Auto-Calibration of CNC Router

[kungfoo]

Hello all,

I'm quite new to machining, and even newer to CNC, and this is my very first post to PM, so please pardon me if I'm repeating a topic.

I'm in the early stages of producing a design concept for building a home CNC router. This design concept calls for cheap ball-screws and the ability to run auto-cal cycles of 10,000 or so calibration points. For example, if I use one ball screw on either side of the gantry, then they will not move at the same rate and so the gantry would never really be square.

Ideally, I'd like to find some physical principle or instrument that I could use to do (preferably sub) micrometer calibrations in an automated fashion over a distance of 1 meter or so. I'm sure I could drop $150k or so and get a laser interferometer that could do the trick, but I'd rather spend a bit less and maybe build it myself.

Anyone have any recommendations?

Thanks,
Harrison

PS: Yes, I am fully aware of the fact that my home-made little CNC router is NOT going to hold sub-micrometer accuracy and that I'm talking about doing static calibrations WELL BELOW the amount of dynamic inaccuracy I'm going to encounter from silly little things like operating the motor and...well...this still sounds like fun.

 

[DDoug]

Hello all,

I'm quite new to machining, and even newer to CNC, and this is my very first post to PM, so please pardon me if I'm repeating a topic.

I'm in the early stages of producing a design concept for building a home CNC router. This design concept calls for cheap ball-screws and the ability to run auto-cal cycles of 10,000 or so calibration points. For example, if I use one ball screw on either side of the gantry, then they will not move at the same rate and so the gantry would never really be square.

Ideally, I'd like to find some physical principle or instrument that I could use to do (preferably sub) micrometer calibrations in an automated fashion over a distance of 1 meter or so. I'm sure I could drop $150k or so and get a laser interferometer that could do the trick, but I'd rather spend a bit less and maybe build it myself.

Anyone have any recommendations?

Thanks,
Harrison

PS: Yes, I am fully aware of the fact that my home-made little CNC router is NOT going to hold sub-micrometer accuracy and that I'm talking about doing static calibrations WELL BELOW the amount of dynamic inaccuracy I'm going to encounter from silly little things like operating the motor and...well...this still sounds like fun.

Try asking over at the zone please
www.cnc-zone.com

 

[kungfoo]

Try asking over at the zone please
www.cnc-zone.com

Forgive the confusion -- isn't that a private company that sells services?

 

[DDoug]

Forgive the confusion -- isn't that a private company that sells services?

it's a forum...much like this one, but more for your project.

 

[EyeEmTrash]

Forgive the confusion -- isn't that a private company that sells services?

he meant cnczone.com....no hyphen

 

[memphisjed]

Why won’t screws be turning at same rate? If you fell for servo sales hype move encoder to gantry so it is measuring real position not motor shaft to coupler to reduction to coupler to screw to nut to gantry stack.

 

[kungfoo]

Why won’t screws be turning at same rate? If you fell for servo sales hype move encoder to gantry so it is measuring real position not motor shaft to coupler to reduction to coupler to screw to nut to gantry stack.

Jed,

Even C1 systems have representative errors that would be quite high on the scale that I'd "want" to say nothing of the fluctuations, and I'm guessing I can't afford C1. My initial thought is to stack up all the inaccuracies in all of the systems and measure both the real-world position as well as real-world backlash in the system some number of thousands of times over the entire range of travel giving me a ground-truth for the 0th order behavior of the system as a function of commanded position.

Also...it sounds like a fun project.

Thanks,
Harrison

 

[Scruffy887]

Have you ever heard the saying "you don't know what you don't know". For starters have you ever heard the term "backlash" used on cheap rolled ball screws? You really need dual glass scales controlled by a turbo encabulator PLC.

 

[Orange Vise]

I'm sure I could drop $150k or so and get a laser interferometer that could do the trick, but I'd rather spend a bit less and maybe build it myself.

They're available for rental.

For example, if I use one ball screw on either side of the gantry, then they will not move at the same rate and so the gantry would never really be square.

This is not an uncommon issue even on commercial machines in the upper 5-figure and low 6-figure price range. Our Multicam Apex 3R uses dual rack-and-pinion drives on the X-axis. When powered off, it takes no effort at all to push/backdrive one side of the gantry, bumping it out of square by a significant amount, i.e. 1" over 72". The machine requires re-homing each time it's powered down and back up again (including e-stop conditions), and it independently finds home on both sides using inductive proximity switches.

 

[kungfoo]

They're available for rental.


This is not an uncommon issue even on commercial machines in the upper 5-figure and low 6-figure price range. <snip>

Thanks, orange!

As I plan out this project, I'll probably assume that I can rent one of these devices that can be used to verify whatever I build. Any recommendations on what to read next regarding how they work?

I don't need to commit to a design just yet, but at this stage of development I DO need to validate the concept of building (or buying) my own calibration system.

Thanks again,
Harrison

 

[Scruffy887]

Hello all,



home CNC router. This design concept calls for cheap ball-screws

Above says it all. Cheap rolled ball screws will either have excessive backlash, or worse. You can check with Centroid as they have a CNC controller that can auto square the gantry. But even glass scales will make you go crazy with cheap ball screws. Glass scales will report exactly where each side of the gantry is. But WTF will they do with cheap screw floating backlash? They will compensate and the backlash will float about and you will hear lots of complaints from servos and drives. Gantry goes to X as glass scales direct it to. But gantry does not stop at X because of backlash. So control moves gantry back. Gantry floats about with backlash and you have the hunting starting. FAIL. I think the word "hunting" will be used to describe that condition.
Sorry, but CHEAP HOME ROUTER screams that the structure of the machine will not support your ideas.
OK, remove backlash from cheap ball screws. 3 pound hammer smack on the sweet spot on the ball nut will introduce pre load, backlash removed.