Does anyone have practical experience with interrupting and resuming FINISHING passes?
My homing sensors are repeatable download to my ability to measure (1/100mm with a mechanical dial indicator ).
Still I see visible lines where one pass ends and the segment of the the next day starts.
Even if I invert the direction, so I resume in the same place where the last one started and move the other way with identical chipload, there is a visible line. It’s to fine to measure but can be seen and felt.
I think the best is to start at the beginning of the finishing pass of you need to interrupt it to obtain the best possible result. still I think you would be able to feel and see 1/100mm anyway. if you have any way of realigning the tool against the item using a probe this would be even better but id’s not a guarantee for perfect result. I have never relied on homing in the middle of a job. Whenever I needed to pause I never homed. I just parked the machine and let mach 3 save positioning. after mach 3 restart the positioning is intact perfectly.
@Henrik_Larsen I used an elecric touch plate but because the axis would never stop perfectly instantly, a dial indicator revealed that this was surprisingly inaccurate.
So I upgraded to photointerruptors and designed it, so they would be shaded from surrounding light before being interrupted. This is repeatable down to less then what my 1/100mm dial indicator can show.
Humans can easily feel features of 1/1000mm and see them if the surface is illuminated from the rigt angle.
My stepper drivers do a random jerk when getting powered on. This also is earthquake territorry. So leaving the machine off but undisturbed, is not an option.
Leaving the stepper drivers powered is too unsafe. I already had one chinese stepper driver catch fire once when I though the machine was off but it was stilll maintaining position.
I should have a chinese, digital “1/1000mm” dial indicator somewhere.
(Probably less accurate but faking it in the display.)
With that I should be able to meassure the actual repeatability of my homing sensors. If it shows anything, I can test adding a schmidt trigger to the sensor output and smooth the LED current with a capacitor against minor flickering in intensity due to other loads.
Any idea how to modify the script MACH3 uses to home a single sensor?
(I only find the “ref all axis” script.)
I could reduce the backoff-speed or make it do 3 or 5 meassurements and calculate the average.
Apart from improving the machine, I see 3 options:
- break up my finishing passes not by distinct region but by accuracy,. I’m currently trying that and will see results next weekend.
- make do with less finishing by sandblasting. I want to but that won’t happen this year anymore.
- less finishing on the CNC and flame polishing POM. Dangerous due to POM burning with an invisible flame, releaseing the pure Formaldehyde it is made up of.
The surface details do not allow for manual sanding with sponges, sticks and paper. I simply can’t get in there.
I get it Marcus. The random jerk could be that the steppers is set to microstepping and the off position will be different from the saved position. you could try experimenting with parking position to hit a non micro stepped position.
@Henrik_Larsen Interesting idea. I guess I’d switch it on, place a dial indicator on it, zero the indicator, home, and find out what position exactly that is (and if it stays constant when parked there).
What’s the temperature difference from one day end to start the next?
CTE on aluminum is in the neighborhood of 25um per meter per degC.
The other possibility is to use ball end or barrel end mills for finishing passes. Is the problem greater in the longer dimension of your setup?
Human finger touch sensitivity of a step largely depends on the sharpness and shape of the edge. Depending on the slope and how gradual the radius is, some profile steps cannot be felt.
Radius cornered cutters with sufficiently fine step-overs might fix your problem.
The other potential issue on repeatability is mounting of your dial indicator. Also, gradiations on the dial is only part of the accuracy picture. Contact force, hysteresis, current state of the bearings in the indicator all contribute to the (in)accuracy of the instrument.
An indicator is a good tool, but measuring with it and being confident of the numbers needs calibration/certification to back that up.
Material:
Why do you say Aluminium?
I’m machining Polyoxymethylene=POM-C=Acetal Copolymer.
Home automation says we are in the area of 24-26°C at all times (cool cellar). It does not matter if I continue a week later or home and start the next section right away without switching anything off.
It MAY be that the cutting speed at the edges of 2 sections differs. I noticed that I had forgotten to switch off a slowdown for high chipload sections and the CAM thinks it’s diving into material while in fact it is not. I’ll have to investigate that. This may lead to a different force acting axial against the cutter and thus material and axis flexing differently.
The steps are clearly visible In a microscope. I’ll try to measure their height with a dial indicator next weekend to know their sizes.
If this is due to the material flexing away from the cutter ever so slightly, I guess the strategy I’m currently testing should do the trick. If it happens every time there is no force against the cutter, something may have too little preload.
Tools:
I do use tampered, ball end mills.
My stepover/WOC is 0.05mm for bollnoses of 2mm radius=4mm diameter. (I would really like bullnoses tapered cutters= flat center and rounded edges but can’t find such.)
Part geometry:
This is a roughly cylindrical part on the 4th axis with a lot of surface details and all of them organic. No flat surfaces or sharp angles.
Strategy:
I’m using parallel toolpathes along the X axis. So in X direction it will always be smooth (actually minor, acceptable tool marks due to only having 12K rpm while moving at 1400mm/min)
I would have liked to use parallel along A and parallel along X but this is my largest 4th axis and it doesn’t like continuous rotation for a long time. The worm gear gets loud when it heats up. I’ll investigate this further after the current part is done. (My second largest with a harmonic drive doesn’t have this problem but is too small for this part.)
Thanks for the clarification about the dial indicators.
I had not considered that at all yet.
I screwed the indicators real tight and am moving towards them with barely 10mm/min. The indicators are new and have not seen much use. Since I am measuring repeatability of the exact same position on the indicator, I guess I can ignore most mechanical errors in the indicator. I am homing with optical photointerruptors. The indicators are just to measure the repeatability of my homing switches and stop at the exact same dial-position every time.
I did find that I seem to still have a cheap, Chinese, digital 1/1000mm indicator in the closet. I’ll mount that parallel to the mechanical one and compare Z. Maybe I’ll still find some non-repeatability that I can improve in my homing setup. If I can measure it, I can improve it and see my progress.
I was thinking of the aluminum in your machine frame expanding when warm.
You mentioned your worm gear making noise when warm. Same issue, thermal expansion. I haven’t looked at your entire setup yet. Do you have images in your gallery?
@Mark_Fuller Not really. Since it is the same when I just stop, home, resume without any pause, that thermal expansion should not be a relevant factor. (The machine is well cooled and the room doesn’t get hot either.)
The 4th axis is a different beast. I’ll take that apart and have a close look some other time. For the moment it works well enough to get the jobs done.
Good thinking.
@Marcus_Wolschon Perhaps microstepping then? Hackaday has a good writeup. https://hackaday.com/2016/08/29/how-accurate-is-microstepping-really/
@Mark_Fuller As does Texas Instruments.
https://e2e.ti.com/blogs_/b/motordrivecontrol/archive/2011/12/16/microstepping-resolution-vs-microstepping-accuracy-clearing-up-the-confusion
Looks like I have a solution.
A lot of computer-calculated finishing passes that complement each other.
(Thank god for script-language support in @DeskProto .)
At the start of each one the machine is homing again.
So errors min out (they don’t average out because we can only remove material).
Seem to work surprisingly fine except for one stupid mistake…
My stock has a circumfence of just slightly >360mm and I did forget to convert my offsets from mm to degrees.