Announcement

Collapse
No announcement yet.

More problems with a touch probe

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • More problems with a touch probe

    Hey folks, I am pulling my hair out with this thing. I thought I had my issues solved with my breakout board and wiring up a conductive touch probe. It does work now, but I am still having issues. I only tackled this for probing PCB stock for isolation routing boards. I bought the autoleveller software and it probes and generates a height map but then it proceeds to etch about 10x deeper than it should be cutting. I have noticed from the first moment I got this working, that the "digitize" LED in mach's diagnostic screen lights up noticeably later than when the probe makes contact and the BOB LED goes off. I think that Mach is noticing the trigger late and therefor running the Z down to far and zeroing late. I have this wired to pin 15 which is supposed to be for high speed triggers.

    The result of this issue is that if probing with my cutting tool I have tiny peck marks on the probe grid locations and the board cuts much to deep. I have tried with a few different shapes of tooling, but they have all had a sharp point. Sometimes if I try to probe a grid location a second time, I cannot get a probe trip, so obviously I am punching through the copper sometimes. I am planning to retry the probing with a blunt or ball end probe, however, the new version of autolevel zeros with the cutting tool prior to starting the cut, so unless I manually edit every g-code file it produces, i'm probably still going to see depth issues.

    The disappointing part is that I bought this software to fix the cornucopia of PCB depth problems and I just ran this job with no autolevel software and got a much better result on the board than I did with the leveled g-code. Is it possible that Mach is triggering late? Is there anything I can do to improve it? MY BOB is a CNC4PC C11. I'm considering inverting my probe output with a high speed small signal transistor just due to how much trouble the C11 board was to pull the one input high. The only thing that is holding me off is that it seems it will take even longer to switch if I add another transistor in the mix...

    Thanks,
    Jason

  • #2
    How fast does the probe move down? Can you slow it down?

    Comment


    • #3
      I've tried feed rates as fast as 5ipm and as slow as 1ipm (which is painfully slow I might add when it needs to do ~400 points). It seems to make no difference. I found something while going through the Mach3 documentation tonight. There is a setting for input debounce in the general settings. I am not home to check it out at the moment, but I wonder if I do not have some value in there and that is forcing the controller to wait a few milliseconds before it takes the probe state as settled. It looks like it is one debounce setting for all the inputs so I may have set something for the limits when I was initially setting it up. It has been years since I had that pleasure, so I am going to have to check to know for certain.

      Thanks,
      Jason

      Comment


      • #4
        The led is just an indicator of the input, it is not an indication of response time.

        One possibility is that you acceleration is slow. The motors can't stop immediately and depending on how the sensing works that could be causing the problem. If the machine stops when it contacts it still decels into the material leaving your divot. Possibly the aoftware is taking the data point when it looses signal on retract which means it is measuring the bottom of the hole and not to top of the surface.

        This is why commercial probes use a spring loaded stylus.

        Posting tech stuff like this specific to Mach is probably better over on the Mach forums.

        Comment


        • #5
          In an effort to understand what is going on, can you simplify the set-up and test?

          Mill a slot into some aluminium (perhaps mounted on a non-conductive layer and with electrically-isolated clamps).

          Note Z-value at which this was done.

          Then without removing the tool or changing anything, probe (first manually by jogging or manually entering G-code, then automatically using the software) the slot just milled. If working correctly, it should come up with the same Z-value.

          Comment


          • #6
            Thanks for the input guys. djc, that was a great idea and something I had not considered yet.

            I actually stumbled on the answer myself up above. It was the switch debounce setting. It was set at 12000, which equates to about .3 seconds. I have no idea why I had it set that high. I did use the xml file that CNC4PC initially provided to get me started with their board. I'm guessing that that was a hold over from when I initially set it up. I have it set at 50 cycles now, which equates to about .001 seconds, and it works much better for probing, but that it a little to low for general noise immunity. I have had the limits trip one time on noise since I changed it. I may have to debounce the limits with a R/C on each input or remember to change the software debounce back after probing.

            Thanks again!
            Jason

            Comment

            Working...
            X