Stabilizer For Less
The Stabilizer for Less has had its first real test: Tribe Nenuphar's ATS-Picnic, August 2015.
A Steadicam[a] rig is about $2,795. There are cheaper ones - for smaller cameras - available for about $300. Being cheapest, I had resigned myself to the fate that I would always have to post-process my video to get it stable.
Then Agnes[b] said that she'd like to get more moving shots of her dance - the fixed camera being just a bit boring - and I started thinking: Could I get a companion to the monopod for less - a stabilizer for less?
1. Problem
Looking through the online retailers of photo equipment, the outlook seemed grim as far as getting a stabilizer off-the-shelf. The only one that looked interesting, the Hague Mini-Motion Cam[c], was only good to about 0.8 kg (2 lbs), and it seemed near impossible to actually point the camera anywhere while using it. Still, it got some fairly good reviews. But that's a stabilizer for $90 when buying directly - and I didn't even know if it would actually stabilize my shots. The Nikon D3200 and a Sigma 10-20mm comes in at 978 grams, or 2.15 lbs, making it almost 10% overweight in its intended configuration. That's not a stabilizer for less, that's just less stabilizer.
I did find the $14 Camera Stabilizer[d], but that whole operation seemed to have gone out of business. Besides, I did want to use my tripod if at all possible. It would save me the hassle of hauling yet another piece of gear around.
I went onto the forums. (Don't go onto the forums.) They said that using a tripod was not a good idea, but I couldn't shake the thought that maybe it was - after all, all you do with a stabilizer is add inertia: weight gets you resistance to translation, and extension in space gives you resistance to rotation.
2. Construction
First a brief recap of how a stabilizer works. A stabilizer stabilizes the camera in two ways - linear movement, which is left-right, up-down, forward-backward; and rotation, which is pitch, yaw and roll. Linear movement is stabilized simply by making the whole rig heavier. More mass means more inertia, which means less movement of the rig when muscle tremors, sudden jolts, or other forces reach it through the handle. According to the same principle, Steadicam rigs can have flexible, spring-loaded arms that soak up jolts acting on it.
The mathematics are quite simple for this one: F = ma , or a = F/m , acceleration is force divided by mass - or, twice the mass gives you half the shake; three times the mass for one-third the shake.
Rotation is stabilized by elongating the mass that is to be stabilized, preferably placing as much mass as possible as far away from the center of rotation.
The mathematics are slightly harder for this case, mostly due to the fact that the tripod itself isn't massless, but the resistance to rotation of two point masses m_1 , m_2 connected by a massless bar of length L is roughly: 1/4 L^2 (m_1+m_2) . In laymans terms: double the masses to double the resistance. However, doubling the length of the bar increases the resistance four-fold. Three times the length gives you nine times the resistance.
I went and bought a pair of 0.5 kg wrist weights ($16). Strapping one to the bottom of the tripod and extending one leg out to the left gave me something that looked like the $14 Camera Stabilizer. For $16, admittedly, but with a whole lot less work.
With both wrist weights, I can even mount the massive Sigma 50-150mm on the camera and get it stabilized. The whole rig then weighs in at 3.9 kg.
Finally, I went and grabbed a pair of bike handles ($4) to make holding it a bit more comfortable. With a total weight of about 2.5 kg, it's easy to end up with a death-grip on the polished metal of the center column to keep the stabilizer from slipping, and the harder your grip, the more tremors are transmitted to the stabilizer. Having a rubber grip there enabled me to keep a more relaxed grip.
In total, the parts came to $20 (excluding the tripod), and the amount of work done can be measured in minutes. Cheap and quick, in other words. But how good?
3. Testing
As they say: The proof of the pudding is in the eating. In the end, what matters is how good the footage looks. I tried walking a bit, holding the camera as steady as I could, with and without the stabilizer.
4. Proper Testing
In order to get some idea of just what I had created, I decided to test this the proper way. I strapped an iPod touch to the camera and recorded the values from the accelerometers while walking a set path with the camera.
The following variables were measured:
-
x - acceleration left - right
-
y - acceleration forward - backward
-
z - acceleration up - down
-
α - pitch acceleration, rotation along the x axis. Camera shake along this axis causes the image to move up and down.
-
β - roll acceleration, rotation along the y axis. Camera shake along this axis causes the image to tilt left and right.
-
γ - yaw acceleration, rotation along the z axis. Camera shake along this axis causes the image to move left and right.
The stabilizer primarily attempts to stabilize α and then β, which are the two great contributors to camera shake and movements which are almost always entirely unwanted. Then γ, which, while a contributor to camera shake, is also an axis that the camera needs to rotate around for certain shots - for example, when walking and changing direction, or when you orbit an object, keeping the camera pointed at the object in the center. Perfect stabilization around γ is therefore not always desirable and is ideally tunable. The linear accelerations, x, y and z have comparatively little impact on camera shake, as the results of linear movement are rarely visible, unless the object being photographed is very close to the lens.
5. Summary of Results
| Variable | Acceleration | Jerk |
|---|---|---|
| x | 45% | 63% |
| y | 43% | 55% |
| z | 39% | 47% |
| α | 88% | 87% |
| β | 85% | 79% |
| γ | 25% | 24% |
All in all, not bad. About 80-85% reduction in acceleration and jerk where it matters to me.
6. Detailed Results
Three measurement series were taken:
Null - the iPod was laid flat on my desk and measurements were taken.
Without - walking as smoothly and steadily as I could, holding just the camera in front of me.
With - walking as smoothly and steadily as I could, holding the stabilizer with the camera attached in front of me.
The measurement values were then analyzed by processing them into the following:
abs - the average absolute amount of acceleration measured, a measure of how much the camera deviates from an absolutely smooth path
dev - the standard deviation of the acceleration measured, similar to abs, but with more weight given to larger deviations from a smooth path
delta - the average of the first derivative of the acceleration, a measure of the "jerkiness"[e] of the movement
| Variable | Null | Without | With | Damping |
|---|---|---|---|---|
| xabs | 0.01 | 0.29 | 0.15 | 47% |
| yabs | 0.02 | 0.28 | 0.17 | 42% |
| zabs | 0.02 | 0.20 | 0.13 | 39% |
| αabs | 0.18 | 3.50 | 0.43 | 88% |
| βabs | 0.27 | 2.30 | 0.38 | 83% |
| γabs | 0.22 | 2.97 | 2.23 | 25% |
| xdev | 0.02 | 0.36 | 0.20 | 45% |
| ydev | 0.02 | 0.37 | 0.21 | 43% |
| zdev | 0.02 | 0.27 | 0.17 | 39% |
| αdev | 0.22 | 4.59 | 0.53 | 88% |
| βdev | 0.34 | 3.08 | 0.47 | 85% |
| γdev | 0.28 | 4.03 | 3.01 | 25% |
| xdelta | 0.02 | 0.36 | 0.14 | 63% |
| ydelta | 0.02 | 0.30 | 0.13 | 55% |
| zdelta | 0.02 | 0.23 | 0.12 | 47% |
| αdelta | 0.20 | 4.58 | 0.58 | 87% |
| βdelta | 0.37 | 2.58 | 0.53 | 79% |
| γdelta | 0.29 | 3.64 | 2.77 | 24% |
6.1. Angular Acceleration
6.2. Linear Acceleration
6.3. Reference: iPod Flat on Desk
All biases were removed in a pre-pass. For example, the z sensor had a 0.33 m/s² positive bias.
