Circles not comming out quite right

[mikec96]

New to the forum and new to cnc plasma, so I am not sure if this is the right area for this type of tech help. My wife's step father (Dave) and I have a table that was built for us by his other son-in-law that lives in California. We have been using the table for odds-n-ends projects here and there, we have started thinking that we would like to start using it to manufacture signs, small SXS parts and some other things. Stuff that could be made as "retirement" work to supplement his income (and maybe even mine).

I am having a little bit of trouble with circles. I have used the 3,4,5 method of making sure the x and y are square and squares look good as far as I can tell, however. Circles on the other hand don't seem to end exactly where they start, there is always a little bit of a lip at the top. My thoughts were perhaps there was a setting in UCCNC that is off. The table is all belt driven except for Z. X axis has 2 steppers (one on either side of the table) and 1 Y stepper. Dave was thinking that the settings for X1 and X2 were different but if I am not mistaken that would cause an issue with one side of a large cut (say a large square for example) to be smaller than the other side. My thought was along the lines of when the table changes direction the X axis needs to be speed up a tad more. I cant remember what this is called but I remember watching a video on 3d printers where they talk about this setting, however I don't know if it exists in UCCNC, I have looked thru the settings but don't see anything that rings a bell.

I have plenty of other questions to ask but I would like to start here for now.
Thanks in advance.
Mike.

[cncdrive]

What you described is a typical sign of backlash, that for example circles which start from on direction and ends to the opposite direction (I mean the movements) have some offset. This is due to backlash on the driving mechanism. It is a usual problem with machines using rack and pinions drive and belt driven machines due to cogging belts.
So, check if your machine axis have backlash.
One way to check it is to move to one point along one axis only then mark it, zero the coordinates out.
Now move towards that point moving the same direction as how you moved to this point and then move back to zero which is the same point.
If there is backlash then there will be offset, position difference, the axis will not reach the same point.
You can then measure the difference which is the backlash and set the backlash value in the UCCNC with enabling backlash.
However using backlash compensation is usually a perfect solution, because the axis might have less and more backlash on different positions of the axis.
The best is if you could elliminate the backlash in the driving mechanism.

[mikec96]

What you described is a typical sign of backlash, that for example circles which start from on direction and ends to the opposite direction (I mean the movements) have some offset. This is due to backlash on the driving mechanism. It is a usual problem with machines using rack and pinions drive and belt driven machines due to cogging belts.
So, check if your machine axis have backlash.
One way to check it is to move to one point along one axis only then mark it, zero the coordinates out.
Now move towards that point moving the same direction as how you moved to this point and then move back to zero which is the same point.
If there is backlash then there will be offset, position difference, the axis will not reach the same point.
You can then measure the difference which is the backlash and set the backlash value in the UCCNC with enabling backlash.
However using backlash compensation is usually a perfect solution, because the axis might have less and more backlash on different positions of the axis.
The best is if you could elliminate the backlash in the driving mechanism.

Thanks Drive for the info. I had backlash in the back of my mind but I didn't think that was the term that was used in the video on 3d printers so I wasn't sure...So I guess then the question becomes, in a perfect world how does one eliminate the backlash altogether? Guess I will have to do some more digging on the old interweb. In the mean time I will run thru this process and see if that makes a difference. I have some large pieces of cardboard and a sharpie that I have used to "draw" a template before with the plasma, I can use it to draw a series of different size circles to test this out.

[cncdrive]

Well, with ballscrew drives there is usually no backlash, I mean that the backlash is so little that it is unnoticable.
This is achives usually with oversized or preloaded balls in the ballnut and trust bearings on one end of the ballscrew, so the screw does not cog when axial force is applied.
With racks and pinions drive the pinion can be spring forces onto the rack, ofcourse for this there is a need for a complex mechanics in which the pinion can move back and forth to the rack. This is nessessary, because it is basically impossible to precisely parallel the linear rail and the rack, so you cannot force the pinion to the rack precisely on all the rack's length. The spring can do it.
But this involves another problem, that how much spring force to apply?
If the force is low then the you cannot use high acceleration on the axis, because then it will apply higher force than the spring and so the pinion will move out causing issues.
If the force is high then the rack and/or the pinion will wear out fast.

And with belt drives there are techniquest to tension the belt, but ofcourse it will be basically never be 100% precise especially if the belt is long and freely hanging on a long distance.

[mikec96]

And with belt drives there are techniquest to tension the belt, but ofcourse it will be basically never be 100% precise especially if the belt is long and freely hanging on a long distance.

