Big screen, BIG Videos! ;)

[Chris]

Hi All

I'm currently looking at a show for a client which requires a 10 projector widescreen blend resulting in a 14880 x 1080 pixel space!
This will be one continuous image which we will be playing back content at the full resolution.
The full screen curved around the audience so screen shaping will be in play here.

I wanted to call on your experience here for some pointers

At the moment my thoughts are to split the load across 3 Ventuz servers.
2 will be outputting 3 x 1920x1080 via ATI V8800
1 will be outputting 4 x 1920x1080 via ATI V8800
Each will have HDSDI input cards for camera images and ATI S400 cards for gen/framelock

Does this seem like the best set up? or with such large content is it better add another server to spread the load?
Has anyone tried to push content this large through a Ventuz cluster?
How good is the edge blend in Ventuz? do you have any control over the gamma in blended regions?

and finally... the manual says that it is better to use a single machine if possible. Is it even conceivable that for this large a screen space with such large content, that i could use a single machine running a V8800 and just add a Datapathx4 to 3 of the outputs and achieve this??
My feeling is that the playback at this resolution would be too much for a single system!

Any thoughts or tips appreciated
Thanks
Chris

[Christian Krix Schmidt]

Hi Chris!
My experience is with nVidia cards but I am pretty sure the same rules apply when it comes to multi system output.

First rule: All systems have to output the same resolution. So your third system with 4xHD out is a no go. Even if that means that one output will be black you should push the same pixel resolution through all machines. We do that sometimes as well when using triple output per machine.

I just had a project with two workstation each outputting 3xHD in a 3x2 projector overlap setup. I did the shaping and edge-blending as well. It all works quite nicely. You also have control over the gamma. But if you never did it before then you HAVE to train a bit before to get to know the system and how it behaves. Doing it for the first time on a job when time is of the essence is a killer. Don’t do it. The best solution for everyone would be to just provide the content overlap for the projector guys and let them do the softedging. They know what they are doing and you have your hands free to do your stuff. Provide them with a proper grid so they can set up using already your output signals and you walk away both happy.

Usually when budget is not preventing you from it a single output per workstation setup is definitely preferable – especially when using GenLock/FrameLock. I don’t even want to think about using multi-output in a synced environment. If you are using three or even four outputs from one machine your performance will be severely limited. Not only by content but especially by the shaping. If you do Shaping via Ventuz you have to go through Render Targets. With 15K x 1K you are alrady past any sane Render Target sizes to do it in a brute force approach. Your content needs to be shifted per system using ProjectionCenterX/Y. And each system needs a separate RenderTarget and for each output a separate Shape Node. Maybe it would even make sense to have a RenderTarget and Shape Surface for each individual output. Then you are not limiting your performance that much. If the projector guys would do the shaping and softedging you would only have to provide the content overlap. That might mean no render target and better performance. But again, that greatly depends on the content and overall setup and such. Usually when you do not go via Render Target you actually have to render the content several times, one for each output to get the duplicated content out for the softedging to work.

I would also highly recommend not use a render target for glow unless you absolutely have to and instead work with Alpha Blending to create your own glows. At these resolutions you also need a high-resolution render target for the glow (think 4K x 512 and a matching Glow Blur Node resolution). Otherwise the quality will be crap.

All of these factors for me are why single output is a better option at a certain resolution and a certain number of outputs in my opinion. Of course it depends on the actual content but usually at those resolutions when you are doing shaping you end up having to deal with a lot of problems.

As for the reduced number of Live Video Input cards. That would work. Only install them where you really need them. But make sure that those systems are your performance benchmark since those machines will have to calculate a bit more than the rest. And if you stay at 50 fps (synced environment) on those machines and then of course on the others as well, the difference in load will be negligible.

It is a complex topic I give you that….

Good luck!

[Daniel Willer]

Another capter for a future Ventuz book.

[Christian Krix Schmidt]

Very true. Now where can I get 48h days? I need more time per day.

[Chris]

Thanks for the detailed answer Chris!!

erm..... where can i get a copy of that book Daniel??

Good to get some first hand info appreciate it! A few things in your reply alone that have caused me to have a little rethink.....

I think that 5 servers each handling 2 HD outputs each will hopefully be the right balance.

Plan will probably be (as you say ) to hand the blending and warping over to the projectors, in order to take the load of the system. Drop the render targets from the scene and instead use a projection node to output 1/5th of the scene from each server, leaving more system power to run the huge video content.
I have some pre production ideas to achieve the video resolution and reduce the system load also.

More thinking to be done on this subject!
Will come back and report later....until then....any more input, from anyone else who's tried this sort of set up, is more than welcome to comment

Thanks
Chris

[joysprod]

I did a 9 wide about two years ago with 3 severs and V8800 GfX and S400 sync cards. Built the servers myself as needed them set up for a good couple of weeks trying various options and x-renting proved uneconomic. In the end did blend in projectors and overlap in Ventuz. Best way to go as projectionists have full control with brightness, contrast, lamp power etc and it would of been a real pain with a short setup time. Luckily the screen was 3 flat areas, the two outer areas being at an angle of 20 degrees to the centre so no curves to worry about. Projectionist are key! I did use Rendertargets but my load wasnt too heavy. Also only outputting to 9 x 1400 x1050 projectors so pixel space was quite a bit less then at 1080. Controlled it all with a Lemur touch via OSC. Now available on iPad but still prefer wired ethernet to wireless. Have you been able to do any tests?

Where are you based? I didnt see a country on your profile.

Regards

Peter

[Chris]

Hi Peter
Not run any tests as of yet, hopefully in the next few weeks

Def think the blend and warp should be done in projectors for now, will make things a lot less complicated!
How have you got on running large videos on your systems, were you running full size footage on the job you mentioned?
Was the whole system generally stable?

Like the Lemur touch option, i've not really used it but heard a lot about it. Is the interface design flexible enough for all your Ventuz projects?

I'm based in the UK.....didn't see that bit will update

Chris