Why is a 3D rig expensive?

[David Quinn Carder]

If I understand correctly, all you need is two cameras with lenses positioned like our eyes positioned on our head, both pointing straight forward; also, the ability to sync the two cameras. I read something about "convergence" in a thread here and that's what threw me off . . . maybe I'm missing something?

Apologies if this has been covered already.

 

[Eren Ozkural]

The RED 3D rig should be rather cheaper than other 3D rigs because it's designed for 1/2 types of cameras only, it looks like a parallel only set up and RED are very aggressive on their pricing anyway.

The reason why a lot of other rigs cost money is if they're beamsplitter rigs. Normally cameras are too large to use attache side by side...the bodies are too large to simulate the 2.5 inch distance between the average human pair of eyes.

To combat this the rigs place one camera above, pointed down at a mirror tha splits the light ont the two cameras. The bodies can be aligned closer together as they're not clashing for space.

The engineering and materials that go into this must be quite costly.

I'm sure that someone here like Mark can clarify better than I can.

 

[Michael Hastings]

David:

It depends on what you want to do.

The issue that comes up is that you need to have very close intraocular distances to shoot things reasonably close. The rule of thumb is to have your subject distance about 30 times the intraocular distance. So for side by side cameras with lenses or body width that are about 3" the closest you can get them is a 3" intraocular distance which dictates a 90" or about 7.5 foot subject distance. The minimum R1 side by side is about 5.2 inches so 156 inches or 13 feet which is not a very normal shooting distance for a lot of things.

Beamsplitter rigs are a solution for small intraocular distances with big cameras.

With that said, my company has been working with some Canon HFS100 cameras which allow about 2.3 to 2.5 inch intraocular and we are able to do some pretty cool things with them. They are very useful for a lot of things and certainly help you learn and understand some of the issues. We set this up primarily for underwater but are planning to make some mounts and brackets for regular use that would cost just a couple hundred dollars (You need to make sure that the cameras are very square and solid or you create major headaches for editing (and viewing!).

Since the cameras (full 1080P 24Mbps H.264) are about $900 apiece you can get into it for a little over $2K. And for super cheap we are playing with some Sony "Webby" cameras since they allow extremely small intraocular and prices for two cameras and a 3D mount for less than $700!

So 3D doesn't have to be expensive, and you can do a lot, but full featured 3D with a lot of creative capability and flexibility probably will be expensive for awhile.

But the trend is in the right direction. The fact that we can get two full 1080P camcorders for $1800 is the first step and if/when perceived market volume is great enough for the video camera manufacturers to incorporate stereo optics with intraocular manipulation there is no reason that that couldn't be done for a few thousand dollars - and a true 1080P/2K full featured stereo camera could be had for what we used to pay for an HVX200.

And Red might be the company that decides to do that...

 

[Jeff Coatney]

When we humans look at something close to us, like a foot or two away from our faces, our eyes actually cross to keep that object centered and focused in our field of view. When we cross our eyes, like when we're making a goofy face or something, you get two discreet views of the world super-imposed on each other. Normally, our eyes sit parallel to each other when we look at subjects on the horizon. When the subject moves closer to us, our eyes begin to cross ever so subtly to maintain the subject in our field of view and in focus. This "eye-crossing" is called convergence.

There's is absolutely no variability in a 3D rig except for the angle at which the two cameras converge. The inter-occular distance is always 65mm. It doesn't matter which format of camera you use, from IMAX to super8mm and everything in between, the inter-occular distance is 65mm because that matches the average position of the left and right human eye relative to each other.

The convergence angle is the degree to which the eyes cross. In a 3D rig, you can have any amount of convergence from 0 (where both cameras are facing forward, in parallel) to about 20 degrees (each camera, pointing towards the other one in a crossed orientation at 10 degrees, giving you a net 20 degree effect or a convergence angle of 20 degrees).

Deviating from these values can induce eyestrain, headaches, vertigo and even a violent, intense desire to look away on the part of the viewer.

