Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

High quality but economical gyrostabilized gimbal for Epic/Scarlet

Cliff Gill said:
Are you thinking specifically about an "A" system or a "B" system in reference to the link and description above?
Click to expand...

Cliff, I'm thinking of an "A" system, i.e., an electrically stabilized system. The goal is a rock-steady system. Constantine has a different approach, which is based on mechanical gyros.

We've discussed with Gilles by email. He has given a lot of useful feedback. I told him earlier today that my engineering intuition suggests that this kind of "A" system does not need to be astronomically expensive. Instead I feel strongly that it should be possible to do the same for gyrostabilizers what Red Digital Cinema has done to the prices of cinema cameras.

The crucial issue is --I preserve the right to change my opinion, if/when find that have made a mistake-- if

1) the gravity center of the camera is precisely the point where the three rotation axes intersect,

2) air drag is not taken into account (in practice, it's a good idea to have a spherical cover around the system)

3) there is no friction in the system (in practice, in the bearings),

then there is no torque that rotated the camera although the 3-axis supporting frame around the camera rotates. In other words, if there is no friction, there is neither forces that rotated the camera. This relies on the basic idea of inertia; if no forces apply, the object remains on its state of motion.

Of course, when the whole supporting 3-axes frame accelerates or deaccelerates, a force acts on the camera as well. But (right now I think) as this force acts on the center of gravity and if the center of gravity is the intersection of the three rotation axes, there simply can't be any torque.

Now, if I'm not mistaken here, all this implies the design should start from the idea of not introducing big motors and gears etc., as then one ends up with compensating mainly the torques generated by the very same motors. Instead, the design goal should be to minimize friction and then to control the rotation due to friction with a kind of 'bended linear motor' that acts the force directly on the frame that should be rotated.

I'm just thinking out loud and have not yet analysed these thoughts in detail. So, feel free to shoot down all mistakes and don't feel guilty to enjoy it!
 
Lauri,

Your concept makes good sense, but I suspect I am not knowledgable enough to comment accurately on it.
In attempting a passive non motorized system over the years, (a type C or C-1 system) using a kenyon gyro like Constatine's, I was never able to prevent drift or remove large oscillations due to perterbation. Achieving your requirements 1 and 3 are close, but never perfect, their will always be friction and always some non coaxial agreement between center of mass and axial intersection, and when one adds in wiring from the camera to points beyond the gimbal, forces are difficult to avoid if not impossible. I am guessing that is the reason that all of these system (that operate remotely rather than hand held) require motorization of some sort. Your blended motor concept is well beyond my knowledge.
Gilles did offer lots of great info when I emailed him earlier, he sounds like he is well along on a similar project, using a microstrain orientation sensor and motors.
Has anyone looked at:
http://triax-aerial.com/triax/products.html
Note that this system reports a drift free controller (the cms) as an option to the gimbal.
 
I have been working on a similar project. You are right it would be great if there was an economical true stabilized head out there. The price of the hardware is not too steep, the real price is the time it takes to research, and develop a solution. I have outlined some of the basics below, I could be missing a few details, but it is a start.

I have a Copterworks three axis head at the moment that I fly on my cable camera. I have it modified with R/C gyros, they compensate for tilt sway, and roll sway. It is a decent start in the right direction of a gyro stabilized head, does it really work like a real professional stabilized head, no. You just can't expect to modify some R/C servos add R/C gyros and viola a gyro stabilized platform. There is so much technology in a "real" stabilized head. Do I think it is possible to build a "real" stabilized head via feedback sensors... absolutely!

To start with, you need to think of the platform as self balancing, self thinking, and self compensating object.

Balance

Gyros sensors will give feedback in relation to the axis's g-force they are measuring. Accelerometers will give feed back as to how fast the axis they are measuring are moving. Last a magneto sensor to give you a value of the Earths magnetic field relative to the camera platform, a constant point I believe. You will need three axis of all of the above sensors. The name for these devices is IMU, Inertial Measurement Unit.

Thinking

After figuring out how the IMU works, you need to process the data streaming from it. This is complicated. From what I could process it involves a program to decipher the feedback data from the IMU to give an accurate reference to where the x,y, and z axis of the platform is at. This involves a microprocessor to constantly run a mathematical formula calculating data, and more than likely a smoothing filter of some sorts to get rid of sensor jitter. So we have clean data from the IMU, now the data needs to be referenced to the actual camera axis.

Feedback

You need to be running a closed loop between motor position, and IMU position. The micro controller has to be able to reference the position of the camera platform motors and the position of the IMU, and compensate for the IMU position via the motors. The only data that is added to the loop is input data from the operator. All of this data needs to be processed via stacked programs running on a micro controller.

If I were to make it this far in the project, I would be very stoked! The mechanics of a three axis head are basic, just a few design elements, and some trickery!

