Testing the accuracy of camera stabilizers - RC Groups
Thread Tools
Jul 01, 2014, 08:23 PM
Zaitsevsky's Avatar
Discussion

Testing the accuracy of camera stabilizers


Comparison of camera gimbals

Rules:
1) The method of camera mount swaying must be clearly visible and reusable by others.
2) The critical level of camera vibrations must be clearly visible.
3) It is imperative that the camera films the moon or the angular size of objects is defined otherwise.
4) The purpose of the demo is to show the critical stabilizer operation mode, not just beautiful footage.


Most often, art cinematography is used for demonstrating the work of such devices. However, in this case, it’s hard to tell the proper work of a stabilizer apart from the work of a good cameraman. That’s why we need better criteria for comparing the quality of work of camera stabilizers. This is an important factor when choosing a model and when trying to detect device defects. If the mechanism is dirty and the axes aren’t rotating properly, this can be fixed. But first of all, you should find out whether your suspension is working worse than others’ designs.

Description of a possible test procedure.
The best time for a test is a night with a full moon. Use a telephoto lens (zoom). The use of an optical or digital stabilizer is not allowed.
1. Point the camera at the Moon with max zooming.
2. Tilt the gimbal 45 degrees forward, then 45 degrees backward and repeat it 5 times over 5 seconds.
3. Turn the gimbal 45 degrees left, then 45 degrees right and repeat it 5 times over 5 seconds.
4. This was a test of the stabilizer during fast tilting.
5. Slowly (under 5 seconds) tilt the suspension mount back and forth, then slowly turn left and right.
6. This was an accuracy test for the position controller.
7. Enable recording for 1 second. Leave the suspension motionless for 10 minutes. Again, enable recording for 1 second. There must be any still remote object besides the Moon or instead of the moon in the frame.
8. That was a slow drift test.
9. Disable software limitations on the speed of rotation. Run the camera rotation command and make a quick stop. This test should be filmed with an additional side camera.
10. It was a motion speed test.



We can quickly compare the efficiency of different stabilizers by looking at how the image of the Moon shifts in the frame. More accurate analysis requires a video editor and a graphic editor. We need to measure the size of the Moon in image pixels. If it’s too difficult for you, measure it on the screen using a ruler. The size of the Moon is 0.5 degrees. You can now calculate pixels or use a ruler to set the camera stabilizer variation threshold.
You will need to cut three video segments with camera controlled rotation in a video editor. The first fragment shows the camera going from standstill to max rotation speed. The second fragment lasts shows the camera rotating at max speed. The third one shows it going from max rotation speed to rest again. Assess the angle that the camera turns at in each of the videos. The duration of the fragments lets you estimate the performance of the camera in controlled rotation.

Gimbal test. Pointer-1 gimbal, ZYX-GS controller, 2.5kg load. (1 min 26 sec)


Why the Moon?
Because its visible arc size is the same in every corner of the world: from 29′24″ to 33′40″. The Moon is a nominally infinitely distant object and that is why hand shaking does not affect the accuracy of measurement using a trig function.
Why -45 to +45 degrees?
Because these positions are perpendicular and it’s easier to notice visually.
Why 5 seconds and 5 tilts?
Because it recreates extreme conditions. 5 tilts in a row give you a more accurate period of vibration.
Why are there no tests with sideway tilting?
Because it cannot be detected as a deviation of the Moon. For this purpose, you can shoot any other objects and measure the change of tilt in the video using an angle protractor. However, this measurement will not be accurate.
Why an additional object in the drift test?
Because the Moon slowly moves. For accurate measurements, you need to use a laser pointer that is installed near the camera. When a laser is used, the stabilizer cannot be moved sideways. If you hold it with your hand, deviation measurement using a laser does not produce an accurate result.
Last edited by Zaitsevsky; Apr 13, 2015 at 11:35 PM.
Sign up now
to remove ads between posts
Jul 01, 2014, 08:24 PM
Zaitsevsky's Avatar
If the drone flies smoothly or the operator moves the camera carefully,
it's hard to assess the work of the stabilizer. However, if the operator
stumbles or starts to dance, its efficiency will be clearly visible.

Approximate subjective assessments for similar vibrations:

An amateur-level stabilizer with a direct drive:
vibrations over 1 degree, shifts of 10 degrees or more

A high-quality stabilizer with a brushless direct drive and a rotation encoder:
vibrations below 1 degree, shifts of few degrees

Updated on april 15 year 2015

Please post test results for different stabilizers to expand and refine the conclusions.
Last edited by Zaitsevsky; Apr 16, 2015 at 03:56 PM.
Jul 05, 2014, 06:52 AM
Zaitsevsky's Avatar
This is how CineFlex is testing their equipment. Unfortunately, user can't repeat the same measurements.
Cineflex Stability Test (3 min 31 sec)

During creation I'm using different measuring equipment: gyroscope, accelerometer, laser level, torque, amperage, sound level, piezo noise sensor, wind generator...
F-Servo prototype #11, optical controller, balancing assistant (2 min 18 sec)

But the "moon test" is the universal method to compare different mounts.
Jul 14, 2014, 01:32 PM
Zaitsevsky's Avatar
Short test of the Pointer-1 stabilizer
ZYX-GS controller, 2.5kg payload.