 

[Michael Hastings]

There's is absolutely no variability in a 3D rig except for the angle at which the two cameras converge. The inter-occular distance is always 65mm. It doesn't matter which format of camera you use, from IMAX to super8mm and everything in between, the inter-occular distance is 65mm because that matches the average position of the left and right human eye relative to each other.

The convergence angle is the degree to which the eyes cross. In a 3D rig, you can have any amount of convergence from 0 (where both cameras are facing forward, in parallel) to about 20 degrees (each camera, pointing towards the other one in a crossed orientation at 10 degrees, giving you a net 20 degree effect or a convergence angle of 20 degrees).

I'm going to disagree with you there. The whole point of a beamsplitter rig is to allow the lenses to achieve an intraocular less than would be physically possible with the size of the camera/lenses. Again, I am just learning this stuff but about 99.9 percent positive that I am correct on this score.

One of the big things in underwater is to do macro shooting and that is where the beamsplitter excels because you can actually achieve zero intraocular (obviously not a lot of point in actually getting to zero but in practice you would use very narrow (sub 1 or even .5 inch).

 

[Joey R.]

If you have to ask....

In all seriousness yes, you can absolutely shoot 3D with two cameras, a couple of nails, a 2X4 and some shoe laces and you may even get it to not suck.

We built a beamsplitter rig used by Geoff Boyle for Dark Country. It was the collaborative efforts of a mechanical engineer and a machinist and it took just over a week to create. It was inexpensive and it worked fine. There were a few more made and they have been out on productions ever since.

So lets first look at what makes it cheap:
• Only one camera can be used (SI2K)
• Only one glass type (C-mount primes)
• No convergence
• Small interocular travel
• ~15lbs with cameras
• Low overhead
• Relatively short machining time

So what makes a GOOD rig expensive?
• Convertible from beamsplitter to side-by-side
• Able to mount MANY camera types
• Hard-core rigidity making it damn near impossible to make the two cameras move when doing something as simple as sweeps and moving from a dolly to a crane.
• Alignment that doesn't change with every camera movement
• Motorizing the I/O and C movements
• Having a wireless controller that controls said I/O and C movements at distances as great as 1500ft in both wired and wireless configurations
• Having an alignment process that takes anywhere from seconds to minutes (how fast can you turn knobs?)
• Software that calculates the correct convergence tangent relative to the I/O
• Prototype planning for two smaller sizes
• etc etc etc

Now of course add on to that 6 CNC machines running non-stop, a crew of over 20, a beta testing regimen that gathers hundreds and hundreds of hours of testing footage and data. A few prototypes builds, marketing blah, trade shows blah, advertising blah.

Its expensive and time consuming unfortunately. But when you are in your post facility and you see the type of stereo you get at the end of the day, our goal is that you are pretty happy.

 

[David Quinn Carder]

When we humans look at something close to us, like a foot or two away from our faces, our eyes actually cross to keep that object centered and focused in our field of view. When we cross our eyes, like when we're making a goofy face or something, you get two discreet views of the world super-imposed on each other. Normally, our eyes sit parallel to each other when we look at subjects on the horizon. When the subject moves closer to us, our eyes begin to cross ever so subtly to maintain the subject in our field of view and in focus. This "eye-crossing" is called convergence.

There's is absolutely no variability in a 3D rig except for the angle at which the two cameras converge. The inter-occular distance is always 65mm. It doesn't matter which format of camera you use, from IMAX to super8mm and everything in between, the inter-occular distance is 65mm because that matches the average position of the left and right human eye relative to each other.

The convergence angle is the degree to which the eyes cross. In a 3D rig, you can have any amount of convergence from 0 (where both cameras are facing forward, in parallel) to about 20 degrees (each camera, pointing towards the other one in a crossed orientation at 10 degrees, giving you a net 20 degree effect or a convergence angle of 20 degrees).

Deviating from these values can induce eyestrain, headaches, vertigo and even a violent, intense desire to look away on the part of the viewer.

