I built a plotter based off of the acro system. Im using some standard stepper drivers and grbl to drive the machine. I use LaserGrbl to generate the gcode and ugs to send it. But when drawing circles or any arcs it stutters and jerks around making the arcs look terrible and I have no idea whats going on. It also has trouble with large arcs you can hear it in the video. Here is a picture of one of the better circles it made
What specific drivers and what voltage are you supplying them with? And yeah, a sample of G-code that directly correlates to a video clip is helpful.
Im not actually sure. Im still trying to figure out how all this works. Here is the gcode for the small circles(20_iter 01 Gcode). and the large drawing in the video(Tribal-Kitten-3 mod gcode)
Hmm. Same as in my laser. The TB6600s can be unreliable sometimes, but I haven't seen this out of them before. Oddly, the circle code is all G3s and the video code is all G1s. Usually it'd be one or the other being a problem, not both. Suggests it might be mechanical. Looking at the first images again, all of the wobbles seem to be near inflection points in Y- +X heading left-down toward -Y, and -X heading up-right toward +Y. Once it turns the corner, it smooths out again. Might be something wrong with your Y axis.
right, thats what I thought! But its not. The wobbles are incredibly repeatable, I can stop the gcode and go back and do the same thing and it follows the exact same path. Its really weird. I just loaded a fresh version of grbl and its still having the same problems. Its almost as if its the gcode. I was looking at N_ARC_CORRECTION and changed the numbers with no luck. Know of anything else in the config that could do it?
Hmm, interesting. Yeah arc correction does almost nothing most of the time. I guess this begs the usual obvious question though; what happens if you swap the motor and signal wires between the X and Y drivers?
it does the same thing. just tried it. I had grbl Arduino shield before and had the same issue. I just used a different sender too and same problem.
Update: Changing the Junction deviation from 0.01 to 0.1 made a huge difference but I don't exactly know why and now its cornering really fast.
This page explains all those config options: gnea/grbl (then read the rest of the wiki, since it's even more important for DIY machines) Edit: Just saw this- Did it do the exact same thing at the same points in the same direction, or do the wobbles rotate 90 degrees? If the G-code looks good in the simulation window and the wobbles are only ever in Y regardless of driver used, it has to be the controller- either grbl's setup or the physical board itself. What's grbl running on?
I will mention, changing Arc parameters have "never" been needed - its weird that it did change something for you, but it might be indirect (something still "wrong" but this change hides it)
\ Right, It looked like the wobbles rotated but its hard to say because I couldn't run a full drawing due to how I have the machine set up. I can't swap the motor and signal wires to both y axis motors, just one(unless I rewire it). So I just ran it off the one motor and it still had a the issue but I didn't run it long enough to see if it rotated. Ill do that today. I have grbl running on an arduino uno. I don't think its just the Y axis though, I can hear the x axis making the weird noises that accompany the wobbles as well. I will say this isnt the first driver I've used and the problems persist.
Ya, as far as I can tell its not mechanical. Everything tight. The noise is like the driver frequency or something. like the typical cnc stepper motor sound but the pitch changes in a weird way. Its more chaotic and you can not only see the wobbles but you can hear it in the frequency change.
Pretty sure its the tb6600. Here is the listing (https://www.amazon.com/Stepper-Cont...8627209&sprefix=stepper+motor,aps,168&sr=8-15). But I also had the same issue with the A4988 driver. Its probably something simple and I just have no idea. Thanks for your help so far! Its weird that changing the junction deviation made a difference.
I have the same pack because it consistently got the best reviews. The actual chip is a Toshiba S109AFTG, according to the reviews (I assume he popped the cover, I never bothered) and the consensus seems to be that they're only really good up to about 2A. But that's fine for these kinds of plotter projects. I suppose you could try turning the current down, just in case, mine are running at either 1A or 1.2A, I forget, but that's been plenty good for NEMA 17s. Did you try reflashing grbl already? Sometimes that helps with these weird issues.
Ya I reflashed it with no mods to the firmware and same issue. Right now the way I have it set up, the large arcs look great but small circles and stuff turn out terrible. Here is a photo of a small circle less then 5mm in diameter. Out of all the circles in the drawing, these were the only ones to end up like this, and all of the circles that size look exactly like that. And its not the gcode. Setting the junction deviation to 0.1 I got this result (20200828_115850.jpg) . so clearly not perfect but even that is better then before
I wonder if it's speed or acceleration settings? You have a very high top speed and a very low acceleration. You could try turning both the acceleration and max speed down really low, just to see if the machine's trying to "jiggle" too much. Like, 200 for 100/101 and 20 for 120/121? It doesn't sound like it to me, personally, but it's a simple enough test. Honestly between the lines that don't meet up in the first set of pics and the wobbly small shapes, I'm also wondering if your actual plotter head/Z axis is sufficiently stiff. Those two things are where that shows up the most. Not sure that explains the "impossible wobbles" though. Those are the weirdest part, I've never seen it before. I feel like someone must have had it at some point but what exact search terms you'd use to find it I'm not sure.
