OpenDACT(Delta Automatic Calibration Tool) - For Repetier

[RollieRowland]

I should also mention to those who are still pondering on getting the FSR kit, that it is extremely convenient. They were definitely worth getting.

I'm happy with it. it's a little thin near the Z tower, but that's not a deal breaker.

The layer thickness is excellent - I made no adjustments to the Z height after the calibration process was finished, so the first layer actually was .25mm.

I figured I'd make this a REAL test and print it on cleaned, totally bare glass. No glue, spray, PEI, etc. PLA @ 190C and the bed at 55C. \o/

Now to FSR-ify White MAX and Orange Menace. Poor Red Sonja and Orion Lannister will have to wait until I can get FSR kits for them.

g.

Glad to see you've had success with this! The version that I've posted should give some improvements, I have made a lot of minor improvements over the past two weeks.

I tried the normal manual calibration for ~5 iterations and i still have multiple millimeters in height difference on my measurement points still.

I guess ill try the expert variant tomorrow, but i really don't get why this isn't working for me. Could that mean a stepper is damaged?

I checked the tower angles and they are pretty much 90° each.

If you run the expert mode, the calculation will be specific to your printer, this will more than likely show improvements. Since your towers are nearly 90 degrees, then that is definitely not the issue.

If you could, write down the height-map, diagonal rod length, and horizontal radius for each iteration. This will help diagnose the issue.

[milp]

Milp, Blue MAX took *18* iterations. If you value your sanity, I'd get an FSR kit.

Hmm ok, well my sanity is long gone. Other than messing around with my rostock max i drive a 90s polish/italian car and an old russian motorcycle, I'm getting used to situations like these.

Just to be sure:

- When you say iterations, are you referring to measuring and calibrating, entering the values into eeprom, and keeping those values to do another run, or are you talking about starting from scratch every iteration?

- Is it normal that the values i measure don't necessarily get better with every iteration, but sometimes get worse? There were runs i measured a point at -0.8mm, the next iteration it was at +2.~mm

- What is an FSR kit and what's the cheapest way to get one? I pretty much spent my last cash on new delta arms.

[geneb]

This is the FSR kit I'm using: http://www.ultibots.com/fsr-kit/

I don't know much about how the html version of the calibration routine is used as I've never used it. If you look at the process log I posted above, you can see how the stand-alone application works.

g.

[milp]

This is the FSR kit I'm using: http://www.ultibots.com/fsr-kit/

I don't know much about how the html version of the calibration routine is used as I've never used it. If you look at the process log I posted above, you can see how the stand-alone application works.

g.

This looks really interesting, can you answer me a few questions about this kit please?

What is this doing, where do all the parts go? Does it interface with Rambo and work with the normal Rostock Max repetier firmware? Does it integrate with Mattercontrol and/or does it work with this great autocalibration tool?

Are there detailed instructions included and does it come with all the things needed to do the upgrade?

I might be able to afford it but i'd have to go out on a limb.

Too bad i can't seem to find a kit like this here in germany. Might be cheaper than paying 13$ for shipping and probably another 15€ for customs clearance.

[teoman]

[img]http://cdn.instructables.com/FIU/QI8E/G ... .LARGE.jpg[/img]Just saw this on instructables, and i thought that i would share.


http://www.instructables.com/id/Reading ... duino-USB/


This could be integrated on to the hotend, and the c# program could read the distances from another serial port. This would pretty much solve almost all fiddling with the machine in my opinion.

[RollieRowland]

Milp, Blue MAX took *18* iterations. If you value your sanity, I'd get an FSR kit.

Hmm ok, well my sanity is long gone. Other than messing around with my rostock max i drive a 90s polish/italian car and an old russian motorcycle, I'm getting used to situations like these.

Just to be sure:

- When you say iterations, are you referring to measuring and calibrating, entering the values into eeprom, and keeping those values to do another run, or are you talking about starting from scratch every iteration?

- Is it normal that the values i measure don't necessarily get better with every iteration, but sometimes get worse? There were runs i measured a point at -0.8mm, the next iteration it was at +2.~mm

- What is an FSR kit and what's the cheapest way to get one? I pretty much spent my last cash on new delta arms.

