors must be lined up on the pixel.
Jack

Hello pterodactyl, still hearing?

I think another add-on could be implemented in the box.

As Infitec filters would be implemented outside the projectors, in font of the lens, their efficiency will be closely related to throw ratio.

For those using long throw, this is not an issue. But if you use short throw (I am using 1.5), color shifts may show up in the corners of the image.

A very simple post process after the LUT output could improve the picture in some way, using a mask such as


For example just add some green in the corners if left filter and some red if right filter.

A single parameter (throw ratio?) could be added to the user interface. Could be ignored for long throw users.

Jack

The issue is actually more complex. If the wavelengths do shift due to oblique incident angle:

• they may fall outside of range of filters in glasses
• the brightness of the lamp in this part of spectrum may be somewhat different than color correction accounts for
• at greater incident angles, filters start to polarize light
• all projectors actually already compensate for reduced brightness at the corners of the screen, but different transparency of filters under different angle may throw that off too

I'm not sure just a brightness adjustment could compensate for all that.

However I think in reality none of these effects should be noticeable. Only a very little amount of the light will be lost; at least according to my quick geometry sketch far less than polarized projection "loses" at corners due to increased reflectivity of the screen (and I find even that acceptable). Anyway, I can confirm that we will verify what happens at 2.0 zoom setting and add the compensation if it is needed.

Thank you pterodactyl. Everything you said makes sense.

I was not speaking about increasing/decreasing brightness in the corners. The mask picture is only to shift colors using boolean. For example white means stay with corrected colors as the 3D color cube outputs it. Black means add some red/blue to the output color to make up for the color shift. That should be very easy to do but would require changing pixel color depending on its location in the picture. No need for a complex algorythm here. Any kind of loose compensation is welcome.

This issue happens when the throw ratio is below 1.7. Make your own test with a 1.5 throw ratio. For example projector sitting at 3m from a 2m wide screen.

During experimentation, I have made AVS scripts which were quite convincing but have given up as my CPU was not powerfull enough for real time.

I agree that, although this greatly improves color uniformity accross the screen, it doesn't prevent from the resulting crosstalk increase in the corners because of the wavelength shifts. Anyway, Infitec ghosting is so low that we could afford the increase....

Jack

Thank you very much for the suggestion, I'll make the tests and the correction looks possible to implement in our hardware.
Technically, increasing red in the right channel is easy, but increasing green or blue in the left channel may not work for all colors because green is usually already at the maximum, but it should help... I'll know more after experimenting...

One other somewhat related thing, we might be able to add automatic calibration with a light sensor built-in in the box. If mathematics proves correct, the box will find maximum overlapping hexagon in CIE color space, which should be an improvement over existing solutions that search for somewhat smaller overlapping triangle.

Pterodactyl,
How much of this new functionality is driving cost? I'm concerned that this thing will end up out of my pricerange.

Nothing or nearly nothing, otherwise I wouldn't be so quick to accept these suggestions
We are using a FPGA chip, one of the most recent and efficient, for all of the video processing. So we can add and change functionality (within reason, of course) while using same circuit board. Adding suppression of ghosting for polarized projections, or latest improvement by jack-bauer, just affects internal programming of the chip. Automatic calibration will need a $5 sensor.

Just make sure to stay within reasonable limits, forget any labyrinthine system... I used to work in microchip design. ha!

*whew*
That's a relief.

Understood

Hi Pterodactyl,

Quick question, if you guys will include infitec (dolby?) filters with your box, any idea how big in size these will be?

We don't have that defined yet, but we hope to be able to offer both Infitec and polarized filters. The size should be appropriate for most or all of home theater and smaller business projectors - I suppose filter diameter will be between 1" and 2". The whole mechanism (we plan to have automated mechanism to put/remove appropriate filters in front of projectors as necessary for 2D and 3D) will be thin and occupy minimum needed area of the front of a projector. This, however, may still interfere with heat exhaust vents in some cases, so we are still deciding what is the most optimal and flexible solution.

Why, do you have a particularly large or small projector in mind, for which default size(s) may not be suitable?

Well, I've been building my own projectors and am looking into ways to make the move to 3D. DIY projectors usually have very large projection lenses, so that could be a problem for the Infitec method. The one that I'm planning to use is 4" in diameter. I shouldn't need the full 4" though, but 2" will probably be very tight. I guess I'll have to do some measurements once I build it to see what size I'll need and if it doesn't fit, I'll just have to stick to polarized.

Right now I am using 50x50mm filters from Dolby and they are exactly the size I need (throw ratio=1.5). Smaller ones would not be OK. Filter frame touches the lens mounting when in place.
http://www.avsforum.com/avs-vb/showt...7#post20536167
Jack

Thanks for letting me know, I'll discuss that with our chosen filter supplier (that might be Infitec). From technological point of view, there is no problem, these are all reasonable dimensions. Otherwise, cost of the filters, even without any licensing rights, is directly proportional to filter area - that's why we are trying to "minimize" them.