The Cube

[geneb]

Brad & I (anyone else?) have the unique problem of having to design a cube that's got to fit around 2,000lbs of un-hackable airframe. Rob & I were checking again yesterday and we may end up building a 12' cube to ensure that we clear everything we need to. Even then, the forward screen will have a seam through the middle of it so I can still open the avionics doors in the nose barrel. I'm sure Brad has similar issues.

 

g.

[BHawthorne]

Yep, my reasoning for going pvc pipe is it's dirt cheap and I can make one screen frame for my room and a separate larger screen frame for my F-84F trailer. I'm thinking about using Rose Brand Tendo again for the screen but going with gray instead of white for better black levels and contrast. Gray material will also better eat stray ambient light.

Edited March 4, 2013 by BHawthorne

[memento10]

What do you guys think about using these projectors? First Full HD Short Throw.

 

http://gadgets.ndtv.com/tv/news/benq-launches-full-hd-short-throw-video-projectors-starting-rs-100000-317196

[geneb]

BHawthorne, the problem with using Mylar as a fold mirror is getting it _perfectly_ flat. The best method is to try a regular mirror to test the concept and then replace it with an acrylic first surface mirror for "production" use. However, if the take-off angle is narrow enough, you won't see the ghosting caused by a second surface mirror and can continue to use that.

 

g.

[agrasyuk]

What do you guys think about using these projectors? First Full HD Short Throw.

 

http://gadgets.ndtv.com/tv/news/benq-launches-full-hd-short-throw-video-projectors-starting-rs-100000-317196

 

Hmm. Makes me wonder what looks better : lower detail 1080p or high detail 720p

And also price per projector.

[BHawthorne]

I'm a bit torn on the 1080p vs 800p. I want 1080p because of the pixel density but I want 800p because you could be pushing up to Eyefinity 5x1 on things. It depends on what you prioritize most, FPS or pixel size. Ultimately you can do 800p for half the cost and it'll be faster. If you have budget for it and a machine that can push 1080p Eyefintiy 5x1 at good frame rate in DCS I'd be all for it though. Personally, I go 800p projectors because build cost is exponential with them. 5 x $500 sounds a lot better than 5 x $1200 to me personally. Price just gets outrageous when you go exponentially with 1080p. You could build a great simpit for the screen with the difference in cost.

Edited March 4, 2013 by BHawthorne

[Flim]

No matter what projector you use.... you will have to run 800p for performance! This is for a 4 or 5 projector setup...

[metalnwood]

Personally, I have had enough of hearing about how cheap you guys can get projectors in the states! lol :)

[Deadman]

Sux don't it;)

[memento10]

All good points re 1080 vs 800 but it would be 'future proofing' as you could still run the projectors at a lower than native resolution to get reasonable performance. Since the game is CPU bottlenecked, and only uses 1 core, if ever the code is upgraded to use more cores or CPU's get fast enough, you could then switch to 1080p.

 

I agree that it all comes down to budget but maybe starting with three 1080's for now might be a good compromise. Not sure what you could sell used projectors for a couple years from now if you wanted to upgrade.

 

Just a fantasy for me now though... :)

[Flim]

All good points re 1080 vs 800 but it would be 'future proofing' as you could still run the projectors at a lower than native resolution to get reasonable performance. Since the game is CPU bottlenecked, and only uses 1 core, if ever the code is upgraded to use more cores or CPU's get fast enough, you could then switch to 1080p.

 

I agree that it all comes down to budget but maybe starting with three 1080's for now might be a good compromise. Not sure what you could sell used projectors for a couple years from now if you wanted to upgrade.

 

Just a fantasy for me now though... :)

 

I don't want everyone to think this is easy! It takes a lot of time to get it right...

 

I have spent days tweaking, moving, adjusting, building, and then you will spend hours tweaking the PC trying to get the smoothest possible image.

 

Imho, it will be a long long time before 1080p would work... first you need projectors with the throw of the gt750 at 16:10, if not... it will be very difficult to make it work. Second, you will have to run it all on one card, 2 cards causes image sync issues... lastly running 9900 x 1080 will not work on a single card. You can see there are many things to think about!

Edited March 7, 2013 by Flim

[BHawthorne]

I agree.

