I just had an issue using the openbuilds XYZ probe plate where at the end of the process the machine brought the tool tip over the top of the plate, but then drove the tool into the plate. It was a 2mm drill, which snapped off. The probe plate rotated after the X probe step, so the Y probe must have returned an unexpected value. I think this caused the probe algorithm to produce some bad values, which then resulted in the Z axis being sent to a bad location - driving the drill bit into the plate. Control threw an alarm (NaNhNaNm) While this is an unexpected situation, it seems to me that the XYZ probing process should never send the Z axis into a "bad" location if there's some kind of anomaly during probing. If something goes wrong, it shouldn't send any further position commands. Is this a bug?
If something did go wrong it would have alarmed. Going into the plate and not stopping is more likely wiring/dirty endmill/etc that stopped the electrical signal. Use the Troubleshooting tab to test your Probe
Hello, Or a broken wire, that has happened to me twice already, but I use my machine 6 days a week and therefore very often use the Probe to change my bits. Bart
Hi All, thanks for the responses. Just to clarify, I believe this was a software issue, not a hardware issue or continuity failure. The Z-axis probe was successful, followed by a successful X-axis probe, the gantry then moved to the Y-axis probe position, but in doing so it happened to bump the probe cable (my fault). When it attempted to probe for the Y-axis the plate was not in the expected location. This resulted in an incorrect value being registered for Y, which then resulted in the X-Y-Z location calculation going haywire. Next the machine moved to the final position like it does after an X/Y/Z probe, but then drove Z way down, crashing the drill bit into the plate. Control threw an alarm (NaNhNaNm) So a calculation error was detected (not-a-number). My hunch is that there's an out-of-range or divide-by-zero corner case that's not being properly error-handled. That resulted in bad position commands being sent before the not-a-number alarm was thrown. Obviously, one solution is "don't let the plate move!" ;-) but sometimes things happen.
