Automated Z-offset and/or Bed Leveling

[Tshulthise]

I just wanted to get this discussion restarted in this forum. Please post your best ideas how to approach this issue...

Here's a link to prime the pump. Its basically a mechanism to use the nozzle to define the plane of the print bed. Not the answer IMO but I like that the guy is thinking out of the box... http://hackaday.com/2014/04/15/automate ... /#comments

Other ways that already exist...
1. Use an actuator to lower a switch to detect the bed in three places. Install the firmware to use those three points to create a virtual plane which causes the Z axis to move as needed.
2. Use an actuator to lower a switch to detect the bed in one place to set z offset only. This doesn't require a really tight z axis to work but does require manual leveling of the bed.
3. Use a pressure sensitive resistor to sense when the nozzle touches the bed. Once resistor can be used to set z offset only. Three can be used under the bed or one can be used in the hot end to measure three points for bed leveling and z offset. Right now I only know of this being successfully implemented in delta bots. The resistor output is sensitive to temperature so that has to be accounted for.
ADDED #4 with 4/15 edit
4. Some machines use a touchless sensor (Hall effect, inductive, capacitive) to sense the metal build plate or a magnet instead of a switch that physically touches the platform. The CubeX printer does this but it was very unreliable so I'm pretty sure the sensor they used was not repeatable enough for reliable operation. Not to say that there isn't a touchless sensor that wouldn't be though.

Other related issues...
1. Removing and reinstalling the build platform can cause the M2 to go out of level and change the z offset. A bed mount system that prevents this would make bed leveling a less frequent requirement. I typically check it before almost every print now.

[jsc]

Bed sensing would be much simpler if we weren't using an insulator as a build surface. Jimc prints on machined aluminum, I believe. With a surface like that, you could detect contact electrically with the nozzle. How to easily attach and detach a wire to the nozzle is a different matter, you can't use a printed mount against something that gets up to extrusion temperatures.

[ednisley]

you could detect contact electrically with the nozzle

Remember that the nozzle may be covered with a thin layer of insulating plastic that prevents dependable electrical contact. Depending on electrical continuity may work most of the time, but when it fails, the nozzle will attempt to punch a hole through the build platform.

prints on machined aluminum

Any aluminum surface in air has a very thin aluminum oxide coating that's a wonderful insulator. The nozzle must displace that layer in order to make electrical contact, which also works nearly every time.

Switches may seem clumsy, but dependable contacts make up for almost all their other faults...

[jimc]

yes ed not to mention the nice coating of hairspray.

[jsc]

Very good points, which I had not really considered...

[Lateralg]

Tony,

I'd like to start topic "Optimizing Manual Z-offset and Bed Leveling".
I don't want it to be in competition with Automated method; I believe the two topics can be mutually beneficial. And a refined manual method can help hold the fort until an automated method is developed.

It's important to me that generating friction between/among members of the forum is avoided.

How do you feel about this?

[Tshulthise]

Gary, no issue with that at all. Thanks for asking though.

T

[jsc]

Trawling thingiverse, I came across this: http://www.thingiverse.com/thing:310424. Interesting part is the non contact probe linked in the description. Very cheap. How would it work? Why not just a switch, though...

[ednisley]

How would it work?

Those likely have an oscillator inside that's detuned by the presence of a ferromagnetic target. The gibberish that passes for specifications says:

• Response Frequency : 100Hz
  Detect Object : Iron
  Detecting Distance : 5mm/ 0.2''

A possibly related datasheet (http://www.chinahwe.net/products/lj18-ID208.html) recommends a honkin' big target and suggests that the repeatability could be, maybe, 0.5 mm (10% of the nominal 5 mm).

I think it would work, in the sense that you could hitch it up and the bed would be more-or-less detected, but it's not clear the result is stable enough to be usable. The frequency response in the Amazon listing doesn't match the "datasheet", either; you'd want to creep up on it to avoid overrunning the response.

A simple mechanical switch has a lot to recommend it...