I count 1 iteration for every time the 7 points are measured on the bed. The values might spike at first, sometimes twice or even three times as much added in the opposite direction. It will usually only do this once and will begin to move closer to zero.

I could suggest to just get a simple momentary switch and use it as a z-probe. You don't need FSRs to use this calibration method, the only thing you will have to change is a setting in the program - default is FSR, change it to z-probe.

Also, if you do get an FSR kit, they will act exactly like an endstop switch. You will only need to make a small firmware modification and re-flash.

This could be integrated on to the hotend, and the c# program could read the distances from another serial port. This would pretty much solve almost all fiddling with the machine in my opinion.

Using this could work, it may be a fair bit of work. This would also require making some modifications in the firmware. Definitely a possibility for those wanting to use calipers!

[jason128]

Got a curly one for all you calibration experts out there... this ones driving me dingbats- I've been chasing scale & bed flatness around for 8 weeks now without success, I'm yet to really get a good larger print out of the machine, which is disappointing.

a bit of background- bog standard V2 kit & arms. all built square (i spent the morning with a engineers level (imagine a builders level that's 10x more sensitive), and the bed, & towers are all straight & parallel to each other within fractions of a mm.

If i run the fantastic HTML script i can get level +-.02mm all positions- which is great..... but it adjusts my diagonal arm length to 272.9mm (they are 269 by measure), si I can print to the edge of the bed no worries, but all my prints are 2-3% small- it a 200mm circle comes out around 194mm which is quite a bit.
I've tried leaving the diagonal arm at 269, but then cant be calibrated within around .3mm!

the numbers after calibration are:
horizontal radius 130.802
diagonal rod length 272.9
x tower endstop 55
Y tower endstop 50
Z tower endstop 50

alpha A 209.394
Alpha B 329.363
Alpha C 89.32

Z axis is spot on as far as accuracy- so its not a steps per mm thing

I;m running this version of the firmware

0.91 firmware for ROSTOCK MAX v2

2/16/2015 changed all instances of prog_char to char in ui.cpp to make compatible with arduino 1.6 gcc compiler per GeneB's input

Iv'e tried downloading the firmware again and resetting everything, but still no luck.

What are I missing here?

Cheers

Jason

[geneb]

Jason, I don't know if the html script fixes the x & y scaling issues.

I ran a scale test on Blue MAX this evening with a 25x25 cube. It measures out at 25.08 x 25.15. It was printed with a .35 nozzle @ .175mm layers. (PLA)

g.

[RollieRowland]

Got a curly one for all you calibration experts out there... this ones driving me dingbats- I've been chasing scale & bed flatness around for 8 weeks now without success, I'm yet to really get a good larger print out of the machine, which is disappointing.

a bit of background- bog standard V2 kit & arms. all built square (i spent the morning with a engineers level (imagine a builders level that's 10x more sensitive), and the bed, & towers are all straight & parallel to each other within fractions of a mm.

If i run the fantastic HTML script i can get level +-.02mm all positions- which is great..... but it adjusts my diagonal arm length to 272.9mm (they are 269 by measure), si I can print to the edge of the bed no worries, but all my prints are 2-3% small- it a 200mm circle comes out around 194mm which is quite a bit.
I've tried leaving the diagonal arm at 269, but then cant be calibrated within around .3mm!

the numbers after calibration are:
horizontal radius 130.802
diagonal rod length 272.9
x tower endstop 55
Y tower endstop 50
Z tower endstop 50

alpha A 209.394
Alpha B 329.363
Alpha C 89.32

Z axis is spot on as far as accuracy- so its not a steps per mm thing

I;m running this version of the firmware

0.91 firmware for ROSTOCK MAX v2

2/16/2015 changed all instances of prog_char to char in ui.cpp to make compatible with arduino 1.6 gcc compiler per GeneB's input

Iv'e tried downloading the firmware again and resetting everything, but still no luck.

What are I missing here?

Cheers

Jason

I am assuming the 2-3% smaller scale is consistent across the plate? This(diagonal rod length) is an issue that I have been fighting with.

