Regarding attainable accuracy

I have had my printer for over 6 months now and I have tried everything to get accurate results but my prints are always off by .4mm to .6mm , both oversized and undersized. I am using a prusa i3 with a lite6 and have

- Calibrated steps per mm for x,y,z based on the ideal prusa caculator
- Calibrated steps per mm for extruder
- lowered extrusion multiplier to try and fit parts
- changed default extrusion width and perimeter to 0.4( nozzle size)
- measured filament diameter at multiple points and averaged it

yet I still cant get parts that will fit together. I have even taken it out and reassembled it at one point. I am at my wits end and really have no idea whats next to get accurate prints. Any help would be greatly appreciated .

Edited 1 time(s). Last edit at 04/26/2015 06:11AM by someoneonly.

If you're sure everything else is optimized (especially the extruder setting), try tweaking the steps/mm numbers for the axes and see if you get better results. If your x axis steps/mm is 240 and a 100mm line comes out at 100.6, you need to reduce the steps/mm. The error is 0.6%, so subtract 0.6% off the 240 steps/mm: 240-240x0.006=238.56 steps/mm

What's the source of the error? Maybe belt stretch, maybe inaccurate pulley diameters, maybe innacurate belt pitch The calculator assumes that belt pitch and pulley diameters/tooth counts are accurate, but when have you seen cheap parts accurately made?

My extruder steps per mm is 200 ( direct drive) after testing a few times , it extruded that exact amount . I even tried lowering it but results were almost identical and still off by around 0.5mm.

As for the steps/mm value for x,y,z I literally tried all the extremes setting extreme values to try and compensate but it was way off and the one prusa caculator and the manufacturer`s values were the closest and correct one , after counting each pulley and belt individually and caculating. I aslo tried offsetting the steps per mm a little bit every print for each axis but everytime I printed a new larger/smaller part , the values were completely off , such as Thomas Sanladerer mentioned in his calibration guide.

The printed parts arent top quality , and truth be told theres some slack in it as the screws wont really go in properly. I wanted to reprint the parts so that it can be tightened but at this point with this accuracy...

Edited 1 time(s). Last edit at 04/26/2015 07:03AM by someoneonly.

Do you trust the accuracy of your measurement tool?

Don't used curved objects to try to calibrate. Curves are made up of segments in the STL file and you can never be sure what you'll get, especially if you didn't generate the STL file from a CAD drawing yourself and had no control over the mesh size in the conversion process.

If the problem is consistent- i.e. always off by +0.Xmm- no matter what size object you print, the problem isn't the steps/mm in the axes but more likely over or under extrusion. If the error grows with the size of the object, then it's probably a steps/mm problem in the axes. You may have both problems at the same time. Printing a small object will tell you about over/under extrusion because error in the X/Y steps/mm will be small. Printing a large object will tell you more about the steps/mm error because the over/under extrusion error will be a smaller part of the total error.

When you calibrated the extruder did you also check solid infill? See [reprap.org] A little over extrusion will add up to a couple 1/10ths of a mm in the printed part measurements.

Finally, filament diameter varies. A small variation in the filament diameter will cause over or under extrusion which will cause some variation in the dimensions of the parts you're printing. Once everything is dialed-in, in order to get predictable/repeatable results you'll have to accurately measure the diameter of the filament every time you change spools. I usually measure in at least 20 places over several meters of filament, changing the orientation of the caliper on the filament as I go, then use the average value when printing. I also mark the average value on the spool so that I have the number the next time I print with it. I have been using volumetric slicing (set filament diameter to 1.12379 mm) and entering the actual filament diameter with the LCD/encoder panel at print time. That way one gcode file can be used for different filament diameters (change spool, etc.).

Im using an analog vernier caliper , not sure how hard I need to press but I might need to get a digital one if it doesnt give enough resolution.

I used around 7 different objects from thingiverse to calibrate but none were curves , all different form of calibration cubes.

The errors are pretty repeatble from small to large objects . interior are consistantly smaller by .6mm and exterior are consistantly larger by .6mm . They only vary abit from object to object.As for the steps per mm value being off , what use is altering it if that means I have to change it every print . If it gets to that point I might as well invest in something else.

I have read through the guide before and tried fine tuneing it . However on my first run it already gave desirable results . I aslo halved the esteps per mm value just to try and get results but perimeters were still too oversized , only difference was that the for and object with squares side by side, there was a gap in between.

I measured my filament 5 times , but I will try measuring more times though i doubt that will help .

I have tried setting all values to the extreme , setting 2.0mm for a 1.8mm filament , reducing steps for all motors but somehow the wall thickness seems to be the last thing affected...