This being our setup, So then best course of action for now then is to play with bent tension and the backlash settings in UCCNC and see if er can get most of the odd shape out of our circles as possible. At least until we are ready to build a new table once things cool down. we've designed new gantry brackets that are way more sturdy than the ones we are using now and will let us use the entire table as cutting space. I will post an update as things progress, hopefully we can get this issue squared away. Might be a while before we have an update to post though. Things have heated up here in Phoenix and of course that means we have slowed down for the summer, plus I only go out to Dave's place on the weekends when I am not busy here at home so an update here might be a while...

Thank you for all the assistance so far.
Mike

[spumco]

There are additional, more expensive methods of reducing or eliminating backlash with rack & pinions, but those aren't applicable to the OP. But there is some hope for belt drives.

Assuming we define backlash as a certain amount of play in the drive system where the driven component is not firmly held in place - if the pulley can rotate some amount and the belt teeth are not firmly wedged between the pulley teeth, there is some slip when reversing direction. The pulley rotates and the belt stays put - you see this on circles when one axis reverses direction as CNCDrive described earlier.

Belt drives can also have 'lost' or undesirable motion even if there is no backlash. The belt can flex, the teeth can flex, and the belt acts like a spring under load. All of these increase what's called 'settling' time - the time it takes for the driven component to stabilize in a particular location. Even with no backlash the component (gantry and/or torch head in this case) may not be exactly where you intend it to be if the next move is started before the system settles down.

Extreme example: imagine trying to precisely position an object using a pair of opposing rubber bands on each axis. You can move it, but it'll take a while to stop bouncing around.

So, in my limited experience here's what I've found effective in reducing both backlash and lost motion in belt drives. In order of complexity and expense:

1. Check for loose pulleys on shafts. This is free, and it's common for a set screw to work its way out and the pulley is free to move a bit on the shaft. Common on low-end stepper drive systems with D-cut shafts and set-screw pulleys.

2. Check belt tension. Also free. Does not need to be so tight the motor and idler shafts are bent or the bearings are angry, but shouldn't be loose.

3. Change to a different belt profile. If you're using trapezoidal teeth belts there is a ton of backlash - relatively speaking. Switching to HTD (better), GT/GT2 (much better), ATL (much better), or some other belt profile designed for motion control and not just power transmission.

4. If you're using rubber belts with nylon reinforcement, change to steel reinforced urethane belts. These are much stiffer - less "bounce" under acceleration, stopping, high loads. Kevlar belts are stiffer and lighter, but I've read many reports of them not having a reasonable service life before the reinforcements start breaking down - especially on small diameter pulleys. No personal experience with kevlar, but given the price I'm disinclined to experiment with them.

5. Use a wider belt. All else being equal, a wider belt will be stiffer.

#3, 4, and 5 are very easy and fairly inexpensive to accomplish. Moving on to more complicated solutions...

6. Change to a different belt drive system. A stationary belt with a 'tractor' pulley will be stiffer than a moving belt for a given travel distance - it's shorter. The trade-off is that the drive motor and pulley(s) are now moving and contribute negatively to the system inertia. If you're not moving very fast then this won't matter much. If you want the plasma and head to whip around like a pick-n-place machine this drive system may not be ideal.

[My first plasma build used a 20mm urethane belt, HTD-5M profile, in a 'tractor' arrangement. Nema 23 steppers and a 3:1 belt reduction before the main drive belt. Fairly light gantry and it'll do better than 1000ipm with ferocious acceleration. Backlash is maybe a couple thousandths - less than perceptible on a plasma cutter. That table hit 100k pierces earlier this year with only one issue since it was built - a loose set screw.]

7. Consider using a Bell-Everman style belt drive system. You can find details of this online, but the basic premise is that there is a stationary belt bonded to a surface like a gear rack and a second belt which is facing it. The two belts are held in mesh by two idler pulleys keeping them together. The second belt loops up and over the motor (servo, stepper, whatever) between the two idlers. The huge benefit is that the effective belt length is only as long as the distance from one idler up and over the drive pulley to the other idler.

Backlash is minimized because the driving pulley has around 180 degrees of belt engagement - belt to pulley tooth gap is averaged over a longer arc length. And even though the preferred belt is a fairly high-backlash T5/T8 profile, tensioning the motor pulley causes the driving belt teeth to contact opposing flanks of the stationary belt. The only remaining slop is what little there is between the drive pulley and the driving belt. And instead of having a very long, very springy belt that bounces when you start or stop motion, you have an extremely short belt - only the tooth flex contributes to lost motion. And lengthening the system does not increase moving mass or reduce stiffness.

It's a quite clever system. Downside, of course, is that it'll take some pretty serious fiddling around to build a working system.

-R