Any adjustment in the diagonal rod length, in respect to the z dimension, changes the 3 points at the towers by approximately half that of the changes at the opposites of the towers. So, the correction method involves increasing the length of the diagonal rod to correct any offset between the points at the towers and the points opposite of the towers. This usually will work, unless there are some values that are not equivalent to there theoretical values - printer radius, carriage offset, end effector offset, smooth rod offset, or even the steps per mm. The main culprit here is most likely the steps per mm. When the end effector is moved from a point near the tower to the opposite side of the plate, there is a movement of the linear actuator equal to the length of the diagonal rod. If, for instance, the theoretical steps per millimeter is 80, when the value is truly 79.5, then

I am going to cut midway through that sentence. Just found another step necessary for the calibration. I have tested 3 different values for steps per mm, here is the data after each calibration (+/-0.01):
1. steps:79.75 diag:272.3 hrad:131.1 XYZ offsets: 402,412,138 iterations: 14
2. steps:80 diag:271.7 hrad:130.8 XYZ offsets: 455,467,171 iterations: 12
3. steps:80.25 diag:270.8 hrad:130.3 XYZ offsets: 455,466,177 iterations: 9

As you can see by this data, through increasing the steps per millimeter, the values - minus the XYZ offsets - gradually came closer to their theoretical values. This means that the calibration can correct the steps per millimeter using the theoretical diagonal rod length vs the tested diagonal rod length and some basic linear regression. Interesting...

I do not yet know how this may affect the scaling - whether it corrects it or makes it worse. I will check tomorrow.

[jason128]

Thanks for the reply- you really have this stuff sorted. Ill have play with the steps per mm and check again- I suspect this will go around in circles, the XY and rod length will be right, but the prints will be too tall.

I checked the steps per mm by measuring a steel bar accurately, then manually moved the print head up and zeroed the dial gauge in it- the printer's measured length & the measured length of the rod were exactly the same (within .05 over 150MM) so i don't think that's the culprit.

ill go through everything else and see if i can find anything out of spec. given how tight all the MDF on the rostock max is, i can't believe its possible to get anything vastly different between printers.

[milp]

Yesterday night i spent calibrating the bed flatness again.

Just to be sure, I'm going after multiple iterations here but which one is correct?

Should i do the "expert" version over and over again until it's flat or the normal one?

Because frankly, after one or two iterations nothing is even remotely flat anymore and my motors start behaving weirdly. I had distance differences of up to around 15-20mm occur. The servos jitter around for a second after reaching the endstops and move weirdly just before reaching each measurement point. This doesn't happen if i reset the changed eeprom values back to default.

Is there really hope for me at all with this script?

I could suggest to just get a simple momentary switch and use it as a z-probe. You don't need FSRs to use this calibration method, the only thing you will have to change is a setting in the program - default is FSR, change it to z-probe.

Where would i put that switch? And how would i connect it? When does it have to be triggered?

Thanks for helping

[RollieRowland]

Thanks for the reply- you really have this stuff sorted. Ill have play with the steps per mm and check again- I suspect this will go around in circles, the XY and rod length will be right, but the prints will be too tall.

I checked the steps per mm by measuring a steel bar accurately, then manually moved the print head up and zeroed the dial gauge in it- the printer's measured length & the measured length of the rod were exactly the same (within .05 over 150MM) so i don't think that's the culprit.

ill go through everything else and see if i can find anything out of spec. given how tight all the MDF on the rostock max is, i can't believe its possible to get anything vastly different between printers.

Actually, a mere difference of around 2-3 degrees in temperature is enough to offset my bed by up to 0.2mm + or -.

Since you are saying that changing the steps per millimeter will offset the z - which it will - I decided to do some tests:
To check if the z scaling was consistent, I moved my effector down to zero and measured the depth from the top of the effector plate to the glass. I then raised two increments and measured. Data:
Steps per MM: 80
1. end effector at Z:0 effector plate depth:54.09
2. end effector at Z:50 effector plate depth:103.86
3. end effector at Z:90 effector plate depth:144.31

Steps Per MM: 80.25
1. end effector at Z:0 effector plate depth:53.48
2. end effector at Z:50 effector plate depth:103.23
3. end effector at Z:90 effector plate depth:143.33

