Wednesday, January 22, 2014

Building a panning time-lapse rig out of scrap

Anyone who's seen a Ken Burns documentary, or most any other nature or history documentary in the last ten years, has seen the phenomenon of a panning time-lapse.  A normal time-lapse movie is typically made from a stationary camera taking a photo of the same scene at regular intervals, anywhere between once every couple seconds or couple minutes depending on the time scale of action you're trying to capture.  When you take all these photos (usually a few thousand by the time you're done) and string them together very quickly in a movie, they give an excellent and fascinating perspective of how things change over periods of time that humans don't usually think in.  We usually think of ice as bobbing around in the ocean aimlessly, but when you take a time-lapse of it over the course of a day you see that it has a flow and pattern all of it's own.

 A panning time-lapse ads another element of motion to the equation; by moving the camera very slowly and smoothly over the course of the sequence, you can generate visually stunning videos where the camera appears to be moving at normal human speed while other objects are whizzing by.  Ken Burns uses these to great effect in his documentaries, and while he certainly wasn't the first, he helped bring the technique into mainstream media production.

The outcome seems easy, but to actually pull it off requires more work than you'd think.  The camera movement has to be INCREDIBLY slow; for many panning shots you'll only want the camera to move a couple feet over the course of many hours.  And for the shot to work well, this movement has to exceptionally smooth and consistent.  Any minor hesitation, even the tiniest wiggle of the camera will be visually jarring and ruin the illusion of a smooth pan.

There are plenty of commercial products to accomplish this; slider and tilting mechanisms, dollies with high-precision bearings and wheels and ultra-low-speed motors designed to creep a camera along at a .000001mph.  But they're not cheap; something to move a very small and light point-and-shoot camera starts over $1000, and to move a bigger camera like an SLR can easily surpass $5,000.

I don't have that sort of money.  But I do have a full carpentry shop, access to a very diverse range of scrap materials, and (if I may toot my own horn) a fair helping of creativity.

I dug around in our scrap bins until I found about five feet of fiberglass handrail, clear acrylic sheet, some nylon all-thread, a hobby servo and a few random gears.  The comms shop was able to give me some electronic bits that they'd scavenged off old equipment and a dead radio that had been dunked too many times in the ocean, then from the labs I scored an empty pippet tip box.

None of this was planned ahead; I designed and built on the fly making adjustments and figuring it out as I went along.  I experimented with a couple different methods of moving the camera, including some pully systems and motorized dollies.  They all had their strengths, but eventually I settled on a long jackscrew mechanism with the camera mounted on a carrier that slid along through a channel cut into the fiberglass handrail.

Getting everything lined up so that it could turn freely without binding turned out to be much more difficult than I thought.  I messed around with a bunch of different methods, but eventually I found that a 3/8" copper plumbing tee with a nut smashed into the bottom was exactly the right height.

The nylon all-thread fit snugly into the middle leg of the T, turning easily without having too much play.  To make the shaft rotate I used a cheap hobby servo, reduced through some plastic gears to give more torque and even finer control over the shaft's rotation speed.

For any R/C nerds reading this, the servo came from the factory set up for 360 degree continuous rotation.

Servos are more complicated than they might seem; unlike a plain DC motor where you vary the voltage to control it's speed and direction, hobby servos have a built-in motor controller and gearbox.  This makes them perfect for this application which requires very precise and very slow speeds, but it means that to make it work you need a micro-controller to communicate with the servo's circuitry.

The Ardunio Pro Micro with it's ATMega 328 chipset was vastly overkill for this task, but it was what I had to work with and it had built-in voltage regulation for the servo, greatly reducing the risk of burning it out.  I wired in a three-way toggle switch and linear potentiometer to allow me to control the direction and speed without having to re-program the chip every time, and wrapped the whole thing up in a pippet tip box that I got from our science labs.

For power, I took apart an old radio that had been subjected to one too many dunks in the ocean.  I ripped out all it's guts and used it's back for a battery mount; this allowed to to use rechargeable long-life Li-Ion batteries from our radio stockpile, rather than being reliant on AAs or an A/C adapter.  I dabbed on a little bit of RTV sealent on the penetrations to give a hope of rain-resistance, and the rig was ready to shoot.

I've been shooting with this thing for a few months now, and the results have been very good.  I can turn the servo's speed down so low that the shaft rotates at less than 3rpm, and with those red gears reducing the speed there is plenty of torque to move the carrier in one continuous motion without it being jerky.

It's not foolproof; there's still more play in the carrier movement than I'd like and the weight of the camera makes it a top-heavy, so it's very susceptible to wind.  Anything more than a gentle breeze will give enough shake and movement that while you might not be able to notice it from looking at the contraption, the images from the camera are all over the place.  Many days are too windy for the rig to be useful and I have to use a normal tripot and keeping the camera stationary, because those windy days are usually the ones that blow the ice around the most and result in the most interesting shots.

Last week I finally sat down and went through everything that I've shot so far this season.  I've collected over 100,000 individual images, totaling close to 280gb.  Editing this sort of thing is a nightmare; even the most powerful video editing software has a tendency to barf on your face when you dump ten thousand full resolution images into it at once.  I eventually resorted to rendering each clip as a full-resolution uncompressed video (at roughly 3gb/minute), and then editing those videos together into the compliation you see below.

It was an insane amount of work for what amounts to a three-minute clip, and I hope you like it.  Turn on your speakers and click through to youtube and watch this at the highest resolution you can.  I uploaded it in uncompressed 1080p, and if your monitor supports that maximum resolution, I think you'll find it worth the extra few minutes to buffer.

(And if you recognize where the music is from, congratulations.  You're a nerd.)


Anonymous said...

Excellent video...impressive pan and lapse clips in there.

I am an Electrical Engineer, so I understood most of what you said, hats off building the thing in Antarctica!

Not like you can drop into a Fry's to pick up components and stuff.

Chris said...

Nice job B, I really like the last shot with the railing from the balcony, as it pans from left to right, keep up the good work and evolving your time-lapse making. -Chris "The WAS"