Why no Camera Arrays?

[Nathan Solomon]

I'm new to digital cinema cameras (started my worklife as a cinematographer, but have been focused on game and immersive tech for a while), so this may be a dumb question, but I can't find an answer anywhere, and this seems like maybe the place to ask.

In projection, there are a number of ways to use arrays of projectors in combination to get a higher resolution or better quality image.  -Correcting for anamorphosis, parallax, etc. I've worked with Unreal Engine in projects doing this.

It's definitely technical possible to do the same thing with cameras, using the same tools we use for projection. -Either employing multiple cameras, or a single lens/camera with an array of inexpensive sensors. It seems like it could be quite useful, especially in SFX.

Is there a reason why this isn't done; or is it in fact being done, and I just can't find it?

Thanks, very much,

Nathan

[Snowfun]

Phil Holland uses several REDs to get high resolution images. Over on reduser I’m sure he has described this and stitching process in some detail. Also:
http://www.phfx.com/articles/Filming100MegapixelLargeFormatAerialsOverNewYork/

[BTM_Pix]

This video from the excellent Media Division channel is a good run through of doing it purely as an experiment with a single camera.

The challenge to do it with multiple cameras is the rigging and sync of the controls.

Small sensor in a box cameras like the BMMCC and the Z camera would be ideal as they are compact enough and are easily controlled remotely so that the controls can be ganged.

The Sigma fP would be absolutely ideal if it weren't for the shortsighted decision to make the only way to remote the camera controls being through the same port as the recording media so you can do one but not both. 

Pocket 6K in vertical orientation would be workable as they can be timecode synced with each other and a single control solution to control multiple cameras simultaneously would be doable by someone industrious if there was a need.

[Andrew Reid]

The 3-4 cam arrays on smartphones show the strength of the future for camera arrays.

As long as they don't take it a step too far!...

[BTM_Pix]

6 minutes ago, Andrew Reid said:

The 3-4 cam arrays on smartphones show the strength of the future for camera arrays.

As long as they don't take it a step too far!...

 

That thing reminds me of the propensity that old Mods had for maximising forward illumination and rear view coverage for their Lambrettas.

[Geoff CB]

Time and more points of failure are the reason this is not done. It's pretty much the same reason 3D fell out of favor. 

+2 Cameras = More grip gear and power
+You have to sync the cameras
+The settings have to be exactly the same
+Your LENSES have to be exactly the same (AKA you need high end matched sets) 
+The cameras and Lenses have to be aligned properly and at the same focus.
+You have to pay an extra person on set to make sure all of that is working properly. 
+Good luck changing lenses fast!

It's useful on maybe landscapes and TV demo/stock footage and like you said VFX (Example being the incredible rig used on The Irishman, I wish someone had come up with before 3D died) but it's a nightmare for simple narrative. 

[Nathan Solomon]

Thanks! Those are some great answers. I really appreciate your insights.

The Media Division clip is inspiring; their take on the potential is pretty weird. Which is probably appropriate, given the level of creative excellence in how they carried out the test. It would not have occurred to me to keep a lens static and move only the sensor in quite that way. Perhaps that was driven by their decision to merge the elements in two dimensions within After Effects?

I think that I can test the concept by building a frame that will hold six inexpensive 1080p cameras, and then working with my collaborators to develop a workflow that merges and corrects their output in real time. -This would be in contrast to the way Media Division simply ensured that the camera maintained perfect alignment. Does anyone have suggestions for a camera to use for this? I'll freely share here whatever I may learn from this.

Ultimately, I think it would be useful to create a structure similar to an analog (5"x7")view camera, using a number of large inexpensive sensors, with automated controls moving front and back panels, as well as the lens. The two things that make this model especially appealing to me are: First, while it would be large, it could be quite lightweight, as the lenses required for this configuration can be very small, and lens' coverage of the sensor panel(s) would never be the issue it is with conventional digital cinema cameras. Additionally, the full control of optics planes would be fun, especially in integration with realtime 3D SFX.  Second, it would be a lot cheaper than a conventional 6k+ camera.

[Snowfun]

1 hour ago, BTM_Pix said:

That thing reminds me of the propensity that old Mods had for maximising forward illumination and rear view coverage for their Lambrettas.

Was that your first ever selfie? 

[BTM_Pix]

2 hours ago, Nathan Solomon said:

I think that I can test the concept by building a frame that will hold six inexpensive 1080p cameras, and then working with my collaborators to develop a workflow that merges and corrects their output in real time. -This would be in contrast to the way Media Division simply ensured that the camera maintained perfect alignment. Does anyone have suggestions for a camera to use for this? I'll freely share here whatever I may learn from this.