Thank you all for your answers, but Jeff, this is what I was looking for. I understood that eyes converged but did not know it was necessary for cameras as well. This kind of shoots down my hopes of run-and-gun 3D shooting. If I understand correctly, this means that it is intrinsically impossible to create 3D video that allows the viewer to focus on whatever they want in the image, even if the depth of field is deep, because of the convergence factor required when dealing with nearby subjects.

Is this correct?

 

[Ryan S]

Great post Joey, very informative! Thanks.

 

[Ketch Rossi]

If you have to ask....

In all seriousness yes, you can absolutely shoot 3D with two cameras, a couple of nails, a 2X4 and some shoe laces and you may even get it to not suck.

We built a beamsplitter rig used by Geoff Boyle for Dark Country. It was the collaborative efforts of a mechanical engineer and a machinist and it took just over a week to create. It was inexpensive and it worked fine. There were a few more made and they have been out on productions ever since.

So lets first look at what makes it cheap:
• Only one camera can be used (SI2K)
• Only one glass type (C-mount primes)
• No convergence
• Small interocular travel
• ~15lbs with cameras
• Low overhead
• Relatively short machining time

So what makes a GOOD rig expensive?
• Convertible from beamsplitter to side-by-side
• Able to mount MANY camera types
• Hard-core rigidity making it damn near impossible to make the two cameras move when doing something as simple as sweeps and moving from a dolly to a crane.
• Alignment that doesn't change with every camera movement
• Motorizing the I/O and C movements
• Having a wireless controller that controls said I/O and C movements at distances as great as 1500ft in both wired and wireless configurations
• Having an alignment process that takes anywhere from seconds to minutes (how fast can you turn knobs?)
• Software that calculates the correct convergence tangent relative to the I/O
• Prototype planning for two smaller sizes
• etc etc etc

Now of course add on to that 6 CNC machines running non-stop, a crew of over 20, a beta testing regimen that gathers hundreds and hundreds of hours of testing footage and data. A few prototypes builds, marketing blah, trade shows blah, advertising blah.

Its expensive and time consuming unfortunately. But when you are in your post facility and you see the type of stereo you get at the end of the day, our goal is that you are pretty happy.

This is why I have been doing business with ET for about two years now, and why I'll be there next week to see the rigs.

Quality of the Rigs built is not all, they must also have Precise and full control of both camera in absolute Sync, and all controls must allow for a simple use in the complexity of the set up, making it as easy as possible for its operators to pay more attention to the shot, instead of the rig, and even if they are now more and more emerging 3D rigs, only few, and only few are worth the prince, as just about any good engineer and or machinist can build one, but only few can master the electronics that go with in order to fully control all aspects of the otherwise incredibly difficult set up that a S3D set up presents, and this is why the old 3D was just not good, they didn't have the technology on their side, but some of us, now will have it.

 

[Joey R.]

Thank you all for your answers, but Jeff, this is what I was looking for. I understood that eyes converged but did not know it was necessary for cameras as well. This kind of shoots down my hopes of run-and-gun 3D shooting. If I understand correctly, this means that it is intrinsically impossible to create 3D video that allows the viewer to focus on whatever they want in the image, even if the depth of field is deep, because of the convergence factor required when dealing with nearby subjects.

Is this correct?

HUMAN eyes converge AND focus on the same object. Your eyes are always converging. If they do not converge they become diverged and it causes general uncomfortableness. If they converge too much you become cross-eyed. For 3D you can have it where the convergence and the focus do not have to be the subject. Or maybe you want it to.

In 3D shooting you have two methods of shooting, parallel and beamsplitter. You shoot beamsplitter in order to replicate as close as possible human interpupillary distance. For close up shooting you may want less than that, say .75" or 1.5". There is no camera out there that you can slap side by side that will allow you to get two camera bolted together that will give you the optical centers .75" apart. (Checking Iconix as I write this... ). Thus the beamsplitter. It allows both lenses to occupy the same space for a 0 interocular or a slight shift for your .75" I.O.

