Welcome! Log In Create A New Profile

Advanced

Kossel "semi-auto" calibration with(-out) RRFW

Posted by giampa975 
Kossel "semi-auto" calibration with(-out) RRFW
January 19, 2016 01:14PM
Dear All,
Let me first state that I have not designed a new calibration algorithm(!), but I used DC42's excellent tools.
I would just share my hobbyist experience in calibrating my second delta bot.
Calibrating a delta is always a tricky process, especially for newbies.
I thought that with a 32bit board (Alligator), RRFW and a probe it would have been super easy, but things were not so linear.

It is since November that I was trying to properly calibrate my mini Kossel 15-15 with T3DP plastic parts (R2) without success.
I started with a pibot IR probe; the measurements were consistent, but I was not very satisfied because the pibot is too sensitive to the light source: with natural light you get some height, with halogen another one and so on. The reflectivity of the bed also matters. This makes it impractical to recalibrate at every print of course.

Then, I decided to go for an inductive probe - fotek ps-5n - which I believe is light enough and I build and tried different beds: (a) a thin alu sheet between glass (4mm) and mirror and (b) a 0.8mm steel sheet on top of the 4mm glass.

While with the pibot I had to fine tune the z height suggested by the RRFW calibration to print fine, with the inductive probe I had my prints not sticking to the bed as the height was too high in some parts of the bed.

Measuring the distance between diagonal rods joints I found 0.2-0.3mm differences and also the following might apply:
+ diagonal rods length might be not identical, no matter if I made them pair by pair in a jig
+ play in joints (traxxas...)

Thus, for sure my effector is tilting significantly. Unfortunately, the scenario after adding the inductive probe became even worst as it introduced another source of error represented by the far from perfect planarity of the metallic layer.

I did not want to compensate the height with G30 Hxxx command.
I decided to simplify as much as I could and to go for a sort of "semi-auto" calibration.
I went back to the glass bed and I removed completely the probe and its voltage regulator.
I probed the 10 bed points with the nozzle tip by hand and put the heights in DC42's javascript calibration tool and -BOOM- the result was a perfect set of calibration parameters! It worked like a charm, no need to fine tune anything.

I thought that since the effector is tilting, then the hotend tip is nutating, but this effect is so small that it cannot practically affect the measurements.

@DC42, thank you very much for this tool! I believe it represents a tremendous value.
I have been very surprised of how simply I could overcome the errors in the spider construction by using the nozzle and piece of paper.
The probe in my case proved to be counter productive.

I understand that it is cool to recalibrate at each print with a probe, but if the machine has a solid build, it is not really required IMPO.
In addition, by associating each probe point to a macro button in RRFW, taking the measures takes no more than 5 minutes.


------------
+ Repstrap Schifanoia Delta - AMEGA/RAMPS 1.4, 4x4988 stepper drv., running RepetierFW 0.92 -
HE IeC 1.75 bowden, nozzle 0.4 on a cold bed (300mm)
+ MiniKossel - Alligator board running RRFirmware, HE E3D-lite 0.4 bowden on a cold bed (170mm)
Re: Kossel "semi-auto" calibration with(-out) RRFW
January 21, 2016 09:50AM
This sounds like a really effective method, and you've explained it well. Calibration is getting the bed relative to the moving effector square, probing the bed to determine various values and making either software adjustments or physical and software adjustments to the printer, then re-probing to confirm the changes. Whether you do this by using sensors or by just manually raising and lowering the effector and recording the heights is a matter of choice and the availability of reliable, reproducible and cost effective sensors.

I am firmly in favour of the FSR bed probing system for deltas. ([www.ultibots.com]). Probing both at and with the nozzle itself removes many irritating features of any offset probe such as reflectivity, ambient light, and the offset itself which means you can never fully probe the entire bed surface adequately. I have not encountered any issues with heat but then I mainly use my delta for PLA printing as it is not enclosed.