Just as a proof of concept you could just go with six lowest budget action cameras that you can find that have real time HDMI output.

You won't get the simultaneous control but neither will you need it because they are all fixed focus autoexposure models anyway and it will be adequate to check your realtime stitching process.

I'd be inclined to get one with a narrower FOV and whilst I can't give you specific models there are plenty of people on here who use action cams who are up to speed with what the latest and greatest clone ones are.

EDIT>

If you want to step it up to a level above action cameras once you've got the stitching working then you could go with a set of used Panasonic LX100s which you can do full image and zoom lens control by wifi with the added bonus of 4K. They are readily available for under £200 used these days so £1200 for a large format video camera with 3x integral zoom isn't unreasonable !

I've just remembered that the LX100 does not have a live HDMI output so would be OK for recording but not for live stitching.

Something like used Panasonic FZ1000s would do the same job albeit with a smaller sensor though.

 You'd need something custom to control six of them simultaneously from one base unit but thats not insurmountable if you know someone who's into that kind of thing...

Above that, then six BM Micro Cinema Cameras would be an option but you'd be then into finding six matched lenses etc so you'd be looking at about three times the cost of using something like the FZ1000s.

[Nathan Solomon]

Thanks, BTM! It'll take me  a while to pull this together, but I'll definitely post when I get it to next steps. I may seek out those Panasonics, or I may swap out lenses on a few action cameras (I have an adapter design I made for that being fabricated now).

[BTM_Pix]

Have a look at some of the Sony,Canon and Panasonic HD palmcorders as well as many of them have LANC controls which will make the simultaneous controlling easier and they are usually very cheap used.

The form factor is quite useful too as they all tend to be quite narrow which could be a benefit in placing them for overlapping coverage.

Good luck with it.

[IronFilm]

Look up photogrammetry (or more specifically: videogrammetry ), I've done a little of this in the past.  That's when you're using heaps of camera angles at once to capture a scene. 

Has uses for VR. But in conventional filmmaking? Nah. 

There is a LOT of work that goes into making a shot happen, throwing an extra camera angle into that means massive massive compromises if you want both to be equally good and not just a waste of time. 

Of course often people do shoot with two (or even three) cameras at once, and they live with the compromises for the sake of time gains / efficiencies they (arguably) get in return. 

Sitcoms with their 3 or 4 camera set ups look drastically different to single camera shows exactly because they've taken the opposite side of this compromise balancing equation in # of cameras vs quality. 

There are very niche cases for running high numbers of cameras for a conventional (non-VR) film:

#1 Special effects, such as explosions / car crashes / stunts / etc. As you really really really  want to minimize how many times  you do a take! Thus throwing 10 cameras at a scene to see what sticks, and gets semi usable footage out of it (even if only a few frames worth from a crash cam, that might be all you need!), is totally worth it if you don't have to re-do that car crash stunt!

#2 & more recently, VFX work, say you want to add something into a scene like a spaceship, or modify something already in a scene (like age-ing an actor, or turning them into an alien). You still treat this as a normal single camera shoot (or perhaps double cam), with you just focusing on one camera angle/shot at a time. But all the extra cameras (which don't need to necessarily match perfectly, could even be from a different brand) give extra data for the post teams to help work with in terms of integrating their VFX into the main shot.  To give just one example, I, Tonya (2017) used a bunch of Blackmagic Micro 4K cameras to help put the actor's face onto the stunt performer's face. (as obviously the actress wasn't capable of pulling off the same level of ice skating skill Tonya herself did!) 

https://www.creativebloq.com/features/top-blockbuster-vfx-youre-not-meant-to-see

https://www.eoshd.com/comments/topic/26411-blackmagic-micro-camera-was-used-on-i-tonya/

 

[Nathan Solomon]

Thanks, IronFilm. I'm familiar with photogrammetry, and I know that up to this point, using multiple cameras in this way is not practical or worth the effort; and yes the concept for multiple cameras is for VFX. Also, I was a cameraman for long enough to know the reasons for shooting multicamera, even if it was long enough ago that I only shot 35mm film. 😂

I've been doing projects using Unreal Engine since since 2014, including full MR greenscreens compositing 3D interactions before they were enabled by the engine. The basis  of this exploration is that advances in the Unreal Engine change the viability of using cameras in this way. I'm just not sure that it changes that enough to be worthwhile, so the question is more subtle than whether it has been done and whether it is viable under past conditions. It may well not be worth doing, but I don't think you've yet gotten closer to answering that.