MK3S: recent prints are about 0.8 % smaller in X and Y than they should be, no change in Z
 

MK3S: recent prints are about 0.8 % smaller in X and Y than they should be, no change in Z  

Page 1 / 2
  RSS
stephen.bannasch
(@stephen-bannasch)
Active Member

Recent prints with my Mk3S printer are coming out about 0.8% smaller in X and Y dimensions. The Z dimension is correct. The problem occurs using both 0.15mm and 0.20mm quality layers. The modified PLA  (Ingeo 3D870 resin) is the same material I've been using for the last 3 kg of printing.

 

To make the problem simpler I created two 5.0 mm tall cylinders. One is 50.0 mm in diameter and the second is 49.7 mm in diameter. When printing using 0.15mm layers I make the wall thickness 1.7 mm. When printing with 0.20 mm layers I make the wall thickness 1.67 mm. These are based on the thin wall thickness recommendations in prusaslicer.

 

The stl models as rendered in prusaslicer are also 50.0 and about 49.7 mm in diameter (estimated measurements by aligning with the 50 mm grid and zooming way in).

 

Models like this printed accurately in the past (within about 0.1 mm).

 

A possible correlation though I can't figure out how it could be related: I recently had two prints in effect crash partially through. The prusa screen reported at the bottom: "MINTEMP BED fixed" after both failed prints. After pressing the button on the printer the screen then says: "Please restart".

 

After two failed prints I turned the printer on its side and checked the thermistor cable. The cable had too much slack in the corner of the hot bed. It appeared to be enough slack that it's possible it got hung up during print bed movement. After removing the slack I no longer have any issue with failed prints. I'm not certain when the issue with XY scaling started. Though I have earlier parts I printed that are accurate to within 0.1 mm. I just measured a part with a 40 mm outside diameter minus a friction fit allowance of 0.3 mm. On my cheap digital caliper it measures 39.8 mm diameter. That's basically what I intend it to be and it provides the correct friction fit when a 40.0 mm ID part is slid onto it.

 

The test cylinders are 5.0 mm tall and when printed they measure the same so there appears to be no issue in the Z axis.

 

I have the Extrusion Multiplier set to 1.01. I tried setting it to 1.05 and didn't see any difference in the diameter of the test cylinders.

 

The printer is using firmware 3.9.3-3556.

 

I'm mystified. I need this to be accurate to print the parts I am making.

 

I've attached an stl file for the 50 mm test cylinder with a pipe shell thickness of 1.67 mm which should print correctly with four lines using 0.20 Quality printing. On my printer after printing the 50 mm test cylinder the diameter is 49.7 mm. The ID is 46.4 which is approximately correct for a 49.7 OD and 2 * a 1.67 mm shell thickness.

 

Posted : 22/04/2021 6:54 am
stephen.bannasch
(@stephen-bannasch)
Active Member

Another clue. I just printed the precision cube. This is a vase mode 40 x 40 x 40 mm cube. The printed base measured 40.0 x 40.0 mm.

 

Now am wondering if selecting a shell/thin-wall thickness that matches the suggestions in prusaslicer causes an issue. Earlier prints that were accurate had thin-wall thickness values of 1.0, 1.5, and 2.0 mm -- measurements that didn't evenly match projected thicknesses when printing an integer number of lines.

 

 

 

 

Posted : 22/04/2021 8:42 am
Diem
 Diem
(@diem)
Honorable Member

The prusa screen reported at the bottom: "MINTEMP BED fixed" after both failed prints.

Has the temperature of your printing area changed recently?  Central heating changed with the season?  Air circulation changed?

The temperature and speed of cooling can have a noticable effect on print dimensions, usually we hear about it as warped prints.

Things may stabalise with the new season or if it's really important consider an enclosure.  You can use a large cardboard box as a temporary test enclosure if you are unsure.

Cheerio,

 

Posted : 22/04/2021 9:33 am
Ringarn67
(@ringarn67)
Active Member

@stephen-bannasch

I have the same problem....
Stock Prusa i3Mk3+ kit, Prusament Orange PLA, 0,20mm Speed.
Everything, except modified prime line (bobstro), set to standard 

Printed a pipe with OD 51, ID 41, height 14, and it came out 50.5, 40.65, 13.97 (Mitutoyo caliper)
Created a box in PS, 50 x 50mm height 4, came out 49.75 x 49.7 height 3.97

Same result with Prusament Black PETG

Did you find a solution?

This post was modified 2 weeks ago by Ringarn67
Posted : 27/04/2021 10:25 am
Diem
 Diem
(@diem)
Honorable Member

So better than 1% error on a small, hollow part and 0.5% on a solid?