Steps Per MM: 80.375
1. end effector at Z:2 effector plate depth:54.65 (Hit the plate so I sent it to Z:2)
2. end effector at Z:52 effector plate depth:104.75
3. end effector at Z:92 effector plate depth:144.73

80 steps:
Difference from 0 - 50(50): 49.77mm
Difference from 50 - 90(40): 40.45mm
Difference from 0 - 90(90): 90.22mm

80.25 steps:
Difference from 0 - 50(50): 49.75mm
Difference from 50 - 90(40): 40.1mm
Difference from 0 - 90(90): 89.85mm

80.375 steps:
Difference from 0 - 50(50): 50.1mm
Difference from 50 - 90(40): 39.98mm
Difference from 0 - 90(90): 90.08mm

Results(theoretical - actual):
80 Steps: -0.23, 0.45, 0.22
80.25 Steps: -0.25, 0.1, -0.15
80.375 Steps: 0.1, -0.02, 0.08

From what I could tell yesterday, my theoretically perfect steps per mm is around 80.3-80.4, which allows my diagonal rod to be set at the correct value. Given the data, the correction in respect to the bed will fix my Z scaling, not offset it.

Yesterday night i spent calibrating the bed flatness again.

Just to be sure, I'm going after multiple iterations here but which one is correct?

Should i do the "expert" version over and over again until it's flat or the normal one?

Because frankly, after one or two iterations nothing is even remotely flat anymore and my motors start behaving weirdly. I had distance differences of up to around 15-20mm occur. The servos jitter around for a second after reaching the endstops and move weirdly just before reaching each measurement point. This doesn't happen if i reset the changed eeprom values back to default.

Is there really hope for me at all with this script?

I could suggest to just get a simple momentary switch and use it as a z-probe. You don't need FSRs to use this calibration method, the only thing you will have to change is a setting in the program - default is FSR, change it to z-probe.

Where would i put that switch? And how would i connect it? When does it have to be triggered?

Thanks for helping

To run the expert mode, what you need to do is first, probe the bed normally and then enter those values into the normal X-ZOpp tower heights. Also, enter all the values you can pull from your EEPROM - don't enter defaults. Once done, open up the expert panel. Leave the percentages and horizontal radius displacement blank. You will now need to do some testing. For the first test - diagonal rod percentages - increase (in your eeprom) the diagonal rod by 1mm, then remeasure the bed and enter the height values into the boxes corresponding to the diagonal rod. Repeat this for the horizontal radius displacement, and XYZ offsets. There is more information available when you hover over the test in question. Once all of these text boxes are filled, then you can press calibrate. And after the first test, all you have to do is continue the calibration as if you weren't running the expert mode - only change the tower heights.

[RollieRowland]

I've just tested with the steps at 80.75:

Steps Per MM: 80.75
1. end effector at Z:2 effector plate depth:55.02
2. end effector at Z:52 effector plate depth:105.12
3. end effector at Z:92 effector plate depth:145.07

Difference from 0 - 50(50): 50.1mm
Difference from 50 - 90(40): 39.95mm
Difference from 0 - 90(90): 90.05mm

Results:
80.75 Steps: 0.1, -0.05, 0.05 Offset from theoretical: 0.05 Percent error: 0.0556%
Diagonal Rod length: 269.055 Offset from theoretical: 0.055 Percent error: 0.0204%
Horizontal Radius: 129.512 Offset from theoretical: 0.012 Percent error: 0.0093%
Measurement from X0 Y0 to X0 Y90: 90.04mm Offset from theoretical: 0.04 Percent error: 0.0444%

[bvandiepenbos]

I've just tested with the steps at 80.75:

Steps Per MM: 80.75
1. end effector at Z:2 effector plate depth:55.02
2. end effector at Z:52 effector plate depth:105.12
3. end effector at Z:92 effector plate depth:145.07

Difference from 0 - 50(50): 50.1mm
Difference from 50 - 90(40): 39.95mm
Difference from 0 - 90(90): 90.05mm

Results:
80.75 Steps: 0.1, -0.05, 0.05 Offset from theoretical: 0.05 Percent error: 0.0556%
Diagonal Rod length: 269.055 Offset from theoretical: 0.055 Percent error: 0.0204%
Horizontal Radius: 129.512 Offset from theoretical: 0.012 Percent error: 0.0093%
Measurement from X0 Y0 to X0 Y90: 90.04mm Offset from theoretical: 0.04 Percent error: 0.0444%

This is all very interesting data.
I was wondering if you could make a option check box to NOT change the arm length when calibration is calculated? This may help analyze the data or adjust some printers easier. For example, I know for sure all my CF arms are precisely 269 but the HTML script says to make it 271.5 ...that can't be right. I would like script to leave arm length alone and try to solve the calibration by adjusting other values.

