A few people have asked about linearity calibration on the EX controls, so here is a video on the process, as well as some data on my system before and after the calibration.
1 Like
Jim, this was really good going a step further in the error analysis than I have done in the past. Like you I use a 12" caliper and then a very good 4’ rule graduated in 1/64" for the full travel. Using a laser center finder that has a very small dot in the chuck I can get a pretty good idea of the performance over the full travel. With my old eyes I have to use a lighted magnifying glass but I can get well within .005". I use a similar method to check for gantry squareness by laying the rule down to measure four point box diagonals.
Here’s another way you can check for squareness: https://youtu.be/qz6VrsEq_Xc
I’ve been thinking about getting a scale like one of these just to play with
for these kinds of measurements. Certainly overkill, but looks fun. I see that the Acorn has a encoder input, so maybe its compatible and then I maybe wouldn’t have to buy a dro for the scale if there is a soft one on CNC12?
Ha, that reminds me, I have to go back and check my square since I’ve upgraded to servos and all the squaring stuff got messed with the upgrade.
This is something that would be of value if you wanted to implement the Acorn linear compensation tables capability. I have seen a single 51" glass scale that could be used to quickly and easily collect error data for calibration measurements for the full XY axis but it seems to be kind of expensive for just a one time use. A .005" accuracy is certainly good enough for any wood project but I work with aluminum and brass on occasion where better accuracy is required, however, it definitely is not needed over the full 48" of travel. I have used the digital caliper method to get a .002-.003" accuracy over a specific 10"x10" area that I have set aside for metal work. It would be nice to boost that to .001" using the Acorn compensation tables over a small area.
I’ve been meaning to do this over the 6x6 inches or so of my metal working vise setup.
One thing I did notice doing the backlash compensation to temper expectations. The Clearpath motors have 800 physical steps per revolution. They are programmed with 2000 virtual steps, but in reality, they only move at the 800 steps per revolution level. This means that the smallest physical step that happens on the rack and pinion, and the smallest correction to backlash, is 1/800/3.2*pi = 0.00127". You can verify this with a good 0.0001" dial indicator or test indicator.
For backlash correction, the result is that I see I can be either 0.0005" over or under correcting on my X axis. My Y axis happens to have backlash that lines up well with the step size, so I am getting a better result there.
It also means that at extremely slow speeds (like you might encounter say using a fixed tool to do engraving on metal), the machine will move in descrete jumps of 0.00127". I can actually see this in cutting samples under a microscope. I can also see the step aliasing in angled lines cut with fixed cutters. All this is irrelevant for woodworking of course, and in fact I think you could tune up to do better than the Avid specs in small regions, probably getting close to 0.001" accuracy, but I don’t think you can do better than that.
You could swap the Servo’s for the 6400 count Servo’s (enhanced model), and that would get minimum physical step size down to 0.00015". This might allow you to get a bit more accurate, and have smoother lines and arcs at a micromaching scale.
For myself, I decided to get a small metal specific mill and set that up with the Centroid recommended 40,000 steps per inch configuration to enable smooth micro engraving on small parts. Next step is to actually build the control box, probably canabilizing my old Avid NEMA23 box for parts where I can.
I expect I’ll still used the Avid for larger alunimum and brass machining where the finish and small scale machining doesn’t matter so much.
Cheers,
Eric
You know, I should have known that because when I was reading the manual on the motors I noticed that 800 step resolution, but then when I saw the 2000 in the CNC12 software, I just went with that. This would explain why the linearity is so rough on the X and Y, and why Z is a lot better (it has 2.54 rotations per inch due to the ballscrew).
I didn’t see anything in CNC12 about this, so I’m assuming the clearpath is actually setup to receive 2000 steps per rotation, but then it just counts the extra steps in the controller, and when it goes over a 1.27mil increment, then it performs an actual step?
I’m a little dissapointed now with the resolution compared to my old steppers.
Thanks for pointing that out.
Yeah, I noticed it when I tried to do some small scale metal engraving and saw I was not getting as good a result as I used to get with steppers.
Of course, the steppers don’t step cleanly at microsteps with large forces, but I’m dealing with 0.005" deep engravings here, so the force is pretty minimal.
The servo’s are programmed to 2000 counts per rev, and the hardware ticks over whenever enough revs pass it must take a step. I think that also means with things set to 2000 counts, there is some additional jitter, because 2.5 counts isn’t a thing. You will get something like a step after 3 counts, then after 2 counts, then 3. I was considering shifting the servos to be programmed to count 3200 steps instead to at least avoid that.