Side-by-side is a must when using huge zooms and great distances. There is obviously more to it but that is the jist of it.

 

[Markus Stone]

AKA the difference between S3D and real life

AKA the difference between S3D and real life

It's counterintuitive to think that you'd have to vary the interaxial (or interocular, if you like to call it that) for each shot when in real life our eyes are fixed, but you do.

Consider if you were doing a 'matrix' shot where the camera travels 'through' the phone lines. Another camera placed 64mm away would be outside of the cable; ie, capturing a completely different image altogether = major eyestrain for the audience.

The human eye has a remarkable ability to deal with things that would otherwise cause eyestrain if caught on camera. I can focus on my finger when it's about 10cm from my nose. I would not shoot the same closeup with a 64mm interaxial/ocular.

My eyes are happily converging on my finger and although the background has way too much parallax due to being overconverged, it's okay because it's way out of focus, and I'm looking at my finger anyway.

The human eye is not like a camera, capturing all parts of the frame equally. We tend to 'see' mostly with the 5 degrees or so of central vision - there was an article on this some time back in New Scientist. If something interesting happens in the peripheral vision, we turn our eyes to look at it. That's why our eyes 'scan' the frame, rather than remaining static, like the camera that recorded it.

So if I’m looking at my finger, I’m pretty much not looking at the background anyway. As soon as I look at the background, the parallax is fine because I have reconverged. Instead, now I have 2 blurry fingers obscuring part of the background.

Stereoscopic cinema presents the world as we would like our audience to see. If I shoot my finger with a 64mm interaxial interocular, the audience may look at the horribly over converged background and accordingly vomit into their popcorn.

This is also the reason why you tend to get a better 3D effect with deep focus rather than shallow depth of field – that, and the fact that your eyes have sharp edge detail to work with throughout the scene to calculate where objects lie in 3D space. BUT it means that we have to create an image that works, no matter where in the frame the viewer decides to look - which is not the same as real life and part of the reason why we shoot with differing distanced between the cameras.

It also has to do with the breakdown of accommodation (focus) and convergence. Even though in real life, your eyes converge AND focus, in the cinema, you only ever focus on the screen (or possibly, your date), while it is solely the convergence that gives you the 3D effect. In real life, focusing on my finger throws the background out of focus, which helps soothe the overconvergence of the background; in the cinema the background stays exactly as the cinematographer shot it.

As an aside, there are various things that, while we accept them in the real world, don't work in 3D. For example, my monitor has a shiny bevel on it that reflects it's environment. Because it’s only 15mm wide, each eye gets a totally different reflection. I don't get a headache from looking at my monitor all day, but if I shot it and projected it, the reflection would probably be annoying.

The reason is that a stereoscopic LR projection is not the same as actual 3D. If I decide to resolve it, I focus and converge on the reflection, move my head, and the reflection moves and shifts. From this I can get a sense of the reflected object. Try this in the theatre and the reflection will NOT shift, in fact the whole world you are looking into will appear to revolve to keep you in the exact same position relative to the objects in the scene.

So I guess in a 3D theatre, the world really DOES revolve around you - at least for 2 hours :smile5:

Cheers
Markus

rig except for the angle at which the two cameras converge. The inter-occular distance is always

65mm. It doesn't matter which format of camera you use, from IMAX to super8mm and everything in between, the inter-occular

distance is 65mm because that matches the average position of the left and right human eye relative to each other.

 

[Pedro Guimaraes]

There's is absolutely no variability in a 3D rig except for the angle at which the two cameras converge. The inter-occular distance is always 65mm. It doesn't matter which format of camera you use, from IMAX to super8mm and everything in between, the inter-occular distance is 65mm because that matches the average position of the left and right human eye relative to each other.

umm....if I understand you statement correctly, you are totally incorrect. This is a ton a adjustments on 3D rigs. Alot more than just convergence.

