Not to my eye, and watching the gauge change as it rotated seemed pretty smooth, more oval like, but I could be wrong.
yeah mine is more like a cam falling into place. pretty abrupt
Thanks for the info.
So new info to add to the discussion.
In cura I turned on the alternate wall direction option. thinking it would average out the issue, well it compounded it.
Now rather than only the one defined vertical shallow section i have two 180 deg apart in parallel with the Y-axis. Which sounds like backlash i must admit so ill have to re do some testing tomorrow or this weekend. is there fault with my assumption that if you move away from a zero 20+mm and back 20+mm to it and it stops at a <.01mm delta that you have low backlash? or could it be hiding the backlash some how?
After some more reading, it appears that we need to make sure the backlash is measured and compensation activated via the M425 G-code command before the axis steps are calculated and changed, as the backlash will throw off the step counts, and changing the steps first with cause the backlash to measure “weird.” Not sure if this is what you have done already, but if not, and you did it in some other sequence, you may want to consider starting over by resetting everything back to defaults using the M502 command.
Thanks for linking my findings.
Last year I needed to print some socks for ball bearings and hat problems with accuracy. The reason was: Retraction and linear advance at the seam. So try to disable retraction for movements smaller than 1 mm, keep the seam on the inside (hide) and always on the same position ( so that no retraction is necessary).
After each retraction the wall was slightly thicker.
That measurement that you are describing is repeatability, it is not backlash. Backlash is measured in a completely different manner, has 2 directions, and thus has 2 measurements. Based on the numbers that are talking about, I would guess that you backlash numbers are around 0.12mm +/-0.05mm. I would suggest to you do more research on backlash, and be aware that the spring in your dial gauge can affect your readings.
Best
That compensate wall overlaps sounded like something definitely worth trying, but it’s no longer available in cura5.
I’m going to try a couple of different sizes/bed positions, just to see if there’s any difference.
So after a nice Friday discussion with the old machinist on the floor.
Backlash is really simple. Its the adjustable wrench problem. Everything has backlash, but technically its the the rotational movement of the power axis that is needed to start movement of the sled after a change of direction.
Easy test:
With the power on and drivers engaged, do you have movement you can push by hand up and down the length of the axis. That’s backlash
Also my test of:
move +10mm
set dial zero
move +10mm
move -10mm
and read the measurement would measure axial backlash in the original + movement direction.
You will have to repeat the test with the measurement taking place on the opposite end and direction to get both measurements. for your total backlash.
example lets assume you have 1mm of positive dir backlash.
command +10mm (axis moves 10mm) but you don’t know this yet because you haven’t set the dial
set dial gauge to 0.00
command +10mm (gauge should read 10mm in our perfect world example)
command -10mm (gauge will now read 1mm in our world example) why +10 -(10mm command - 1mm backlash) = +1mm
Repeatability is more of a test like:
set dial to zero
move +10mm
move -20mm
move +10mm
read dial. that is your repeatability measurement.
Everyone follow?
Do you know what snapmaker default settings are? Ill have to read up on this a little. But makes sense if we are battling a adjustment factor that we didn’t know about in the machine / code.
Yes. I do. On the x axis. There’s just a slight click of movement. Possibly (by some fairly crude measuring) around.0.2mm. Can’t feel anything on the y, so not sure if that’s not a problem or just harder to tell because of the two modules.
I was just starting to think it must be hardware related, as tonight’s tests have all produced the exact same result. Those were:
9. No retraction
10. 40mm diameter with thicker walls
11. Z seam moved to the front (this proves its nothing to do with the seam, as suspected)
12. 3 in different positions
14. Just half a circle
So I will have to look up how to fix it. Cnc maker did mention above that snapmaker have some guides on it somewhere. I think if rather try a firmware fix first rather than a physical one.
Depending on the version of the linear rails, the 8mm lead screws are a 400step by default, and default backlash for all axes is set to 0.0 with compensation turned off. I am not sure what the steps for the 20mm lead screws are set to by default.
That would make a lot of sense that you have backlash or bearing slop in the x-axis since the flat spots are parallel to the y-axis.
Try this page 10-11
i have like no movement of the white nylon nut down the length of the axis. But i was under on the 34.29mm measurement so putting back together now see how it goes.
I see how they have an adjustment screw to lock in some back pressure for low back lash. But i also see how it would work nicely at first but wear in really easily.
If you adjusted the nut to give more pressure on the white nylon it would deform the nylon nut a little. BUT I AM NOT RECOMMENDING YOU DO THIS just saying what would happen if you adjusted the lock nut to pull that end of the nylon nut closer together would make the thread spacing of the two thread pitches closer together. Im assuming this is how it is adjusted at factory.
just curious try “alternate wall direction in cura”
See if you get the same results i did.
I’ll try the wall direction tonight if I get time.
That guide sounds more like it’s meant for vertical looseness, rather than along the rail, and I’m a little reluctant to dismantle stuff at the moment. I still have a good few months of warranty left, so I think I’ll try the compensation code (M425) and see if that improves anything first. If it doesn’t, or if it does but only for a couple of months, then I guess I’ll send the linear module back.
Just out of interest, according to this - https://snapmaker.github.io/Documentation/gcode/M425, the G425 command automatically measures backlash. Anyone know how exactly it does that, before I try it?
