Auto-bed leveling with piezo sensors on bed

Thanks I read about FSR some time ago, but was put off by the low sensitivity (100g weight to trigger I think). But that was with passive electronics. And they're temperature dependent as well. I would be interested in your experiences with them.

Status update:

As of now, it works. Kind of As the video shows it works quite well when everything is set up properly, but there are some issues yet that I'm working on (I have at average 30 min tops to work with this, between all my other duties, so it might take some time). I am positive I'll solve them though, but I'd rather try and fix them before making a circuit and BOM available, to prevent people wasting their time on an unfinished design. For people more technically inclined I can make an exception

* Sensitivity. They are very, very sensitive. Right now the circuit is just an op amp as charge collector, with high and lowpass filtering, with output fed into a comperator. The comperator uses a 10k potmeter as a voltage divider between VCC and GND. With lowest possible threshold, the LED flashes when I breathe at the bed surface. Obviously, this triggers the circuit whenever a stepper moves, so the threshold needs to be adjusted up. I'll add some resistors so I can fine tune the threshold voltage more, because now it's very coarse, making it very random if a small change in the potmeter will make the system better or fail spectacularly.

* Noise. Without any low-pass filtering, the circuit would trigger all over the place when the steppers were energized. After setting the low pass freq to ~300hz, these problems went away (as the steppers PWM freq is ~20 kHz). The problem now, however, is that the hotend pwm output runs at 63 Hz, with wires directly next to my thin wires to the piezo discs. At high current. So the circuit freaks out when the hot-end is turned on. This can be solved in a couple of ways. I guess I'll try first to shield the wires, and perhaps also the circuit (which is on a breadboard now, making it difficult). If not, I can just make a script to heat up the bed and hot end, turn off the hot end while probing, and back on again.

* Left side of bed was less sensitive than the right at some times. I've made new ABS mounts and done some changes to try and prevent any binding, I'll see if they fix this. There shouldn't be any way for a force to be directed towards the bed without going through the piezo discs now.

* Sensitivity again. I want the sensors to be sensitive, but at the same time I want to move the z at about 5mm/s to get a good signal (2mm also works though, I guess one should experiment a bit). Since the circuit has a high pass filter of about ~3 Hz, any static or slowly increasing force will be attenuated. However, the acceleration to 5mm/s can cause the circuit to trigger on the force from the inertia of the bed. This can probably be fixed with a script setting the z (and xy) jerk to zero, and acceleration low like 30 mm/s^2. And after probing, setting them back to their normal values.

I did a test run just now when printing out new improved holders for the piezo elements:
20:50:34.083 : Z-probe:2.88 X:50.00 Y:50.00
20:50:37.361 : Z-probe:2.83 X:150.00 Y:50.00
20:50:40.895 : Z-probe:3.24 X:100.00 Y:150.00
20:50:41.592 : Transformation matrix: 1.000000 0.000002 0.000437 0.000000 0.999993 -0.003862 -0.000437 0.003862 0.999992
20:50:41.592 : Info:Autoleveling enabled

As can be seen, the table is really off (by purpose). I made a small video showing that it works fine. It There's a small amount of under-extrusion at the right side, but this is not caused by the bed leveling but is caused by the bowden setup, which is a separate issue I'll tackle some other day.

Crappy video showing the z moving while printing, and that the first layer isn't completely hopeless.

RG174?

I've added it to thingiverse

Thanks, had no idea for names of any small coax. I'm used to work with them, not buying

Found some cheap rg316 cables with male SMA and female panel mount SMA connectors pre-installed for cheap on ebay. I guess it'll be some weeks before I see them. Although, after cleaning up the breadboard and switching the hot-end pwm frequency from 15 to 61hz I do not have problems with interference (right now at leasT). Maybe I'll get around it without shielding As I want the design to be as simple as possible so people can easily adapt it, this would be a good thing.

Anyway, back to progress. I made the breadboard layout a bit less like a nest of wires and resistors, and drew the schematic up in cadsoft eagle. It's far from a perfect design, with dc bias drift, no adjustable amplification and so forth. I want to add a second dual opamp to use for bias reference voltage and adjustable amplifier between the charge collector and the comparator. But I'm short some capacitors for AC coupling each stages I'll see what I can do. Without an oscillioscope it's a pain to debug as well.

Also found a flaw in my mechanical design. I use a single bar in front for both piezo discs. Depending on how much I tighten the screw holding it down, one side becomes more or less sensitive than the other. The lone disc in the back works fine. So I'll print out some new ones keeping the two front discs independent of eachother. In the meantime, I'm using the auto-leveling as is

After cleaning up the breadboard, setting max z travel acceleration to 30 mm/s^2 and 0 mm/s z jerk, it seems to work fine.

It's still very much a work in progress though. But I'm confident it's at the point where it works well enough that any issues should be rather easy to overcome with some research

I love playing with sensors and saw this article on hackaday that may be useful for this project.

http://hackaday.com/2015/03/16/instrume … le-change/