Force Servo camera gimbal thread

Pointer-1 gimbal, ZYX-GS controller, 2.5kg load (1 min 25 sec)
Last edited by Zaitsevsky; Jul 14, 2014 at 07:09 PM.
Sep 16, 2014, 12:03 PM
Registered User
sooo when can i toss some of these on my old av130????? loooks promising
Apr 06, 2015, 04:11 PM
Zaitsevsky's Avatar
SimpleBGC based NoName Gimbal
SimpleBGC based NoName camera Gimbal (0 min 45 sec)
Apr 06, 2015, 04:32 PM
Registered User
Well, as we are comparing systems, check this out:

TestFlight2 (16 min 22 sec)


65mm effective lens, no user tuning whatsoever, on a multicopter (meaning high frequency vibrations your handheld tests can't recreate). Compare to something your size, not a SimpleBGC

With a handheld setup, you can expect 2x-3x usable range over this video.
Apr 07, 2015, 05:19 AM
Zaitsevsky's Avatar
Quote:
Originally Posted by SamurAchzar
Well, as we are comparing systems, check this out:

65mm effective lens, no user tuning whatsoever, on a multicopter (meaning high frequency vibrations your handheld tests can't recreate). Compare to something your size, not a SimpleBGC

With a handheld setup, you can expect 2x-3x usable range over this video.
On your footage a critical level is not shown and can't be measured. So it is not possible to compare properly. Angle can't be measured properly if platform is moving away. Use a longer lens and repeatable vibrations. Smooth video is useless here.
Last edited by Zaitsevsky; Apr 07, 2015 at 05:24 AM.
Apr 07, 2015, 05:30 AM
Registered User
I don't think what you did is a proper comparison either. Who tuned the gimbal? Why do you repeatedly throw it until a gimbal lock occurs (by inducing large roll angles which of course a gimbal in this mechanical configuration can't support)? We all have an agenda; only in my case that was a customer made video showing the performance.

As your thread is called "Testing the accuracy of camera stabilizers", we might as well make it a fair comparison.
Apr 07, 2015, 06:42 AM
Zaitsevsky's Avatar
Quote:
Originally Posted by SamurAchzar
I don't think what you did is a proper comparison either. Who tuned the gimbal? Why do you repeatedly throw it until a gimbal lock occurs (by inducing large roll angles which of course a gimbal in this mechanical configuration can't support)? We all have an agenda; only in my case that was a customer made video showing the performance.
As your thread is called "Testing the accuracy of camera stabilizers", we might as well make it a fair comparison.
We have different stabilizers, but comparison method must be based on measurable values and repeatable conditions.
Please, show clearly visible critical level.
Mark an angle or use a moon as an universal reference.
45 degrees - is not a gimbal lock, it is a usual tilt limit for a drone.
Apr 07, 2015, 06:52 AM
Registered User
1. What are the measurable results in your video?
2. How repeatable is your test, being that you just grabbed it in your hand and violently jerked it around?
3. What is the exact version and tuning of your gimbal?
4. If you want the comparison to be taken seriously, you need to do much better than showing a completely random video of a gimbal jerked around violently.
5. No one does a 45 degrees roll tilt in aerial filming.
Apr 07, 2015, 07:33 AM
g0t rabb1t?
ABLomas's Avatar
Quote:
Originally Posted by SamurAchzar
1. What are the measurable results in your video?
Moon?
Seriously, what are you trying to archieve?
Apr 07, 2015, 11:41 AM
Zaitsevsky's Avatar
Guys, explanation is up there, first post.
Rules:
Swing method must be shown and repeatable.
Critical vibration level must be clearly visible.
Angular reference is obligatory.

I can't spend 16 minutes to watch your smooth flight via fisheye.
Don't tell me how to fly. I'm flying 45 degree with a telephoto and it is not a problem for a gimbal. Here we are talking about gimbal only and only measurable values are acceptable. Nice picture is not a subject, but the limit is the subject of demonstration.
Apr 13, 2015, 11:06 PM
Zaitsevsky's Avatar
SimpleBGC 8 bit (BaseCam, AlexMos), Custom gimbal

Custom gimbal, SimpleBGC 8-bit controller (4 min 40 sec)