Building a gimbal (Part 7 Conclusion)

The new gimbal from the back

We have now increased the load the gimbal can take by using much larger and more powerful motors. These motors are much larger than our previous one. The are the same configuration of our smaller motors, 14N12P.  These newer motors have much higher torque to handle larger cameras, including professional DSLR cameras.

This gimbal has a yaw motor with a slip ring so that we can allow that camera to rotate as much as it wants around the Z axis. When suspending the payload from the balloon, it is connected to the balloon from a single string so it will rotate around that string.  We got a 50 mm diameter brushless gimbal motors with a 12mm hollow shaft where we can put the slip ring into. We got the slip ring together with the motor just to make sure it would fit. We bought it from Aliexpress.  Using the fedex express service we got it in a week.

We got this setup in a few hours, lasercutting all the arms of the gimbal and bending them the usual way. (with a heat gun). Onboard we have a cannon powershot camera on board.

It worked as expected, with the storm32 board controlling all the motors.

The three motors drew a current of around 0.9 amps. The previous motors drew a current of around 0.5 A.


Here is a video of the whole setup in action. We have not yet wired up the slip ring and it is a job for another day.

Here is some footage taken from our camera from the Imperial College Robotics Society’s lab.

In this video we did not zoom in. We do still have some jitters in the footage. Note that we were moving the base of the gimbal around quite rapidly, like in the previous test video.

In this video, we used the optical zoom of the powershot to see if the jitters were a problem when zoomed in. Unfortunately, the jitter is more obvious when we zoomed in.

I have simulated some long exposure shots from these videos by averaging successive frames. The source code I used is here on github. The program uses opencv to view the frames of the video and averages the frame data using python.

Here are some images from the simulated long exposure shots.

The zoomed in shots are quite blurry because it appears to have moved quite abit. It looks relatively fine when zoomed out.

Leave a Reply

Your email address will not be published. Required fields are marked *