Nearing 3D printer need for institutionalization.........

Turbojoe

Elite member
I know, I know, quit trying to obtain that perfect print but I'm just trying to get "closer". I spent an hour or so this morning with my Ender 3 printing a 14 x 3 mm test part with a 6 mm hole in the center and multiple times and inputting several M code changes using Pronterface I finally got it to 13.99 mm on X axis and 14.01 on Y axis. Z axis is right on the money and 3.00 mm hole diameter is right at 6 mm. Great! I thought I had it all worked out. Then I printed a 20 mm "CHEP" test cube that I sliced in CURA 4.6.2 just like I used for the 14 mm x 3 mm part. WTF!!! Now I get 19.84 mm on X axis and 19.89 mm on Y axis. Z axis is 19.98 mm. I created the same size in TinkerCad and got the same results so it wasn't an issue with the "CHEP" file.

What the heck am I doing wrong? Why the big change in final print size differential between two parts sliced at the very same time using the very same slicer? If I wasn't already nuts this one problem is going to push me over the edge and require a few years in the rubber loony bin wearing a strait jacket............ I've replaced all the belts and they're tight. I've replaced all the rollers too and they're adjusted for snug but very free movement. Glass bed is level. All extrusions are painstakingly set at 90 degree angles.

Is this a CURA issue or is there something else with the Ender 3 that I need to address? I'm going nutso here guys. Please help save me from the rubber room.....

Joe
 
You shouldn't try to chase dimensional accuracy by changing the steps/mm. The steps/mm for X, Y and Z should always be set to a calculated value depending on the per-step angle of your motor, the number of micro steps configured, the number of teeth on the pulley and the pitch of the belt (or the thread pitch of the Z-axis lead screws). Unless you're skipping steps or your belts have stretched (common with cheaper belts) this should give the exact number of steps per mm. If it is not then you need to take a look at the mechanics of the printer, specifically the belts. Gates 2GT belts are the gold standard - they are not inexpensive but they are much more flexible and have a lot less stretch than the alternatives, and last a long time: https://www.aliexpress.com/item/32853559638.htm . If you look for something cheaper, find a belt with glass fiber reinforcement that is flexible. Before I moved to the Gates belts I used some Hictop belts that were also pretty good but do not seem to be available any more.

The dimensions of the finished part will never be "perfect" because of the width of the extrusion. This is the least accurate part of the process, and unlike the steps/mm which have a linear effect depending on the size of the part, they have a constant effect. So if, for example, you print a 20x20x20 object and get 20.1x20.1x20.1 and you blow this up to 200x200x200 you wouldn't get 201x201x201 as a result as you would if the steps/mm were wrong, instead you'd get 200.1x200.1x200.1. So trying to adjust for an extrusion width issue by adjusting the steps/mm is always counterproductive. You should do a test where you print a 20x20 square, measure, then scale it up to 100x100 and measure. That should let you figure out whether you have a stretched belt situation or it's just an extrusion issue.
 

Turbojoe

Elite member
Many thanks for the in depth description! I've been wrongly assuming that being a CNC machine that if you enter a dimension it should print at that dimension. Obviously there are many factors that control the final output. My problem is that parts print SMALLER as they are enlarged though... That's what confuses me so.

To be clear my as new before any adjustment steps were 80.00 on both X and Y axis and 400.00 on Z axis. E steps were 100.00. Z axis is still 400.00. X is still 80.00 and Y is now 80.03. E steps are at 99.00.

I bought THIS kit from Amazon. It "says" that it's 2GT belt but there are no markings whatsoever on the belt. The kit does have mostly 5 star reviews. I did NOT replace the gears and idlers though. Maybe I'll tear it down next weekend and give it a good going over WITH the new gears and idlers. They all looked good is why I didn't replace them. Maybe one of them is still a bit wonky though....

Joe
 
Many thanks for the in depth description! I've been wrongly assuming that being a CNC machine that if you enter a dimension it should print at that dimension. Obviously there are many factors that control the final output.

Yeah. Because you can only extrude a line about the size of the nozzle opening (0.4mm) there are limits to how close you can get to "perfect". The slicer will attempt to offset the centerline of the outer layer such that the edge of the extrusion matches as closely as possible to the design value, but there will always be a bit of squish that will affect this. Tuning the extruder will help a lot, and you might want to take a look at reducing your flow rate in the slicer. I find that parts look much better if I set the flow rate to 95% of the theoretical value, but I always need to adjust things by a few tenths of a mm to get holes the right size.

Another test you can do is to print out a square 30x30 with a 10x10 hole in the middle. I expect you'll find that the outer square is slightly oversize and the hole is slightly undersize.

My problem is that parts print SMALLER as they are enlarged though... That's what confuses me so.

