Full Dome Hardware Set Up

Hi There,

I am new to the forums but am excited about vvvv. So please bear with me.

I am intending to set up a Projection system for a full dome. This will entail 5 projectors, four set a quarter way around the bottom of the screen beaming to the other side and one filling in the section at the top of the dome.

The dome will be modelled as a mesh in 3ds max and then be imported into vvvv. We will be using an Black Magic Inetnsity Pro capture card to get video into the system. The reson for doing it this way is it allows different vj’s to plug into the system and just feed any video they like in (Fisheye or Flat). This will then be applied to the dome mesh and filmed by cameras in vvvv. This will be processed in vvvv and sent to the beamers. I have seen other people doing it this way (u7angel) using vvvv, so i belive it should work. Do you think this system is viable with vvvv?

My question is is about the computer hardware to run this system. I have 2 options, which do you think would be the better way to run the beamers.

  1. Core i7 server - 6 go RAM 3 - 3x nvidia gtx260 cards

6 dvi out puts, 5 to the projectors and one to the server monitor.

  1. 1 lower spec pc with each projector and 1 server controlling them all via boygrouping. Connected together by gigabit lan

Questions;

With option 1 would there be any issues with delay between the projectors
with option 2 would the content be able to be sent over the ethernet or would each pc require it own capture card.

Any help or ideas would be greatly appreciated!

We have done almost the same project with 5 beamers.

Our solution was 1 Server and 5 Clients.

After it would depend of which content ( generative, video …) you would broadcast.

With this setup you surely have potentially best perfs even if it is more expensive.

With the i7 setup note that most of mainboard which are running with 3 GPU, have different timings for PCI-E slot like 16x / 16x / 8x.( i think there is none with 16x 16x 16x)
I guess is not a problem with i7 but perhaps it can be tricky.

Note that Gigabit is not a must too.

Note that it is quiet tricky with camera and edges, and in our setup we were using a DX9texture which had to be in a high resolution for not being aliased, but that seems to be fixed with Beta20 ( not tested in that setup).

The content would be video, 720x720, most of my video is .mov with Photo-JPEG compression.

What sort of content are you using? Is it sent over the LAN?

The main board is an Asus P6T Deluxe, thanks for the heads up, the slots are

3 x PCIe 2.0 x16 (at x16/x16/x1 or x16/x8/x8 mode)

Could you elaborate on your last comment

“Note that it is quiet tricky with camera and edges, and in our setup we were using a DX9texture which had to be in a high resolution for not being aliased, but that seems to be fixed with Beta20 ( not tested in that setup).”

We are using a mobile dome, do you think this will cause problems with calibration of the system during set up?

If you uses video you need to have at least 1 server and 2 nice PC , I would say, with something like RAID.

I would recommend to forget about sending video (streaming) over LAN.

The most complicated into the patch would be developing the camera patches and developing the softedge.

If you are sure about the projector position and still the same it’s ok, even if you would have to tweak that each times.
But if sometimes you have different position , it adds some extra work each time.

Also video playing over multiple outputs of gfx cards can be difficult, Digitalslaves tell me they have done it but I have never got it working without separate video streams to each head. The master and 5 slaves is a better solution for this reason IMHO.

yes. i would try to go for the following setup

server -> 1 ontrol monitor
client 1 -> projector 1+2
client 2 -> projector 3+4
client 3 -> projector 5

you could even try to save the last client by taking a dualhead 2 go vga splitter on client 2. this is much depending on the content. if you “only” want to pass a video signal through vvvv, i would say you could even do it with only 2 computers with projectors also connectec to the server…

have a look here
http://www.matrox.com/graphics/en/products/gxm/

i would NOT go for a video signal over ethernet (streaming) solution. hence, each computer needs a video capture card.

good luck.

OK

I seem to be getting mixed up.

If I apply a Video texture from (Intesity Pro)capture card to my dome mesh on the Server and then have a Boy Grouped PCs rendering each section of the mesh for the beamer to project vvvv will transfer all necesary data to the render PC’s for them to render.

Is this correct?

Note that you need 1 videograbber for each PC .
You need all sources, pictures, meshes, shaders on each PC.

For my part the Intensity Pro was not really a stable grabber with vvvv. Even with 2.x drivers. Perhaps it is better now ??

OK

We have had issues with the 1 Intensity Pro before also, I couldn’t deal with 6 of them.

Are there any other Hi-Def capture cards any body could recomend?

To be honest I might have to consider other software to run the whole thing…

As cheap as Intensity , i would say no …

OK,

I have been looking at Chromium (WireGL) which can be used for distributed rendering with a Node setup. However I would face the same problems with putting so much information onto the LAN proably wouldn’t able to get acceptable frame rates.
It is still very important to us to have an input for live video so we can have many different artist input video into the system.

So,

I am now looking into the single server option again. I have been reading about the VideoIn node seen all the posts about only being able to output to one renderer. This normally takes the form of a very wide screen outputed from 1 card to a Dual or triple2go. So now I am thinking

Corei7 Server
GTX295 Graphics card
Each DVI port feeds Dual/TripleHead2go
4 Projectors

This is fine for a flat screen, is there anybody who has used this setup for a dome screen? I am having trouble visualising how to get the image bent around the dome as it would be shorter on top than on the bottom?

Could it be done by out putting one long rectangular renderer which has 4 quarter circles next to each other with the video texture on which when the projectors are positioned around the dome they cover the screen?