Add a load cell to the roller

This is just an idea that I want to put out there so it’s public domain.

Using the motor current to measure tension works pretty well, but it’s not a perfect system. If we decided that we really need higher precision tension measurements we could add a load cell to the roller that the belt passes over. That should give quite an accurate measurement of the tension in each belt all of the time.

3 Likes

This is a great idea. Probably would make accounting for stretch a lot easier right? @bar how would something like this get integrated into the machine? I have no idea how these work…

Would love to be the guinea pig with testing this!

I’d like to also put my idea here as well, where it might be to better use than in my head.

Reading the different accuracy topics. I was wondering if separating the load bearing belts and the positioning system would be an option. Having independent position data is essential for analysing the exact problem and help immensely to debug and check potential solutions (also the one Bar considers).

The goal would be load/stretch independent position data. Additionally it would detect movement even when the belts/motors “think” they are stationary. (I would want this on my device all the time, but at first would be most critical in updating calibration on all existing machines)

This could be a crude gantry style setup, where a belt connects in a loop to a carrier and an encoder. Having two would give you X,Y data. (This would be a one time analyse and debug setup)

Another system that came to mind would be more in line with the Maslow, but more custom. I’m imagining an thin dyneema (or other non stretch line) running parallel to the belts. A passive spring retraction mechanism (ski-pass or vacuum cleaner cord roller type mechanism) could pull it taught and by a roller that could be checked with an encoder. Here three of these lines should give you all the data for X and Y.


(please excuse my drawing skills, hoping it clarifies my writing and train of thought)

1 Like

What about suplimenting the position data with some external time of flight sensors, place one on each arm aimed at a target on that arms anchor?

Example: Adafruit VL53L1X Time of Flight Distance Sensor - ~30 to 4000mm [STEMMA QT / Qwiic] : ID 3967 : Adafruit Industries, Unique & fun DIY electronics and kits plus a 3d printed mount to hold it in an exact position on each arm, then feed the data back to the esp32.