Let me know if this helps you out. I apologize if it sounds condescending, I guess it was in the tense I wrote it.

It is a job for a team of people!
 
There is so much technology in a "real" stabilized head.
Click to expand...

So much so, that the technology is considered militarily sensitive, and subject to export restrictions. Not hard to imagine the weapons targeting precision achievable from a highly stable mobile platform.

Still, I think a reasonably workable and low-cost system can be developed for mobile cam apps, hopefully without attracting the interest of the Men in Black. Lots of enthusiasm and great ideas already on this thread.
 
Tom, most of the information that I have acquired has been over the internet. It is just being able to know how, and where open source code is place in a string. Its all about software.

I am will keep you guys posted!
 
About these IMUs, I have been designing such systems especially ones that are meant to high accuracy navigation. We have also tested quite many sensors. The typical problem is that MEMS gyros drift, and second, if one rotates a full turn slowly or quickly, the sensor shows something quite different than a full turn. The output of the gyros depend also on the temperature. The first aid is to calibrate every sensor in a rotating table. Magnetometers are helpful assuming the field does not vary that much. Summing up, there's no trivial solution here, but still don't feel there's a dead end either.
 
Thanks Alex, amazing really, what's available on the internet just by poking around.

Lauri, on the aircraft I fly, laser IRU's are installed for navigation and attitude reference, costing many hundreds of thousands each of course. As Alex rightly points out, the software is key in developing a reliable system using low cost components.

The MEMS gyros being substantially cheaper, would it be possible to employ multiples of sensors in different alignments, output averaged in order to reduce error to acceptable limits? Perhaps GPS input for gross error check (the 'full turn' problem you mentioned) in order to derive a best computed solution?
 
I was at REDucation (the first one), and at Community Day there was a Tyler Mount mini-gyro there. (Torrey Loomis from Silverado was showing it along with Tyler Berry).


Tyler Mount mini-gyro there seems like a good fit for the RED on stability shots on the back of a motorcycle. (although I now have an Actioncam / Steadicam ) But for under $17,000 it seemed pretty nice. - I think that was the price .
 
Tom Hamilton said:
The MEMS gyros being substantially cheaper, would it be possible to employ multiples of sensors in different alignments, output averaged in order to reduce error to acceptable limits?
Click to expand...

Definitely, the average of several sensors makes the measurement more accurate. Not all axis need multiple sensors, the roll axis is the most critical one. Maybe also tilt axis require more than one sensor.

The whole thing should also be as light as ever possible. It would be nice to use the same gimbal on the top of a crane without a ridiculous amount of counter balancing mass.
 
Unfortunately, multiple redundant IMU's will not help eliminate the major errors, they will all have the same error, but measure it more accurately. On a fundamental level, the IMU's measure and correct for gravity. If an IMU is sitting stable on the ground, it will get an accurate measure of gravity, if an IMU is placed in an aircraft, it will measure gravity with the motion of the aircraft superimposed on the gravity, and its purpose in our case will be to try and remove the confusing accelerations of motion from the reference of gravity. If the aircraft enters a turn (and especially stays in the turn) the IMU will create a new reference of 'gravity' which will not be straight down, but sticking out the side due to the turn acceleration, and eventually start tilting the mount. The quality of the IMU will dictate the rate of this drift, a very high quality IMU will remember the correct orientation for longer than a cheap one, which may get confused after just a few seconds or maybe minutes. High quality IMU's acquire heading reference from the rotation of the earth, keep it stable on the ground for a few minutes, and it will know which way it is orientated due to earth rotation, they are that sensitive.

Has anyone experimented with triax aerial's cms product? triax-aerial.com
This CMS add on to the gimbal is apparently using an IMU (and not simply accelerometers, this from triax:
"The cms solution does use absolute orientation, using data from
accelerometers, gyros and magnetic compass to compute the control feedback corrections, which effectively eliminates drift.
The calculated corrections (not raw data) are fed directly to the
continuous rotation servos"
If anyone has experience with Triax's cms stabilization, let me know, as it seems economical (about $700, but optimized for their $1700 mount).
 
Cliff, you say multiple sensor improve only random errors?

And yes, in my understanding Cliff points to a major issue, which is about detecting gravitation. In an accelerating or deaccelerating object it's not possible to distinct gravitation from other forces. For, the force acting on a moving body is "mass times (acceleration + gravitation)", and there's no trick to say which acceleration is about gravitation.

If one had ideal gyros, it would be enough to first set camera horizontally, and then measure angular velocities and rotate the camera with respect to the gimbal such that all three angular velocities remain null all the time. But, in practice, when the camera is stationary one needs also to detect the direction of gravitation. Then the idea is to keep track of gravitation, and exploit this as a reference for vertical lines. And whenever the gimbal moves with constant velocity, in principal, accelerometers yield a new updated reference for vertical lines. The practical difficulty is to know, when the velocity is constant, that is when accelerations are zero.

