High quality but economical gyrostabilized gimbal for Epic/Scarlet

[Eki Halkka]

In short, the challenge of stabilizers is this: There is no way to distinct gravitation from other forces acting on an object.

I'm trying to solve a problem rather similar to this... what i am trying to come up with is not a stabilizer, but rather just a 3D camera tracker using sensors. Kinda the first half of what you wish to do. There are a few affordable "9 degrees of freedom" daughter cards for Arduino microcontrollers that might be up to the task.

The challenges we have are very close to those with UAV's, unmanned aerial vechiles. There are some affordable solutions for those that look very promising too: https://store.diydrones.com/category_s/1.htm

[Lauri Kettunen]

All we need is a line between the centre of gravity of the earth and the centre of gravity of what we want to stabilize. Once we have that line established we can find a horizon (90 degrees to that line) and all other angles in relationship.

That's correct and is precisely the same as knowing the direction of gravitation. In fact, knowing any line with respect to gravitation is enough. This is trivial when the camera gimbal is in rest or moving with constant speed. For, in such case gravitation is the only force acting on the camera. But when there are accelerations, then it's impossible to distinct between the accelerating force/torque and gravitation. Instead, one employs motors to hold angular velocities in zero. With time small errors accumulate, and that's the challenge.

[Lauri Kettunen]

challenges we have are very close to those with UAV's, unmanned aerial vechiles.

Eki, yes, it's precisely the same challenge. Motion tracking is very sensitive to accumulating errors. If you have a 1/1000 error in gravity, in 30s the position is 9 meters off. Stabilizing the camera is more foregiving, as one needs to compensate only for the rotations. Translation forces have no effect.

[Eki Halkka]

Eki, yes, it's precisely the same challenge. Motion tracking is very sensitive to accumulating errors. If you have a 1/1000 error in gravity, in 30s the position is 9 meters off. Stabilizing the camera is more foregiving, as one needs to compensate only for the rotations. Translation forces have no effect.

Yep, drift is one challenge... but those boards should be corrected for drift at least to some degree, using sensor fusion. I think i will buy one of those to see how well or poorly they work in practice.

One possibility might be always starting the recording at "zero" position, and returning back there at the end, then compensating with the help of the known parameters: take length and the amount of drift accumulated during the take. That might work if the drift rate is constant (which it probably isn't).

In my case, mainly camera tracking on a greenscreen stage, it should also be possible to add an ultrasonic rangefinder's sender parts ("loudspeaker") to the camera, and three receivers ("microphones") to known fixed locations in the studio - the position could then be calculated from the delays / distances, pretty accurately AFAIK. But of course it would be nicer if the system was not restricted to the studio...

[Jason Sinclair]

From my understanding drift over say one hour is negligible. Is that true to your experience? In this case would not the solution be to reset the gyro in-between shots? What are the problematic to this solution? Obviously wescam and the american military have it worked out. I think one answer would be to have vision reading equipment, that measures pixel movement. One must remember that operator control doesn't need to be in the equation of fault as encoders can read all those movements, only vibrations and aircraft movement. But once the operator aims at something, it's fairly well in the operators eyes and for the most part he is going to get a shot just based on what he wants from viewing. Now all that is needed is for the vibrations to be taken from that shot and aircraft movement. (Object tracking) So you can use pixel recognition to track the object and keep it steady from there? Theres a big difference between operated camera and other technologies such as missile guidance i think, but even then, it's probably why they use cameras on drone missiles. I think there is a two part system, the gyro to get a rough look in, which allows for a small amount of drift, and then some other system to get precision such as pixel movement. Either that or there is someway to reset the gyros. For instance, the camera man could center an object and reset the gyro. A horizontal line through that object would give you it's horizon and if you circled around that object it will remain steady, if all readings regard that as the horizon?