Most Critical Measured Dimensions

Technical Details
1

Can I get some help with listing the most critical measurements that need to be made as accurate as possible that directly effect the accuracy of your cuts?

I’ll try from most important to not as critical (please correct or provide input):

  1. Distance between motors (highly important)
  2. Measurements asked during calibration cuts (highly important - used to calculate chain sag)
  3. Aligning sprockets to one tooth straight up (highly important though not sure how much the smallest increment plays into accuracy)
  4. Distance from top work piece to motors (medium importance - only effects vertical center of work piece and not accuracy)
  5. Distance from center bit to ring (not sure, seems like the default ring value is good enough?)
  6. Making sure frame is level

My first time through I measured close enough for most measurements. Now I’m wondering which measurements I really need to take my time with and get as accurate as possible.

2
  1. Distance between motors (highly important)
  2. Measurements asked during calibration cuts (highly important - used to calculate chain sag)

yes

  1. Aligning sprockets to one tooth straight up (highly important though not sure how much the smallest increment plays into accuracy)

the chain length will be wrong by the amount the tooth moves.

  1. Distance from top work piece to motors (medium importance - only effects vertical center of work piece and not accuracy)

isn’t this calculated by calibration

  1. Distance from center bit to ring (not sure, seems like the default ring value is good enough?)

this is calculated by the calibration, it may help to have this reasonably close
to start with, but in theory, any value should work.

  1. Making sure frame is level

this will make a little difference, but it should not make much.

Can someone test what the effect of this is? do a calibration, then raise one
end of the machine by several inches (put 2-3 2x4s under it), do an accuracy
benchmark, then move the supports to tilt it the other way and do an accuracy
benchmark, how much do the benchmark numbers change?

My first time through I measured close enough for most measurements. Now I’m
wondering which measurements I really need to take my time with and get as
accurate as possible.

the motor distance and the calibration measurements are the big ones.

2 Likes
3

To achieve the Zipper Tree Challenge, I analysed extensively the list of sources of errors, found some missing parameters, then searched ways to set parameters right.

Below I summarize my findings about chain length calculation for accurate sled positioning

  • the most critical machine parameters,
  • the missing parameter in the current official firmware (items 4 and 7),
  • the target accuracy needed for each parameter,
  • and my advice about parameters quite impractical to measure directly and requiring a “calibration routine” help :

ChainCalculation-Parameter-accuracy-measurability
ChainCalculation-Parameter-accuracy-measurability.png788×549 132 KB

Now, either the firmware uses the actual sag correction value , or gets improved with accurate catenary correction using weights, but either way weights are missing and necessary in the firmware to identify chain stretch.

(edit) I invite you to look at the Wiki page showing the sled position accuracy vs precision where the sled position and chain properties are highlighted.

7 Likes
4

Can you comment on how critical it is to set the sprockets at 12 o’clock at the beginning? I’ve printed an aid to help with that but I’m wondering if it is worth the effort to go through another calibration afterward.

I imagine ideally the sprocket would be slight pre-loaded in the normal direction of the chain pull (top vs bottom).

5

Calibrating 12OClock before going through parameter optimisation is the intended way.
Optimisation does not correct 12OClock afterward.

I did not evaluate the effect of slack in the gearbox. Before reading the level, I would recommend preloading slightly with your finger in direction of the sled pull to see if it makes a difference, and how much it does. If you do, please repport about it :slight_smile:

I invite you to review the error plots to help your evaluation: