To understand the basics of stereoscopy, depth, parallax, the inter-ocular distance, convergence, fusion and the challenges in reproducing stereopsis in cameras and displays, please read Driving Miss Digital. For an excellent primer, take a look at A Beginner’s Guide to Shooting Stereoscopic 3D, by Tim Dashwood.
Setting the Interaxial distance
So the inter-ocular distance is 2.5″, or about 63 mm. This means that the distance from the vertical optical axis (the middle of the lens) to the edge of a camera body should be no greater than 32mm. From the chapter on Ergonomics, you’ll notice that none of the cameras in this guide can be placed side by side to be within this distance.
The closest are the Red Epic and Scarlet, which can be placed about 3.5″ apart. At this interaxial distance the closest object you can shoot at will be about 10 to 12 feet away from the camera. If you’re shooting narrative fiction, this is tough to pull off.
In any case, the first thing you’ll need to do is figure out what the distance of the subject will be in relation to your camera. For more than one setup, you need to figure out the closest and the farthest distance and focusing points. From this data you should be able to calculate the interaxial distances you’ll need to set for each setup depending on the 3D effect required.
This is critical. Once you set the interaxial distance, you can’t change it. There are some productions that try to shoot with three cameras, but I’m not covering that in this guide. I won’t be covering 3D attachments either.
Convergence or Toe in
The next thing you’ll need to figure out is whether or not you’ll be toeing in for each setup, and how much you’ll need for the effect you want to create.
Toeing in is a controversial subject. Not everyone agrees whether or not this is necessary on set. Some feel it’s better to make convergence adjustments in post.
It is critical for good stereoscopy to have perfectly synchronized frames. The underlying technology that allows this on the cameras in this guide is genlock. Sometimes this might be known by another name. The cameras in this guide that have genlock were covered in the chapter on Making the Connections.
Obviously, I don’t recommend shooting stereoscopy with any camera incapable of receiving and locking on to a genlock signal. However, if you desperately want to shoot 3D with DSLRs, here’s an ingenious solution from Peter Clark at Attic Studios.
HFR – High Frame Rate
It is a well known fact that increasing the frame rate increases the perception of resolution. This is covered in Driving Miss Digital. Keeping its aesthetic merit aside, it is extremely beneficial in avoiding temporal artifacting and motion blur in stereoscopy.
Some cameras in this guide only go up to 30 fps, while others go to 120 fps and beyond. If HFR is part of your workflow, you need to know the limits of your sensor at the resolution you are shooting in.
If you’re using a zoom lens on each camera, they’ll need to have synchronized movements so they are at the same focal length.
For lenses that need focusing, a follow focus system that simultaneously controls both lenses precisely is extremely important. A slight variation in the speed or accuracy of the focus or focal length will result in almost unusable stereo images. It is also important to get lenses that are as close to each other in all respects, if possible with consecutive production batch numbers (thought that is not a guarantee of anything).
For a brief overview of 3D solutions pertaining to zoom and focus control, check out this PDF.
Rules of thumb:
If you are happy with the interaxial distance in a side by side configuration, you should shoot on a side-by-side rig, as it is the most simple to set up. If you need to shoot at interaxial distances shorter than what is physically possible side by side, then opt for a beamsplitter rig.
A 3D rig needs to offer the following:
For greater control, you’ll need a setup that gives you maximum flexibility:
For a top of the line system that can take the Alexa and F65, check out the:
Beamsplitter rigs are more complex beasts. The most important component being the beamsplitter mirror itself:
It is critical for each camera to be on the correct optical path, at exactly the same distance, precisely leveled and genlocked.
To change the interaxial distance, one camera is moved laterally. A beamsplitter can hit 0mm and above, depending on its specifications. The 0mm setting won’t result in stereoscopy, but it can be used for HDR imagery.
To toe-in, the camera is turned at an angle.
For most prosumer setups, this is an excellent rig:
For the Red cameras, the 3alitytechnica Atom is perfect. For the best possible flexibility, the top of the line solution is:
To take the beamsplitter underwater, check out the Gates Deep Rig for the 3alitytechnica Atom/Pulsar systems.
Monitoring is a very frustrating problem for stereoscopy. The interaxial and convergence values are selected based on a certain display size assuming a fixed viewing range. If your final delivery is television, you can opt for different sized 3d monitors on set. But if your final delivery is for cinema, it’s tough to gauge the actual effect on set.
Then there is the problem of what viewing method to adopt: Autosteroscopic (no glasses), anaglyph, polarized, etc. Every method has its pros and cons, and slightly alters the perception of stereoscopy on the set itself.
Finally, there is the issue of data conversion. If one is recording on RAW, how does one pull data in real time for monitoring?
There are solutions to many of these problems (but not all of them). Usually all of them get in the way of the creative effort, and are not fun to use – unless you throw a lot of money and hire people to do the dirty work, and trust their judgement. But that can throw up nasty surprises in the viewing room, and by then it’s either too late, or you’ll need additional money or time to fix it in post (assuming it can be fixed at all).
You can appreciate why not many professionals shoot narrative 3D. It just feels like a chore.
For autostereoscopic monitoring, try this:
For a general system that is versatile, try the:
If you want to process HD-SDI Left/Right data in real-time and create dailies, take a look this solution:
The MC-100 multiplexes dual streams to one SDI stream and one HDMI stream that can be fed into the:
This system can be used for monitoring or creation of dailies. For an excellent calibration and analysis solution, take a look at Stereo3D Cat from Dashwood Cinema Solutions.
For a top of the line image processing system, try the:
We are done with rigs, believe it or not. Next we’ll look at the particular problem of streaming.