[DC-Modder]

First off, Hi! I'm new to this forum and new to CNC. I'm not even sure I'm posting in the right section. I'll get right down to it. I bought a home built CNC with a cutting area of 2' x 4'. I wish I'd done more research because I wouldn't have bought this one after all I've been through with it and what machines I found afterwards. After setting it all up and running test after test, I have what I would consider major accuracy problems on the X and Y axis. It's a rack and pinion setup. X axis has a 3:1 pulley with a YA20 boston spur gear. Y axis has 3:1 and a YA15 gear. Stepper driver is a Gecko G540 and 1.8 degree stepper motors. I guess my question and fear is, what amount of accuracy should I be getting with this machine? With the current settings in EMC2 (set to everything it's "supposed" to be according to everything I've read). It is only really accurate at long distances of 36". The less I go the less accurate. For example, I set it to cut 3.5" on the Y and it cut 3.6". I have Onsrud bis and a precise bits collet. I would just make adjustments to the measurements if I could, but since the accuracy changes over the coarse of length then it seems impossible. So is this what I should expect or am I doing something wrong? The Z axis has a 16 TPI lead screw and has no issues. Any help would be much appreciated.

[LeeWay]

What kind of tensioning do you have on the gearing? I recently completed a 4' by 4' plasma table that uses R&P and it is far more accurate than that. I got it to about .005" per foot and called it close enough. You can calibrate the steps too I think. It may actually be loosing steps on the long tries and why it seems to work correctly, though the step count may actually be off. This might be why you get so much error on a short burst. I would dial the steps back a hair until you get the short tries correct. Then worry about the overall. Do you have a dial indicator?

[DC-Modder]

I'm so new I don't even know what a dial indicator is lol. So I probably don't have one? The belt tension is very tight with no slack. How do I calibrate the steps? If it were losing steps then wouldn't it be more off the further the distance? I calibrated it with calculations and not by measuring it since the first time I calibrated it by measuring short distance moved with a caliper, but then it was falling very short on long runs. I'm at a loss right know as to what to adjust or troubleshoot.

[LeeWay]

I don't mean your belt tension. I mean how is the gearing held tightly into the rack? A dial indicator is just a little more accurate than calipers. If it is too long on a short run, then you decrease the amount of steps just a bit each time and recheck. That is called calibration. First though, I would make sure there is enough tension on the spur gear to rack. If it is loose, it can easily skip a tooth. You only need to calibrate once you are sure the rest of the mechanicals is working as it should.

Here is the type dial indicator I have. I have both cheap ones and nicer ones, but I generally use the cheap ones.
G9849 Magnetic Base/Dial Indicator Combo - President's Special

Can you post a few pictures of the machine?

[DC-Modder]

I will post some pics when I get off work tonight. I work 12hr. days on the weekends so it will be around 8pm. Sorry for the misunderstanding on the tension. I'm an electronics tech not a machinist. The gear shafts run through bearings that are pressed into the gantry and z axis's. The gear track or whatever it's called is bolted to the gantry and to the x axis where they are meshed tight enough to have practically 0 backlash. The x axis has spur gears on both sides connected by one shaft and driven by one motor.

[DC-Modder]

Here are a few pics. Hope this helps. I can upload more specific if you request.
Attachment 187370Attachment 187372Attachment 187374

[P.Passuello]

It is normal practice for the pinion gear to be spring loaded against the rack to take up any variation in the rack along its length.
Any variation in the rack / pinion tension can cause backlash and inaccuracy at that point.
I can't quite tell from your photos but I suspect they are not.

If you look at this video about the 4.30 minute mark you will see what I mean

R&P Assembly Instructions - YouTube


Having said this, if your issue was backlash then the item you cut would be smaller rather than larger.
The one thing I would question is have you calculated the steps per inch correctly?

Boy do I hope I have done this right.
I have done some calculations for the YA15 gear and come up with 2246.85 steps per inch.
Likewise the YA20 I get 1736.24 steps per inch.

Does that match your settings or have you got something different.
I am not familiar with EMC2's required settings.

