How-to: Dimensional Accuracy Calibration for Delta Printers

[IMBoring25]

Diameter is a measurement. Multiplier is already a fudge factor. Not sure why you'd change a value you measure in lieu of a value that's already kind of arbitrary.

Also, extrusion multiplier is linear with the amount of filament extruded. Changing diameter alters the amount of filament extruded as a square function.

[626Pilot]

So I've been wondering this for a long time, but I'm always too lazy to ask...
Isn't changing extrusion multiplier the same as changing filament diameter? don't they modify the same thing?

In theory, you could set the filament diameter to 1.75 and only mess with the extrusion multiplier. I do both because I figure having the filament diameter properly measured will give me a starting point that's as close to accurate as possible. If you try only modifying the flow multiplier, please let us know how well it works.

if the overall squareness is off a bit it could just be squishing down the nozzle further on those sides causing a wider wall thickness.

That would count for the first layer, and maybe second one a bit. But the instructions specifically talk about NOT measuring those first layers. Will see, have very little time this and the next week.

My instinct is that the mashing effect is more pronounced in lower layers because (especially with PLA) you have to mash it a little to get it to put down the first layer without anything coming loose from the surface. The first few layers can therefore be flanged out somewhat. On the higher layers, if the nozzle is rotating marginally with XY position, it may wind up lower in one place than it is in another, and the same thing happens - the plastic has to flare out in the XY plane because there's no other place for it to go, and you get a thicker wall.

[CodonExe]

Very Nice Work! You helped me with the how 2's of calibrating my BlockHead(my printers name) you also helped me better understand why. Thank you very much 626Pilot!

[dajay23D]

My Delta build is very accurate. All angles are nearly perfect. Effector is parallel to the printbed. Arms are all within 0,05mm Tolerance.

But why are all prints in Y-Direction 0,1mm to small, but in X-Direction nearly perfect?
No matter what I do, Y-Direction is 0,1mm smaller than X-Direction, independant from size of the Testparts?!

Any tips to solve this little issue?

Aside from my system, how are you calibrating the printer? GeneB's 4-point method, auto-calibration firmware, or something else?

what's geneb's 4-point method? I should check mine again after this thread, but last time i check, oddly, my boxes were fine. Length, width, and extrusion rate was also good. My biggest problem I have is the location of holes. They always seem to be off! so I also end up trying to slot a hole or enlarging it to make things fit.

[626Pilot]

That's in the Rostock manual, somewhere in the calibration system.

[thingismith]

Anyone have any other solutions for fixing the side-to-side/back-to-front size ratios? I've followed the directions...my side-to-side is at 50.00, back-to-front is 50.18. Switched out my .4 nozzle with another .4, no difference. I doubt tilt is the problem, the difference in the sides of my single-wall box is only .01mm.

I just installed my accelerometer, but I was having this issue pre-installation as well. I was hoping the upgrade would fix the issue.

Is there another setting to consider?

[626Pilot]

Consider that a 50x50mm solid box has 122 times more XY area than one side of a single-walled 50x50mm box. I would expect to see more positional errors pile up given the increase in area. In Delta printers, there is generally more positional error between two points the further they are from each other, and the map of that error often has a weird, non-uniform shape.

It could be effector tilt, or some slight misalignment. The towers may not be exactly 120 degrees apart. They may not all be exactly the same radius from dead center. One or more towers could have a tiny amount of lean. The delta radius or endstop settings might be off by a small amount.

Unless you can find and correct the mechanical issue, your best option is to get the calibration so that the length and width of the 50x50mm solid box is as close to 50mm as possible. Maybe 49.9x50.1 or something like that.

Also, make sure your slicer is set to draw the outer loop first. Slicers usually do the outer loops from the inside out in order to avoid zits. Going from outside to inside on the loops can help with dimensional accuracy, as the outer loop of filament doesn't have anything pushing against it from the inside when it's laid down. I would also make sure the loops aren't being drawn any faster than 30mm/sec.