When everything is moving fast enough, you won’t see this, because momentum takes over, but at very low speeds the RAS can’t adapt and keep things moving (I think the lower limit here is about 4.6 IPM, so we are talking very slow here).
Or you shell out the $$ to get the enhanced servo’s with 6400 counts per rev…
I just went out and looked at it more closely and even with my not so expensive 0.0005" per line dial indicator if I set the pendant to 0.001 and click through it for a while I can see that the step size is slightly over 0.001, and every 3 or 4 steps it takes an extra “click” on the pendant to get it to step.
I agree, for cutting wood this is no big deal, but I hate to pass up a chance for more optimization. I might not be able to sleep now until I get the higher resolution motors on X and Y 
Do you know of any downsides of the high res motors (obviously cost would be one), like top speed or ability to hold postion? I would think that since both models have the same 12800 resolution encoder and probably the same driver (but different firmware settings?) that they would be about the same in most other respects. I wouldn’t thing the motor itself would be any different, would it?
I think there probably is no benefit for me though, because I can definitely see the motor hunting for the final position for at least a second after making a discrete step, so as you say, it never really gets to a “final” position as its moving at any reasonable speed.
So how did you determine the 4.6 ipm value, just cut an arc or circle at different speeds and look at the smoothness?
That is the reason I got enhanced motors three years ago when I first put the system together. You know you can go into the motor software and set different RAS values.
Yeah, if you are doing something very slow, you don’t see as much overshoot hunting, but at speed you can start to see it. That said, I’ve succesfully done things like bored holes in brass with a helical path that ended up within about 0.002" or so tollerance as measured by go/no-go pins, so for some use cases it works out. In real tool paths you are not stopping much, you keep moving.
As far as I know there are no down sides for the enhanced servo. Same motor, same encoder, just better station keeping in the software and maybe some of the driver circuitry is different.
Your speed limit goes down in CNC12 a bit if you have a higher step count, but I’m not sure that would be limiting enough.
I did notice something odd when I switched the servos to 6400 counts / rev as an experiment to see if it would smooth anything out. The MPG ended up behind the movement of the machine at low speeds, and would take a second or two to catch up and reach the rest position after the machine was actually stopped.
I get the 4.6 imp value two ways. One I did a series of straight line cuts with a fied bit tool mounted on a lathe tool post on the Z axis. I did these spaced about 0.4mm apart at speeds increasing from 10mm/min and up by 10mm/min increments. Looking under a microscope I can see when the cuts transition to a smooth motion pretty clearly. On the other hand, at the point the microscopic steps disappear, ringing of other kinds appears, and the cuts start to show features on the mm scale that leave them less clean that I would want. Since these occur at the same spot, I think it might be down to various microscopic discontinuities in the mechanical system. These are very small, probably less than 0.0001", but visible to the eye as changes in reflectance in the lines being cut.
You can also get to the same number with some simple math. The RAS setting on the regular servo is 16ms, which is the highest you can set it on this model. (You can also choose 0ms and 4ms, but that would make things worse!). At 16ms for RAS to induce smooth motion, you have to have at least 1000/16 steps per second. Given the step size of 0.001227", the rest is just multiplication. 1000/16 * 0.001227 * 60 = 4.6 IPM or 116.9 mm/min.
I have gone back and forth on getting the enhanced servos after discovering this to be honest.
Right now I decided that for the small scale use case I have it’s better to get a small and much more ridgid mill for the engraving I’m interested in doing, and use the Avid for my wood parts, and larger scale metal work.
The tricky part now is finding room to set the new tool up in my shop! I was really hoping to get by with just the one tool.
Cheers,
Eric
Hi Jim. I just got my Ex Controller upgrade completed and I am trying to calibrate the linearity. How did you setup your caliper to make Z axis measurements? No problems for X and Y as you show in the video but not much to clamp onto for Z.
Theres a couple ways you can do that.
I have a spoilboard with 0.75" holes in it, so I can drop the long end of the caliper down into one of those holes and then clamp the bottom jaw of the caliper to the side of a block of wood that’s clamped to the spoilboard, and the top jaw to the side of my dust colletor boot, or just come in under or on top of the jaw with that boot frame and do a lift or push down (whichever is easier) to get the movement.
The other thing I’ve done is just chuck up a good dial indicator in a collet (mine fites a 3/8" one just right), or use a holder and point the indicator up. I only get a little over 1" of travel that way, but that indicator is more accurate than my caliper, so its probably almost as good as the caliper over 8" of travel.
Thanks, Jim. I have pretty much decided to use the dial indicator.