Camera / Projector Revolve

This was a really interesting project. For an exhibition in Project Arts Centre called A Landscape of Known Facts by artist Katya Sander, I built a machine that rotated first a camera then a projector, continuously, accurately and at very low speeds.

The revolve was to be ready before the piece was made as footage needed to be shot on a DSLR camera using the revolve and then projected in a gallery using a modified version of the revolve at exactly the same speed. The required speed of rotation was 1/3 of revolution per minute or 0.333 RPM, which is very slow indeed.

Briefly. I used a motor, bearing and mount assembly from a dead Martin moving light fixture (I had to replace the actual bearing, motor and belt). Moving heads (movers) are amazing contraptions to disassemble and a dream for inventors and fabricators. I think it’s a marvel that often 20 or more of these things hang over a stage, doing their thing, with hardly anyone noticing. They are among the most sophisticated off-the-shelf ‘robots’ out there.

Mechanical power is provided by a stepper motor and transmitted via synchronous belt and pulley train. The pulley configuration provides slightly higher gear ratio, but not enough in hindsight – if I had anticipated a video stuttering problem (see below) at this stage, I probably would have used a gearhead stepper motor with a much higher ratio, (75 or 100:1 maybe) and driven the stepper at a higher speed.

My design placed the projector and video playback device on the revolve and the audio playback device off the revolve. That the image was to continuously rotate meant that power to the projector (and playback device) needed to pass through a slip ring as cables would have coiled and jammed. This was the reason for having the playback device (a MacBook Pro) on the revolve (after the slip ring) as high quality slip rings with enough circuits and noise shielding to allow for video or data to pass through were super expensive, but they do exist. The slipring I chose was a mercury disc type one – pretty amazing things. Instead of rings of metal (gold in the high end ones) and contacts, this one had discs of liquid metal (mercury) so that continuity with the rotating contact in each circuit couldn’t break – so no noise!

I used a three circuit slip ring to get power and earth to the revolve platform to power the projector and laptop (care must be taken to select a slip ring that has a sufficient current capacity). I wish I had photos of this, but the Mac transformer was attached to the back of the projector using sawn-off RCA jacks – glued to one side of the transformer which then attached to the back of the projector using it’s RCA sockets. nerdy triumph.

The stepper controller I used was from Jaf Motion (I fried an Arduino Stepper shield) and was controlled over USB from a Processing application. The Processing app was on a machine off the revolve – it played the audio and controlled video playback on the laptop over an ad-hoc WiFi connection.

The video below show a test of the mercury slip ring – the projector can rotate continuously while powered on. Note the mount with shelf for laptop under the projector. Mount was made in sheet metal and spaced using PCB standoffs. (apologies for terrible video quality)

The video below shows some footage captured using the revolve using a Canon DLSR camera. It may not show here but viewed at full resolution some slight but regular jitter was visible on all of the footage. I gave some possible reasons for this to the artist during shooting (see copy of suggestions below). The main reason (I think) was down to the individual steps of the motor registering on the footage due to the low speed – this is why a geared stepper would have been better (more steps per revolution). I removed all of the jitter completely using the amazing Virtualdub (had the best stabilisation results for this particular problem out of any software I tried, including After Effects, Premiere and Final Cut Pro). It takes a while to set up suitable passes but it was worth it.

Some suggestions I gave while we were trying to figure out the jitter, one or two are pretty ‘out there’ but plausible all the same:

Something else that jumped to mind was whether you are using image stabilisation on the Canon or not. There’s are reasons I can think of why it may be creating this unwanted effect by being either on or off. It would be worth trying to shoot some footage to test this. If image stabilisation is on – which it probably may be – what could happen is that the camera could be trying to compensate for very regular ‘noise’ on the y axis (i.e. the steps of the stepper motor), since image stabilisation is designed to reduce unwanted movements produced from organic, irregular human movements, it may not be very sucessful at reducing movement from the revolve and might actually magnify or even be responsible for some of stuttering.Another possibility is that the stuttering may be a product of constructive interference between the stepping frequency of the motor and the framerate of the camera – the two frequencies could be out of phase in a way that produces unwanted motion in the video footage. As far as I know, the Canon shoots progressive frames at a fixed framerate of 30 FPS. Progressive meaning that each frame is captured by the CCD and stored in one pass, as opposed to interlaced where each frame is captured as two fields and mixed together alternatly during playback – resulting in 60 half frames. If this is happening, a possible (hack) solution might be to very slightly change the speed of the revolve, plus or minus a 1% perhaps. If you do try this, make sure to note the exact speed you used for each shot, as syncing up the video upon playback will rely on knowing this speed.As I am sure you are already aware, video compression adds another layer of complexity to figuring out the cause of certain video artifacts, so too do the physical attributes of the camera. The Canon 5D mkii seems to have issues with what is known as rolling shutter, where unwanted movement is untroduced during panning shots, it may be more apparent with a motorized panning head.

there seem to be some solutions to reducing or removing this in-camera artifact online

Compression could definitely be an issue with this problem you are having – especially since you have noticed a reduction when using a different compression type. The Canon 5D shoots video natively in H264, that while being a sophisticated and very high quality codec, is a ‘lossy’ compression technique that uses intra-frame compression as part of its suite of compression tools. Intraframe compression uses data from preceding and/or following frames to apply compression to the current frame. This allows for more intelligent and usually higher compression ratios but may also discard information that is crucial to particular types of motion, and panning seems to be one of these.

One test you could try would be to shoot in 720P 60FPS – this will double the number of frames shot which, while reducing the resolution of the image, will increase the temporal resolution which is important to capturing motion. The footage could then be pulled down to 24 or 25 fps in post-production for playback purposes if needs be.

machine design, motion control, moco, stepper motor, motor controller, mechanical design, machining, visual arts