Having problems with Anycubic Kossel Linear Plus calibration
-
@veti
Current Z probe offset is Z15.791, do you mean i should change it to 15.891 ?
It was
5 : 100K thermistor - ATC Semitec 104GT-2/104NT-4-R025H42G (Used in ParCan & J-Head) (4.7k pullup)
in Marlin, but that was original Anycubic thermistor. Same one for the bed. -
that is significantly different.
semitec is B4725 C7.060000e-8
-
@veti Yes, you are right, looks like i really need to change B from 3950 to 4725 at least for the bed thermistor, but what about the hotend thermistor? It is different from the bed one. And should i use С parameter?
And about Z offset for the probe: should i change it to 15.891? -
@snakesp said in Having problems with Anycubic Kossel Linear Plus calibration:
And should i use С parameter?
yes without that the B parameter is wrong.
for the offset. what i do is do a first layer print. if i need to adjust it with microstepping to get a good first layer i update the offset with the microstepping used.
-
The firmware upgrade guide tells to update not only the firmware itself, but DuetWiFiServer and DuetWebControl.
WHATS_NEW.md for 2.04RC1 says, that compatible files are:
DuetWiFiServer 1.23
DuetWebControl 1.22.6 or 2.0.0-RC6 or 2.0.0-RC7But i cannot find those files. Can you point me where to get those?
And also a question: what is an appropriate order to upgrade firmware? Upload all 3 files to /sys through Settings-General-Upload files declining update offers and run M997 S0:1:2 after that, correct? Will it preserve my network settings or i'll need to connect using usb and terminal to restore those?
Nevermind. Found those:
WiFiServer in 2.0.3 release
WebControl -
Results so far:
updated firmware
changed bed thermistor to B4725 C7.060000e-8
recalibrated with 15.881 (recalibrated Z probe, at Z=0 no paper can be placed under the nozzle)
tried printing- 190C 50C babystep=0 - no sticking to bed at all
- 190C 50C babystep=-0.5 - partially sticking, corners not sticking
- 190C 50C babystep=-0.8 - smashed into the bed
- 200C 60C babystep=-0.5 - better sticking, some corners not sticking
- 200C 60C babystep=-0.6 - same
But when trying to print at center of the bed - no sticking until babystep=-0.12 at all
Looks like the calibration is still not consistent, although with S-1 it reports height deviations less than 0.07:
G32
G32 bed probe heights: 0.056 0.037 -0.068 -0.010 0.041 0.026 -0.060 -0.038 0.058 0.032 -0.059 -0.034 -0.000 0.021 0.029 -0.031, mean -0.000, deviation from mean 0.042Thoughts:
either calibration is still incorrect or temperature is still incorrect. -
can you check with 7 factor what the rod length is calculated to be.
and maybe try 8 factor.
https://duet3d.dozuki.com/Wiki/Calibrating_a_delta_printer -
@veti Will do a bit later.
For now i've got some more interesting findings. Near the morning i've got crazy and ended up with Hotend at 235C and bed at 80C. And suddenly PLA sticked to bed even without any babystep. The print quality was bad, but plastic sticks. So it looks like more likely the problem is not with calibration itself, but with thermistors showing fake temperatures.
Hotend thermistor was set to B3950 (as on Biqu site where i've bought it).
Bed was B4267 as in Marlin.
I took a multimeter and tried to measure the temperature with it. For bed it was about 5С under on multimeter till 60C on Duet and about 8-9C under from 60С till 85C on Duet.
But with hotend this is a different story. I tried to put a thermocouple from multimeter to the heating cube near the thermistor shaft and the results on multimeter were absurdly low:
173C when it was 235C on Duet.
So the problem for now is: how to make printer thermistors show actual temperature (or at least close to real)? -
@snakesp said in Having problems with Anycubic Kossel Linear Plus calibration:
Hotend thermistor was set to B3950 (as on Biqu site where i've bought it).
the chinese B3950 are a bit of a pain. that beta is only accurate for the range of 25-50C.
There seem to be different variation of these that behave differently at higher temperature ranges.
I had some trouble with those and in the end got a pt1000. -
@veti Don't have PT1000 right now, but looks like i'll end with it.
But i need to make prints by the end of this week and won't be able to get PT1000 by then.As for the current thermistor, i've managed to get almost acceptable results up to 100C so far by using boiling water and measuring thermistor resistance (B3988). Now thinking how to get resistance at 200-250C. Possibly can steal curling iron from my wife and use it to get needed temperature
-
@snakesp No luck, curling iron won't give steady temperature...