I do think it is possible that the actual steps/mm varies from the theoretical value. Belt tooth and pulley size is not necessarily precise. Belt tension will effect it also.

[milp]

To run the expert mode, what you need to do is first, probe the bed normally and then enter those values into the normal X-ZOpp tower heights. Also, enter all the values you can pull from your EEPROM - don't enter defaults. Once done, open up the expert panel. Leave the percentages and horizontal radius displacement blank. You will now need to do some testing. For the first test - diagonal rod percentages - increase (in your eeprom) the diagonal rod by 1mm, then remeasure the bed and enter the height values into the boxes corresponding to the diagonal rod. Repeat this for the horizontal radius displacement, and XYZ offsets. There is more information available when you hover over the test in question. Once all of these text boxes are filled, then you can press calibrate. And after the first test, all you have to do is continue the calibration as if you weren't running the expert mode - only change the tower heights.

Ok, thanks! So i run expert mode once and then just continue going through the normal X-ZOpp tower heights until i get a good result?

[RollieRowland]

I've just tested with the steps at 80.75:

Steps Per MM: 80.75
1. end effector at Z:2 effector plate depth:55.02
2. end effector at Z:52 effector plate depth:105.12
3. end effector at Z:92 effector plate depth:145.07

Difference from 0 - 50(50): 50.1mm
Difference from 50 - 90(40): 39.95mm
Difference from 0 - 90(90): 90.05mm

Results:
80.75 Steps: 0.1, -0.05, 0.05 Offset from theoretical: 0.05 Percent error: 0.0556%
Diagonal Rod length: 269.055 Offset from theoretical: 0.055 Percent error: 0.0204%
Horizontal Radius: 129.512 Offset from theoretical: 0.012 Percent error: 0.0093%
Measurement from X0 Y0 to X0 Y90: 90.04mm Offset from theoretical: 0.04 Percent error: 0.0444%

This is all very interesting data.
I was wondering if you could make a option check box to NOT change the arm length when calibration is calculated? This may help analyze the data or adjust some printers easier. For example, I know for sure all my CF arms are precisely 269 but the HTML script says to make it 271.5 ...that can't be right. I would like script to leave arm length alone and try to solve the calibration by adjusting other values.

I do think it is possible that the actual steps/mm varies from the theoretical value. Belt tooth and pulley size is not necessarily precise. Belt tension will effect it also.

Well what I plan on doing is to not change the values for the diagonal rod, since it is static. But using it as a way to adjust the steps per millimeter. If I removed the diagonal rod calibration in its current state then the build would be offset and the scaling would be disastrous.

I will be adding a step to the calibration to set the steps properly. Once I have a version, I will test it, then upload it.

With the current steps per millimeter, 80.75, I am printing a tube 1 layer thick with a radius of around 133 mm. So far it is great, except for having the bed temp to high. I will post some pictures in a bit.


And milp, yes, you are correct!

Edit:

Here is a picture of the cylinder, it is one perimeter layer thick - 0.4mm.

File size: 252.45mm
Actual size = 252.413mm

[mhackney]

I am concerned about XY bed jiggle if you have the pucks sized to freely move in Z to trigger FSR's correctly.

Yes, indeed. Building a three-point floating support system that has no perceptible slop yet doesn't bind is quite a trick. If you get it right on the first try, you must be living right. It's the key to a good FSR installation, though.

Hey gents, it isn't as difficult as you might think. Any "slop" can be removed by slightly repositioning the base/cylinder part of the mount. It is quite easy to get a smooth Z axis movement (which is over something lime <.01mm) while having no perceptible slop in X and Y. And consider, there is virtually no force in X Y from the filament bead being laid down. On a milling machine, there is a lot of cutting force so things must be much more rigid.