Ya I bet someone else has dealt with it before. I changed the acceleration and speed and im drawing good circles now kinda. Its super jerky still on some arcs even when going really slow. In the video you can hear it slowing and speeding up. is that normal like that? Update! might be narrowing it down. I changed the acceleration to 500 and now its supper jerky moving though each arc segment here is a video, this cannot be normal. (arc jerk.mp4)
Specs of the PC up to task to send the short moves without delay? Try OpenBuilds CONTROL (OpenBuilds Software - FREE Software for CNC Control: OpenBuilds CONTROL and OpenBuilds CAM) ?
Thats what I was wondering. that was my next change. Im using a 2015 macbook air. Ive used the openbuilds cam sofware and sent the files using it with the same result. You can really see the issue when you turn up the acceleration
There's a distinct chance that's the problem. A non-Windows laptop that's now a few years old, that's the trifecta of less-than-stellar grbl performance. If you have a Windows desktop, try with that. Otherwise, whatever the next best thing is. USB on mobile devices is not the most reliable for real-time usage.
Nope... that wasn't it either. I used my surface with no luck. The reason it was jerking around was actually just the junction deviation was set to 0 for some reason. So far the only thing to make a difference is that number. the higher I set it the less "wobble" I get. Ive switch boards, divers, loaded fresh grbl, changed usb cables, used a different computer, used different software. It seems to be like the issue is with how grbl is interpreting the gcode, but that seems unlikely because no one else seems to be having this issue. I guess im just gonna have to play around with it. Thank you guys for your help! really appreciated!
Oh, sh*t, really? I have the same pack but it's on a router not printer. I may have a problem now since the motors are rated 3A and I may need it (when the spindle arrives from China ). It may be the reason I had the noise at high stepper revolution.
Yeah I don't think I'd use these for NEMA 23 or above, but if I were to do another NEMA 17 project I'd buy them again in a heartbeat. $50 for drivers, Elegoo Uno and a terminal shield can't be beat for low-power projects. For NEMA 23+... Well, I'd probably go closed-loop because you can get a 2Nm/280ozin motor/driver/cables kit for around $90 - $100 these days. Four motors + drivers + controller + cables, whether that's DM542s or a BlackBox, are still gonna be around the $350 region. Closed-loop is a no-brainer to me at this point. It's quieter, higher-power, cooler-running, can't lose steps... And around the same price. If you limit the acceleration, any plunges, and overall MRR, you can probably get away with setting them to 2A, but in that situation I'd probably just treat them as a temporary thing to get the machine functioning and prepare to upgrade six months down the line. Which if you have plans for the machine to make money, should be justifiable by that point anyway, in theory. I'm also getting ready for a spindle purchase... Man, those things are hard to buy. Half the time it seems like nobody gets a good one. I've settled on G-Penny, I think. $100 more thanks to shipping and I have to import it, but I haven't heard a bad word about them so far.
Closed loop is, well, literally what the name implies. It closes the feedback loop at the driver (much easier than at the controller), which technically makes them a type of servo motor. The motors come with encoders built in, and there are two cables going from motor to driver- one power, one encoder signal. If the motor loses a certain number of steps (not sure how many, I guess more than two?)- the encoder says it's moved without being commanded (like if you twist the shaft really hard) or hasn't moved when it has been commanded (like if you turn your accelerations up too high)- the driver will throw an error (and has an error alarm signal that you can send the controller if you want) and become unresponsive until you power cycle it. Which means that if the machine crashes or loses steps, the driver knows, and cuts power to the motor, even if it's still receiving step signals from the controller. This also means that the driver only sends a large amount of current if it actually needs to- most of the time, it'll use far less current than it's set to, until the encoder tells it that the motor's met resistance (because it steps, but the encoder doesn't, so it boosts the power). Which allows it to use significantly more current for short periods of time, so the peak torque for a given physical size is generally higher than open loop steppers. It's not quite as good in terms of accuracy as glass scales on the machine reporting back to the controller, but it's a huge amount less processing as well. You can use them with grbl without changing anything, no need for LinuxCNC or Mach 3 plugins. I think they're also always 32-bit digital drivers, so just as smooth and quiet as the expensive "standard" drivers. They're also quiet because they only need to bother sending current to lock the motor if something's trying to move it, so there isn't a bunch of wasted power on standing still and overheating everything. Basically, for anything over NEMA 17, I don't use anything else these days. They generally look something like this: (That looks like a 12Nm/1700oz-in NEMA 34 kit to me, but they come in NEMA 23 and 24 as well) You'll usually see them advertised either as "closed loop steppers" or "hybrid steppers" or sometimes "hybrid servos" - since they are technically servos, being closed-loop. But not what machine builders think of as servos, which are way higher performance and three times the price.
