PDA

View Full Version : A Shenhui accuracy problem



Greg Facer
10-07-2013, 11:14 PM
Hello all,

It's been a while since I've dropped in since I'm not doing a lot of the laser work myself these days, mainly on the CNC router. My signature explains why that suits me just fine.

Anyways, we've noticed that some identically drawn parts vary about 1.5mm from point to point on the table. Doing a quick test now of 300mm rectangles shows the X low side (ie opposite the X motor) is out 1.5mm and the other two are about the same (actual is about 2mm too small and .5mm too small, no kerf compensation done). I've read Rich's posts about the timing belt stretch and even saw the new belt in person when I saw his machine prior to purchase....but I can't quite wrap my head around a size that is a) too small and b) only in one part of the machine as a stretching issue. It seems to be more of a crappy belt mold issue (assuming they use molds to make belts?)

Anyways, rather than changing out the belt and pully, I was thinking of maybe getting a better belt to try first. But, I can't seem to find one on-line long enough or continuous. (looking a little more before posting) I think I see why Rich did the change out he did.

So, two questions: Any other ideas that might be causing the problem? and any idea where to get a 3MM HTD Timing belt in the 2800mm plus length? Or suggestions on alternative setups (like Rich's) and suppliers for that?

Also, I tried the math Rich did for my machine's set up, and came up with a value of 4.8 um per step, based on stepper having a 24 cog gear, meshing with a 72 cog gear and then a 24 cog gear of 3mm pitch. Stepper driver at 5000 steps per revolution. So, 72mm travelled for 15000 steps = 4.8 um per step, the software has 4.79639. But that only changed the theoretical result on my 300mm rectangle by about .22mm. So, I don't particularly think it's the settings.

As always, I appreciate the help. If I can return the favour, let me know!

Dan Hintz
10-08-2013, 8:21 AM
I don't know about the Shenhuis, but on western systems there is a GUI setting for such a problem... basically it gives you a percentage of increase/decrease to X/Y runs. Normally it's set to 1.000, but setting it to 1.001 would give you a 1% increase in length of any run, so a 10" box that always comes up short by 10 mils will have the error removed.

Anything like that?

Greg Facer
10-08-2013, 6:14 PM
I don't know about the Shenhuis, but on western systems there is a GUI setting for such a problem... basically it gives you a percentage of increase/decrease to X/Y runs. Normally it's set to 1.000, but setting it to 1.001 would give you a 1% increase in length of any run, so a 10" box that always comes up short by 10 mils will have the error removed.

Anything like that?


There is, sort of, in the 1 step = x um value. Change that and it will be adjusted and that's what the factory must have done as the math puts the value at 4.8 and the current value is slightly off that. But, changing that doesn't solve the issue I have, which is specific to areas of the table (or areas across the length of the gantry anyways).

I have at least wrapped my head around the facts that a) belt stretch could make a part shorter rather than longer, and b) the belt stretch symptoms could in fact be worse if the bearing is further from the motor. I just wish I actually saw a problem that could account for the amount of error, 1.5mm is a lot.

Dave Sheldrake
10-08-2013, 6:58 PM
Crab gantry Greg?

What are the corner angles like? 1.5mm is half pitch so if the belt jumped 1 tooth one side you would be 1.5mm out (roughly)

just a thought

cheers

Dave

Greg Facer
10-09-2013, 12:29 AM
I haven't even really looked at the y axis with re two belts, this is just on the x across (or along) the gantry. So one tooth skip would equal 3mm and not too likely to repeat easily either.

Allen Rawley
10-09-2013, 4:04 AM
Hello,

Not sure of the problem exactly but perhaps this might assist in debugging.

If it appears like a belt stretch, then it may be tested by reducing the acceleration (setup parameters) and velocity (job parameters). Different acceleration and speed should provide different results with a stretching or skipping issue.

