Why not use steppers for the belts?

Hi :slight_smile:

Just asking out of curiosity, why not use steppers on the belts too? those might have enough torque to not need a gearbox, and you can skip the encoder and count steps instead, I understand it is currently magnetic encoder so we don’t have an absolute encoder anyway.

resolution wise you can have 200*32=6400 micro-steps per rotation which is way more then enough, maybe you can even use less ,micro-stepping to get more torque

Is this something about needing the belt motors be complicit sometimes?
Or maybe the torque is too low?

I also recon it might be cheaper and simplify the assembly

Really interested to know,
thanks,
Eldar

Eldar wrote:

Just asking out of curiosity, why not use steppers on the belts too? those might have enough torque to not need a gearbox, and you can skip the encoder and count steps instead, I understand it is currently magnetic encoder so we don¢t have an absolute encoder anyway.

resolution wise you can have 200*32=6400 micro-steps per rotation which is way more then enough, maybe you can even use less ,micro-stepping to get more torque

Is this something about needing the belt motors be complicit sometimes?
Or maybe the torque is too low?

lots of reasons, among them

wihtout an encoder, you would not know if steps were skipped

if you used steppers, you need to know how much belt is on the spool, and what
tension was used when it was spooled, because the more layers of belt on the
spool (and the less tension on the belt while it was spooled), the bigger the
effective diameter of the spool, and so the more the belt will be pulled for the
same amount of rotation.

with the current setup, the encoder is not on the motor, it’s on a sprocket that
the teeth of the belt rotate, so you measure the belt movement, not the motor
input. This means that the diameter of the spool doesn’t matter.

the current encoder is able to report 80 steps/mm. if you used a standard 1.8
degree stepper (200 steps/rev) with a 20 tooth gear (fairly typical), you only
have 5 full steps per mm of belt movement.

Yes, you could do 16x microstepping to get to 80 microsteps/mm, but these
microsteps will not all be 1/80mm, each step will move the motor more, but you
can’t really know how much more. And the torque rating would be 1/10 of what the
full steps would be (see

) for an overview (just a top hit from a google search)

while you can get steppers that are powerful, they are big and heavy. the
current motors can pull with something like 20-30 kg of force. Doing this with
full steps on a stepper with a 20 tooth pully would require a stepper with 1N-m
of torque (~150 oz-in). a typical nema 17 cube stepper (like you see on low-end
3d printers) has about half of this. If you were microstepping, you would need a
stepper rated at ~10nm, a quick search shows that this would either be a very
long nema 34 (3.4" square) or a cube nema 43 (4.3" square). These are much
bigger and heavier (and more expensive) than the dc motors and encoders.

David Lang

1 Like

Thank you David for this great explanation, I learned a lot!
I though too about some of those problems and I think a lot of them are solvable, I’m thinking about researching this in the future.

Do you, or anyone else, have a recommendation to a 5+ axis stepper ESP based fluidnc board?
I could use it as a start

Thank you,
Eldar

Eldar wrote:

Thank you David for this great explanation, I learned a lot!
I though too about some of those problems and I think a lot of them are solvable, I¢m thinking about researching this in the future.

Do you, or anyone else, have a recommendation to a 5+ axis stepper ESP based fluidnc board?
I could use it as a start

I don’t know about fluidNC supported, the current maslow codebase is a pretty
significant fork of fluidnc. There is work being done to split things so that we
can rebase from the current fluidnc. But at the moment, it would be a lot of
work to try and use a normal fluidnc board with a cable driven robot like the
maslow (I know that klipper has cable driven robot support)

what I would suggest that you do is to start in CAD and create a new version of
the arms that fit your stepper, make them, then do some tests of your arm vs the
stock arm. Hang them on a wall with a long digital position sensor, say
something like these
https://www.aliexpress.us/item/2251832737680593.html

measure the accuracy of your stepper version against the standard arm then start
adding weight to both until they fail (either mechanically, or more likely are
unable to move as instructed, skipped steps or can’t lift the load)

I would only start looking for a 5+ axis controller after you have tested the
arm.

David Lang

1 Like

Sounds reasonable enough, I think I will try to make it my finals project at my Uni next year
Thank you,
Eldar

1 Like