I would like to develop a similar system for my i3 printer also but it is more difficult as the bed would not be stable balanced on the FSR's in motion. I did buy from Aliexpress a hot end with an FSR built into it but I have not had chance to wire it up and try it yet. I am also concerned it does not seem to be rebuildable. There are scheme using electrical contact from nozzle to a metal bed but this seems almost certain to be problematic when the nozzle is always covered in insulating plastic.

Edited 2 time(s). Last edit at 01/21/2016 09:56AM by DjDemonD.
Re: Kossel "semi-auto" calibration with(-out) RRFW
January 21, 2016 03:18PM
Quote
DjDemonD
I am firmly in favour of the FSR bed probing system for deltas. ([www.ultibots.com]). Probing both at and with the nozzle itself removes many irritating features of any offset probe such as reflectivity, ambient light, and the offset itself which means you can never fully probe the entire bed surface adequately. I have not encountered any issues with heat but then I mainly use my delta for PLA printing as it is not enclosed.

My differential IR sensor is designed to eliminate the effect of ambient light and minimise the effect of varying reflectivity. However, any Z probe that is displaced horizontally from the nozzle is susceptible to build inaccuracies that cause the effector tilt to vary as it moves in the XY plane. This tilt causes the relative heights of the nozzle and the sensor to vary, causing the nozzle height at which the sensor triggers to vary too. One defence against this is to position the probe as close to the nozzle as possible. That is why I made my IR sensor small enough to fit below the heatsink of a genuine E3DV6 hot end. The Allen-key probe used on the original Mini Kossel is also very close to the nozzle. Other types of sensor such as inductive are larger and have to be positioned much farther away from the nozzle.

Using the nozzle itself as the probe avoids probe height errors caused by varying effectir tilt - although you need to minimise tilt anyway, because it causes other problems such as inaccurate XY scaling and long straight lines not being completely straight. However, FSRs under the bed also tend to give different trigger heights at different XY points. This is because the distribution of the force of the nozzle over the three FSRs varies with XY position. Probing at the centre of the bed is likely to need more force to trigger the FSRs than probing close to one of the FSRs.

Whatever the cause, in RepRapFirmware you can measure the trigger height at each probe point and correct for it in the bed.g file. Tedious, but it works.

An FSR mounted in the hot end mount should avoid the problem of trigger height varying with XY position. The remaining issue is that probing the bed with a hot nozzle may damage it, for example if the bed surface is BuildTak, or it is PEI and the nozzle is heated to ABS printing temperatures.

So I don't think we have a perfect Z probe yet.

Edited 2 time(s). Last edit at 01/21/2016 03:19PM by dc42.



Large delta printer [miscsolutions.wordpress.com], Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].
Re: Kossel "semi-auto" calibration with(-out) RRFW
January 21, 2016 04:34PM
Quote

Whatever the cause, in RepRapFirmware you can measure the trigger height at each probe point and correct for it in the bed.g file. Tedious, but it works.

Isn't this a bit of a chicken and egg problem though... before you calibrate you can't be sure that the effector is going to be moving flat as you go to the different probe points in order to determine what the correction should be in order to calibrate properly... It all gets a bit circular! Or is there some sort of cancelling out going on somewhere?


Cheers,
Robin.
Re: Kossel "semi-auto" calibration with(-out) RRFW
January 21, 2016 05:28PM
Changing the calibration parameters does not affect whether or how much the effector tilts. It does move the coordinates of the probe points, but by such a miniscule amount that the effect on the trigger height at each point is negligible. Therefore, if you are affected by trigger height that varies with probe position and you want to.correct for it, you need only measure the trigger height at each probe point once, until you either move the probe points or rebuild the machine in a way that affects the nozzle tilt.



Large delta printer [miscsolutions.wordpress.com], Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].
Re: Kossel "semi-auto" calibration with(-out) RRFW
January 22, 2016 02:08PM
[Lightbulb!] Ah, I see, we're just looking at the difference between the trigger point and when the head touches, so the starting Z height before the Z move down until trigger happens doesn't matter.
Sorry, only registered users may post in this forum.

Click here to login