Is the table square? In visiting many factories, including Shenhui, I see the workers eye-balling the part alignment then drilling holes. Measure the distance between the linear rails for the Y axis. For the x axis, it may be skewed. Techno has a pattern that they run to check for table alignment, it looks like a circle in a box and has cross-hairs in it. You could get that pattern, run it on the Shenhui, then have it diagnosed by Techno if you are unable to determine what is going on.

Sometimes the stepper motor can have a problem. Stick you hand on it while the X axis is moving. Do you feel vibration? (I have a bad motor on a new machine right now and it is making a noise). Will replace it before shipping in the next few weeks.

Regarding other issues, is this a test using your laser? That is a rather large table and the laser beam may not be aligned. Take a piece of acrylic and put it before the final beam bender and do a raw beam shot into it, at the four corners of the table. If the beam is moving around then that can contribute to the accuracy problems.

Khalid Nazim
10-09-2013, 4:17 PM
I have exactly the same problem - a rectangular box changes its dimensions at different positions on the tabel. I have tried all adjustments in software and tightening belts etc to no avail. This is really frustrating as I cannot get predictable results from the machine.

I know that this does not help but you are not alone in having this problem.

Rodne/George/Rich - Have you observed this problem? Can you cut a rectangle at different locations on the table and see if all them measure the same.

I would be happy if all them have the exact same measurement even if the measurements are different than what the graphic has as I can then use compensation. I cannot compensate for different measurement at different positions on the table.

Regards
Khalid

Rich Harman
10-09-2013, 7:07 PM
I have noticed the problem and have written about it before. I replaced the X axis belt and enjoyed a significant improvement - but the problem did not disappear. I now run my belts with low tension which seems to help. My theory is that tightening the belt stretches it unevenly. When you have a cheap belt to start with then the effect is more pronounced. I think the tendency to tighten the belts makes the situation worse and is irreversible. Also be sure that the X and Y axis are perpendicular, when they put the machines together they do not seem to care much about getting things lined up perfectly.

I have considered a couple of other solutions but the reality is that for the work I do, a small fraction of a millimeter variation does not make a difference. The first is to put a stepper motor on each end of the gantry driving the same belt - this should at least halve the errors, but I would expect the errors to be nearly eliminated. The second is to use a ball screw. There are some light duty ones that should work well for cutting but they would not do well for rastering at high speeds - too long of an unsupported shaft. The last is to use an encoder. Not on the motor but a linear encoder on the gantry. It would require replacing the stepper drivers with ones that support closed loop.

You can get open ended belts from polytechdesign.com - http://shop.polybelt.com/15-3M-Open-End-Belt-Roll-Polyurethane-with-Steel-Cords-B15-3M-MPS.htm

Greg Facer
10-10-2013, 5:27 PM
I have noticed the problem and have written about it before. I replaced the X axis belt and enjoyed a significant improvement - but the problem did not disappear. I now run my belts with low tension which seems to help. My theory is that tightening the belt stretches it unevenly. When you have a cheap belt to start with then the effect is more pronounced. I think the tendency to tighten the belts makes the situation worse and is irreversible. Also be sure that the X and Y axis are perpendicular, when they put the machines together they do not seem to care much about getting things lined up perfectly.

I have considered a couple of other solutions but the reality is that for the work I do, a small fraction of a millimeter variation does not make a difference. The first is to put a stepper motor on each end of the gantry driving the same belt - this should at least halve the errors, but I would expect the errors to be nearly eliminated. The second is to use a ball screw. There are some light duty ones that should work well for cutting but they would not do well for rastering at high speeds - too long of an unsupported shaft. The last is to use an encoder. Not on the motor but a linear encoder on the gantry. It would require replacing the stepper drivers with ones that support closed loop.

You can get open ended belts from polytechdesign.com - http://shop.polybelt.com/15-3M-Open-End-Belt-Roll-Polyurethane-with-Steel-Cords-B15-3M-MPS.htm

Thanks for the link Rich, especially knowing you went with a different belt geometry!. I am not quite ready to pull the trigger, but just ran some thicker pieces that happened to be close to my earlier test, with the same results. I realized that my next step should be to dived the 300mm test area that had the error into 3 100mm areas and see if I can narrow the problem further.....and then hopefully find the defect on the belt or some other clue that gives me something concrete to fix.