[dc42]

Those using Duet electronics can use the M579 command to scale the axes, see https://duet3d.com/wiki/G-code#M579:_Sc ... esian_axes. But as it's already been pointed, geometrical errors on deltas generally have non-uniform results, so use this facility with caution.

[Jimustanguitar]

Any chance that the error aligns with your hotend fan? I'd seen the draft from my E3D shrink ABS that's 'downwind' from it. You might be seeing the results of something similar.

[thingismith]

Any chance that the error aligns with your hotend fan? I'd seen the draft from my E3D shrink ABS that's 'downwind' from it. You might be seeing the results of something similar.

I assume you're not talking about the squirrel fans for PLA (which I don't have attached atm), I suppose it is possible, my fan is pointing away from the Z tower.

Those using Duet electronics can use the M579 command to scale the axes, see https://duet3d.com/wiki/G-code#M579:_Sc ... esian_axes. But as it's already been pointed, geometrical errors on deltas generally have non-uniform results, so use this facility with caution.

But this code is only for Reprapfirmware and not the Repetier that SeeMeCnc instructs us to use?

I ended up using this procedure: print an Dodecagon (12-sides) with a 30mm inner diag, altering the arm length according to the length of the sides along the Z tower, running the G29(after it cooled down), and checking the EEPROM for results. Eventually at arm length 288.6, the endstop offsets for the Y and Z towers resulted in 0 (X was 47), and the sides of my dodecagon on the X and Y (Ztower) axis were equal. The sides along the X and Y towers were off by .1mm, but by that time I was willing to accept it.

[gajtguy]

Sorry for being a complete moron, but can someone validate my "lingo" conversion to cura 15.04.6 settings?

- Layer height: 0.1 - 0.2mm is fine, the shorter the better. Same in Cura, LAYER HEIGHT
- Extrusion width: Your nozzle width plus a little extra. Is this SHELL THICKNESS, or NOZZLE SIZE?
- Extrusion multiplier: Start with a value between 0.9 and 1.0. Seems like FILAMENT FLOW % is the right choice here.
- Loops: 1. This is what is really throwing me off. I have no idea where this option is in Cura.
- Skin thickness: 0. Is this the BOTTOM/TOP THICKNESS? Cause 'skin' goes all the way around doesn't it?
- Infill: None. Pretty sure this is FILL DENSITY.
- Speed: 20mm/sec outer loop, 30-40 for everything else. PRINT SPEED is the 'everything else' value and on the Advanced tab the OUTER SHELL speed is the same as 'outer loop' speed?

Thanks for the help. I'm trying to dial this sucker in now that I've got the new firmware uploaded.

[geneb]

Layer Height == Layer Height

Shell thickness is basically the extruded plastic width for a single perimeter. For example, if you've got a 0.4mm nozzle, you can set your shell thickness to automatic (if available) or 0.45mm if not. Nozzle size is going to be the actual diameter of the nozzle.

Loops may refer to the number of times it loops around the part to prime the hot end. Does it include a distance from the part?

Skin thickness likely covers the whole model.

Infill == Fill Density

The outer loop and outer shell speeds likely refer to different paths.

g.

[Harblar]

Quick question.

I've got my printer pretty well, to the best of my current ability, configured so that the nozzle is moving flat and the only errors in first layer height are likely due to bed warp when heated. I want to start getting the printer calibrated for x/y accuracy, but noted that the guide says to recalibrate every time you make a change to the diagonal rod length. I've spent probably 15-20 hours this week just getting it calibrated to this point and I'm a bit gunshy of messing with it too much more.

I've got a V2 that I've recently upgraded to the new carriages and ball cup arms and added the accelerometer board for the auto calibration. After a lot of meticulous measuring and adjusting of the physical machine (and countless methods of performing the autocal, manually calibrating, or any combination there of) the only way I've gotten a good single layer test print has been by the following procedure:

1.) Run geneb's V3 Probing gcode on a cold hotend/bed at least twice.
2.) Perform the online Delta wizard calibration (also on a cold hotend/bed) https://www.seemecnc.com/pages/delta-calibration-wizard
(This took me from an error of .26 down to .04)
3.) Heated the bed and hotend to full printing temps (90 and 230 respectively) and then ran the V3 probing code again.

