Delta arm length discrepancy

Nxt-1

I have a modified Rostock delta printer. I have no way of accurately measuring the entire arm length, but measuring with a simple tape rules comes out to about 290mm. The CAD drawings confirm this as well.

If I let auto calibration do a 7 or 9 factor run, the arm length gets 100% consistently resolved to 303.314mm. What might be the cause of this very large discrepancy?

bot

@nxt-1 I think the common advice is to not use 7 factor. Essentially, there is not enough info to accurately guess the arm lengths unless you probe a larger area than most print beds allow.

The reason it's picking a definitely wrong number, is because it's trying to fit everything into a model. Using the settings you otherwise use, with those arm lengths, would cause the probed points to, in theory, be more close to each other -- however, there is likely other distortions being introduced that are not captured by simply probing Z height, or specifically at the selected probe points.

Nxt-1

@nxt-1 said in Delta arm length discrepancy:

I have a modified Rostock delta printer. I have no way of accurately measuring the entire arm length, but measuring with a simple tape rules comes out to about 290mm. The CAD drawings confirm this as well.

If I let auto calibration do a 7 or 9 factor run, the arm length gets 100% consistently resolved to 303.314mm. What might be the cause of this very large discrepancy?

Yeah, I ways refrain from using 7/9 factor when I am doing anything but testing and tinkering. I was just wondering whether the discrepancy is meaningful in any way.

dc42

@nxt-1, one reason for the discrepancy could be that your tower steps/mm setting in M92 is higher than the actual value by the same factor. For example, if your belt tension is low then the steps/mm will be lower than the theoretical value, which assumes a belt under a lot of tension. If you reduce your M92 setting by a factor of (290/303.4) then I think auto calibration will leave the rod lengths at 290mm. But you may then find that your sizes are off. One way to establish the tower steps/mm is to compare the heights of 2 different calibration cuboids or towers, one short and one very tall.

Other factors that might cause it include:

Bearing spacing at the carriages being not quite the same as bearing spacing at the effector
Z probe sensitivity (and hence trigger height) different between the edges of the bed and the centre - this may happen if you use FSRs under the bed as the Z probe
Using strong springs between the arms (to attempt to remove backlash), which can pull the arms together by varying amounts depending on the XY position (pulling them together alters their effective length).

There are certain to be other causes that I haven't thought of.

Interestingly, when 7-factor calibration is used, the rods always seem to be adjusted upwards.

To get good results from 7- or 9-factor calibration, use as large a probing radius as you can. Between the towers you can even probe at a larger radius than the printable radius, if the bed is big enough and the joints have enough movement.

Nxt-1

@dc42 I've never checked the steps/mm from the default 80 the printer came with. Do I understand correct that steps/mm in a delta printer, for the carriages and not the effector? If so, is there a way to move a tower individually, so I can easily use calipers to tune the steps/mm?

I have always find belt tension a difficult subject. On my printer they are tensioned by turning the steppers of center (which seems to work fairly well, but it's subject to user error I guess). Is there a rule of thumb or guide to eliminate user variation?

Danal

@nxt-1 said in Delta arm length discrepancy:

@dc42 I've never checked the steps/mm from the default 80 the printer came with. Do I understand correct that steps/mm in a delta printer, for the carriages and not the effector? If so, is there a way to move a tower individually, so I can easily use calipers to tune the steps/mm?

On a Delta, a Z move which starts at X0Y0 (and doesnt change X or Y) moves all three towers, and the effector, the same distance. Use that to check steps/mm. A possible sequence is G28, mark a tower, then G1 Z-100, then measure a tower.

NO SPRING TIGHTENERs in the belts. Those things are horrible and should be shoved under the fingernails of whomever thought they were a good idea. (No strong opinions here, eh?)

I have always find belt tension a difficult subject. On my printer they are tensioned by turning the steppers of center (which seems to work fairly well, but it's subject to user error I guess). Is there a rule of thumb or guide to eliminate user variation?

Tight enough to "twang" when "plucked", not so tight that the "twang" would EVER be audible...

Nxt-1

@danal I should have thought of the straight Z move, thanks for mentioning it.

As for the belt tensions, plucking them is kind of difficult since the only sit about 1 mm from the tower extrusions. I guess I keep using my gut feeling for now