Greg

Greg Facer
10-10-2013, 7:38 PM
Yep, Rich is right, wonky belt.

Rather than 3 rectangles 300mm wide, I made 9 rectangles 100mm wide. Results vary across all the pieces, only about 3 were within .1 mm of what I am guessing was "right"....keeping in mind the kerf is no accounted for and the math works out that they all should be short about .08mm anyways. The X low (beam) side was consistently low on all three pieces, so that's why I noticed the issue there, but it wasn't a completely progressive issue from one side to the other.

So, I will try Rich's belt source and suggestions of low tension (how low are we talking about?) and correct the math issues and see what happens.

Thanks everyone, I will report back when I have installed the new stuff.
Greg

Rich Harman
10-11-2013, 9:45 PM
If you purchased spare belts with your machine I would try that first - but tension them only enough to take the slack out.

Rodne Gold
10-12-2013, 12:29 AM
I have not noticed anything untoward , albeit I dont use the Shenuis for anything requiring major precision.
I dont think the belts the chinese lasers use are great tho , nothing like kevlar reinforced stuff etc.
I havent replaced any and not sure at what tension they should be , we just flog the machine on a daily basis and it seems to work...

Kay Bengtson
10-16-2013, 11:26 AM
Do these Chinese lasers have enough accuracy and reproducibility to do multiple passes? I cut thick balsa parts for model airplanes and often make 4 passes to cut through a hard section running through the sheet. My ULS 25ER with a Testra Accelerator controller does this easily. And the size of the part doesn't change no matter where it is cut on the table. is this not the case with Chinese systems? If so, what applications can they be used for?

Kay

john banks
10-16-2013, 11:39 AM
Closer to the front of the machine, the more stretched things get in our y axis. It is noticeable on some jobs where parts cut in different areas of the table are overlaid. At the time I just changed the artwork, but did think it was related to Rich's previous descriptions of belt problems.

It is a few mm of an error, but seems consistent and unrelated to speed or direction of travel, so it must be stretch?

Khalid Nazim
10-16-2013, 11:57 AM
So this issue of having small differences in dimensions of cut parts at different locations on the bed only visible in Chinese machines or this issue is there on Trotec/ULS/Epilog as well?

Would a rectangle cut at different locations on a large Trotec/ULS/Epilog bed be exactly the same size or would be out by 1 or 2mm? I am trying to understand if this accuracy issue is in all lasers or is it only with the chinese machines.

Regards
Khalid

Dan Hintz
10-16-2013, 12:45 PM
The accuracy specified by western machines is across the entire working surface. If it wasn't, there would be a lot of cranky people out there.

Rich Harman
10-16-2013, 3:47 PM
It is a few mm of an error, but seems consistent and unrelated to speed or direction of travel, so it must be stretch?

A few mm??? Holy cow that is a lot!

I have cut a lot of parts that are 30 inches long at all locations on the table. Symmetrical parts that share a cut line that interlock - box joints. When assembling if one piece gets flipped over it will still fit but will be a bit tight in places. If the error were were that big it would be impossible to assemble.

273108

john banks
10-16-2013, 4:30 PM
The back 100mm of the table has the cable chain hitting the cross member of the rear lid hinge, and the front 200mm have this error. So I guess we have a 1250 not a 1280. I must get it sorted, just haven't had the time and nothing has forced it to the top of the pile to sort. Embarrassing for a perfectionist, but it is my wife's machine and I only go in her workshop about once a month so I forget about the issues!

Rodne Gold
10-17-2013, 2:33 AM
We cut lots of circles and squares and oddly shaped badge backs on both our machines , using the whole table , I just measured some of the stuff we do and there is no appreciable dimension change at all , no matter where the piece was cut on the tables. I would also imagine belt slop or stretch would show up as misregistration at the start / stop point of a cut?