This got my first layer thickness, when set at .3mm, to vary no more than from .2mm to .35mm along the outer perimeter of the bed. (I'd love to see that down to +/- .05mm, but I can live with that, providing it doesn't get worse on additional layers.)

So, my question is, do the V3 probing code and Delta Calibration wizard automatically change the Diagonal rod length setting? If so, is there any way to prevent them from doing so?

If it does change the diagonal rod setting and can't be disabled is my only recourse to do the calibration manually? I've not done the 50mm and 100mm test prints yet, so Maybe everything is perfect already and I'm worrying about nothing! (hahaha... try typing that with a straight face! )


I've done a lot of reading over the last few days and all I know for certain is that I wish I would've known about and gone the Duet Wifi/ FSR route instead! ( I still might... Just irks me that I've already thrown a couple of C notes at this thing in the last two months and I'm not really any closer to having it where I want it to be)

[geneb]

The Octoprint version of the auto-cal doesn't mess with the arm length.

Also, once you've used the calibration wizard, do NOT run the g29 (or any other script) process after that - it'll smash the calibration that was just performed.

g.

[Harblar]

Also, once you've used the calibration wizard, do NOT run the g29 (or any other script) process after that - it'll smash the calibration that was just performed.

g.

True enough, though, I did do a test print after running the wizard and the results weren't overly great. Running the follow up autocal at temp got me pretty close. Damned if I know why.

Anyway, I did the calibration prints and think I've got my extrusion multiplier pretty close. Also noticed it was really flattening out the first layer (near 1mm wide on the single wall box). I took .05mm off from my max Z height and I think I've got that really close now. +/- .05mm on the layer height (inside the delta triangle at least.) and just slightly wider than the additional layers. Just out of curiosity, when setting Nozzle size in the firmware is it best to use the stated size of the nozzle (.4mm in my case) or the measured width of the filament actually extruded from the nozzle (which I think is around .46 in my case)? Right now I have .4 plugged into the slicer and while my printed extrusion width is really close to .45mm I'm noticing some pillowing, zits, and gaps when attempting to print a solid layer indicating that I might be under-extruding.

As for the 50mm and 100mm boxes, it's pretty obvious my diagonal arm is off. 50mm= about 49.1mm and the 100mm= about 98.4mm. I don't currently have Octoprint installed since I print direct from a dedicated laptop 6 inches away. I see now there is a way to install the print server onto a windows pc, so I'll go give that a try in a bit here.

Thanks for the help! hopefully I'll be able to report back with some good results.

[Harblar]

Got Octoprint Installed. Working good for the calibration, but so far I'm not a huge fan and have been going back to matter control for the test prints. Once I'm fully calibrated and running some big prints I'll revisit it and try to get it working with the camera on my laptop so I can monitor things when I'm working. On the auto calibration end of things I'm still not 100% certain it's working completely right. I run G29 a couple of times and then run the plugin a couple times till I get little to no change. The first few tries had me at around .09 to .1 error. After printing several tests and changing the diagonal rod as needed, The plugins reported error seems to be getting worse. I was up to .17 and this last time it settled right in on .2.

The weird thing, after printing a single layer, measuring, and then tweaking the z height I've been noticing that things aren't that far off... even on the 100mm box. After a few slicing tweaks and a couple mechanical tweaks (my x tower belt got a little loose and I was getting lots of zits on my layers. Tightened the belt and, after some reading, ran a sanding pad over my nozzle to smooth it up a bit.) I'm getting some of the nicest single layers I've ever printed! That being said should I be worried about the auto cal bed error being up to .2mm. I'd love to get it down to .04mm again, but have no idea where to start there.

I've been using Cura 2.5 for prepping the gcode (since the built in cura in matter control leaves out half the available options). All I can say is... Nice! Much improved since the last time I used it a couple of years ago. I think I've got most of my print settings pretty well straightened out. No detectable stringing, zitting is gone, and (if I get a layer fan setup right) I should be able to get some decent bridging.

