On Saturday 06 October 2018 00:56:58 Gene Heskett wrote: > On Friday 05 October 2018 23:24:38 Chris Albertson wrote: > > A third option is a REALLY long belt. It is long enough to make a > > double pass on each side and needs about 8 pulleys but you only need > > one motor. I saw one of these once but it used a kevlar cable, not > > a belt > > > > I think the shaft is move simple but but you need precision > > mechanics to adjust it. It is ever comes apart for service you'd > > need some good measuring tools to align it. But with two motors > > the adjustment misdone in software and the mechanic doing the > > service only needs to get it "close" by eyeball with a tape measure. > > find adjustment is later in software. > > > > My dirt cheap 3D printer uses two motors and parallel screws and I > > use basically a sheet of paper as a feeler gauge and I can reset > > alignment in abut 30 seconds to about 0.005 mm. > > > > Any design can work as long as you have a designed on squaring > > method that can be done with simple/cheap tools > > > > All the talk about the long axis but I think the REALLY hard part is > > the Z axis. Lets say you make a cut that is 0.15 mm deep and 3 > > meters long. Will it remain exactly 0.15mm deep over that long > > distance? I bet you a buck not. > > No takers on that bet Chris. > > If indeed its that critical, one would need a dynamic distance > detection method of some sort riding the work pretty close to the > tool, which would be outputting the instant offset, and would be > driving the Z to maintain that distance in real time. And the success > of that would be in building it to not be material sensitive, or > subject to getting tangled with the cuttings. > > Because it has to operate in a pretty noisy environment, I'd design it > with a hardened ball pointed lever, sitting the ball quite close to > the tool, and pivoting a couple inches away from the cutter, with the > far end moving a ferrite slug in and out of a coil which was part of > an oscillator running at 100+ kilohertz, and some sort of a frequency > discriminator, possibly a crystal referenced phase locked loop, using > the pll's error as the signal into an offset module. With the correct > gain, I'd say you could make a working correction circuit that would > reduce a 5mm error to +- .01mm, which might be close enough. That > wouldn't care how long or wide the bed and workpiece are but the lever > should be constrained from dropping over the edge of the workpiece and > being caught and destroyed. That implies a bigger ball. The ferrite > slug could be moved by a small wire thru a hole in the shielding so > other electrical noises wouldn't bither it too much, or were fast > enough that the drive motor will function as a noise integrator. This > could be centered in the control range by useing a screw the pivot > point up and down with a hand knob, on the levers pivot point height. > With the tool stopped, adjust the slider with the rest of the circuit > live, so the tool touches the work or holds a paper slip, then dial an > offset in with a sum2 to drive the tool to the desired depth. And > apply just enough shop air to keep the contact area 99% clean. > > Theres probably a few more ways to skin that cat, as all of this would > have to be moved with the z drive, but thats what I'd try since the > electronics are fairly well understood here at the coyote.den. Making > the hardware would take me some longer than some of you. But I'd get > it done as long as I'm still looking at the green side of the grass. > ;-) Not having the real estate for a big gantry machine would slow me > down I expect.
Nother idea. Put the air into a small pipe that nearly touches the work, feed it enough air to be self cleaning, and the air pressure backup because of it close proximity to the work would be fed to a barometer chip to measure the back pressure, and feed that to the offset module. That would be more sensitive to the clearance and air gap from previous machining it was crossing over, so that would likely be impractical. Also highly non-linear because of the square law as it applies to that sort of thing and moving air. Not well taught is the fact that a motorized louver is flowing 50% of its wide open flow when its opened 10%, or only 8 degrees from completely closed. The only other idea I could come up with would be to map the face of the work area prior to doing any cutting. If the code was generated by a cad/cam proggy, the resultant gcode could have its z modulated with an intermediate processing utility using the previously mapped data for that xy point. I've not played with the file idea because I've not figured out how to use the file method cross axis. LinuxCNC has not been setup to use correction files for anything but correcting its own screw errors for that same axis. To use the lincurve and offset modules for something like this would need a lincurve with 64 to 128 points on a sagging 3 meter machine. Andy might be able to build it that long, but what would be the execution time per loop pass? Scary. For a big bed, it would need a huge xy array of data if the correction grid was in 2" squares. How have others arrived at something that works in this scenario? -- Cheers, Gene Heskett -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene> _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