[mhackney]

RollieRowland,
I'm *finally* done with my annual fly reel building workshop in the Catskills and can spend time on some other projects again. Send me a link to the latest sources and I'll compile things for Mac users to try out.

cheers,
Michael

[geneb]

mhackney, here's his github repo: https://github.com/RollieRowland/Delta- ... ion-Source

I spent some time over the weekend adding data validation to many of the input fields as well as changing the user interface to be a bit more friendly. RollieRowland was kind enough to merge my patches. \o/

g.

[mhackney]

Great, thanks Gene.

[RollieRowland]

RollieRowland,
I'm *finally* done with my annual fly reel building workshop in the Catskills and can spend time on some other projects again. Send me a link to the latest sources and I'll compile things for Mac users to try out.

cheers,
Michael

I've just committed another update. It includes the untested steps per millimeter calibration and the changes Gene has made. I am fairly sure that there are still a couple of checks that I need to do to complete the steps per millimeter calibration but it is able to be tested at least.

[626Pilot]

Where would i put that switch? And how would i connect it? When does it have to be triggered?

You can use my mount for that if you want: http://www.thingiverse.com/thing:1000936 - the microswitch itself will be very, very cheap.

For example, I know for sure all my CF arms are precisely 269 but the HTML script says to make it 271.5 ...that can't be right. I would like script to leave arm length alone and try to solve the calibration by adjusting other values.

My calibration routine for Smoothie comes up with similar results. It produced an arm length of 270.883 the last time I ran it. If I don't let it adjust the arm length, it comes up with a worse calibration. The firmware increased the length from 269.0 to 270.883 simply because that happened to produce less average Z error across the bed.

As far as I can tell, two factors are at play here. One, the joints are not zero-tolerance - they will have some slop. Maybe only a fraction of a millimeter, but still measurable. Two, the calibration routine (Rollie's or mine) will do better when given more kinematic variables to simulate. Neither of us is simulating any sort of joint slop. He is doing some analysis on tower lean, but I don't know that this is correctable in Repetier. I don't mess with tower lean at all (although I want to).

The per-tower steps/mm adjustment sounds like an interesting avenue for further research. I think someone suggested that ages ago, but I didn't think it would pan out. Since there is now some experimental data supporting that exploration, maybe it's time. This is a good example of cross-pollenation between open source projects.

One more thing. I did have per-arm length adjustment as an option months ago, but it never seemed to yield useful results. Even large changes in the arm length would produce only minor changes in accuracy.

[jason128]

This may be a silly question? but what should the correct PRINTER RADIUS be. Reading through repetier- its set as 198.25mm.
From the the carrage offset being set to 38.4, it appears they are using the CL of the extrusion as the datum, which is fine.- all mt carrage offsets read 38.4 within .1mm
However opening the lazercut files, and drawing a circle, the radius of a circle through the CL of all the towers comes out as 200.4? this seems like a big difference to me.


messing with steps per mm...

RollieRowland - since you started playing with them, i went back and rechecked all 3 towers individually, are all accurate to .1 of a mm over 200mm of movement- so messing with this will definitely scale up the Z dimensions for me.


the other thing i have realised, is the perfectly calibrated bed i thought i had i do not... as the diagonal rod length is so far out, i'm getting the 'saddle effect' people refer to- there's a distinct thin point about 2/3ds out between the towers, so thin the nozzle bottoms out and cant print. - but at the perimeter, its almost perfect, i suspect this a result of the diagonal arm length being calibrated out at 272 rather than the actual 269mm it really is.

I'm wondering if the incorrect printer radius in the firmware is for a reason, or a mistake that is not helping the calibration processes at all?

[bvandiepenbos]

For example, I know for sure all my CF arms are precisely 269 but the HTML script says to make it 271.5 ...that can't be right. I would like script to leave arm length alone and try to solve the calibration by adjusting other values.

My calibration routine for Smoothie comes up with similar results. It produced an arm length of 270.883 the last time I ran it. If I don't let it adjust the arm length, it comes up with a worse calibration. The firmware increased the length from 269.0 to 270.883 simply because that happened to produce less average Z error across the bed.