That's what I'd expect if you'd reduced the number of steps/mm in an attempt to adjust for a slightly oversize part, but as you went the other way I'm a little perplexed.

I bought THIS kit from Amazon. It "says" that it's 2GT belt but there are no markings whatsoever on the belt. The kit does have mostly 5 star reviews. I did NOT replace the gears and idlers though. Maybe I'll tear it down next weekend and give it a good going over WITH the new gears and idlers. They all looked good is why I didn't replace them. Maybe one of them is still a bit wonky though....

2GT is the original Gates belt designation. GT2 is the designation used by the generic clones, and quality varies widely. The belt in the picture looks good but it's hard to tell from a picture how good it really is. One thing I'd suggest is to throw away the springs that came with that kit. That's just asking for ringing in your prints. The belts should be tensioned with as little elasticity as possible - there are many designs for Ender 3 belt tensioners available on Thingiverse.

The belt should be taut but not too much - you want it to give some resistance when being pushed but not feel like a steel cable.
 

Turbojoe

Elite member
Awesome stuff! I'm sure others seeing your posts are picking up lots of useful tidbits too. I just did a fresh install of CURA 4.6.2 and set my flow rate to 95%. I'll print out that 30x30x10 and see what happens. I spent 30 years in the automotive industry. Gates likes to print their name on everything they sell. The belt material I bought says nothing. I may go on the hunt for genuine Gates belts just to eliminate possible problems.

Joe
 

Merv

Site Moderator
Staff member
You shouldn't try to chase dimensional accuracy by changing the steps/mm.
The dimensions of the finished part will never be "perfect" because of the width of the extrusion. This is the least accurate part of the process
So if you don't change the steps/mm, is there merit in chasing down the filament extrusion? Will changing the filament steps/mm get a more accurate squish?

What scale of accuracy can one expect to get with an Ender 3? 1mm, 0.1mm, 0.01mm or 0.001mm
 

Turbojoe

Elite member
I just printed a 30 x 30 x 3mm with 10mm hole in the center. X reads 29.84. Y reads 29.91. Z reads 2.96 and the hole reads pretty round for once at 9.98 - 10.01 mm.

Joe
 
What scale of accuracy can one expect to get with an Ender 3? 1mm, 0.1mm, 0.01mm or 0.001mm

I'd say that around 0.05mm is about as good as you can get with the FDM technology, and that depends a lot on the part. Printing interlocking parts the usual limit for a really well tuned printer is closer to 0.15mm - if you're getting 0.05mm gaps to work you're ahead of the game. Angus from the Maker's Muse Youtube channel has a tolerance gauge model that can be used to see how accurate your printer is and help tune it: https://www.makersmuse.com/clearance-and-tolerance-3d-printer-gauge . There are things like the fact that extrusion is not completely linear as the printer accelerates and decelerates. causing bulging at the corners. There's also the fact that filament diameter varies slightly across the spool - good filament varies less than 0.03mm but that's still 2% which can be noticeable on large flat surfaces if the light hits the layer lines just right.

Tuning the extruder steps/mm is always a good thing, however. I use the following procedure to tune my steps:

  1. Heat the hot end to your normal printing temperature.
  2. Make a mark on the filament as close to extruder 150mm away from where it enters the extruder. You want to make sure that the filament is as straight as possible.
  3. Tell the printer to extrude 100mm of filament.
  4. Measure the distance from your mark to the extruder and subtract that from 150.
  5. Set your new steps/mm for the extruder to (old_steps_mm * 100) / measured_distance. So if, for example, your old steps are 99 and you measured 101 mm extruded you'd set the new steps to (99 * 100) / 101 = 98 steps/mm.
This is the theoretical value - as I mentioned above, 5% or so under-extrusion tends to result in cleaner prints.[/QUOTE]
 

Merv

Site Moderator
Staff member
I've had my printer for 2 months now, I'm reasonably happy with the results. Like @Turbojoe, some things happen I don't understand. I print a lot of small parts, a single part will print great, print time say 4 minutes. If I try many of the same part parts at one time, say 10, print time 40 minutes. I have a lot of problems with them coming off the bed part way through the print. Any thoughts on why singles stick but multiples will not stick?
 
I have a lot of problems with them coming off the bed part way through the print. Any thoughts on why singles stick but multiples will not stick?

My guess is that the problem is insufficient part cooling causing filament to curl up. When the nozzle moves from one part to another it hits the curled filament and knocks the part loose. This is most likely to happen when the first solid or partly solid layer gets printed after sparse infill - often you end up with overhangs that can curl up.

Better part cooling can help avoid the curling - the default cooling setup on the Ender 3 is not very good. Another option that might help is to enable Z-hop in the slicer. This will make the nozzle rise a bit when travelling from one object to another so if it moves across an object it will avoid hitting the curled filament.