Workbee - precision

Thanks Benjamin, it was particularly useful to comment out the various attempts at calibration. According to your config.g file you had a perfect square with;

Square perfect: M92 X53.997 Y53.744 Z381

and you are getting a near perfect square now with;

M92 X53.611 Y53.475 Z400

can you confirm that both squares were 100mm?
Alex.
PS haven't had a response from Ryan yet.

 

It shouldn't make any difference Benjamin (the default microstepping should be 16) but try moving the line;

M350 X16 Y16 Z16 I1 ; Configure microstepping with interpolation

between the M584 and M92 lines.

Alex.

 

Struggling to come up with a reason for 1000mm to be accurate, 400mm to be high, and 100mm to be low. Seems impossible from either a motion planning point of view or a machine kinematics point of view.

What happens if you turn your microstepping down to 4x with zero interpolation? Just wondering if that particular drive doesn't like it, for some weird reason and it's having a non-linear effect.

Edit: bear in mind I'm completely unfamiliar with the Duet controller system, I'm just approaching it from a general machine-integrator point of view. Others may (likely will) have better ideas if they're more familiar with the system.

 

Not sure how many belt drive machines with the Duet controller there are out there - most people on here seem to have the screw drive, but Benjamin has the 1500 x 1500. Screw drive use x8 microstepping. Waiting for a response from Ryan to see if he has any clues. Like you, Rob, I can't make sense of the non-lineararity of this.
Alex.

 

No, you'd just change your calibration for that axis. Your steps per mm would be 4x fewer, more or less.

You can set your microstepping to whatever you like. I personally think that 8x is the best sweet spot for a balance between precision and torque- as you increase microstepping (ie. the amount of positions it tries to hold between the true physical magnetic motor steps, usually 200, by "feathering" the current between the two phases) your torque goes down because it has to split the current and can't pile it all into one position. After 8x microstepping the torque curve pretty much falls off a cliff. And often, you're not actually getting the true number of positions it's set to, either; it struggles to precisely hold a commanded in-between position after about 4x microstepping.

Stepper motors aren't really all that great for what we do, they're just really cheap and barely good enough! Anyway, I try to use either full stepping, 200 steps/rev, or 8x microstepping, 1600 steps/rev, on basically all projects.

 

Following as I am having identical issues with the Ooznest 1500x1500 belt driven Cnc. Benjamin, you are not alone if that helps

 

The belt should be OK - brand new machine, but worth checking as it won't take long.
Default microstepping for the Duet is 16x - intended for the light loads of 3d printing - but you are getting inconsistent results when not cutting anything. It certainly won't hurt to try changing the microstepping to 8x, but be aware it will change your steps/mm calibration.
Alex.

 

One thing I have noticed is that the whole z axis beam flexes forwards and up quite easily if you go too deep too fast (everything is tight). Whilst testing I’m only cutting by 1mm increments into pine.

 

Is it the Z C-beam that's flexing, the X gantry that's rotating around the X C-beam, or the X C-beam that's flexing?
What tool, feed rate and spindle speed are you using?
If possible post a video shot from one side showing the problem (but we can't tell much from a shaky video)
Alex.

 

Hi @Benjamin Vg, have you tried swapping motors? If Y axis is OK leave those where they are, but try swapping the Z and X motors.
The only other thing I can think of is reducing the microstepping even further.
Still waiting for a response from Ryan - I guess we won't hear from him until tomorrow.
Alex.

 

An easy way of swapping drivers (for a 4-way test: 2 motors, 2 drivers) without having to rewire anything except plugging motors is just to rotate all your code commands; use G1 Z-400 instead of G1 X400, etc. As long as your limit switches are installed and working, there should be very little danger to this method, and you then know whether everything's good or not. If you have a nice pluggable setup instead of terminal blocks, this might not be as relevant.

I'd also try turning microstepping off completely on the x axis. It's the only way to know that you're getting full torque from the motor and eliminate that variable, at least for the time being. This process is always about eliminating variables, one by one.