We change IO on a per shot basis! Every situation demands a different setting. In fact in most cases it will be less than 2.5"

While I doesn't matter what camera you use, it does matter what your near and far points in the image are. Most of all it matters what you will be displaying it on. If you are shooting for IMAX and you stick with a 65mm IO on a certain kind of shot your parallax on a 75ft screen will be to big.

So IO settings have to change mostly to conform to viewing screen size, near and far points, focal length being used and a host of other conditions.

Can you shoot 3D with a fixed IO? SURE. It been done alot! Is it ideal not for every situation.? No. Adjustablility is key for 3D. if you take care and shoot things correctly when using a fixed IO you can produce good content, but care must be taken.

just wanted to expand and clear that statement up since the original poster seamed to be new to 3D.

If I misunderstood you post, I apologize in advance.

 

[Jeff Coatney]

OK, let's clear this up before we all confuse the hell out of everyone reading this thread. Some of you sound like you're confusing the Inter-Occular distance and the Convergence angle. They are not the same and they must be isolated and controlled independently from each other.

Unless you are a Pirate with an eyepatch and a parrot on your shoulder, you have two eyes in your head. Your eyes are 65mm apart, sitting in your skull. This is the (average human) inter-occular distance. Your eyes never move closer together or further apart because they are fixed in the eye-sockets in your head.

When you adjust your two cameras so that the inter-occular distance is greater than 65mm or less than 65mm, you are changing the scale of human vision. If the inter-occular distance is 130mm, you just made everyone in the audience a 12 foot tall giant. If your inter-occular distance is 32mm, you just made everyone in the audience into a 3 foot tall hobbit. When you vary the inter-occular distance, you change the scale of not only the audience, but everything on-screen in your shot. The behind the scenes of the film Coraline illustrates this perfectly since the inter-occular distance on the camera had to match the relative size of the puppets to scale. The inter-occular is typically not adjusted during a shot and 99% of the time is set at 65mm.

The amount of eye-crossing that we humans do is the convergence angle. Here again the distance between our eyes doesn't change, just the angle at which they're pointed changes. This is typically adjusted during a shot to maintain the subject within the Z axis. This angle can change constantly and dynamically in a shot, yet it should be done within a 20 degree range so as not to induce negative responses from the audience.

There are many rigs you can use, but they all perform the same basic principles: replicate human vision. 3D is an attempt to replicate human vision. That's why it works. You can deviate all day long from this paradigm, but it has consequences with how the audience experiences the movie. Just because you can shoot your subjects outside these parameters, doesn't mean it will automatically translate into effective 3D footage. Stay Human! It works 99% of the time.

 

[Jeff Coatney]

This kind of shoots down my hopes of run-and-gun 3D shooting. If I understand correctly, this means that it is intrinsically impossible to create 3D video that allows the viewer to focus on whatever they want in the image, even if the depth of field is deep, because of the convergence factor required when dealing with nearby subjects.

Is this correct?

Correct. But you could shoot with a fixed convergence angle and adjust your subject to perform or move within that plane, provided you can control your shooting environment. Or maybe its okay that the subject moves in and out of optimal 3D viewing? Your artistic and content choices would have to determine if this is acceptable.

 

[Michael Hastings]

OK, let's clear this up before we all confuse the hell out of everyone reading this thread. Some of you sound like you're confusing the Inter-Occular distance and the Convergence angle. They are not the same and they must be isolated and controlled independently from each other...

.....The inter-occular is typically not adjusted during a shot and 99% of the time is set at 65mm.

Stay Human! It works 99% of the time.

Joey, Markus, Pedro - isn't this still incorrect?

i.e. we aren't confusing intraocular:

".....The inter-occular is typically not adjusted during a shot" (that's probably true) "and 99% of the time is set at 65mm" (and that's probably not).

and that intra-ocular is routinely changed for the reasons Markus explained?.