Magnetometers yield another reference in addition to gravitation.

Notice, in tracking gravitation long enough, one should also take into account the Coriollis forces due to the rotation of the earth.

So, in the end of the day, one has to cook stabilization from imperfect measurement data. In this kind of situation it is common to employ so called Kalman filters to combine imperfect data to gain a better result than what one can get from any single measurement.

Furthermore, I suspect the eye does not easily recognize slow changes of pan, tilt and roll axis. And if so, the feeling of rock solid stabilization is about stabilizing within some limits with corrections slow enough.

Finally, notice that the success of Nintendo Wii has created a mass market for MEMS gyros and accelerometers. Consequently, the cost is expected to decrease while quality improves.
 
Here is a great blog about IMU's and Kalman Filters. It is all in respect to DIY UAV's, bit the theory can be applied to stabilized platforms.

http://tom.pycke.be/

For me the hardest thing to wrap my head around is the feedback from the IMU to the motors.

Any ideas?

Cheers
 
Quote - Has anyone experimented with triax aerial's cms product? triax-aerial.com
This CMS add on to the gimbal is apparently using an IMU (and not simply accelerometers, this from triax:
"The cms solution does use absolute orientation, using data from
accelerometers, gyros and magnetic compass to compute the control feedback corrections, which effectively eliminates drift.
The calculated corrections (not raw data) are fed directly to the
continuous rotation servos"
If anyone has experience with Triax's cms stabilization, let me know, as it seems economical (about $700, but optimized for their $1700 mount).[/QUOTE]

I would be skeptical about an R/C solution from past experience. Not to mention the whole point of having a true stabilized platform is the ability to use a zoom lens. R/C motors do not have the resolution needed for smooth stabilization. I checked out Triax's Vimeo clips, and it was posted that they perform "post-production" on their clips. Thats cheating! There is no "smooth-cam" on live events!
 
There are several reasons for GPS input here, as far as I can tell:
Notice, in tracking gravitation long enough, one should also take into account the Coriollis forces due to the rotation of the earth.
Click to expand...

On the equator, where Coriolis force is precariously zero, change dependent as one moves either south or north (one single step is all that’s necessary) a GPS position input would either negate the need for such sensor input, or at the very least provide accurate database correction wrt magnitude and direction.

In an accelerating or decelerating object it's not possible to distinct gravitation from other forces. For, the force acting on a moving body is "mass times (acceleration + gravitation)", and there's no trick to say which acceleration is about gravitation.
Click to expand...

The three-dimensional accuracy of GPS, given a clear view of enough satellites, is now remarkable. Here in the States, we are performing high precision aircraft approaches, requiring pinpoint accuracy in both vertical and horizontal planes, to small landing targets with nothing other than commercially available GPS units.

Assuming our software knows the weight of the camera system/platform, and which way/how fast it accelerates/decelerates in a three dimensional environment, gravitational forces acting on same can be theoretically identified, separated, and compensated for. I think. The math is certainly beyond me here.

Am I on the right track Lauri, or are GPS measurements not refined enough to contribute to a solution here?

Any solution involving GPS, of course, would necessitate a clear view of the sky at all times, not suitable for some applications.
 
Tom Hamilton said:
Any solution involving GPS, of course, would necessitate a clear view of the sky at all times, not suitable for some applications.
Click to expand...

If you have advanced enough receivers you can rather easily use GPS inside as long as you aren't in a cave.
 
Alex, thanks for the observation about the RC servo systems and the potential for being unsmooth. Lauri, I think I posted simultaneously with you about using multiple gyros, but I think both our points are still valid, multiple IMU's will improve accuracy, but are not very likely eliminate drift over time, they will still suffer the same drift effects.
GPS and IMU together are ideal for correcting each other, but this drives up the cost. The IMU provides for short term corrections, and the GPS provides for long term corrections, if they are well integrated in they reduce or remove the drift. But GPS/IMU systems (INS) are quite a bit more expensive, anyone know of costs?
Lots of people have done this work before, nobody has heard of any open source or maybe master thesis projects in producing a relatively low cost attitude and heading reference system?
 
Check out this forum http://forum.sparkfun.com/

Search under"Kalman" there are many many posts in regards to IMUs and feedback loops.

It seems like there are many people who want to make self balancing robots, and low cost autopilots these days.

Also I believe the accelerometer data vs. the gyro data combined makes up for gyro drift. I think one IMU is all that is needed. They do make horizon sensors, which from a camera perspective is really what needs to be stabilized. That might actually be the only reference data that needs to be stabilized, just getting three axis to work together......
 
You must log in or register to reply here.
Back
Top