I'm sure this sounds obvious to people who aren't as stupid as me, but there is quite a bit of difference between the "z-position" where the needle tip touches the object and the "z-position" where the limit switch triggers. Took me a few hours to realize this was the source of my problems. For hard objects on my machine the difference is 1.343mm, not a small distance at all!
Unfortunately this offset depends on a lot of things. One of them is the tension in the spring that pushes down on the hollow shaft (so it doesn't bounce upwards and trigger the limit switch if there are small vibrations). Sadly this tension amount will change as the spring ages, and is probably dependent on the room temperature.
For hard objects I eventually settled on calibrating my machine with a pressure sensor (I used this one: http://www.amazon.com/gp/product/B00B887CLS) with a piece of microscope slide glass on top of it to make the result less dependent on needle tip diameter. Probe the pressure sensor, record the z-position where the limit switch triggers, then back away from the sensor until it can't detect any pressure. The difference should be valid for a day's work as long as the room temperature stays the same. By using the ADC pin (see thereza's patch) this can be automated, so I added it to my default set of beginning-of-the-day calibration routines.
Unfortunately the offset also depends on the hardness of the object you're probing. I was able to eliminate all such materials from my setup except for gel-pak substrates, which I really love. Gel-pak is great, but unfortunately they have a huge amount of "cushion" -- several mm worth. Fortunately they're manufactured with very uniform height consistency (probably poured as a liquid) and all of mine are of only two different types, so I simply measured the height of each and add that as a hardcoded offset.
- 1.jpeg (52.86 KiB) Viewed 2583 times
- 2.jpeg (46.53 KiB) Viewed 2583 times
- 3.jpeg (49.04 KiB) Viewed 2583 times
- 31kUOn1-onL._SX385_.jpg (7.84 KiB) Viewed 2583 times