3D Stereoscopy – Canon 5D MkII and some DIY tips and tricks

I’ve had some enquiries at work, as well as some recent facebook conversations about low budget DIY 3D, specifically around the Canon 5D/7D.

First off I’m going to talk about the Canon’s and why it doesn’t work… at the moment, and I’ll give a link and pics to people who don’t care and have done it anyway. So read what I’m saying but don’t necessarily let it stop you from experimenting.

In this age of firmware updates and software driven, microprocessor controlled cameras (basically all of them), the camera’s capabilities are typically defined and limited by software rather than by hardware alone. This is fantastic as it allows new features to be added over time in firmware updates that can be released over the net. One prime example of this is the update that will allow 23.98, 24, 25fps recording on the Canon 5D. It didn’t come out of the factory this way, but market demand dictates to Canon that they missed the mark and the demand is for those missing frame rates. They built the camera in such a way that they can change the software, make it available as a download, and keep their clients happy. It’s no longer a case, as it used to be, that you had to buy a whole new camera, or update the hardware inside. Red Digital Cinema has been doing this for years, and with each new firmware update, a Red One owner finds themselves with a brand new camera. The hardware hasn’t changed, but the software and codecs have become more sophisticated and better with time.

Back to the Canon, the two main issues facing anyone wanting to shoot 3D with the Canon’s are sync, and codec. Both of these issues are actually tied to the same underlying need, a need which the Canon’s cannot meet and no firmware update can change this.

For 3D stereoscopy to work, both cameras shutters must be perfectly in sync, in fact, really the sensors must be read at identical times down to the millisecond. Right and left eye must be temporaly identical, exposures taken at precisely the same moment. This is especially the case when the camera moves, or a subject moves through the frame. Any temporal shift that results in a mis-match between the relative positions of cameras or subject due to out of sync shutter will compromise the stereo effect.

At the moment, there is no way to sync the shutters on two 5D’s or 7D’s. The other issue is that H.264 is not recording true frame by frame motion in any case. A frame is an approximation, and may share much of the image data from the frame before it or after it (in image areas with limited or no percievable change). This in fact may not make 3D an impossibility but it is certainly a major compromise.

Here’s a link to a very good discussion on the Cinema5D forum and some pictures of a stereo rig made from 3DFilmFactory’s excellently priced beam splitter rig.

http://www.cinema5d.com/viewtopic.php?f=14&t=4813&st=0&sk=t&sd=a&sid=d774b80aa5ffc8f9d138a737db104c16#p33219

5D MkII 3D Rig on modified 3DFilmFactory beam splitter

another angle of the 5D’s on the beam splitter

Routing of the fibre optics

The 3DFilmFactory 3D BS Pro priced at only $4295

Here you can see the modified rig, and the mothod they have used to sync the “record” start/stop as best they could by splitting the infra red signal from the remote over fibre optic cables to the two respective IR sensors on the camera bodies.

http://www.3dfilmfactory.com/3d_camera_rigs.html#3dbs

This is the best, most cost effective beam splitter you will find, it’s really a indie DIY 3D dream come true.

I want to mention a couple other generic but important considerations.

Inter-ocular distance is the distance between the optical centre axis of the two cameras. This needs to approximate the distance between two human eyes, but the rule is about 1/30th of the distance to the nearest object.

Convergence is the ability to bring the two centre axis together at the point of focus. In this case, convergence is  a matter of an adjustment at the camera base plate, and could be geared to the motorized focus mechanism driving the focus rings on the two lenses.

Some rigs us a parallel set-up where inter-axial distance is changed according to the distances and composition of the shot, some choose the convergence method. Some post adjustment may be required in both cases.

Here is a great discussion on this topic:

http://forums.creativecow.net/thread/268/134

The last thing I think should be a consideration, is the method of motorizing and controlling the focus of two optically matched lenses. ARRI and C-Motion, perhaps others have 3D solutions.