Hope this helps.
Cheers
Peter

[LeeWay]

Yeah, it doesn't look like you have any adjustment on the spur to rack. This is common in a lot of machinery like drill presses, some lathes etc. This isn't common on machines needing accuracy. As mentioned, those are typically tensioned to the rack to take up slack along the way. Even if your rack is perfectly aligned to sit tight against the spur gear along it's path, starting with a higher acceleration in between where your rack mount bolts are has enough torque to deflect and skip a tooth.
For this reason, you should probably reduce your acceleration by quite a bit and use the minimal setting there that you can. That will help eliminate the initial torque jump especially when reversing directions. Then calibrate your steps as mentioned. Just start reducing the number of steps per inch until your travel reaches the desired distance like 3.5". Do that for all three axes.
Then you can check out your long runs for accuracy. Cut some parts and start design a spring loaded tension system.

[DC-Modder]

You are correct that there is no adjustment. The gears are meshed tight and the rack is perfectly straight. I have ran it at the slowest speed possible knowing that it has not skipped any teeth with the same results. I know for a fact that those are not the steps per inch that I have. I will change them when I get home and check the results. However since it is accurate at long distance and the slightest adjustment has thrown it off considerably, I think it will be off again at least the way it sets. Right now it is set at 1/10 microstepping at 200. So to decrease the steps say to 1/5, then do I set it to 400? That is where I'm a little confused. Thanks for all the help. I will update with results.

[LeeWay]

If you are only off .1" in 3.5", then you are already very close on the math. I would say if your steps for instance was as above, 2246.85. then reduce them to 2240.00. Then recheck. Once you get it to where it travels 3.5" when called for, then log those numbers in a safe place as backup.
Do similar on each axis.

You might think it is very tight now, but another thing that spring tension does for you is extends the life and keeps it correctly tensioned even through the wear. Especially if it is very tight now it will wear pretty quickly, so get a bunch of grease in there. If not then you soon will have slop because of wear on the spur, the rack or both.

[Hirudin]

You say it's accurate at long distances but not at short distances, right? You've mentioned that when you tell it to cut 3.5" it actually cut 3.6" but, unless I missed it, you haven't mentioned how accurate it is with longer distances. If the accuracy issue was tied to the steps per inch a call for a 35" move would actually go 36"; I take it that isn't the case though.

When you say you told it to "cut" 3.5 inches but it actually cut 3.6 inches what do you mean? Are you using an actual cutter on a piece of material, or did you chuck-up a pen to draw on paper or something?

Using a pen instead of a cutter might be helpful at this stage. It has a sharp point and will not impede the movement of the machine. It would also be a lot cheaper than using a bunch of material. A downside would be if the pen flexes too much. Try to prevent that by using a rigid pen, making the amount of the pen that is sticking out of the machine as short as possible, possibly using a felt-tipped marker (like my favorite: a Sharpie), or a combination of these things.

Is the inconsistency... consistent? I mean: does the machine move the same amount at different areas of the table?
I feel like I'm being unclear...
Jog the spindle so it's near the south-west corner of the table and tell it to move 3.5" to the east. Measure the actual distance traveled. For argument's sake, let's say it actually moves 3.6" to the east. Now move the spindle to the north-west corner of the table and tell it to move 3.5" to the east. Will it move the same amount or will it move a different amount? Repeat the test elsewhere on the table, is it consistent or inconsistent. It would also be a good idea to try the same motion over and over in the same place. If you move 3.5" to the east then immediately move it 3.5" to the west does the pen return to the starting point?

It sounds like you already had this idea, but I would suggest doing the tests at different feed rates as well.

IF the inaccuracy is inconsistent you probably primarily have a hardware issue. Note that you may still have a software or steps per inch issue, but they'll be a lot harder to diagnose if your hardware isn't working properly.

If the inaccuracy is consistent what happens when you double or halve the movement amount? Let's say you tell it to move 3.5" but it ends up moving 3.6", what happens when you tell it to move 7.0"?
Does it move 7.2"? <- this would suggest a steps per inch problem
Does it move 7.1"? <- I'm not sure what this would suggest, but if you find out it's always off by 0.1" I'm sure someone can help you
Does it move some random amount? Does it move exactly 7.0"? <- These would probably be the least desirable outcomes and will necessitate more testing.

[DC-Modder]

The testing I have done has been both cutting and with an 1/8 sraight non-cutting bit that came with my collet that I can chuck in for measuring. I have tried calibrating using several methods you have mentioned and some more I will try. The measurements are always consistent in the same measurments, but inconsistent as I increase the distance. So right now when I set it at 36" it is dead on. The shorter the distance the less accurate. Now when I make it more accurate at short distance it is less accurate the longer the distance. By a lot. At 36" it will be off by a few inches. I don't remember the exact numbers at the moment, but when I get home I will run several test and post the results. Hopefully they will clear up any confusion. I will also post my current settings and the math I did to get there.