George M. Perzel
10-17-2013, 4:21 AM
I am currently cutting 3 inch ornaments out of 30x30 inch BB sheets. I just checked dimensions of the 4 corner pieces and see no difference in size but noticed small difference in cutting depth at max XY corner (front left).
Best Regards,
George
Laserarts

Greg Facer
11-01-2013, 3:52 PM
Do these Chinese lasers have enough accuracy and reproducibility to do multiple passes? I cut thick balsa parts for model airplanes and often make 4 passes to cut through a hard section running through the sheet. My ULS 25ER with a Testra Accelerator controller does this easily. And the size of the part doesn't change no matter where it is cut on the table. is this not the case with Chinese systems? If so, what applications can they be used for?

Kay

Repeatability isn't the problem, so multiple passes should be fine. but when you buy a chinese machine the cost increase for more power is much less than a ULS would be, so you should really likely only use one pass.

Specifically to Rodne's question: Since the errors seem to be repeatable, there isn't a problem with the cuts lining up.

Greg Facer
11-01-2013, 6:51 PM
Hi Guys,

I finally got to installing the new belt today. Not the improvements I was hoping for. I haven't tested as extensively as the old belt (yet) but I have about .6mm variance on a bunch of 100mm cuts across the width of the table. I guess that is better than before, but not exactly great.

I then figured out a test to check the roundness of the gear wheel. The movement per rotation of the final drive cog is 72mm, so I made some 72mm cuts spaced 72mm apart. If they all worked out the same, I had my problem (not round wheel). But, nothing changed, still about the same variation from piece to piece, so that wasn't it.

Next step I guess is to check the alignment of the beam, as that's about the only other thing that might cause the issue (but I would expect the changes to be progressive across the table, and they are not),

After that I guess it's change out the stepper and then on to the controller. Also on my list, if nothing else works, to make a roller for taking up the slack in the belt that mounts to the bottom of the gantry and goes up scorpion style. That would help with any belt droop issues.

Another idea, I guess I could change the dip switches on the stepper drivers to have less steps per revolution and that would point to electrical issues if that helped (revising the step length in the machine as well). 5000 steps per revolution seems a little overkill anyways, and maybe it's too many pulses for the system?

This is very frustrating as I really can't think of ANY problem that could have a non-progressive error (ie something like -.35, -.25, -.6, exact size (ie too big with the kerf), -.35, -.25, moving left to right on the table) and be as repeatable. Any other ideas? 50hz versus 60hz electrical?

Dan Hintz
11-01-2013, 7:26 PM
I then figured out a test to check the roundness of the gear wheel. The movement per rotation of the final drive cog is 72mm, so I made some 72mm cuts spaced 72mm apart. If they all worked out the same, I had my problem (not round wheel).

If you are testing for eccentricity in the cog, you need to test at more positions that one full revolution... you need to test at least the 90 degree marks of a full rotation, so make a line every 18mm. If any one of those lines comes up as more/less than 18mm, then you have an eccentric cog.

Greg Facer
11-01-2013, 7:42 PM
If you are testing for eccentricity in the cog, you need to test at more positions that one full revolution... you need to test at least the 90 degree marks of a full rotation, so make a line every 18mm. If any one of those lines comes up as more/less than 18mm, then you have an eccentric cog.

True, but I was more checking to see if hitting the same spot on the cog (in this case 1 rotation, but the spacing made the pieces one further full rotation apart) caused the problem to disappear. If yes, that would say the problem was the cog, and I could investigate further as you describe.

Greg Facer
11-09-2013, 7:08 PM
An update:

I have done the beam alignment and I think it helped a little. The beam was consistent before, not quite as centered as would be ideal and is now more centered on the final mirror.

The cut pattern I am now using is 72mm cut, 72mm gap, and repeat along the X axis. I still have the same problem, low in some areas and one or possibly two pieces that are too large. As I have previously mentioned, the 72mm spacing is one rotation of the cogs the drive the belt, 3 rotations of the stepper motor. So, any errors to on eccentricity should be eliminated in this test pattern.

