Tevo LittleMonster calibration issues

Hi guys. I really need your help in getting my Tevo LM calibrated or at least diagnosed.
Sorry for the long post, but I'm loosing my mind after two months of trying to get this beast working.

The machine is a delta printer, 100cm high Cbeams, 340mm diameter bed, 6 and 8 mm thick cnc alu plates top and bottm, 397.95 rod length (measured by me, 397.1073 in original config). Comes with MKS Sbase shitty board and BLTouch.
From the start using the latest edge smoothie calibration G32 and grid levelling G31 had some high and some low spots.
-The table is flat and horizontal, the machine base is flat and horizontal, the beams are square to the base up to 0.2degrees (DXL360S' ability to reliable measure). The machine is built like a tank.
-I'm using a center mounted probe for calibration, a simple microswitch with 6 micron repeatability, tested using 626pilot's smoothie fork (G29 E S15). I've tested with the stock DRV8825 drivers and TMC2100 drivers, both with stock steppers and with Wantai 0.9deg 1.7A steppers, the probe resolution is 1 step (or microstep) 0.006mm (or 0.012 if using tmc 16 ustep/mm with 1.8deg steppers). The probe is attached with some laser cut acrylic parts.
-I have no visible effector tilt detectable by a small bullseye level on the effector.
-I have replaced the optical endstops with mechanical ones.
-All rods have been measured with a jig and are all the same length +/-0.1mm (fit the same 4mm screws on the jig with the same force, but don't have a caliper this long)
-Distance between rods is the same at carriage and effector up to 0.1mm, maybe 0.05mm but I don't trust my chinese calipers that much.

I post some graphs, the measurements are taken using the latest edge build (G29 P1 I15 J155.0), the one from last week since the last one has some debug messages at every move.
1. Machine not calibrated at all.
2. Calibrated only endstops using G32 E0 I0.02
3. Full smoothie calibration G32 I0.02
4. SmoothieLSQ branch from purplefabr (Mark) or David's Escher3d online wizzard. They basically give out the same result, third place after decimal point difference.
I've copied the relevant data and used in the latest edge branch since the LSQ one does not have a good bed mapping G29 (only 7/13 points).
M665 A0.0000 B0.0000 C0.0000 D-0.3775 E-0.0724 H0.0000 L397.9500 R155.8342
M666 X-0.120 Y-0.334 Z0.484

I've tried asking Tevo on their Facebook support group and even MKS and all I got was "format the card" or "use a different slicer". I guess I know more about these machines than they do.
I really need your support since I've bought the machine to be able to print large parts on the whole bed. I'm using a 0.6mm Volcano now with 0.3mm layer height and I still have spots where the extruder eats through the filament and can't push it and spots where is too high to get adhesion.
The grid compensation in the latest edge seems not to have any effect at all, I've tried running the G31 I13 J155.0, M374 saves the grid, M375.1 show the grid as loaded but when printing I still have the same high an low spots...

I've also posted this in the delta bot group and got a lot of help from Arthur Wolf. It seems not to be Smoothie related.
After two months of measuring and squaring and rotating carriages on the towers and rods I've finally rotated the effector plate (which kinda has a "front face") and the pattern changed.
Before I've had low spots on the X and Y towers. I've rotated the effector 60 deg CW and the pattern rotated 60 deg CCW, now I have low spots at Y and Z towers, and the difference high/low decreased. It's almost printable I might say.
Each pair of rods has the same distance at top and bottom +/-0.1mm.

I'm still waiting for ideas.
Thanks.

[fr.aliexpress.com]

dimension are 1200*600

Join the tevo little monster group NOW
Edtit: sorry, did not see that you had been there

Edited 1 time(s). Last edit at 05/27/2017 07:15PM by SJN.

I've got a comparable sized custom kossel Xl, which calibrates to 0.006mm deviation. I haven't used a tevo LM but would quite like to, as I like the frame design a lot, it's simple and in theory the frame should be very capable.

Where they have cut corners as best I can see are the controller, the arms, the joints and using rollers not linear rails (this one really makes a big difference). How much play is there in the effector when holding in the homed position? How much do the arms flex? They're quite thin for their length carbon fibre or not. How similar in length are the arms? Magnetic joints really do offer zero backlash and if well made are all within 0.05mm length of one another.

Your microswitch should be accurate enough to get reliable probing.

If I bought one of these I'd change to linear rails, Haydn magnet joints, and duet wifi/ethernet.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Quote
DjDemonD
I've got a comparable sized custom kossel Xl, which calibrates to 0.006mm deviation. I haven't used a tevo LM but would quite like to, as I like the frame design a lot, it's simple and in theory the frame should be very capable.

Where they have cut corners as best I can see are the controller, the arms, the joints and using rollers not linear rails (this one really makes a big difference). How much play is there in the effector when holding in the homed position? How much do the arms flex? They're quite thin for their length carbon fibre or not. How similar in length are the arms? Magnetic joints really do offer zero backlash and if well made are all within 0.05mm length of one another.

Your microswitch should be accurate enough to get reliable probing.

If I bought one of these I'd change to linear rails, Haydn magnet joints, and duet wifi/ethernet.

so you will change almost everything lol
all ready to use delta are not perfect out of the box but this tevo has a very good design

@DjDemonD
Unfortunately the C beams (8040) do not allow centering a linear guide rail on the tower. I've contemplated the idea, but easier would be to use 2 rails/tower, but with all the aligning issues and costs I don't know if it's worth it.
The arms are 0.1mm equal in length, measured by placing them in a jig (2 screws in a long 2020 profile).
It's got something to do with the effector plate/hotend, I've rotated the effector plate with hotend and fans end the high/low pattern changed.
Joints have no play. they are tight but smooth. There is a youtube review measuring deflection using a micrometer. (https://www.youtube.com/watch?v=refBP9MVO_4)
It's something the probe doesn't detect as I get 15-20 micron errors reading across whole bed.
Maybe the carriage's rod mounts are not level with the bed. Something is definitely tilting the hotend.

@titeuf007
Indeed the LM is a great piece of machinery, built like a tank, good quality joints, didn't expect these woes with it.

Edited 1 time(s). Last edit at 05/28/2017 05:06AM by executivul.

If its effector tilt can smoothie compensate for tilt at each probing point? Is your micro switch sensor positioned directly below the nozzle or offset?

Is the spacing between the rods the same at carriages and effector for each rod pair?

Do you have another printer or can someone print you a new one?

Yes I'd change a lot so I'd probably just make a frame like the LM with top and bottom plates. You could drill the centre of the extrusion and add a single rail.

Edited 2 time(s). Last edit at 05/28/2017 06:11AM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions