Jet: Odd results, I have no suggestion about the possible cause. Hmm.
Wormball: If we "forget" the first result in your Y set (-15, 24, 21, 21, 21, 21, 21, 21, 21, 18), your results are the same as mine.
I put in the vigorous homing when I found my machine showing similar symptoms, becoming more repeatable after first few moves. At the time I suspected that it is caused by wear and tear on my machine, especially as I have crashed it several times, hard, during software development. It seems that there is more in this. I will investigate, and I also appreciate any insight or additional clues you may be able to throw into this.
Backlash and Slack Compensation [SOLVED]
Re: Backlash and Slack Compensation
Finally, I’ve got to the bottom of this, it’s a combination of problems not one issue which has had me going round in circles, I’ll deal with these over the next few replies.
First the easy one, belt. Different belts account for different amounts of backlash. First, don’t use steel this is the worst as it can’t be tensioned high enough without distorting the machine, and it wouldn’t last too long because of the bend radius breaking the steel. I tried it in desperation, so you don’t have to
So that leaves Fiberglass (the belt that ships with Liteplacer) and Aramid (Kevlar). I haven’t tried Aramid, but I’ve tried the stock belt and one I purchased on Amazon. The one from amazon accounts for a 0.02-0.03 mm improvement in backlash over the one that shipped with Liteplacer. This is what I used:
https://www.amazon.ca/dp/B078X8NCF5/ref ... 70_TE_item
First the easy one, belt. Different belts account for different amounts of backlash. First, don’t use steel this is the worst as it can’t be tensioned high enough without distorting the machine, and it wouldn’t last too long because of the bend radius breaking the steel. I tried it in desperation, so you don’t have to
So that leaves Fiberglass (the belt that ships with Liteplacer) and Aramid (Kevlar). I haven’t tried Aramid, but I’ve tried the stock belt and one I purchased on Amazon. The one from amazon accounts for a 0.02-0.03 mm improvement in backlash over the one that shipped with Liteplacer. This is what I used:
https://www.amazon.ca/dp/B078X8NCF5/ref ... 70_TE_item
Re: Backlash and Slack Compensation
Now the variable backlash. This one I tracked down to the ends of the belt where they are zip ties, and I suspect most machines have this to variable degrees depending on the zip ties used and how they are placed.
If they’re placed as per the instructions (e.g. X axis belting in the build instructions), there’s a fatal flaw that causes the attachment point to look firmly attached, but it’s not due to the zip tie orientation. On the top it looks perfect, but underneath it’s not. It’s very easy for this to be not detected visually as it’s undernearth and it accounts for about 0.1-0.3mm of error that appears like backlash, but is variable according to the tension in the left/right sides of the belt depending on the direction of travel... essentially the end points shift and need a very large movement to correct. The simple action of rebelting can change the amount, but it’s still there.
It’s due to the zip tie, if the clamp part of the zip tie is put on the bottom of the belt (the thinnest part) as per the instructions, that part doesn’t bend and the tie coming out of the clamp doesn’t bend enough to keep the underneath of the belt clamped. This picture shows 2 different mounting operations, the one on the left is as per instructions (looks good from the top, but from underneath it’s not), the approach on the right is better.
(Click on picture to see)
Now, it’s still not perfect, the zip tie doesn’t bend at 90 degrees, so there will still be some flex. So find a bigger zip tie (about the width of GT2 belt), and cut some pieces off to act as a clamp. This has the effect of distributing the force over the clamp (via the corners) rather than on the corners of a flexible belt (which distorts the belt and doesn’t clamp fully).
It also enables it to be tightened up better as the zip tie slips against the plastic better than rubber. Finally I use small needle nose pliers on the free end of the tie up against the head, and twist 90 degrees against the head to tighten.
Here’s how it looks:
Finally, if you have a longer length of excess belting you’re clamping, you may need more cable ties to keep the clamp down for the full length of the tag end.
Juku, if you’re in agreement with this, I suggest at least changing the mounting orientation of the zip ties.
I “suspect” that the reason “Vigourous Homing” is required on some machines, or machines that have crashed are no longer repeatable, may be due to this end point mounting issue.
Simple and obvious when you think about it, embarrassing it took so long for me to figure out. The problem is you can’t see it as it looks good on top or detect that small amount of movement by pulling on it.
If they’re placed as per the instructions (e.g. X axis belting in the build instructions), there’s a fatal flaw that causes the attachment point to look firmly attached, but it’s not due to the zip tie orientation. On the top it looks perfect, but underneath it’s not. It’s very easy for this to be not detected visually as it’s undernearth and it accounts for about 0.1-0.3mm of error that appears like backlash, but is variable according to the tension in the left/right sides of the belt depending on the direction of travel... essentially the end points shift and need a very large movement to correct. The simple action of rebelting can change the amount, but it’s still there.
