How much backlash should I expect to live with on a UM1

[aviphysics 9]

According to my careful measurements, I currently have about +/- 0.25 mm of backlash on the X axis and about +/- 0.1 mm on the y axis. I plan to tighten the belts some, but don't want to over do it.

So basically, how much backlash should I expect to live with?

 

[Dim3nsioneer 557]

[...] I plan to tighten the belts some, but don't want to over do it.

[...]

 

Smart! :wink:

How did you measure it?

 

[aviphysics 9]

Smart! :wink:

How did you measure it?

 

I used calipers and measured the distance between the edge of the top opening and the top of the hot end. It is a very nice set of certified calipers that were given to me as a graduation gift, so I am inclined to trust their stated precision and accuracy.

Process:

0) After homing, lower the bead out of the way move the print head to 90 mm along the x-axis.

1) Run the head out to 100 mm from left to right and measured the distance.

2) sent the head out to 110 mm, then back to 100 mm and measured again.

3) repeated about 5 times

4) repeat process around 50 mm

5) repeat steps 0 through 4 with y axis

From the results I computed backlash as well as steps/mm.

I was happy to at least find that the measurements were very repeatable (all within +/- 0.02 mm ); the backlash measurements at 50 mm and 100 mm gave very close to the same results; and the computed steps per/mm were essentially the same for both the y and x axis.

 

[aviphysics 9]

edit: By "hot end" I mean the edge of the wood square at the top of the assembly.

 

[gr5 2,176]

Very cool test.

I haven't measured my backlash (aka play) but .1mm sounds about right to me - but again - I don't know.

Usually most of the backlash is in the short belt which is also the easier belt to tighten. If you tighten the long belts too much you will increase friction so much that you will start to increase backlash again (bizarre but true!). So as you are tightening the belts, push the head around to make sure the friction doesn't increase drastically.

There is a nice video that explains how tight to make the long belts based on pitch made by Erik himself (one of the founders):

 

 

[gr5 2,176]

By the way - another way to measure this is to print a wall where the lower part of the wall is approached from one side of the bed and the upper half of the wall is approached from the other side of the bed. Very similar to your test but you end up with a part where you can use calipers.

One way to do this is to print two square columns where one of the columns stops half way up and hope cura slices it in such a way that when it switches from the short column to the tall column it starts on the edge closest to the other column. You can force cura to do this by messing with thickness of skin (2 versus 3 passes or .8mm versus 1.2mm thick).

 

[aviphysics 9]

I saw that video in the guide. I was not able to get my pitch as high as his, so I think I should probably add one of the belt tensioners. It does seem like he moves his hot end more easily then mine, but maybe his is more broken in.

I tried tightening the short belts, but it didn't make any difference.

BTW, when I say +/- 0.1, I mean that one of the measurements read something like 100.1 and the other was something like 99.9. So a total difference of 0.2 mm. Would be nice if there was a setting for backlash compensation in Cura. At least for handling this last little bit. I really want to get my precision to better than +/- 0.5 mm.

 

[aviphysics 9]

I mean 0.05 mm

 

[gr5 2,176]

>I mean 0.05 mm

Wow. I can't even measure that with my calipers.

I think you can get pretty close to .05mm but it will probably require tweaking the model. The biggest error at some point is going to be shrinking related (unless you buy PLA45). Because the shrinking is happening as you print it is complicated and changes the shape of the print in a complicated manner. Similarly with play (backlash). It matters which direction the head approaches from.

In general, to get .05mm accuracy you will need to print everything twice. The first time you print it "close" to spec. Then you get out your calipers, measure all the errors and adjust in CAD and then print again. That's really the only way I know to do this. After a while you get good at predicting how much an error will be and your first print will be very very close.

For example my parts are usually a little thicker on the bottom 3 or so layers. Some people call that the "elephants foot". You can measure this and add a chamfer (correct word?) in CAD and end up with a perfectly vertical part with no elephants foot.