 

[J.J. Blumenkranz]

I think there is another point to be made about convergence. It can be faked in post by shifting the Horizontal Offset of the images. (I didn't see that mentioned in this thread; If I missed it, apologies.) So, shooting parallel can give you a little flexibility down the road and let you focus on the inter-axial/inter-ocular (same thing). Shooting parallel also has added benefits such as no keystoning artifacts. Of course, it also implies that you have to shoot with some padding to give room for the shifting.

 

[Mark L. Pederson]

Joey, Markus, Pedro - isn't this still incorrect?

i.e. we aren't confusing intraocular:

".....The inter-occular is typically not adjusted during a shot" (that's probably true) "and 99% of the time is set at 65mm" (and that's probably not).

and that intra-ocular is routinely changed for the reasons Markus explained?.

Sometimes you will rack IO, convergence & focus - all within a shot. But I'd agree that you will typically rack convergence more than IO while rolling.

"99% of the time is set at 65mm" is not a correct statement. You COULD do that - if you want to shoot simple, safe 3D - but I'd be shocked if any folks doing 3D features are doing that - we are certainly not.

 

[Michael Hastings]

Sometimes you will rack IO, convergence & focus - all within a shot. But I'd agree that you will typically rack convergence more than IO while rolling.

"99% of the time is set at 65mm" is not a correct statement. You COULD do that - if you want to shoot simple, safe 3D - but I'd be shocked if any folks doing 3D features are doing that - we are certainly not.

Mark, thanks for clarifying.

I think there is another point to be made about convergence. It can be faked in post by shifting the Horizontal Offset of the images. (I didn't see that mentioned in this thread; If I missed it, apologies.) So, shooting parallel can give you a little flexibility down the road and let you focus on the inter-axial/inter-ocular (same thing). Shooting parallel also has added benefits such as no keystoning artifacts. Of course, it also implies that you have to shoot with some padding to give room for the shifting.

I touched on your point in another thread. But it is an important one.

I think in essence it boils down to this:

3D adds a whole 'nother level of complexity to shooting movies.

As I stated in post #3, you can do some very interesting things - and get a lot of practical experience with a simple side by side rig (It can be useful just to see how things work as subjects/objects move in and out of the frame/side to side, etc.). And it can be pretty useful for simple stuff: like my underwater nature shoots where you are basically just assembling clips of interesting stuff and throwing out whatever doesn't work; a short tradeshow commercial piece where you are just going for the gimmick effect; 3D porn; etc.

But it would also be mistaken to think that you would try to shoot a full length feature type movie these days without the capabilities you get from using a full-featured rig like ET makes.

 

[Jeff Coatney]

Sometimes you will rack IO, convergence & focus - all within a shot. But I'd agree that you will typically rack convergence more than IO while rolling.

"99% of the time is set at 65mm" is not a correct statement. You COULD do that - if you want to shoot simple, safe 3D - but I'd be shocked if any folks doing 3D features are doing that - we are certainly not.

So if simple and safe 3D can be achieved using a 65mm IO, how is it not a correct statement? You're implying that you shoot complex, unsafe 3D. What distance do you set your IO typically?

 

[Jeff Coatney]

As I stated in post #3, you can do some very interesting things - and get a lot of practical experience with a simple side by side rig (It can be useful just to see how things work as subjects/objects move in and out of the frame/side to side, etc.). And it can be pretty useful for simple stuff: like my underwater nature shoots where you are basically just assembling clips of interesting stuff and throwing out whatever doesn't work; a short tradeshow commercial piece where you are just going for the gimmick effect; 3D porn; etc.

But it would also be mistaken to think that you would try to shoot a full length feature type movie these days without the capabilities you get from using a full-featured rig like ET makes.

Michael, this is a complex topic. Its difficult to discuss even when you have images to illustrate it. I have no doubt that you are aware of the parameters of 3D. If you're building rigs, then you know what you can get away with for the look you're after. There are many 3D rigs that have been utilized in recent history, yet there's a dearth of real understanding of the underlying complexities. Don't take my comments personally. I just think its important to make a clear distinction of these concepts for people reading this thread that haven't built and experimented with their own rigs yet.