Not bad on a printed part and it's in the nature of 3d printing that the error dimension stays about the same size so with scale the percentage improves.

You have the option of printing outside perimeters first which will help with some objects, not all.

And the height can only be a multiple of the layer height above the thickness of your first layer - the stated first layer is a nominal value but in practice it varies with each individual printer & steel sheet combination.

If you need absolute precision print oversize and machine to fit.

Cheerio,

Posted : 27/04/2021 11:46 am
Ringarn67 liked
Ringarn67
(@ringarn67)
Active Member

I am perfectly fine with the height, that is sooo close to perfect. Guess it comes down to first layer height.

I haven't done any parts that should fit to the "outer world" before, just parts that should fit to each other. Therefore I haven't noticed it.

Will try with outer perimeters first and see if it gets closer 🙂 

As long it is consistent it's easy to adjust scale in slicer 

Thanks

Posted : 27/04/2021 5:07 pm
Diem
 Diem
(@diem)
Honorable Member

@ringarn67

You always have to allow a little tolerance, good luck.

Cheerio

Posted : 28/04/2021 6:23 pm
Ringarn67 liked
Tim
 Tim
(@tim-2)
Illustrious Member

It's probably not a good idea to use percentage for checking accuracy.  After all, a 0.1 mm error on a 5 mm part is huge percentage, but rather meaningless. 

The XY calibration is based on embedded bits in the heater, and these have some tolerance for manufacturing. Z is mechanical threads turned by a stepper, so is generally quite accurate. I don't know what the heater manufacturer has for a spec, but +/-0.25mm seems to be common figures most folk report for a 100mm part. Once known, pretty sure PSlicer has a method to set the correction factor (or just scale accordingly), or maybe it was in Pronterface by changing a setting.

It is always wise to get more than one opinion...
Posted : 29/04/2021 10:51 am
Diem
 Diem
(@diem)
Honorable Member

@tim-2

It's probably not a good idea to use percentage for checking accuracy.

Youre right - but it *is* a good way to reassure an inexperienced user who is well on the way to good, practical printing but who is prematurely diverting effort into an area that will come good with experience ... Encourage them to get on with it and the problem fixes itself.

Cheerio,

Posted : 29/04/2021 4:17 pm
Ringarn67 and Tim liked
w88thpro04
(@w88thpro04)
New Member

go

Posted : 29/04/2021 4:18 pm
Ringarn67
(@ringarn67)
Active Member

It IS percent, 20mm is 19,9, 40 is 39,8, 60 is  59,6 and so on, haven´t tried anything bigger .
I have over 30 years in the cnc industry, milling and turning within a couple of hundreds of a millimeter.
Guess I have too much demand on the printer 😀 
And, I am a newbie and learning the machine and PS bit by bit, and this forum is an ocean of knowledge 👍 

 Thanks

Posted : 29/04/2021 10:09 pm
stephen.bannasch
(@stephen-bannasch)
Active Member

Precision is now better. 

 

Changes:

 

1. I re-did the whole XYZ calibration and used the wonderful 75 mm square @jeffjordan made to calibrate live-7 (changes as little as 0.005 make a difference).

 

2. The thermistor cable seems to have been problematic. Until I get a new one I fixed it by removing the cover and using electrical tape to hold it and the heat bed power cable in position. 

 

3. I also updated to firmware 3.10.0-RC2 but don't suspect that made a difference (not certain though).

 

In my case the parts I'm printing are alignment tools for holding other parts in place for gluing. I'm printing the alignment tools in plain PLA and the parts that are being aligned with a modified PLA using Ingeo 3D870 resin.

 

The parts of the alignment tools need a precise ID for cylindrical elements that need to slide together with a mild friction fit. They should stay stuck together when assembled but come apart with a mild-to-medium amount of force. 

This was working and then a bit later it wasn't. The problem with the thermistor cable occurred between when it was and wasn't working.

 

The parts being assembled come in different sizes and have a relatively precise ID. Here is a set of ID values for a random set of the parts being assembled that have a nominal ID of 45 mm: 44.89, 44.88, 45.03, 44.98, 44.93. Am measuring using a good digital caliper which is not a tool designed to measure ID so actual values might be a tiny big larger due to not measuring across the largest chord.

 

I also measured the OD of two assembly tools printed in the last couple of days that are designed for gluing parts with a 50 mm ID : 49.63, 49.65. In the model these have a 49.7 mm OD so this is actually pretty good.

 

I'm printing more of these alignment tools now and will collect a larger sample of data.