It’s due to the zip tie, if the clamp part of the zip tie is put on the bottom of the belt (the thinnest part) as per the instructions, that part doesn’t bend and the tie coming out of the clamp doesn’t bend enough to keep the underneath of the belt clamped. This picture shows 2 different mounting operations, the one on the left is as per instructions (looks good from the top, but from underneath it’s not), the approach on the right is better.
(Click on picture to see)
- 6B5D54BC-8DBE-4E6E-A47A-1B9CA48440BB.jpeg (185.55 KiB) Viewed 9527 times
It also enables it to be tightened up better as the zip tie slips against the plastic better than rubber. Finally I use small needle nose pliers on the free end of the tie up against the head, and twist 90 degrees against the head to tighten.
Here’s how it looks:
- CCE52C9A-3FE3-48AD-9AD5-D14ED2023ABC.jpeg (142.7 KiB) Viewed 9527 times
Juku, if you’re in agreement with this, I suggest at least changing the mounting orientation of the zip ties.
I “suspect” that the reason “Vigourous Homing” is required on some machines, or machines that have crashed are no longer repeatable, may be due to this end point mounting issue.
Simple and obvious when you think about it, embarrassing it took so long for me to figure out
. The problem is you can’t see it as it looks good on top or detect that small amount of movement by pulling on it.Re: Backlash and Slack Compensation
A bit a about testing for repeatability.... It seems a repeatability accuracy of 0.02mm - 0.03mm in X an Y is achievable, but to get there, you have to have some way to test and improve....
This is how I ran the tests (with Slack Compensation on)
1. First get camera homing measurement running reliably and accurately, it’s the only way to test accuracy reliably.
2. Put a very small clean dot (I printed mine out on a brother label printer), in the middle of the machine, after homing move the camera to sit over this dot.
3. Move Left 100mm, Right 100mm, Right 100mm, Left 100mm, Left 100mm, Right 100mm and measure the dot with the UI, and write down the X value. (This sets up the direction and loading on the belt correctly).
4. Move Right 100mm and Left 100mm and measure the dot again, subtract the X values and that’s the repeatabilit error.
5. Test can be repeated by moving Left, Right, Measure, Right, Left, Measure
There are some advantages to switching Slack Compensation off during this process. To get a true view of how much backlash you’re removing as you refine the accuracy, slack compensation off will give a better view. I.E. say you get 0.2mm backlash with compensation off and 0.09mm with compensation on, improving the 0.2mm will also improve the 0.09mm, but it will be easier to see changes on the 0.2mm.
Y seems to exhibit more backlash than X, presumably due to the greater complexity of the axis (2 belts, 2 gears and 4 more pulleys). 0.4mm slack compensation removes the backlash on X for me, but Y needs about 0.8mm to bring it in line with the accuracy in X.
This is how I ran the tests (with Slack Compensation on)
1. First get camera homing measurement running reliably and accurately, it’s the only way to test accuracy reliably.
2. Put a very small clean dot (I printed mine out on a brother label printer), in the middle of the machine, after homing move the camera to sit over this dot.
3. Move Left 100mm, Right 100mm, Right 100mm, Left 100mm, Left 100mm, Right 100mm and measure the dot with the UI, and write down the X value. (This sets up the direction and loading on the belt correctly).
4. Move Right 100mm and Left 100mm and measure the dot again, subtract the X values and that’s the repeatabilit error.
5. Test can be repeated by moving Left, Right, Measure, Right, Left, Measure
There are some advantages to switching Slack Compensation off during this process. To get a true view of how much backlash you’re removing as you refine the accuracy, slack compensation off will give a better view. I.E. say you get 0.2mm backlash with compensation off and 0.09mm with compensation on, improving the 0.2mm will also improve the 0.09mm, but it will be easier to see changes on the 0.2mm.
Y seems to exhibit more backlash than X, presumably due to the greater complexity of the axis (2 belts, 2 gears and 4 more pulleys). 0.4mm slack compensation removes the backlash on X for me, but Y needs about 0.8mm to bring it in line with the accuracy in X.
Re: Backlash and Slack Compensation
Belt Tension: Belt tension in CNC/3D printers is always a subject about guessing, using thumbs etc., the reality is it doesn’t matter too much.
Too tight and the end plates start to distort in this case, too loose and you loose accuracy. Just a bit tighter than loose seems to be the best approach.
The bonus here, using the above test with slack compensation off, it’s easy to find the good range of tension because it shows up in repeatability.
Although I didn’t find an evidence for this, keep an eye on the belt movement across the gears and pulleys, they tend to work up and down the pulleys/gears according to direction changes. It’s possible this could cause a difference in accuracy. The solution here is to apply a bit of force in a particular direction by adjusting the height of the gear to keep the belt in a particular direction.