Ok, back on to the actual topic of this thread. After several print measure tweak cal repeats, I've managed to get my 50 mm and 100mm to an average side of 49.82mm and 99.82mm respectively. That points to under extruding since both are off by the same amount, though .18mm seems like an awful lot of error to make up considering the single wall test was averaging out to about .45 - .47mm. I currently have my nozzle diameter set to .46mm (since that's what I've measured the actual extruded filament at), so should I be looking for a single wall thickness of .5 - .52? And even if I bump it up to say .52mm, wouldn't that only get me halfway to where I need to be?

Finally, I'll just throw in that I am also getting uneven sides as well. The 50mm is pretty good and only varies by .1mm from longest to shortest side. The 100mm box varies from right at 100mm down to 99.65mm. It's kind of like the lower right corner of the box has been slightly tweaked in towards the center of the box. My guess is that this translates to something mechanical (as well as the recent increase in bed error seen the calibration plugin). Any Ideas on what to check and how to properly correct it?

[626Pilot]

If you're using a 0.5mm SeeMe nozzle, I recommend setting extrusion width to 0.55. You may measure that it's less than that, but that could be due to contraction from cooling. It's usually not a good idea to use any value even slightly less than the nozzle diameter. If you do, you may see weird artifacts as the nozzle opening isn't fully charged with melted plastic. Going a little over is fine (and a good idea, IMO) because it can always handle a little more volume.

[Harblar]

I'm currently running an e3d V6 hotend with a .4mm nozzle, or at least I was till last night (see Here:http://forum.seemecnc.com/viewtopic.php?f=62&t=11936). Up until that major muck up on my part I was noticing that with the .4mm nozzle extruded filament had a measured width of about .46-.47mm. Not sure if this was the actual nozzle width or the size the filament was expanding to after leaving the nozzle. Is there any good chart to show recommended extrusion width versus nozzle size? If a .5mm nozzle works best around .55mm width, what would be best for a .4mm nozzle or a .25mm nozzle? Is it a proportional difference that can be easily calculated for non standard nozzle sizes, or is it just trial and error?

Anyway, beyond that, I thought I had my accuracy dialed in very close, but when I remeasured the part yesterday morning it had shrunk up .1-.15mm from the night before and then it dawned on me, "Duh! ABS shrinks you dummy!" On that note I decided to try out a roll of the Maker Geeks Raptor PLA I had sitting around (since PLA has minimal shrinkage). That did the trick. The 50mm box was within .05mm of being perfect and still was several hours later. I think I found a great easy to print alternative to ABS! I did a write up on it in the filament section.

Once I get the replacement parts for my Hotend I'll run through the process again just verify I'm right on. That only leaves me with some mechanical irregularities. Is there a good guide somewhere on how to precisely measure, tweak, and adjust the mechanical structure for optimum results that doesn't require $10,000 worth of measuring tools or completely rebuilding it as a Metal Max?

[geneb]

Look up "die swell".

Also understand that perfection is the bitter enemy of good enough.

g.

[Harblar]

Looked up Die Swell and it makes sense. Think I'm just going to split the difference between measured and stated nozzle width and call it a day.

On your second point, I know all too well! lol

Too that end, what is typically considered "good enough" when it comes to the auto cal? Right now I'm stuck at .2mm error across the bed. Sure, that doesn't sound like much, but at first layers and nozzle diameters at/or below .25mm it could be a big deal. Aside from bed warpage (I'm waiting on a Geckotek buildplate and magnetic base/heatspreader which I'm hoping will be somewhat flatter than my current glass) I'm currently stumped as to what else I can do to get things more accurate. Towers are trued up to the best of my ability. Tower heights are as even as I can get. Top of the melamine base is as flat and level and as close to perpendicular to all the towers as I can make it. Belts are all tightened up niceley and to similar degrees (don't know for sure since I don't have a tension guage. Would a fish scale work and About how many lbs should one look for?) Anything else I could be missing?