The weird part is that the error seems to be good (or slighly high), undersized, good, undersized, good undersized. I don't have the exact numbers with me right now or I'd post. The difference between pieces on the low or high "set" is only about .2mm (or plus/minus .1mm I suppose one might say). The difference between the high spots and low spots are in the range of about .4 best case to .6mm worst case.

I believe the issue would be compounded if the shape was spanning two low areas and only one high, esp. if compared to another part that spanned 2 high areas and one low.....assuming that the problem is tied (somehow) to the placement on the machine or the belt. That is to say, I think it's about as bad as I first determined, but I haven't directly compared that test pattern.

Obviously, what is weird in the above is that it has an larger/smaller/larger pattern at all. Progressive from one side of the table to the other, or progressive to a point in the center of the table? I could see both happening. But, the L/S pattern was unexpected since I had taken steps to remove any effects of the stepper cogs or the reduction cogs.

Anyways, if that gives anyone ideas, please share!

I haven't tried the stepper drive reduction of # of steps per revolution yet, ie changing the dip switches, but that's the next thing to try I guess.

I think I'll do some more tests as well, maybe moving the origin over by 36mm or similar and see what happens.....reading my earlier posts, I'm not sure it had that same pattern.

Rich Harman
11-09-2013, 7:58 PM
What is your step length set to? 6.000?

Greg Facer
11-12-2013, 11:43 AM
What is your step length set to? 6.000?

4.8, my calculations are different than yours, although I can't remember where the difference was. Shenhui had it set to 4.7932 or something similar. Close, but not quite the mathematically correct answer.

They said themselves via email that they cut, measure, and adjust the value (I think the software has a function for exactly that process as well.

Greg Facer
11-12-2013, 9:58 PM
Another update (no success yet)

I have tried 10000 steps per rev, and 1000 steps per rev......figuring out how to set the "Y" axis value in the meantime (I had set both drivers the same but didn't realized there was a "Y" setting separately). Exactly the same results as the regular settings with either change.

My next step I guess will be to change the control board to the spare I had, as changing the stepper is more work :-), then the stepper / stepper driver in turn....that will be for another night.

Oh, I also ran the program after downloading to the machine as I thought I might as well rule out a windows "hiccup" of some sort from the USB connection, but the same results again. I haven't been measuring tonight, just placing the pieces on end and comparing one set to another.

And just for kicks, and because I've been avoiding it, I tested the Y axis. Same issue (not really the up and down part), but to a lesser degree than the X.

jordan matthai
03-06-2014, 8:07 PM
Hey Greg,
Did you ever get the issue fixed? I've been having the same issues and can't seem to come up with a solution.

Greg Facer
08-21-2015, 12:22 PM
Hey Greg,
Did you ever get the issue fixed? I've been having the same issues and can't seem to come up with a solution.

Short answer is no. we've been having worse problems recently as the cuts aren't lining up. Tried changing the stepper, but didn't seem to help. And, actually, the other axis (Y) is worse apparently.

So, I've ordered up some closed loop steppers and drivers to try ($400 us approx) and will try and re-seat wiring and grounding at the same time. Maybe change out the controller box if those steps don't help, as I have a spare, but don't expect that is the issue.

Will update thread if I have a fix.

Clark Pace
08-21-2015, 1:29 PM
Short answer is no. we've been having worse problems recently as the cuts aren't lining up. Tried changing the stepper, but didn't seem to help. And, actually, the other axis (Y) is worse apparently.

So, I've ordered up some closed loop steppers and drivers to try ($400 us approx) and will try and re-seat wiring and grounding at the same time. Maybe change out the controller box if those steps don't help, as I have a spare, but don't expect that is the issue.

Will update thread if I have a fix.


Have you checked your gantry alignment. Mine was of a bit, and it was causing issues. Another thing I'm seen is belts moving up and down on the gear.