 

Price vs Resolution and Performance vs Resolution are both very valid issues at the heart of this kind of build. Most of you simply have not used a wall sized projection before therefore do not understand that regardless of the resolution if you have full visual coverage at what you're looking at you get quite the immersive experience. That is the ultimate goal. Splitting hairs over pixel density also shows a fundamental lack of understanding on just what it takes to push 5x1 Eyefinity at a reasonable FPS. Most of you need to realize that while 1080p projectors sounds appealing in theory there are a lot of issues that they bring that are not there with 800p. For one, aspect ratio of projectors is one thing not mentioned at all so far in this thread. In an ideal world we'd have 1600x1200 5:4 projectors. The wider the aspect ratio the less suitable the projector is for doing a symmetrical cube. 5:4 <- 16:10 <- 16:9 in order of preference. As is, 16:9 is a squat rectangular cube. The wider the aspect ratio gets the shorter your vertical fov is for the room. If anything 16:9 projectors are the least ideal of the available aspect ratios for something like this. The 800p 16:10 is a compromise that works primarily because of lens throw and price point. All of that aspect ratio issue can be resolved quite easily with pre-warp correction of the projection width to a more 1:1 aspect ratio, but many people do not want the added cost of pre-warp software. For me I prefer using pre-warp software because I want control over the projection not the projection have control over my design. I can force projectors to do more atypical configurations with pre-warping on my terms. Instead of spending frustrating hours on alignment location for mounting I just ensure full panel coverage with slight projection bleed and dial it in over the course of 10-15 minutes using pre-warping software. For me the price is worth the lack of frustration involved. It's trading convenience for money and man hours though. Depends on your build motivations and budget.

Edited March 6, 2013 by BHawthorne

[memento10]

Thanks for the explanation BHawthorne. I didn't know the aspect ratio preference for 5:4. I'll have to get my head around why a more symmetrical cube is preferential...

 

Is it correct that in a perfectly symmetrical cube, the top projector would completely cover the ceiling as opposed to the rectangular projector only covering about half the ceiling?

 

Using the more rectangular projection of 16:10, do the side projections 'match up' with the ceiling projection?

 

Sorry if my comments or questions are trivial to you guys... just want to learn.

[PeterP]

Thanks for the explanation BHawthorne. I didn't know the aspect ratio preference for 5:4. I'll have to get my head around why a more symmetrical cube is preferential...

 

Is it correct that in a perfectly symmetrical cube, the top projector would completely cover the ceiling as opposed to the rectangular projector only covering about half the ceiling?

 

Using the more rectangular projection of 16:10, do the side projections 'match up' with the ceiling projection?

 

Sorry if my comments or questions are trivial to you guys... just want to learn.

 

 

Have a look at this post:

 

http://forums.eagle.ru/showthread.php?p=1641832#post1641832

 

Watch the video and compare it with the setup-lua "plan B". and the following post that shows the model.

 

You can see that each view-port has a aspect-ratio of 1 in the lua file.

 

If it is not 1:1 (Width /Height) you wouldn't be able to 'fold' them to a cube and the image would be distorted/not fitting at the edges.

 

But because the Projectors are not able to show everything of the view-port we have to drop parts of it.

 

In other words : DCS is rendering a full 1:1 image in 1280x1280 for each view-port but Flim is only able to project 1280x720 of it with the old projectors .

Edited March 7, 2013 by PeterP

[BHawthorne]

Thanks for the explanation BHawthorne. I didn't know the aspect ratio preference for 5:4. I'll have to get my head around why a more symmetrical cube is preferential...

 

Is it correct that in a perfectly symmetrical cube, the top projector would completely cover the ceiling as opposed to the rectangular projector only covering about half the ceiling?

 

Using the more rectangular projection of 16:10, do the side projections 'match up' with the ceiling projection?

 

Sorry if my comments or questions are trivial to you guys... just want to learn.

 

It typically isn't the case with multi-projector because most people using them in a 3 panel curve or a uniform curve design which in those cases to get more FOV you'd want the wider aspect ratio projectors. It's opposite the case with a cube configuration. You want a 1:1 if possible.

 

Yes, in a 1:1 aspect ratio setup the top would be 100% covered. I'll be able to cheat though with pre-warping and force the full ceiling projection.

 

The questions aren't trivial at all. I find them important to screen design. No worries. :thumbup:

[memento10]

Have a look at this post:

 

http://forums.eagle.ru/showthread.php?p=1641832#post1641832

 

Watch the video and compare it with the setup-lua "plan B". and the following post that shows the model.

 

You can see that each view-port has a aspect-ratio of 1 in the lua file.

 

If it is not 1:1 (Width /Height) you wouldn't be able to 'fold' them to a cube and the image would be distorted/not fitting at the edges.

 

But because the Projectors are not able to show everything of the view-port we have to drop parts of it.

 

In other words : DCS is rendering a full 1:1 image in 1280x1280 for each view-port but Flim is only able to project 1280x720 of it with the old projectors .

 

So if it's rendering 1280x1280 but projecting 1280x720, is anything being cut-off from the view or is the graphics card just downscaling the vertical resolution from 1280 to 720? If so, I see the performance hit this setup takes just to stay symmetrical...

