Piezoelectric disks for Z contact detect and bed levelling

With my fingers crossed to ward off Murphy and other malign spirits, the present setup looks good as-is with Delta printers and hopefully should work with Cartesian printers.

Piezoelectric materials are a lot more sensitive than I thought before I tried the disks. Below is a table from Wikipedia showing a comparison between Peizoelectric and Piezoresistive (FSR) sensors.



Mike

Nice info, except that FSR are not piezoresistive sensors at all!
They really are resistive surfaces pushed against each other.

---

Blog: [tridimake.com]
Designs: [www.thingiverse.com]
Profile: [www.linkedin.com]

But given these values.. well... I am more than inclined to switch to something else.
I did preliminary tests with load cells which I think are promising, and have better thresholds than FSR.

---

Blog: [tridimake.com]
Designs: [www.thingiverse.com]
Profile: [www.linkedin.com]

I just looked at both Force Sensitive Resistor and Piezoresistance on Wikipedia and I stand corrected. Load cells would be nice but where can you get such a beast at a hobby price?

Mike

I bought small ones on ebay. But I bet the deal is to use the film sensors themselves on a 3D printed part (they are VERY cheap then)
Looks like this: [plus.google.com] (but not these! [plus.google.com] )

Edited 1 time(s). Last edit at 03/18/2016 03:34PM by jeremie.

---

Blog: [tridimake.com]
Designs: [www.thingiverse.com]
Profile: [www.linkedin.com]

I was quite intrigued by reports that piezoelectric disks would not be suitable at the temperatures of heated beds. The temperatures given in various data sheets were 85C for both operation and storage, but this is just the "industrial" temperature classification. RepRappers have used wirewound resistors and thermistors way beyond their datasheet values for years. Looking at the construction of piezoelectric disks, the ceramic used in Murata disks (and the no-name coppies are probably the same) is almost certainly lead zirconate titanate (PZT) which Murata gives the designation P7 for use in buzzers. This seems to be good to over 200C, especially if there is no long-term applied voltage.

A quick test with a piezoelectric disk with a PEEK pressure pad bonded to it, recording the output when a screwdriver handle was dropped on the pad at both 18C and 110C gave the following traces.




Preliminary I know - it being hard to calibrate a dropped screwdriver; but I think it likely that piezoelectric disks can be put directly under the plate.

Mike

I can't claim any expertise, but my quick googling suggests that high temperatures have a cumulative affect on lifetime, as does high humidity. Given your design allows the piezos to be mounted relatively far away from the heatbed I think I would stick with that layout unless there's a convincing reason to do otherwise.

One of the things I looked at was a Murata document at [www.farnell.com]. There is a bunch of graphs on page 9 - although the aging is at 85 degrees I assume that they had no interest in looking higher for their needs. I did find something in Wikipedia about time dependent dielectric breakdown, but that implies a sustained applied voltage but no voltage is applied in this kind of sensor.

I am fitting both my Delta and Cartesisn printers with the type of sensor that I have designed, but I will build a little micro oven with a piezoelectric disk under load at say 120C with an outside vibrator and check it every couple of weeks.

Mike

That would be great. I think the only question left about this approach is the robustness of the sensors.

The piezo disks are now working very sweetly on my Mendel Prusa. In order to get it immune to the crosstalk from the X and Y movements I had to reduce the gain of the amplifier from about 5 to 2 and put a bit of hysteresis on the comparator.

The red trace below shows the output of the amplifier op-amp showing the deceleration noise. The small sineywave stuff from about half way is Z motor artifact then the rising voltage is good signal, followed almost immediately by the falling voltage of the nozzle lifting off. The yellow trace shows the signal that is fed to the Arduino. And yes, I am also startled by how quickly the motor reverses - from nozzle contact to nozzle lift is 10ms.



A last trace showing the leveling actually working. When I looked at the Z step trace it seemed to be always running but a close up look showed that that is just the tweaking that Repetier firmware is doing to keep the nozzle the right height from the bed as the X carriage moves.



I have tried enough on my Delta to be happy that it will work even better there.

