Potentiometer noise


Still on my motion control arm project, I am using the mini maestro 24 to register position of several POTs. All works great, but as expected there is always a slight noise going on, which makes the result oscillate */- 3 or 4 values. That is no big deal, but as I’m trying to increase the precision of the system, the pivot POT oscillating from 762 to 766 (for example) induces precision errors in the capture at the end of the arm.

The POT has the signal and ground twisted together, with ferrite beads at both ends, and the 5V power is running separately. It is only setup like that cause I’m still running tests, I am planning on having all 3 wires twisted together (or maybe should I twist only 5V and GRND together, and run the signal separately?)

My question is, what’s the best way to reduce the noise from a POT reading to a minimum. Higher quality POT? Thicker wires? I am trying to gather as precise data as I can…



Higher quality pots might help some. Try not to run the signal wire next to any servo power wires.

There are a couple methods of averaging the output which would make your result less noisy. The simplest is to put a small capacitor from the signal line to ground. You can also do it in software by averaging each reading with the last (weighted averages too, not just 50-50), or taking multiple samples for each measurement and averaging those.
With either method you have to balance smoothness with response time. An average that is weighted much more to previous readings, many sample averaged together, or a very large cap will all make the time it take to respond to a real movement go up. A perfectly noise immune system would not respond to any change on the signal at all, so you have to find the thin line of acceptable noise filtering though testing and/or math.