Filament extrusion is based on the volume calculated from the diameter of the filament and the diameter of the extruder nozzle. A seemingly small change from 1.8mm dia (where do you get 1.8mm filament?) to 2 mm makes a big difference in the extrusion calculations because it makes a big difference in the volume of each linear mm of filament. That's about a 20% change in volume and that will have a very noticeable effect on the print.

I don't think you understood what I said about setting the steps/mm. It isn't something you'll have to change for every print. Once you find the correct setting, that's the one you'll use.

I think you're going to need a better measurement instrument. You need to be able to read 0.1 mm accurately. If the device can't read 1/100ths you won't be able to get accurate 1/10ths.

I tried changing affected settings to the extreme such as estep and filament diameter to the min and max respectively. The walls I got were still .6mm thicker but the infill and perimeters had gaps inbetween them.

I tried calibrating the steps per mm as you said but the moment i printed a different object there was a huge error. With the ideal set of values at least the error is repeatable.

I will get a digital caliper asap and report back to the thread . But my issue is that no matter how much my estep or filament diameter changes , the outer walls are still oversized but the insides are just less filled

Hi,

Dimensional accuracy I've learned so far.
Before you start the calibration of dimensional accuracy,
Bed level calibration has to be done first followed by
Feed rate calibration of the extruder
Temperature calibration of the filament
Flow rate calibration.

Set the target size that covers 70% to 80% of print area.
Dimensional calibration done by small target wouldn't guarantee consistent of accuracy over big objects.

Prepare an hollow test object, i.e. 100mmx10mmx10mm cube, to save filament.
Slicer setting:
Vertical perimeter: 2 to 3
Solid infill: Top 0, Bottom 0.
Infill: 0%

Formula of diagonal rod length
(Increasing diagonal rod length decreases object size.)
new_diagonal_rod = measured_length / target_length * old_diagonal_rod

Example:

Iteration 1

target_length = 100.00
measured_length = 100.35
old_diagonal_rod = 213.00
new_diagonal_rod = 213.74 (100.35 / 100.00 * 213.00)

Change the diagonal rod length to 213.74 in Configuration.h and upload firmware or M666 D213.74

Iteration 2

target_length = 100.00
measured_length = 99.90
old_diagonal_rod = 213.74
new_diagonal_rod = 213.53 (99.90 / 100.00 * 213.74)

Keep repeating above procedure until measured length of test object is within the accuracy range you set, i.e. 0.05 mm.

Calibration cubes I printed last time. It turned out square cubes waste filament. Long narrow cube is good test object.


Magnetic ball adapter I printed yesterday.
Printing accuracy was 0.0 mm to 0.1 mm.



M6, M8, M10 metric bolts and nuts




Edited 1 time(s). Last edit at 04/27/2015 09:27PM by janpenguin.

Hi , thanks for your suggestion . I have calibrated the above except flow and feedrate which i left at 100. I read about them but I didnt seem to suffer from any of the effects shown and there wasnt a clear writeup on it.

When I try fine calibrating using your method , any other object I print gets thrown off. The dimensions get off and I have to individually set the esteps/mm everytime I print something new which is unacceptable.Just to clarify , Im having issues with walls being excessively thick (not so much in single width walls but rather anything with infill).

"I have to individually set the esteps/mm everytime I print something new"

you have to change the firmware once you get the right number download arduino software here to access the firmware and change the esteps/mm settings then you will never have to change it again.

What I mean is that changing the esteps/mm to some value other than the ideal value derived from the prusa calculator causes inconsistency for different sized print. For example , calibrating the esteps/mm value to compensate for a larger print causes a medium size(or for that matter any other sized) print to be sized differently and completely off in my case.As much as I wish that fine tuning the esteps/mm could get me usable results it only throws off the dimensions each time I print something new. Its something Thomas Sanladerer mentioned in his calibration guide which I though I could ignore but it is a huge issue for my printer. Im not saying that the ideal values work for me but its the only value that give me a consistent ~0.5mm over-thickness of the wall. I use the firmware eeprom to set values in repitier host . Once im conviced the values are correct I reupload the code to my melzi board via arduino.

if you are designing your own parts you can do the ghetto method and design them around the problem by removing 0.5mm from the edges, sucks but should work.

Yeah but I print most parts off online. Not to mention I have tried making parts thinner by 1cm models that usually throw an error in both the slicer and the 3d modelling program as the wall gets too thin ._.

Edited 1 time(s). Last edit at 04/28/2015 07:50PM by someoneonly.

If you can read a vernier caliper I would trust it over a cheap digital. vernier can be read to 1/100'th of a mm, same as a $300 digital caliper.
and you can turn off 'check for thin walls' or some such in Slic3r to remove the error (don't know about whatever modeling program you use though)