 

 

 

Posted : 29/04/2021 10:48 pm
Tim
 Tim
(@tim-2)
Illustrious Member

@stephen-bannasch

44.89, 44.88, 45.03, 44.98, 44.93

Expecting these to all be the same value is asking too much because the printer and materials used just isn't that good. The motor step values of the extruder or gear tooth effects can make changes larger that this.  Not to mention moisture content of the plastic being used, harmonic issues in the frame and chassis.  

Print a single line, and look at it microscopically and you may begin to grasp the issue. Here's something from way back when I was looking at print variation causes. 

Some of it is extruder, some of it is X and Y movement. But it is virtually impossible to get a smooth even line with the Mk3. When these features combine it is like a rogue wave in the ocean: random things can make fairly large effects.

 

This post was modified 2 weeks ago 4 times by Tim
It is always wise to get more than one opinion...
Posted : 29/04/2021 11:07 pm
Tim
 Tim
(@tim-2)
Illustrious Member

@ringarn67

lol - yeah - but with percent you need to specify "percent of  what" to make sense.  Percent of reading? Percent of full scale? Percent of error (are we talking monotonicity and linearity, too)? So a hard number, like an error of 0.1 mm is a lot easier to fix to the real world. People can then decide to convert to percent or something, if that helps.

 

It is always wise to get more than one opinion...
Posted : 29/04/2021 11:15 pm
stephen.bannasch
(@stephen-bannasch)
Active Member

@tim-2

Maybe you missed my first post in this thread. The test stl file I attached to my first post has a 50.0 mm OD in the cad model and when printed the diameter was 49.7 mm. The 0.8% refers to that difference. That test stl file is meant to represent a simple element of the aligning tool I'm making in regular PLA.

I also explained in the first post that I had been regularly getting much better precision on the equivalent of a 50 mm ring and there was a sudden change for the worse.

To get the friction fit I want there needs to be between 0.15 and 0.20 difference in radius between two cylinders that slide over each other. I am now getting that precision.

When measuring the ID of the parts to be glued (not the alignment tool parts) which have an ID of 45 mm in the  CAD model I measured : 44.89, 44.88, 45.03, 44.98, and 44.93. These are great data and I am quite happy with how the parts are printing now.

And the fact that I am measuring ID with a regular caliper and not a real bore gage means that some of the actual inner diameters for these parts are slightly larger (am only using my eye to get the jaws lined up directly along a chord tangent to the ID circle) which makes the actual ID of these parts even closer to my goal.

 

Posted : 29/04/2021 11:48 pm
stephen.bannasch
(@stephen-bannasch)
Active Member

Just finished a print of the alignment tool parts

 

In the CAD model this part of the alignment tool has a 49.7 mm OD. I took five OD measurements at different rotational orientations around each of the two parts I just printed. When measuring I avoided the layer seam which is set to "back" in the gcode for these parts.

Part 1: 49.66, 49.65, 49,64, 49.62, 49.64

Part 2: 49.72, 49.64, 49.77, 49.69, 49.64

 

[caption id="attachment_85294" align="alignnone" width="600"]regular PLA parts for aligning other parts to be glued 50 mm Alignment tool parts[/caption]

 

Going to try adding a short movie too:

 

[video width="640" height="360" m4v="https://forum.prusaprinters.org/wp-content/uploads/2021/04/alighment-tool-precise.m4v"][/video]

 

Posted : 30/04/2021 12:44 am
Diem
 Diem
(@diem)
Honorable Member

Those without an engineering background often have unrealistic expectations of 3D printing  - mirror surfaces, backlashless gear trains, fully finished products and even parts printed in mid air...

It's better to think of prints in the same vein as sandcastings or rough forgings: as parts ready for finishing whether by machining or just sanding and painting.

3D printers don't cut out *all* of the hard work.

Cheerio,

Posted : 30/04/2021 1:30 am
stephen.bannasch
(@stephen-bannasch)
Active Member

@diem

I'm quite confused about some of the responses to this thread.

 

Models were printing precisely and then they weren't.

 

My expectation was that something changed.

 

After fixing a number of things and adjusting settings I'm now getting OD and ID values that are within 0.1% of the dimensions in my CAD models when printing items with a diameter of about 50 mm. I think that's great and certainly results in parts that work well for my purposes.

 

Posted : 30/04/2021 2:02 am
Dan Rogers
(@dan-rogers)
Honorable Member

Is just math (well, actually simple arithmatics) so if you don't want to calculate accurately, you'll have to settle for close.  No problems here to solve folks, move on

 

Posted : 30/04/2021 2:43 am
stephen.bannasch
(@stephen-bannasch)
Active Member

@dan-rogers

What in the world are you talking about???

Posted : 30/04/2021 2:51 am
Page 1 / 2
Share:

Please Login or Register