Yet looks like it is possible to print using temperature approximations with current beta: +5-10C on hotend and +7C on bed. Not best quality, but sufficient so far. Will get PT1000 or PT100 after a while.Now back to calibration. The outcoming print does not have accurate dimensions. Z is less for 0,05, which is good enough.
But X and Y are off +0.3 - +0.4 mm. Which is not good. Changing rod length should resolve this i assume, but the question is: should i use autocalibration with S7 or just leave the calibrations as they are and only change the rod length in config-override.g? -
@snakesp said in Having problems with Anycubic Kossel Linear Plus calibration:
PT1000 or PT100
PT100 will need a daughterboard.
should i use autocalibration with S7
i would do it once and see what the result is. you dont have to save it.
But X and Y are off +0.3 - +0.4 mm.
i have read that even though marlin uses 267 for rod length, some users say its 266.5
-
You said in your first post "no change to rods and carriages", and "changed the heating cube and nozzle". Do those changes affect Joint-to-Joint distance of the rods in any way? If not, stick to the length that worked for you in Marlin. With "worked" defined as "produced correct X & Y measurements".
-and-
Consider upgrading to Haydn Huntley magnetic-ball-end rods. They are the single biggest upgrade you can do to this printer; they will result in more noticeable quality than the change of controller board (I have one of these printers). And... Haydn's rods are marked for length. If you use his rods, always use that marked length.
Delta / Kossel printer fanatic
-
Oh, and I'm running an E3D hotend in a Smart Effector. Running E3D's thermistor with no problems, and I regularly print anywhere from 190 to 240.
-
@veti Yes, i know PT100 requres daughterboard, but it should be most accurate. I'm really tired of temperature sensors calibrations
PT1000 is cheaper, but i will need to change the heating cube at least because i haven't found PT1000 in 3 mm capsule, only in 4 mm.Did calibration with S7:
6:14:15 PMM666
Endstop adjustments X-0.72 Y1.04 Z-0.32, tilt X-0.03% Y-0.20%
6:14:07 PMM665
Diagonals 274.863:274.863:274.863, delta radius 136.095, homed height 287.654, bed radius 115.0, X 0.469°, Y -0.907°, Z 0.000°
6:13:58 PMG32
Calibrated 7 factors using 16 points, deviation before 0.132 after 0.024
Don't think that 274.863 for rod will do any good, too big difference with reality.
Stayed at 266.5 for now. X is 100.32 mm (for 100 mm in model), Y is 100.97 mm. Will try M579 to compensate. -
@danal In fact Marlin never gave really accurate dimensions. Y and Z were fine (less than 0.1 mm deviations), but X was always 1% more than it should be. This was with 267mm rod length. I assume that was due to inaccurate printer build. That's why a have rebuilt whole frame from scratch. And at the same time changed MB to Duet to move to 32 bit and 24 V for faster printing.
Now the X and Y dimesions errors are close, so looks like this can be compensated with M579.
As for Haydn's rods: too expensive right now for me. But i'm working on selfmade magball arms to test (with cylinder magnets in rod and steel balls on effector and carriages). -
Unless you've posted newer files that I missed...
Your config-override contains:
M665 L271.500 R135.224 H287.541 B115.0 X0.505 Y-0.998 Z0.000
@snakesp said in Having problems with Anycubic Kossel Linear Plus calibration:
@danal In fact Marlin never gave really accurate dimensions. Y and Z were fine (less than 0.1 mm deviations), but X was always 1% more than it should be. This was with 267mm rod length. I assume that was due to inaccurate printer build. That's why a have rebuilt whole frame from scratch. And at the same time changed MB to Duet to move to 32 bit and 24 V for faster printing.
Now the X and Y dimesions errors are close, so looks like this can be compensated with M579.Got it. Sounds good.
As for Haydn's rods: too expensive right now for me. But i'm working on selfmade magball arms to test (with cylinder magnets in rod and steel balls on effector and carriages).
Cool! Magnetic joints really do help this printer. Of course, measure them as best you can after you build them (maybe even take them to a local school, or maker space, or...).
-
@danal
config-override.g now has the following:
M665 L266.500:266.500:266.500 R133.832 H287.529 B115.0 X0.460 Y-0.913 Z0.000
M666 X-0.711 Y1.035 Z-0.324 A-0.03 B-0.20The main thing for rods is to build them at the same jig, so they are totally egual in length. But yes, length measures are important too.
-
@snakesp said in Having problems with Anycubic Kossel Linear Plus calibration:
because i haven't found PT1000 in 3 mm
-
@veti Thanks