 

Here are the videos of the Z/X flexing which is probably contributing to inaccuracy. The X beam itself also looks to be twisting. I've checked all screws for tightness etc. @Benjamin, it could be something else for you to watch out for when you put your machine under load.

 

I'm afraid that is the limitation of C-beam - 1500mm is probably too long for the X axis in particular. If you search the forum you will find different ways of stiffening the C-beam, but otherwise you will have to restrict yourself to light cuts. It would be a good idea to always do a roughing cut first, leaving say 0.2 to 0.5 mm of stock to take off in a finishing pass.
Alex.

 

A belt-driven 1500mm machine is a near-zero-force laser platform, not a milling platform, but all the flex in the world doesn't adequately explain these results; if the motors aren't losing steps and the spindle mount is stationary, the machine frame would absorb force and tense up at the beginning of a cut, then release the torsion at the end of the cut and still more or less come to the same spot. Anything other than a straight line would be inaccurate, though, because it wouldn't be able to "catch up" in the correct plane. It also can't deflect past the stopping point unless there's some really heavy climb cutting going on, and that seems unlikely here.

Lost steps also seems a little dubious because we're talking hundreds of steps difference, unless the lost steps are based on jerk or acceleration, perhaps? With such a low acceleration, there are hundreds of mm for it to come up to top speed, and at certain points (resonance, perhaps) it's losing steps.

What motors, drivers, and current settings are on this thing? I've been assuming it's NEMA 23s with discrete drivers and around 2.5A, but if it's not, that could be part of the problem. The Duet appears to have on-board drivers, only good up to a maximum of 2A with adequate active cooling. I see 2400mA in the config, but I'm not sure if it's actually putting that out.

Try no microstepping, like I said before, full steps. Bring your top speed down to 500mm/min for testing, probably around 1500mm/min once you're going, because 10k isn't gonna happen without a CO2 tube laser anyway. Also try setting the acceleration to 1000mms^-1 or as high as it'll go without stalling the motors out, see if we can force it to error within a specific section of motion. At that point, 100mm should be way out, but 400mm and 1000mm should agree.

The only thing that makes me dubious of the missed steps hypothesis is that they sound awful, you'd be able to hear if it was grinding away without going anywhere, even for just a couple steps at a time. But it's worth chasing down the rabbit hole just in case.

 

The steppers are; 230HS280AW - 175oz -2.80A.
Judging by the torque I get on my (leadscrew) workbee I guess the drivers are supplying 2.4A (ripped 18mm mdf spoiler board apart when I accidentally collided with a clamp).
I was wrong about the microstepping on my leadscrew machine - it is 16x.
Alex.

 

Very true - but it looks (from my set up) as though the controller/motor combination are supplying the torque - whether its getting through to the X and Y carriages is another question. I limited myself to the workbee 1010 because I wasn't sure that belts would transmit the torque adequately/consistently. I am also (with the same motor/controller combination) seeing no signs of missing steps even with some heavy cuts at times.
Alex.

 

It is not clear to us what you are doing with each test Benjamin. Could you;
Without cutting;
A); tell the machine to move 1000 mm and tell us how far it really moved
B); tell the machine to move 500 mm and tell us how far it really moved
C); tell the machine to move 100 mm and tell us how far it really moved

Then repeat those tests but cutting into the surface of the board 0.2 mm (no deeper)

The results from those 6 tests might tell us where the problem is.
Alex.

 

Does that mean we have been sold something that is not fit for purpose as I (can't speak on benjamins behalf) purchased my machine as a CNC router kit?

 

I think Rob's comment comes under the heading of personal opinion - if you look at his build (Grizzly G0758 Benchtop Mill Conversion) you will see that he goes in for really sturdy stuff! Hobby machines built using aluminium extrusions are always going to have limitations which we have to accept unless we can afford tens of thousands (dollars or pounds).
Alex.

 

I wonder if that 100.3mm is statistical noise or not, because that's the direction I had expected it to go with more torque capacity.

The reduction in max speed was intended to be done after setting the microstepping down and whilst also increasing the acceleration (150mm/s/s is pretty low).

