HELP! Trying to Level Bed, Y Tower Driving Me Crazy!

[lbarger]

I have now spent a couple of weeks trying to dial in my RMax V2 bed leveling in order to print large objects I need for my business. I've gone through the procedure in the manual and a few others methods found on line. The obvious problem is the end effector rising at points opposite the three towers. However, in trying to diagnose what is happening (and what to adjust) I've come across another issue that is particularly troublesome.

My most recent print consisted of 1.95mm tall, 5mm diameter cylinders printed at 90, 210 and 330 degree positions on a 129 mm radius as well as at 0,0. (Base of Z, X and Y respectively) The print also includes a 0.75 tall, 1mm wide triangle connecting each of these posts. Measuring the posts, my calibration looks pretty good...
X 1.86
Y 1.85
Z 1.84
0 1.82
At all test points inside the towers I see a variation of just 0.08mm from lowest to highest points. Yes I should raise the end effector 0.1mm and could bump the horizontal radius down just a hair, but that is not the problem I am seeking advice on.

If I measure the thickness of the traces between each post, I find a much larger variation. Furthermore, that variation depends on the direction of motion of the end effector. The problem is most noticeable at the Y tower, but in double checking everything for this post, I see it is present at the other towers as well. The line connecting the X and Y tower is 0.77 thick at X and just 0.41 at Y. The line connecting Y and Z is 0.81 at Y and 0.62 at Z. The line connecting Z and X is 0.75 at Z and 0.59 at X. Stated another way, at X one trace(Y) is 0.77mm thick while the other (Z) is 0.59, at Y one trace (X) is 0.41mm while the other (Z) is 0.81 and at Z one trace (X) is 0.75 and the other is 0.62mm.

I'm guessing the cylinders are closer to target because the end effector does a 360 at each one while the line is printed in just one direction.

What have I checked?
1) All skates run smooth and have no play.
2) All belts have about the same tension.
3) Towers appear perpendicular to hot bed when using framing square.
4) All three towers are parallel. Diagonal measurements from base of one tower to top of adjacent all within 1/16".
5) Verified stock arms are indeed 269mm in length
6) End Effector has some wiggle which has been reduced by rubber bands between parallel arms, but is still present.
7) All screws are tight. Found a loose screw on the top of the Z skate on the Y tower side. Tightened it which seemed to help for a while. Issue came back when I started the bed calibration from scratch yesterday.

Unless some of you can offer any additional suggestions as to what to look for, I'm thinking my first problem to solve is the apparent free motion of the end effector. Any thoughts on Trick Laser's CF arms vs magnetic ball joints? Can you get the CF arms with magnetic ball joints? Could this be a tower rotation/angle issue?

I've seen the paper online where a person looked at effect of x, y and z position based on carriage displacement error for a couple different geometries. I have not found one document that discusses many different error possibilities such as arm length, horizontal radius, tower angle, and end stops and what effect those variables have on the position of the end effector. Has anyone seen such a document?

Any and all suggestions are welcomed.

[lbarger]

Recap - A minor displacement in X and Y near the base of the Y tower would cause the print head to raise/lower nearly .5mm. Makes it impossible to calibrate properly.

I kept going back to the play I was seeing in my end effector. Initially, it did not appear seem to be a big issue as it did take 'a little pressure' to cause the end effector to shift and/or rotate from horizontal. But how much pressure is too much? Needing to print some large pieces in ABS for work, I broke down and ordered carbon fiber arms. Immediately noticed a difference after installation. Play is gone. My whole printer/stand will rock on the carpet/padding before the end effector moves. (The motion of the printer due to carpet pad is barely noticeable.)

I calibrated the printer per instructions except using points on 129mm radius. First print of my calibration star had high low variation of 0.3mm where I was only to 0.35mm after weeks of tweaking the original arms. I created as spread sheet that calculates where the skates should go based on the X,Y, Z coordinates and firmware parameters. The spread sheet then allows you to enter different values for diagonal arms, horizontal radius, tower rotations and a few other variables. Using the new numbers, the spread sheet calculates where the end effector would land based on the introduced 'errors'. Playing with these numbers led me to rotate the X tower 0.4 degrees, the Y tower .1 degrees and modify my horizontal radius. Second print had variation down to 0.16mm. New numbers tells me minor tweaks to the end stops and one more adjustment to the horizontal radius should have the variation below 0.1mm across the full bed.

It appears I am finally getting this machine to the point I need it to be. Yea!