Greg Facer
09-01-2015, 2:25 AM
Have you checked your gantry alignment. Mine was of a bit, and it was causing issues. Another thing I'm seen is belts moving up and down on the gear.

Can you describe that a bit more (belt up and down on the gear).

As a continuation of the thread, I put in the hybrid closed loop steppers tonight, put on a new belt (well, old belt but since the one on the machine hadn't changed anything, and was a little melted, I went back to the old belt that came with the machine), and tweaked the X backlash slightly to .1mm.

Circles are coming out OK now, but cutting a line of 72mm wide rectangles still shows a fair bit of variation in length, although not as bad as before. I am about .5-.6mm from smallest to largest. But, the effect is somewhat random, or partly is. I cut two rows this time, They get larger and smaller together, but are about .1mm different from each other. So, I'm trying to nail down .4-.5mm of error, belt wandering might account for part of that. I'm assuming some error I will never eliminate and some error is probably beam related...different angles at different parts of the table.

I'll have to do more testing, but if I end up with .5-.6 error down from 1.5mm error previously, I suppose I'll be somewhat happy. Might still change out my controller too. I'd like to get my pieces within .3mm ideally.

Good thing was that the hybrid steppers was more or less a drop in fix, save for a little hacking metal holes to allow the encoder connector to pass through.

Bill George
09-01-2015, 7:40 AM
Is this a new machine or is this something that just started? If from day one it sounds like a design/setup problem unless the gantry as someone said is out of square or loose.

Greg Facer
09-01-2015, 8:53 PM
This has been from day 1....really, I don't use the machine for stuff that needs absolute accuracy anymore, use the CNC for that.

Sad to re-read my posts and realize that maybe current results are same as before, but I was "sure" that I had 1.5mm total variance, but it's been a few years too :). Regardless, our "sister" laser (ie same machine bought at the same time, though different tube setup over time) is still much better, more consistent on same-position cuts (.17mm variance over 5 samples) and across the table cuts (.25mm variance over 5 cuts).

I think I might cut a series of circles next. Any electrical noise type issue (introduced before the closed loop at the stepper driver), or crappy controller, should show up in circles with a offset start stop line that varies. My square cuts would mask that somewhat. Anything that is just due to positional variance in the belt or gear performance would presumably still line up, which has mostly been my experience. Then I'll probably re-set all wires and change out the controller too, and re-align the beam. If that doesn't work, it is what it is.

I will check the squareness of the gantry and test bolts for looseness too. Hard to say a design problem when I have an identical machine that works acceptably. Still much worse than the machines spec's, but a spec of .002" repeatably on a belt drive should be ignored anyways!

Bill: right now (and from day one) I have the machine grounded directly to the building's ground rod, so directly grounded but not to a separately driven rod that I understand could cause a ground loop problem in certain cases. That should be good, yes?

Bill George
09-02-2015, 8:14 AM
In the US if you have properly installed electrical service, grounded per NEC and your outlet is properly connected to either the bare or green grounding conductor that is all you need. Added a separate ground otherwise can cause ground loops. Building and earth grounds are not always at the same potential as the electrical service ground. But you said connected to the exact same ground rod as the electrical service so that should be ok.

I would check your machine for square and otherwise as you have planned.

Rich Harman
09-06-2015, 6:17 PM
Regarding this accuracy problem... a few years ago I was cutting out 7" belt sprockets out of 1/4" ply for our FRC robot. I had to make dozens of them. Four would stack together to make a single pulley. I found that they had to be all rotated to a certain position for the teeth to line up perfectly. There was some inaccuracy in the way it was being cut.

So I sent the file to a friend that worked at very well equipped R&D place that had a 100W Universal laser. We went there after hours to get these parts cut. The parts all had the same problem. Obviously at this point we suspected it was an error in the file. It wasn't, we checked several times. The file was perfect.

The point here is that it may not be unique to Chinese lasers - others can have the same problem.