Other manufacturing processes have the same issues. For example injection molding has worse problems. For example if you want a 90 degree corner you have to make the mold have a slightly different angle and then when it shrinks it ends up being 90 degrees. Also the entire mold has to be about .3% larger than final part for injection molded PLA.

 

[Dim3nsioneer 557]

I used calipers and measured the distance between the edge of the top opening and the top of the hot end. It is a very nice set of certified calipers that were given to me as a graduation gift, so I am inclined to trust their stated precision and accuracy.

Process:

0) After homing, lower the bead out of the way move the print head to 90 mm along the x-axis.

1) Run the head out to 100 mm from left to right and measured the distance.

2) sent the head out to 110 mm, then back to 100 mm and measured again.

3) repeated about 5 times

4) repeat process around 50 mm

5) repeat steps 0 through 4 with y axis

From the results I computed backlash as well as steps/mm.

I was happy to at least find that the measurements were very repeatable (all within +/- 0.02 mm ); the backlash measurements at 50 mm and 100 mm gave very close to the same results; and the computed steps per/mm were essentially the same for both the y and x axis.

 

I hope you will not get me wrong, I don't mean to offend you.

But if you really can measure a distance of around 10cm with a (calibrated?) set of calipers between two wodden parts at an repeatability of 0.02mm, then I really have to congratulate you. :wink:

You should be aware that with repeated measurements you may reduce the statistical error but the systematic error of your measurement device stays the same. If you compare two measurements then it is even worse: no matter if you add or subtract values, the errors of the two measurements are correlated. So you have to add the individual errors which means in this case to double the error.

Realistic accuracy of a caliper measurement lies somewhere in the range of the backlash you actually measured. To my opinion, most people overestimate the accuracy of calipers. Having a number of digits on the display doesn't mean the tool measures with the accuracy of the last digit.

To get a real feeling about how large the backlash is, one has to compare it with the spacing of the MXL belts wich is to my knowledge about 2.03mm. So you have a backlash of around ten percent of the spacing. I think this is a bit larger than normal even when considering you have the play of two belts in series. You may try to tighten the long belts a bit. BUT:

 

... If you tighten the long belts too much you will increase friction so much that you will start to increase backlash again (bizarre but true!). So as you are tightening the belts, push the head around to make sure the friction doesn't increase drastically.

...

 

@gr5: Where do you know that from...? :rolleyes:

 

[aviphysics 9]

@Dim3

Whether or not you believe the measurement of my calipers, which I have used in enough situations to feel reasonable confident and were certified to a certain degree of precision, they are certainly able to measure a difference of 0.5 mm without a problem. I can almost measure that with a ruler. At any rate, Starrett claims the accuracy of these calipers is +/- 0.02 mm in the 0-100 mm range.

In this case, I tried to guard against systematic error as best as possible by taking all my measurement from between the same two points in the same way and taking the difference. With any luck, most of the systematic errors should cancel out.

In general you are correct about people over estimating the precision of their tools, and without knowing me or my tools, I wouldn't ask you to. I am also firmly aware of the pitfalls of systematic error. I would list my credentials for understanding error, but that doesn't mean a lot on an internet forum. All I can do is report the results of my measurements and that I am inclined to believe them. You are free to decide for yourself whether or not to believe me.

 

[gr5 2,176]

@gr5: Where do you know that from...? :rolleyes:

 

LOL! Most of my knowledge comes not from direct experience but from other posts!

 

[gr5 2,176]

By the way avi, if you are planning to print very small, very precise details, at some point you might want to get a smaller nozzle diameter. It sounds like you might have the equipment to make your own nozzles. Some people have printed with .25mm nozzles and you get much better detail but they print much slower - the volume of nozzle is going to be related to the square of the change in diameter so .4mm to .25 will have a max print speed 2.56X slower.

Another disadvantage of .25mm nozzles is they clog much more easily. I have a .65mm nozzle and love it. But I switch back and forth depending what I print and how much patience I have.