Mike

Job well done. Congratulations and thank you

Nice

---

Blog: [tridimake.com]
Designs: [www.thingiverse.com]
Profile: [www.linkedin.com]

A quick passing note. I noticed this on a forage for information [www.robotoid.com]

I don't know how well such a direct input would work but if it does then using piezo disks for bed leveling is potentially the cheapest accurate bed sensor around (about £2 in parts)

Mike

Really interesting thread, and definitely something that will end up on my to-to list.

I was thinking that it may be possible to simplify the mechanical implementation by moving the piezo sensors to the z-carrige suspension, I recall someone a coupple of years ago implemented ABL this way with microswitches.

This sort of worked, but most of the weight of the z-carrige had to be transfered to the bed, to get the switches to trigger, causing potential damage to the bed surface. This wouldn't be the case with piezo sensors, as much less force is needed to detect contact. Also the polarity of the signal would be predictable (i would guess), no matter where the nozzle touches the bed.

Of course this would only be easy to implement if your z-carrige is hanging freely.

How about using a Piezo ring instead of a plate, mounting that "inline" to the bed with the spring and measuring "only" the change of pressure ?
that could reduce the mechanical construction to a minimum ?



Just an idea, have not much knowledge of piezo electrics anyways ^^

Chri

---

[chrisu02.wordpress.com] Quadmax Intel Delid Tools

Here's my current progress with implementing piezo sensors into a delta effector.


Pressure on the nozzle causes (minute) flex in the arms. The piezo sensors are super glued rigidly to the upper leaf of the arms so when the arm bends so does the sensor. I'm using 10.5mm transducers but they're only about 0.15mm thick, much thinner than the 20mm ones I have.

I'm still working on tuning the sensor conditioning circuit at the moment. I'm getting reliable triggering so far but I haven't yet installed it in my printer.

Moriquendi

I tested putting a piezo disc in the z suspension, and it didn't seemed like a usable solution, even thoug I use trapezoidal leadscrews with almost no wobble, the "almost no wobble" was enough to false trigger on every revolution on the leadscrew.

It's possible that one could overcome this issue via a better mechnical design, but for now I'll move on with the confirmed working solution, of incorporating the discs in the bed.

For those (like me) who doesn't want to mess around with SMD I have dropped a leaded version of the electronics here: [github.com]

This is really interesting stuff, but I still don't see how these are better than FSRs.

FSRs run about $15 each, where as piezo disks are about 18c each in quantities of 10. Call me cheap, but that's not a bad start.

Quote
JamesK
FSRs run about $15 each, where as piezo disks are about 18c each in quantities of 10. Call me cheap, but that's not a bad start.

I head no idea they were so cheap! I ordered a kit of 3 FSRs and a controller board for $30 which I see as very affordable, but if you could cut that price in half, that would a very attractive option.

I have been on holiday for the last couple of weeks and am pleased to see that this thread hasn't died.

@Moriquendi, Nice job. How sensitive is your setup? (light finger tap or rap with the blund end of a screwdriver?)

@Ralf, Thanks for putting up the leaded circuit. I used SMD as I thought it may be an idea to put the amplifier close to the piezos (not necessary) and selected the semiconductors because I have used them before and had or coule easily get them. The excessive sensitivity that you mentioned threatens to be the most significant problem with piezo sensors. More work may be needed on frequency response and hysteresis to cover a wide range of printers.

@MrBaz, There are actually quite a few ways that Piezoelectric sensors MAY BE superior to FSRs. I say MAY BE as I spent many years working in product design and can appreciate that an entire room full of bright cookies can often not see a fatal flaw until it bites them - which happens much more often than not. So, Ways in which Piezoelectric disks MAY BE better than FSRs:

• Much cheaper
• More commonly available
• More sensitive (Maybe too sensitive)
• More robust once installed
• Better high temperature performance. I am investigating but they look good to over 120 degrees C.

Mike

Ralf, don't suppose you could post the Gerber and NC drill files as well please? I don't want to have to buy Eagle to get the PCB made. I want to get a few PCBs made and the ones that I don't use, I'll make available for sale at cost through this forum thread.

PS Had a bit of an accident with ordering and now have 100 27mm Piezos!