[626Pilot]

GeckoTek won't be flatter than glass. There are tiny undulations on the surface. I got sick of how easily the plates get scratched and have taken to clamping the glass disc to the GeckoTek aluminum base, as it works very well as a heat spreader.

[Harblar]

We'll see how I like it. I get what your saying, but understand that my glass isn't what I would call flat and seems to warp pretty good at a few spots. putting down a .25mm layer it will touch at some spots and be as much as .35mm thick at others. I'm hoping the geckotek will be a bit flatter than that. The other day I had a hotend muckup (damn cable to the acceleromter board came off mid print. Grrrr...) After creatively applying a rubberband to hold it in place I redid the calibration plugin. it came out pretty nice, so I let it go only to come back 5 minutes layer and see the hotend scraping tape off the glass on a part of the bed that is usually low! Uggh... recalibrated several more times, did a few manual checks and printed a test circle. I then marked the high and low spots on the bed. I then applied clips to all the 3 main high points and began sticking a few layers of tape under the glass at the low points. I printed several test circles and added tape as needed till most spots showed high. I then bumped up the z height and managed to get a somewhat decent circle. I reran the calibration plugin, tweaked the z height, and printed a final circle that came out pretty nice. It's the flattest I've managed to get the bed so far, but it's still off from where I'd like it to be. Oh well. I'll report back when the geckotek arrives. (if nothing else I'm looking forward to easier print removal with it. Glue Sticks and ABS never gave me a problem, but I can't believe how tight this Raptor PLA is sticking to the tape on a cold bed!)

[hectori]

14 april 2018 I found: https://www.thingiverse.com/thing:1274733/

Hi Harblar,
When I start calibrating my rostock with the box 50x50x10
I noticed some distortions in my printer.

Hex50x50x2HOK1.stl

(1.05 MiB) Downloaded 522 times

I would like you to use a hexagon better than the cube so you can play with.
repetier
#define DELTA_RADIUS_CORRECTION_A -0.9
#define DELTA_RADIUS_CORRECTION_B 0
#define DELTA_RADIUS_CORRECTION_C 0
#define DELTA_DIAGONAL_CORRECTION_A 0
#define DELTA_DIAGONAL_CORRECTION_B 0
#define DELTA_DIAGONAL_CORRECTION_C 0
Marlin
#define DELTA_TOWER_ANGLE_TRIM { 0.0, 0.0, 0.0 } //{ 0.25, -0.17, 0.0 } //? X-0.22 Y-0.32 Z0.00 get these from auto calibrate
// delta radius and diaginal rod adjustments measured in mm
#define DELTA_RADIUS_TRIM_TOWER {0, 0, 0} //{0.15, 0, -0.13}
#define DELTA_DIAGONAL_ROD_TRIM_TOWER {0, 0, 0} //{0, -0.5, 0} this change the X to 49.61 49.61 I need to go other way

this it´s my case

[626Pilot]

I get what your saying, but understand that my glass isn't what I would call flat and seems to warp pretty good at a few spots.

I don't think glass can be warped unless it was handled very roughly during production, jostled badly before it was done cooling. During production, it's in a molten state on top of a metal with a lower melting point (mercury, I think). Gravity ensures the liquid metal is perfectly flat, and thus the bottom of the glass, resting on that metal, is also perfectly flat. For the same reason (gravity), the top of the glass should be perfectly flat as well. As it cools, the liquid metal solidifies first, and the glass solidifies last. It should be flawless when it's taken off the metal layer.

My guess is that the effector is not tracking parallel to the glass because of whatever difference exists between how the printer really is, and what your firmware thinks it is. For example, the towers are all supposed to be exactly 120 degrees apart. If one of them is 119.5 degrees from one neighbor and 120.5 degrees from another, it will throw off the XYZ positioning. You won't easily notice the error in XY on the first layer, but you will definitely notice the error in Z because it will be too high in one place and too low in another. That's why more advanced firmware like Smoothie and RepRapFirmware has, for example, per-tower positioning, and heuristic calibration systems to try to guess what the true values are.

[geneb]

Float glass is made using tin I think. It's typically made in huge sheets (4x8 I think).

g.