Too tight and the end plates start to distort in this case, too loose and you loose accuracy. Just a bit tighter than loose seems to be the best approach.
The bonus here, using the above test with slack compensation off, it’s easy to find the good range of tension because it shows up in repeatability.
Although I didn’t find an evidence for this, keep an eye on the belt movement across the gears and pulleys, they tend to work up and down the pulleys/gears according to direction changes. It’s possible this could cause a difference in accuracy. The solution here is to apply a bit of force in a particular direction by adjusting the height of the gear to keep the belt in a particular direction.
Re: Backlash and Slack Compensation
So with the above techniques and 0.8mm of Slack Compensation, the repeatability error on my machine now is:
0.021mm for X
0.032mm for Y
0.4mm is fine for X, but not enough Y.
Success !
I’ve exposed the Slack Compensation Distance setting in Liteplacer to allow it to be adjustable, I’ll upload it as a Pull request to the main github:
0.021mm for X
0.032mm for Y
0.4mm is fine for X, but not enough Y.
Success !
I’ve exposed the Slack Compensation Distance setting in Liteplacer to allow it to be adjustable, I’ll upload it as a Pull request to the main github:
- Attachments
-
- E767ED3F-6D70-48D7-8CFE-0803CD7BE077.jpeg (502.36 KiB) Viewed 9527 times
Re: Backlash and Slack Compensation
Maybe this is a better solution?Jet wrote:Now the variable backlash. This one I tracked down to the ends of the belt where they are zip ties, and I suspect most machines have this to variable degrees depending on the zip ties used and how they are placed.
If they’re placed as per the instructions (e.g. X axis belting in the build instructions), there’s a fatal flaw that causes the attachment point to look firmly attached, but it’s not due to the zip tie orientation. On the top it looks perfect, but underneath it’s not. It’s very easy for this to be not detected visually as it’s undernearth and it accounts for about 0.1-0.3mm of error that appears like backlash, but is variable according to the tension in the left/right sides of the belt depending on the direction of travel... essentially the end points shift and need a very large movement to correct. The simple action of rebelting can change the amount, but it’s still there.
It’s due to the zip tie, if the clamp part of the zip tie is put on the bottom of the belt (the thinnest part) as per the instructions, that part doesn’t bend and the tie coming out of the clamp doesn’t bend enough to keep the underneath of the belt clamped. This picture shows 2 different mounting operations, the one on the left is as per instructions (looks good from the top, but from underneath it’s not), the approach on the right is better.
(Click on picture to see)
Now, it’s still not perfect, the zip tie doesn’t bend at 90 degrees, so there will still be some flex. So find a bigger zip tie (about the width of GT2 belt), and cut some pieces off to act as a clamp. This has the effect of distributing the force over the clamp (via the corners) rather than on the corners of a flexible belt (which distorts the belt and doesn’t clamp fully).
It also enables it to be tightened up better as the zip tie slips against the plastic better than rubber. Finally I use small needle nose pliers on the free end of the tie up against the head, and twist 90 degrees against the head to tighten.
Here’s how it looks:
Finally, if you have a longer length of excess belting you’re clamping, you may need more cable ties to keep the clamp down for the full length of the tag end.
Juku, if you’re in agreement with this, I suggest at least changing the mounting orientation of the zip ties.
I “suspect” that the reason “Vigourous Homing” is required on some machines, or machines that have crashed are no longer repeatable, may be due to this end point mounting issue.
Simple and obvious when you think about it, embarrassing it took so long for me to figure out
https://openbuildspartstore.com/belt-clamp-crimp-style/
Re: Backlash and Slack Compensation
Yes that would work well.ds23man wrote:https://openbuildspartstore.com/belt-clamp-crimp-style/
Re: Backlash and Slack Compensation [SOLVED]
Jet said: I’ve exposed the Slack Compensation Distance setting in Liteplacer to allow it to be adjustable,
Is this an adjustable parameter? I have looked at the tinyg parameters and the Liteplacer configuration and cannot seem to find this value. I am assuming that Jet modified the code to be able to change the distance setting on the fly. I would be happy to just change a parameter if possible. I have my liteplacer accurate to approx 0.3mm (with the slack compensation box checked) when approaching my target from two directions and I would like to reduce this offset if possible.
Keith
Is this an adjustable parameter? I have looked at the tinyg parameters and the Liteplacer configuration and cannot seem to find this value. I am assuming that Jet modified the code to be able to change the distance setting on the fly. I would be happy to just change a parameter if possible. I have my liteplacer accurate to approx 0.3mm (with the slack compensation box checked) when approaching my target from two directions and I would like to reduce this offset if possible.
Keith
Re: Backlash and Slack Compensation [SOLVED]
KeithM: Check the GitHub for the liteplacer software, the change is a Pull request