[jimhurlburt]

To kind of jump into this discussion -- The way I would attack this problem.

From your description, I would suspect that the step size is wrong, or the rack/pinion clearance is too tight, is binding up, and you are losing steps. There needs to be just a little clearance between them at the tightest point unless you have a spring loaded setup so the clearance can float as the machine moves. It's also possible that the roller setup for the gantry is binding and causing step loss. Or your motor/driver/power supply combination isn't a good one.

Put some sort of pointy object in as a tool. One of the laser edge finders would be excellent, but almost anything that has a point that is small compared to the accuracy you are trying for will work. For the moment, don't worry about anything smaller than 1/64 or so. Move the z axis till the point is a little above the bed but not touching.

Using emc2, move the tool to the center of the bed and zero the x/y axis. Mark the location of the tool. Placing a 12" steel rule so the tool is at 6" would be a good start. If you can find/borrow/buy four of them it would be excellent. Check them against each other. On the scale of 1/64 they should be the same and the zero/12" points should be at the ends of the rulers -- not inset from the physical end of the steel.

Keep in mind that tape measures from Lowes or Home Depot are not guaranteed to be real accurate when you get to the little numbers. Good enough for cuting sheetrock, not so good for machining.

Setting the travel speed to slow, very slow, move the tool some not large distance. Perhaps 3". Check the tool against the rule. Record the measurement.
Repeat going the same direction till you run out of ruler or table travel. At this point, don't fret about what the measurement is. We are trying for reputability for the moment.

Return to zero using the same increments. Again, record the results.

Take note. You should return to the zero point +- backlash in the gears -- what you can move the tool by pushing on the gantry or table, first one way, then the other.

Take note. The measurements at each movement should have been the same (again +- backlash). At this point don't worry about what the backlash is, even if it is large.

Repeat the above several times without changing the zero on emc.

If the zero is moving, you are losing steps. Either the motors aren't set up correctly with the drivers/power supply/emc, or there is too much friction in the roller/rack setup. Unless too tight on the hardware setup is pretty certain, I would go into emc and reduce the acceleration to pretty slow as a first test.
Don't worry at this point about setting it too slow, you can increase it later when you get the rest of the problems sorted.

Repeat the above going the opposite way from center, then switching to the other axis (x vs y).

Fix any binding issues till you can get repeatable results. At this point, don't worry about what they are, just that you can move the machine back and fourth in both x and y and return to zero -- the same zero every time. If you have a mismatched set of motor/driver/power supply, it may be harder, although if they work at all, moving slow and under no load, you should get consistent results. You may need to adjust the pinion/rack engagement and or the roller track clearance to looser to get consistent results. Bind is bad. For now, don't worry about getting them too loose so long as the pinion can't jump a tooth or the gantry fall off the rails.

After you can get the machine to give consistent results, start back up on checking for bind and backlash along the length of travel on both axis. I would move an inch, then check for excessive play or binding, searching for the tightest point and adjusting for minimum at that point. You need a smidgen of backlash in the gears at that point, and for the rollers (probably a different point), no bind. For the rollers, you probably want to take the belt off and move the gantry back and fourth by hand. It should move freely at all points in the travel with no detectable tight spots when you are pushing it by hand. And no, or almost no slop at the tightest point.

When you get that working OK (It's your call on OK. I don't know whether you consider .032 OK or if you need .0005. If you need less than .015 you're going to have had to find a better measuring system than the steel rules or tape measure I described.)

Then in emc, adjust the step/inch till when you tell emc to move some distance, your measuring tools agree.
If you have mismatched motors /drive sprockets they may not be the same numbers for x and y.

After you are satisfied with that, start working speed and acceleration up. At some point, you will start losing steps again.

Back off some and build something. Keep in mind that cutting something will load the steppers, and the speed/acceleration where you start losing steps will be lower than with a marker tool in the tool holder.

