I am trying to improve the resolution of my Shapeoko 3 in order to minimize the stair step I see in aluminum on faces that are not parallel to either X or Y. I have thought about it before and it was brought up in another thread that the first option may be to swap out my steppers for 0.9 degree rather than 1.8. My issue is that I do not know enough about the electrical considerations when picking a stepper. I know that the existing steppers are 1.8 degree 125 oz-in NEMA 23’s and that is about it. Here are my questions:
Do we know the max amperage of the controller? I vaguely remember it being 2A
Do I have to worry about inductance or voltage specs on a stepper?
I have looked at this stepper before:
I didn’t take the time to determine if it was the right choice then. I would appreciate it if someone would take the time to explain the electrical specs to me. I have a slightly better than a novice level of understanding when it comes to this kind of thing.
Here’s the info I was able to dig up on the base stepper from various posts by Will, links to the old Shapeoko forums etc.
It’s nominally
2 Amps
9 kg cm
4 mH
1.3 Ohms
As compared to the stepperonline alternate you linked
1.26 Nm (12.6 kg cm)
1.16 Ohms
4.2 mH
and it’s rated up to 2.8Amps
It’s probably worth a go as a 0.9 degree alternate, the resistance, inductance and torque are all pretty close and there’s nowhere near as much choice in 0.9 degree motors as 1.8 degree.
I asked the manufacturer about the Nomad 3 motors and they were rated for 1.67A, so I’d expect ~1.3A from the Carbide Motion board (same board is used for the SO3 I believe).
Not really, no. Fancier drivers like the Trinamic ones will do fancy stuff with the back EMF but the drivers Carbide 3D uses are simple beasts and don’t care about such non-worldly things.
Have you tried measuring it and confirming that the resolution of the stair steps are in the same ballpark as the resolution of the stepper motors? That should be a decent sanity test.
The spacing is extremely regular, and I counted 31 marks for a total length of 45mm, so each one is around 1.45mm
This 45mm segment is at angle of 1°, so at the end of that 45mm segment, the displacement in the other axis is 45mm * sin(1°) = 0.7853mm. And 0.7853mm divided by 0.025mm is…31!
Basically we’re seeing the quantization steps.
Interestingly, 0.025mm is the Shapeoko’s resolution with 8x microstepping. That means the quantization here is microstep quantization.
That means that another way of improving things here could be to increase the microstep resolution.
I think I could actually test that on my Nomad (I can reconfigure my microstepping configuration at will)…
Microsteps are not the same as full steps, and even if the step size is the same for two motors, one which is 0.9 degree, the other 1.8, the former should have more torque for holding on each step, and better positional accuracy.
There seems to be a point of diminishing returns for micro-stepping, since the positional accuracy is somewhat reduced as the step size increases — was experimenting w/ that on my Ordbot when I was first setting it up — also, there seems to be a weird interplay w/ torque, since the step positions are closer together as micro-stepping gets finer it’s easier for the machine to gain/lose a step since less force/energy is needed to cause the shift in position.
I agree that there’s definitely a limit for microstepping, the incremental torque dropoff can really kill it.
But Julien’s example, to me, showed that the 8x microstepping is still working pretty okay on the Shapeoko, so maybe pushing it to 16x would still work.
There’s also the prospect of closed-loop steppers. With a ~$20 16-bit encoder from CUI, a decent closed-loop driver should be able to hold the position to within ~1/6554 of a revolution, which would be 0.006mm on a Shapeoko, a ~4x increase in resolution.
If only the Carbide Motion board exposed step/dir outputs. I wonder if GRBL could be hacked to multiplex step/dir outputs onto one of the exposed pins…
The battle is over, and the acrylic won
It will be quite interesting to see if “just” upgrading to higher resolution steppers helps significantly (I have a feeling that it won’t suffice and milling a “perfect” curve would also require moving to a 32bit controller)
If perfect is your goal, you’ll need to go even further. Here’s a guy who has a LinuxCNC machine. He still had issues as shown in his first cut. He resolved it by going closed-loop with linear scales.
I’m hoping closed-loop with a rotary encoder is enough…
I will probably be getting these 0.9 degree steppers next month. I will update here with how it goes. It is definitely the cheapest and lowest effort of the options I see. I will only be replacing X and Y with these. I will be doing some test objects now so I can compare to the same objects after. I will be going over all the V-wheels and belts prior to cutting anything. If just a drop in new stepper doesn’t provide any real improvement, I will look at messing with the microstepping on these 0.9 degree steppers.