[BHawthorne]

So if it's rendering 1280x1280 but projecting 1280x720, is anything being cut-off from the view or is the graphics card just downscaling the vertical resolution from 1280 to 720? If so, I see the performance hit this setup takes just to stay symmetrical...

 

Actually, those projectors are 1280x800, not 1280x720. There shouldn't be any real hit vs running 1280x1280 vs 1280x800 other than pixel count difference. Less pixels = faster FPS.

 

The GT720's I have are capable of being fed non-standard resolutions like 800x800 if you create a custom resolution with NVIDIA cards. AMD cards are a bit more difficult to do custom resolutions though. I'm sure you can do 800x800 on AMD with a display inf hack though. It's something I need to experiment with. I have more experience with custom resolutions on NVIDIA than on AMD, although I have more spanning experience with AMD.

Edited March 10, 2013 by BHawthorne

[memento10]

Don't want to take this thread off topic but I see many simulator companies quoting the vertical field of view in degrees. For example, 270 degrees horizontal, 80 degrees vertical...

 

http://www.simpit.co.nz/

 

Is the vertical field of view determined by the number of pixels in the vertical resolution? Or a ratio of the horizontal to vertical? What would be the vertical resolution of this cube?

 

Does the 'zoom' level not affect this number?

Edited March 14, 2013 by memento10
added question

[geneb]

Pixel counts and FOV don't go hand in hand. There's no reason a scene generator couldn't provide that same FOV with a resolution of 640x480. You'd have a massive "screen-door" problem (from the pixel sizes), but it could be done.

 

g.

[geneb]

Here's a question - is there any reason a rear-projected display cube couldn't be built to a very small scale? Say 60" to a side? 48"? As long as your head is in the physical center of the cube, it should work. Correct?

 

g.

[BHawthorne]

Here's a question - is there any reason a rear-projected display cube couldn't be built to a very small scale? Say 60" to a side? 48"? As long as your head is in the physical center of the cube, it should work. Correct?

 

g.

 

No real reason other than you have to be mindful of the material you use for the cube. Maybe some sort of fabric and you have to be doubly mindful of how the fabric is tensioned on the edges and make sure the frame does not get in the way of the projection.

[agrasyuk]

measured the to be sim-room in the new place: room for the Cube, couple desks and storage for electronic junk and some more. awesome! but vertically it is exactly the standart 8ft high... remembering that the "native" or recommended dimensions for this setup are 8.5ft i tried to model/scetch 8x8x8 cube to see what it will impact.

 

Peter, Flim,

not a question of feasible/not feasible, but rather what it is that i'm missing without this 0.5 ft? on paper it looks fine and the beams clearing my future cockpit contraption. is it 0.25ft lower then natural seating position? i'm afraid that from paper i just can't grasp what it is being impacted. i will be projecting directly onto front wall and ceiling (prepped and painted with fittin paint).

thanks!

Edited March 17, 2013 by agrasyuk

[cichlidfan]

measured the tobe sim-room in the new place: room for the Cube, couple desks and storage for electronic junk and some more. awesome! but vertically it is exactly the standart 8ft high... remembering that the "native" or recommended dimensions for this setup are 8.5ft

 

Now's the time to raise that ceiling, before you move in.

 

How much work can six little inches be? :D

[PeterP]

measured the to be sim-room in the new place: room for the Cube, couple desks and storage for electronic junk and some more. awesome! but vertically it is exactly the standart 8ft high... remembering that the "native" or recommended dimensions for this setup are 8.5ft i tried to model/scetch 8x8x8 cube to see what it will impact.

 

Peter, Flim,

not a question of feasible/not feasible, but rather what it is that i'm missing without this 0.5 ft? on paper it looks fine and the beams clearing my future cockpit contraption. is it 0.25ft lower then natural seating position? i'm afraid that from paper i just can't grasp what it is being impacted. i will be projecting directly onto front wall and ceiling (prepped and painted with fittin paint).

thanks!

 

 

Please read : #201 / #211

 

Summary:

Working with 8x8x8 will give you also good results,

It really depends on the throw-ratio of your projectors and if you plan to put a full cockpit mock-up in it.

 

The Throw ratio determines your head/cockpit clearance - keep in mind that there is only one ideal view-point for a distortion-less view in a cubic projection.

 

My recommendation :

Get a CAD program to visualise your plans in a 3d space (Google-sketchup is free and fully sufficient for this task) and get the data of the exact Throw-ratio and down-angel of the projectors you plan to use ,

and don't work with the absolute limits ! - always count in a 5% calculation error,

so you won't have any 'surprises' or uncertainness while you build your cube.

Ideally you would make you own measurements with a projector of your choice to get the absolute limits - you can't thrust the calculation-tools and manufacture data !

 

 

Here is a example how I did it for Flim :

 

Edited March 17, 2013 by PeterP

[jctrnacty]

PeterP you are real genius