PDA

View Full Version : Calibrating my new machine - "never the same twice" woes



HankMcSpank
22-11-2011, 11:26 PM
I recently purchased a nice little CNC on Ebay (to my eyes at least, it really is well made)...the maker quotes repeatability to 0.05mm - so I'm trying to calibrate it.

The leadscrew pitch is 1.5mm, I'm 1/2 stepping it, so I reckon that's a 'steps per unit' setting of 266.6666 steps per unit (ie 1mm)

Ok, so I whip my dial indicator out - pretty much bang on, backlash is about 0.05mm which I dial out via software - all is good so far.

The main purpose of this machine is to depanel multiple repeated pcbs on one board ...and this is where it's all going 'Pete Tong'.

On my pcb I include some crosshair registration marks of known set distance ...but when I mount this 'etched pcb' onto my machine, and jog between registration marks....my cnc DRO READOUT doesn't marry with the ('known') distance between the registration marks....the pcb is a fairly intricate design with no room for error (it has PCB tracks are near the edge of the board outline)...so I got to get to the bottom of this.

Now bearing in mind I etch my pcbs from a laser 'printed' onto 90gsm tracing paper, I'm thinking this is likely to be an error somewhere between the laser printer itself & the paper stretching while being heated. So I therefore frig my cnc macines 'steps per unit' so that my CNC marries with the registration marks exactly. But then I come back next time & it's out of whack again.

I'm of the opinion that this is now likely to be thermal expansion of my machine - basically my CNC machine sits in an unheated workshop ...if it's Baltic outside then so is my machine - if its a warmish day, then so is my machine. (the cnc machine is small and made of ali)

Now short of fitting heating to control the ambient temperature of the room & keep it constant (that ain't gonna happen), i'm not sure what my options are here? Is there a quick way of calibrating the machine to cater for these possible expansions/contractions?


Edit: The error I'm talking about is over a distance of about 150mm ....probably an error of something like 0.75mm-1.0mm over that distance. Now when I calibrate my machine with the dial indicator...that's only using a distance of about 10mm to calibrate ...so what do folks use to calibrate their machine for larger distances? (or can I safely assume if it's bang on @10mm distance, then it's going to be bang on at 150mm too?!)

Web Goblin
23-11-2011, 06:15 AM
Try to calibrate the machine over the longest distance you can so that any errors are as small as possible. I normally set the machine to zero and run it as far as I can and simply measure with a tape, check the travelled distance and alter it if required and test again. I did my Z axis with a dial guage and steel rule. Good enough for me. I wouldnt have thought that the machine would expand or contract enough to cause any movement. I have checked some of our big machines at work during warm summers and winters when its regularly below zero and the machines are always inside 1mm over a distance of 10mts.
Have you checked over the mechanics to rule out any backlash?

Regards

Ian

HankMcSpank
23-11-2011, 09:16 AM
Hi Ian,

It's *definitely* not backlash - how do I know? Well I make sure not to change direction on the leadscrew when aligning all this up ....ie to set this all up, I initially I shoot past my registration mark, then bring it back (therefore take up any backlash slack) ....then use a 'G0 Y -150' command all travelling in the same direction to move to my other registration mark.

I've since learned about the G51 'scale' command, so I reckon I can use this little procedure

Set my steps per unit to mate with the leadscrew pitch (therefore 1.5mm @ 1/2 stepping = 266.6666666666667 steps per unit)...always leave it on this (I've been adjusting the steps per unit to dial out the pcb registration mark aligment errors)
Align over my first registration mark - zero the DRO.
Use a G0 command to move the Y axis 150mm (this being the distance between my PCB registration marks)...it won't totally align
Now jog the Y axis until it does align
Enter the actual DRO reading (eg 150.7 etc) into a spreadsheet ....the spreadsheet calculates the 'scale' adjustment
enter scale value as a G51 command.
Once finsihed clear the scaling with a G50 command.

Jonathan
23-11-2011, 01:22 PM
Thermal expansion will be negligible. The PCB will change dimensions slightly when you hit it with the iron to transfer the design.

Standard 10mm dial indicator is 0.01mm resolution, so if you calibrate from the full range of that it should be maximum 150/10*0.01=0.15mm out over 150mm. It sounds like your error is more than that which implies it is an error in the screw pitch. What screws are you using, ballscrews/ACME?

Have you got a digital caliper? Standard ones measure up to 150mm so you can use them to check over that distance. The resolution is 0.01mm, but the instructions state the accuracy is worse however it's significantly better than about 0.15mm over 150mm from a dial indicator. Just be careful with either of them that you mount them carefully, if the indicator is not precisely parallel to the axis then the reading will be out.

I take it it does move back to 0,0 precisely?

HankMcSpank
23-11-2011, 04:54 PM
Thermal expansion will be negligible. The PCB will change dimensions slightly when you hit it with the iron to transfer the design.

Standard 10mm dial indicator is 0.01mm resolution, so if you calibrate from the full range of that it should be maximum 150/10*0.01=0.15mm out over 150mm. It sounds like your error is more than that which implies it is an error in the screw pitch. What screws are you using, ballscrews/ACME?

Have you got a digital caliper? Standard ones measure up to 150mm so you can use them to check over that distance. The resolution is 0.01mm, but the instructions state the accuracy is worse however it's significantly better than about 0.15mm over 150mm from a dial indicator. Just be careful with either of them that you mount them carefully, if the indicator is not precisely parallel to the axis then the reading will be out.

I take it it does move back to 0,0 precisely?

I'm not using any iron to transfer...I knock up my pcb design in eagle, print onto tracing paper, expose my presensitized photo board to UV through the tracing paper, develop, etch. So the only potential for error is the laser printer itself & tracing paper stretch (due to the heat from the laser)

Re the hmm screws, not 100% sure(!) ...the maker call them 'zero backlash' screws (yeah, right!) - the advert is here ... http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=230689354532

I've set up all the axii (??!!) with my dial indicator...like I say I was getting .05mm backlash, which I dial out in software - yes it does return to zero.

I do have a digitial caliper, but then need to be sure it was totally square to the plane of travel, else I reckon I could be introducing more errors just using it to measure over such distance (then as you say there's the digital caliper error itself!)