Those are somewhat promising numbers, though! I think it can get there in the end.

I was wondering about that earlier this morning. It's probably a grey area. I would say that 1500mm is much too long a span to rely on the positioning reliability of GT2 belt under relatively high, fluctuating loads. A lightweight laser head, 3D print head, maybe a PCB spindle, no problem. I'm sure you can mill foam, acrylic and other nice, homogeneous polymers, and some lower-density homogeneous composites like MDF just fine. Can you make new aluminum plates for your machine or a nice oak cabinet door? I find that unlikely. Could you legally prove it's unfit for purpose? I doubt it. Would I buy one as a milling platform? Not a snowball's chance in hell. It's basically an ACRO as far as I'm concerned.

And yeah, Alex is right; I generally approach machines from a "higher standard" than extrusion machines are typically capable of, and it does cost a bit more (though less than you'd think, at least before import mills got expensive). I'm not opposed to extrusion machines in any way though, I've built one already (Portable Diode Laser Cutter) and have plans to build at least one more at some point; they're great at what they do and can allow a pretty nice entry price point to the hobby/industry. But yeah, as Alex and Peter are aware, there may be a little bit of opinion-based hyperbole involved. I still wouldn't run a mill spindle without screws though.

 

bit disappointed really as I sold my smaller Chinese machine that worked perfectly, however, I needed a bigger work area so opted for the Ooznest 1500x1500 - just feel like I have wasted my money and am going to have to look elsewhere for another machine when the one I have isn't 2 months old.

 

As an open platform, converting it to screws shouldn't be enormously expensive or complex. Not great for something sold as ready-to-run, I know, but I've seen a lot of what I would consider... Elastic morality and less-than-complete explanations. Looking at their description page for the WorkBee, they don't technically lie about its capabilities, but they also don't really break it down clearly what's reasonable to expect out of a given configuration. You really have to go in knowing and understanding the components and machine-building and the history of the open hardware machine designs to be able to accurately parse that page. I don't like it at all, but it seems to be de rigeur to grow the market of hobbyist machinists... Or at least, people who buy machines. Feels a bit like the 90s home gym equipment craze.

Looks like the WorkBee's specifically designed to be usable in either screw or belt configuration, so a bit of chopping down (8mm screws aren't usable over 1000mm, too much whip) the extrusion as they say on that page, or using a different screw is in order. Leadscrews are available in larger diameters from places like McMaster, I'm sure there are EU equivalents, or you could try ballscrews (a 1500mm ballscrew can be had for about fifty quid or so if you look around enough in the usual places). You're not locked into anything if you don't want to be! The whole point of this is to empower people to make their own stuff, learn about the technology, and grow their abilities. No need to buy a whole new product.

Thinking about this some more... Skip the acceleration test for one second.... Keeping it at current settings but sticking with 1x/full stepping, if you calibrate it to 1000mm, what are your 100mm and 500mm marks? If that 100.3 number is consistent, you should see 99.9, 49.9-50.0, and 1000.0.

If so, it may be that your microstepping is either a) losing steps at resonant nodes, b) drawing insufficient current to properly "click over" to the next microstep and your actual motor poles are having an outsized effect, c) trying to reach too high a speed (probably not this one, based on your new testing data), d) drawing insufficient current to actually move the weight of the router (unlikely, but not impossible, if there's an issue with the Duet), e) some other microstep-based issue I haven't thought of yet.

 

Well, I am pulling my hair out like many. I tried using a sharpie to do the Hello World and it was a disaster so I thought lets try something simple like a 50mm diameter circle. Wow.. did that show how out something is. I have checked my connectors and verified they are tight and that each y axis is moving exactly the same and that there was no binding on the X axis. When I told the X axis to move 100 mm it actually move 98.425mm. I am using the standard 1040mm Tr8*8-2p (4 starts) Lead Screws. I have literally torn this thing apart and put it back together and it still prints the same. Does anyone have any ideas? Here is the circle that was created in
OpenBuilds CAM - GCODE Generator.