Arm diagnostics via telemetry

Technical Details
,
1

Hi all - sharing some interesting (to me) data generated while working towards high precision cuts with the Maslow4. (I also care about high accuracy - but I’m more concerned about precision for now.) Below includes telemetry data for belt motor currents during retract all (presumably the closest thing to a baseline/“no load” condition). A big thanks to @ronlawrence3 whose telemetry additions in the firmware made this possible!

This started as an exercise to understand differences between each of the arms and their potential impact on calibration. Given the single setting for calibration current, I was curious to know if calibration might be destabilized by baseline differences in current demand across the arms (because current is the proxy measurement for belt tension). At this point, I think there are more significant factors impacting calibration (for example, see the discussion in @Ulmair 's Loose belts during calibration). However, I do still believe arm differences can be a secondary factor and, if nothing else, I think this kind of data can be useful to check on quality/consistency of the arm assembly process and machine health in the field.

Here’s what I got from running retract all across several dates (all with firmware v0.84):

  • 2024-10-02: Original field test run after a handful of test cuts; there are clear periodicities (pretty sure this aligns with spool revolutions) and the relative range of average currents is appreciable (~1.5x from smallest to largest)
  • 2024-10-19: Bench test; run after rebuilding arms with new JST-XH encoders - tried to pay more attention to uniform and firm-but-not-aggressive bolt tension when assembling the arm halves; still some periodicities, but smaller currents and smaller relative range (~1.4x)
  • 2024-10-20: Bench test; no deliberate changes from 2024-10-19 test, but things mysteriously improved; smaller periodicities, ~same small currents and smaller-still relative range (~1.3x)
  • 2024-10-21: Field test run after a failed calibration attempt (fitness too low at 0.44); clearly something went very wrong with the BL arm but I haven’t had a chance to look into it :frowning:

1_2024_10_02_retract_all_arm_currents
1_2024_10_02_retract_all_arm_currents859×547 63.9 KB

2_2024_10_19_retract_all_arm_currents
2_2024_10_19_retract_all_arm_currents859×547 54.2 KB

3_2024_10_20_retract_all_arm_currents
3_2024_10_20_retract_all_arm_currents859×547 52.3 KB

4_2024_10_21_retract_all_arm_currents
4_2024_10_21_retract_all_arm_currents859×547 61.1 KB

If anyone wants to look at the raw data (including many more telemetry fields) you should be able to access it in the following Jupyter notebook (Google Colab): maslow4_retract_all_telemetry.ipynb

2 Likes
2

That is very interesting! I love the idea of using this for diagnostics and as a way to detect issues. I am hoping that the switch to using bearings on the idler gear will help to bring these into being more consistent because in my experience the idler gear seems to be the largest source of friction.

It would be super interesting to track if the periodicies (is that a word?) line up with idler gear rotations or full spool rotations.

3

According to Merriam-Webster, yes. :smile: (Periodicity Definition & Meaning - Merriam-Webster)

1 Like
4

Funny English teacher quirk: I read “periodicies” as “periodicities.” :man_shrugging:t3:

1 Like
5

You are welcome. I am so glad someone is using the data! I knew it would help find issues, but I’m more a “code-monkey” than “data-monkey” so was hoping someone like you would come along and use it :smile:

1 Like