Good luck,
Jim Hurlburt
Yakima, WA

[DC-Modder]

I will try those steps as well. When I was originally trying to calibrate it I was using the ruler that came with my caliper. I would assume that it is accurate. During test it would always go the same distance and return to exact 0 spot. From what I can tell by googling. The driver, power supply and motors were bought as a set. A nema 23 setup. At least the driver and power supply because they are the same as what you see online. The motors I'm not 100% sure. I was able to get a spec sheet on them by googling the model #. I will also post those specs later tonight as well. If to tight can cause loss of steps then it may be an issue. They are snug but not binding. With the motors on I cannot move either axis. With them off even with the motors still connected to the drive I can slide both axis's around. Can't wait to get home to try this stuff and feed you more info.

[DC-Modder]

Ok, I'll start off with a little info for you guys to think over while I run some test.

Settings in EMC2 are as follows

X Axis:
Steps per revolution: 200
Driver Microstepping: 10
Motor:Leadscrew 1:3
Leadscrew Pitch: .31847133758
Axis Scale: 1910.8 Steps/in.

Y Axis:
Same as above except
Leadscrew Pitch: .4246284501
Axis Scale: 2547.8 Steps/in.

Equipment:
PC: HP Pavilon a1517x with 2 gigs of ram and pci-e video card. All other options stripped out and/or turned off for performance
Motors: labeled as OEMOTOR model 57BYGH433-06BS http://www.cncrouterparts.com/specs/57BYGH433-06BS.PDF
Driver: Gecko G540
Power Supply: S-350-48

With belts removed both axis's slide with no binding. I also previously installed a resistor on the driver to limit the current to the motors to 3.3 amps intstead of 3.5 as I thought they were getting a little too hot and I wanted to help prolong their life by not maxing them out. I'm going to move on to some testing again and get back with some more details. Hope to hear back from you guys soon!

[DC-Modder]

Well after moving the X and Y axis around the table checking for backlash and how much I could move the gantry and Z axis by hand while locked into place I did find some troubling issues/weak spots. While the X axis was at 0 in the middle of the table there was backlash on one side of the gantry. I was able to fix that for now. The most troubling though was the Z axis mount to the Y axis. No backlash on the gear, but one of the mounts for the bottom rail is stripped out causing the z axis to wiggle. It is fixable, but not tonight. However since I have been running the test from the middle of the workspace while not under load it is pretty accurate under the same settings with no change. It is starting to look like if I want this thing to be accurate I'm pretty much going to have to rebuild or modify a lot of it . So any suggestions on what I should do and how much it's going to cost to rebuild the X and Y or should I take this to another part of the forum?

[MooseCat]

I would cut/draw a square wave right across the x axis in 1 inch steps (after calibrating for 1inch). Then draw/cut a straight line next to it. Final result should look something like the attached picture.

Are all the square waves exactly 1 inch? Is the total length of the square wave pattern equal to 1inch x number of waves? Is the line the same length?

This could highlight a few problems. If problem is just at a certain point along the axis, one or more of the squares will be wrong size. If total length is not the sum of the smaller measurements this would point to measurement errors rather than the machine being at fault.

I might well be wrong on this, but if I was there with you, the first thing I would be checking is that the width of the cutting/drawing tool has been correctly accounted for. If it hasn't, this would give the symptoms you describe. ie the length you calibrate for will work perfectly but all other cuts will be incorrect on X and Y. Z will be unaffected.

Attachment 187532

[DC-Modder]

I will try that as soon as I figure out how I'm going to fix the issues with the gantry. The tool was accounted for when the cut was made. That is what I originally thought was the problem was at first until I went back into cambam and double checked. The tool has also been checked with a caliper and it is precise. I'm beginning to think that with the slight play in the bad sections was causing most of the issue. I'm just at a bit of a loss now on what I should do. When I bought the machine I expected to have to put some time and money into it to get it where I wanted it. I just under estimated how much. If I had known it was going to get to this point I would have just bought a zenbot.

[P.Passuello]

Really embarassing!

I did get my calculations wrong because I read the gear specs wrong. I must need my eyes testing.
The X axis YA20 is a 1" spur gear not a 1.1" so it should be 1909.86 steps per inch
The Y axis AY15 is a 0.75" spur gear not 0.85 and should be 2546.48 steps per inch

This is a lot closer to your settings and with the setting you have on your machine for the Y axis.
I reckon it would cut a 3.5" piece to 3.502" which I don't think you would quibble over and is a long way from 3.6" you are getting.

Cheers
Peter

[DC-Modder]

So what settings did you come up with to get those steps?