As far as I can tell, two factors are at play here. One, the joints are not zero-tolerance - they will have some slop. Maybe only a fraction of a millimeter, but still measurable. Two, the calibration routine (Rollie's or mine) will do better when given more kinematic variables to simulate. Neither of us is simulating any sort of joint slop. [/quote]

The Traxxas joints are a spherical joint, with the "ZERO LASH STRAPS" pulling the arm pairs against each other (against the spherical shape)... there is no joint slop.

[RollieRowland]

My calibration routine for Smoothie comes up with similar results. It produced an arm length of 270.883 the last time I ran it. If I don't let it adjust the arm length, it comes up with a worse calibration. The firmware increased the length from 269.0 to 270.883 simply because that happened to produce less average Z error across the bed.

As far as I can tell, two factors are at play here. One, the joints are not zero-tolerance - they will have some slop. Maybe only a fraction of a millimeter, but still measurable. Two, the calibration routine (Rollie's or mine) will do better when given more kinematic variables to simulate. Neither of us is simulating any sort of joint slop. He is doing some analysis on tower lean, but I don't know that this is correctable in Repetier. I don't mess with tower lean at all (although I want to).

The per-tower steps/mm adjustment sounds like an interesting avenue for further research. I think someone suggested that ages ago, but I didn't think it would pan out. Since there is now some experimental data supporting that exploration, maybe it's time. This is a good example of cross-pollenation between open source projects.

One more thing. I did have per-arm length adjustment as an option months ago, but it never seemed to yield useful results. Even large changes in the arm length would produce only minor changes in accuracy.

Yes, the tower lean is not able to be corrected in Repetier. Is this a modifiable value in Smoothieware?

The steps per millimeter isn't specific to each tower, sadly. It isn't possible to do so with the firmware. If it were specific to the tower, to use my current method to find the actual value, it would require the diagonal rod length to also be specific to each tower. Even so, it is possible to at least partially correct the error that is created from having this value significantly off.

...Just read the last line of your post. I am assuming the diagonal rods/stepsPerMM are able to be modified per tower then? If this is the case, it is theoretically possible to find the perfect steps per millimeter for each tower.

This may be a silly question? but what should the correct PRINTER RADIUS be. Reading through repetier- its set as 198.25mm.
From the the carrage offset being set to 38.4, it appears they are using the CL of the extrusion as the datum, which is fine.- all mt carrage offsets read 38.4 within .1mm
However opening the lazercut files, and drawing a circle, the radius of a circle through the CL of all the towers comes out as 200.4? this seems like a big difference to me.


messing with steps per mm...

RollieRowland - since you started playing with them, i went back and rechecked all 3 towers individually, are all accurate to .1 of a mm over 200mm of movement- so messing with this will definitely scale up the Z dimensions for me.


the other thing i have realised, is the perfectly calibrated bed i thought i had i do not... as the diagonal rod length is so far out, i'm getting the 'saddle effect' people refer to- there's a distinct thin point about 2/3ds out between the towers, so thin the nozzle bottoms out and cant print. - but at the perimeter, its almost perfect, i suspect this a result of the diagonal arm length being calibrated out at 272 rather than the actual 269mm it really is.

I'm wondering if the incorrect printer radius in the firmware is for a reason, or a mistake that is not helping the calibration processes at all?

If that is the case then try changing your printer radius to 200.4, this should help!

Is your scaling consistent on the opposite of towers as well? This is where the effect will be the greatest (farthest point from the top of the linear actuator). Either way, a change of 0.1mm over 200mm is enough to say that this would not correct your printer. I do not know where your issue may be coming from in this case. It may be your plate radius, or it may be something entirely different that is offsetting your diagonal rod. I should also ask what your horizontal radius is set at? And if the calibration code finished properly with your machine, then the bed should not have low or high points.

By two thirds of the way out, do you mean from the center of the printer to the edge it dips in the middle or?

The Traxxas joints are a spherical joint, with the "ZERO LASH STRAPS" pulling the arm pairs against each other (against the spherical shape)... there is no joint slop.

I can agree here, I have the Traxxas joints as well, and there is no slop.