So to me it seamed to “average it out” and i had two flat spots. Which makes sense.
So I believe in conclusion: yes in conclusion…
and thanks to everyone who has / had input on this.
Precision is a never ending task or goal. Items wear and the environment changes. But with work and time you can get to a very reasonable precision with the snap maker system.
If you personally are okay with +/-0.25mm dimensional results then this write up / experience isn’t for you.
While I believe precision and absolute accuracy are a defined item. How and what is defined as precision is up to the user. Going from +/-0.4mm to +/-0.2mm to +/-0.1mm is not linier in the difficulty, the time or the effort it takes, its exponential.
Only you the user of your device can decide what is “good enough” personally I want to do some work prototyping on my machine so I need repeatability and accuracy to what has been designed at a relatively low value of error. For me my goal is <+/-0.1mm (rumor has it the machine mathematically is capable of +/-0.04mm)
At the beginning of this thread I was at .32mm. I am now very close to the <+/-0.1mm My circular error on the last test was .11mm and my last 20mm x-y cube was 19.98mm x 20.01mm with no flow or slicer modifications.
First thing to note. The machine is mass produced by people who are assembling them to get product out the door. Everything is built to a tolerance in the world. Just because something is in tolerance doesn’t mean it cant be better. To them if it passes the companies tolerance requirements its good enough.
Not for me.
I recommend following this thread lots of good info in it but I will summarize it here.
Nothing that i am writing down here is new info. I didn’t find some magic secret setting or device that helped me.
Absolutely Everything matters! all the test points all the calibrations they are so intertwined together with one effecting another. that all i can say is everything matters. Environment, initial set up, material, age of machine, speed you test at. It all matters. So all you can do is be very methodical and go over and over the calibrations and tests that everyone says to do.
First thing, get a decent reliable print going with a material of your choice. Simple PLA that you have a consistent first layer with Low to no elephants foot, and overall dimensional accuracy in the +/-0.25mm or better range. I think most snapmakers that are put together correctly will fall into this category. They may be some outliers and if you have one I’m sorry.
next go to https://teachingtechyt.github.io/calibration.html bookmark it, read it, watch the videos.
he put allot of work into it and it shows.
third invest in the level of tools for the level of precision you want. If you don’t have a decent caliper and dial gauge and base. You are not going to have a fun time.
So what matters? Make a plan, do it, test, make a plan, do it, test, repeat. Kinda of simple when said like that.
mechanically i think the big 4 are:
Physical setup
XYZ step calibration
extruder e-steps
and Backlash.
Testing, measuring, calculating, programing and testing again of these 4 things i think will get you the biggest bang for your time in getting your machine physically / mechanically setup the best.
Physical setup.
Hate to say it but you will need to check the rollers on each axis.
page 10-11 like i said everything matters. everything compounds any error. all results are inter connected.
after that is done for all 5 axis 
When you put the machine back together. is everything square? don’t be afraid to shim things to be square. a multi pack of brass shims in the .002"-.016" range is easy to find on amazon https://www.amazon.com/dp/B09WJ6Y9YL?psc=1&ref=ppx_yo2ov_dt_b_product_details
Once your machine is square and parallel as much as you can.
My test process was:
1: heat to constant enclosure temp.
2: Test XYZ stepping at the physical axis mounting plate. here the cosine error of your dial gauge is your #1 enemy.
3. test backlash with new steps programed by printing a backlash test.
I personally liked Printables
and Backlash calibration by marams - Thingiverse
all credit and thanks to the two modelers for the designs.
4. data collect all your xyz steps your math and your new e-steps. you may need to go back a step or two at any time.
5. print a 20mm dia round cylinder and a 20mm cal cube
6. document results.
7. use the teaching tech websites calculators and the marlin backlash calculator.
8. change firmware as needed M92 and M425 commands.
Go back to step 1 and repeat until you get the results you want. If you change XYZ e-steps then backlash may change so check it. If you change M425 Backlash M92 steps will change. Its a circle do one check, program other, check, do first one, check again. each time if you are doing everything correct you should get closer and closer to you goal.
After this you machine is only physically setup. now you need to get your profile better, and you may notice through out all of this that your prints have seamed to change. you will need to recalibrate you profile also.
things also to consider doing after your machine is physically / mechanically correct are
Linier advance or k-factor.
acceleration tuning. all though snapmakers are heavy and run slow it never hurts.
then and only then start looking at slicer flow adjustments, and slicer fixes.
So that’s what i have learned. hopefully this helps someone else out also.
@Spaced - that it some amount of effort you’ve gone through.
You gotta show us what kind of print deserves this.
P.S. I’m inspired to go through this now. Thanks to all the people who contributed to this thread. Getting my dial indicator now.
Well to be honest no specific print so far.
I mean I would like to print some of the nice print-in-place stuff and have decent results on the tolerance tests you can print.
But its more I want to be able to have confidence in when I design something that has to fit with other sourced cnc / sheet metal / or printed items. That I have confidence that my part will fit and act the same as others. I don’t want to have to design in a specific way for just my machine. I want to design what the part needs to be and have my printer print it accurately.
nweolu, beware precision is never ending…