Edited 1 time(s). Last edit at 04/13/2016 08:35AM by MotoBarsteward.

---

Using ABSPrusa Mendel Zaphod with Pronterface and slic3r 1.3.0. Printing well with 3mm PLA and ABS through 2 x J Head Mk IV b and Wade Geared Exruders. Controlled using RAMPS1.4 board running Marlin_v1.1.9

You won't be missing piezo discs anytime soon..

I have added gerber and drill files.

I have tested that I can open them with FlatCAM, though I had to modify the excellon_zeros parameter, as per instructions from this page: [flatcam.org]

Edit: Also added the circuit diagram

Edited 1 time(s). Last edit at 04/13/2016 02:10PM by Ralf.

@Leadinglights, its plenty sensitive.



with R6 at 10M the slightest touch on the heatsink causes a trigger. I've replaced R6 with a 1M trimmer, reducing the value reduces the sensitivity of the circuit. I found this app note useful. I can reduce sensitivity as much as I like, its worth noting that the speed of deflection is much more important than the displacement. I think this may be the hardest part of using piezo sensors as a sharp movement (vibration?) can cause a spike that's very difficult to differentiate from a true contact. I still haven't got the contraption mounted on my printer, I need my printer functioning at the moment and I haven't got time to go bug hunting right now.

Moriquendi

I was hoping that any false triggers from the probe are ignored by the firmware unless actually probing - is that not the case? If the probe is only monitored during probing then false triggers shouldn't be a problem as the probing motion itself can be relatively gentle.

@Moriquendi, Although the lower value resistor across R6 would lower the sensitivity, it will also increase the lower cutoff frequency, particularly as the smaller piezo sensors will also have a lower capacitance. My guess - and is is just a guess - is that you need a frequency response <1Hz to >500Hz. I found it necessary to reduce the gain to only about 2:1 and also incorporate some hysteresis - most recent circuit below.



The gain is set by R7 and R8 in this schematic and the hysteresis is set through R12.

@JamesK, Repetier - and probably most/all other firmware ignores false triggers until the nozzle starts moving down in a probe check. This does not mean that we are out of the woods however as the whole printer will shake for a short time after an X or Y move. What is needed is a short delay between X/Y moves and the beginning of probing. Sadly, my high level language programming skills are pretty much pants so I reduced the sensitivity and the speed of all movements. Bed leveling on my first printer is now accurate and repeatable.

Mike

OK. If a software change will help that shouldn't be a problem - that's my day job.

I'm having a very hard time stopping myself from ordering a shiny new oscilloscope. I haven't had an interesting electronics project in years and my excitement is bubbling over Of course, it may be an equal number of years before the next one, which makes buying another scope when I could make do with the old one a bit unjustifiable! But on the other hand, shiny new tools!

(Logic analyzers look pretty fun too...)

I've had problems previously with my Zprobe triggering from vibration during probing, that's why I'm concerned about vibration. The two firmwares that I have experience with (marlin and smoothie) both ignore Zprobes if they're not actually probing by default so that's not a problem.

I'm not sure how travel speed affects vibration at the effector (looking at a delta printer). With a Cartesian printer the Z axis typically has much higher steps/mm than X and Y axis so vibration from moving in the Z direction should be lower and you're not moving the X&Y so there's no vibration from there. With a delta this isn't the case steps/rev for all three axis are the same and a roughly equal to a Cartesian X&Y axis so you're stepping three motors, in unison, for any Z movement. I've started thinking that having the piezo elements in the effector may not be ideal so I'm adapting the design to support the bed instead.

This is all theoretical for me at the moment, if anyone's got piezos working with a delta I'd love to hear about it.

Moriquendi

I have got it working on my Prusa Mendel with Marlin firmware.

The only real problem I had was that I needed to introduce a small delay in the probe routine, between positioning X and Y, and doing the actual probing. Without it, it false triggered, because of residual vibrations after moving the y-axis.

I have been following up some earlier trials I did with piezoelectric disks at higher temperatures and the results are not as good as my earlier "thwack with the blunt end of a screwdriver" tests seemed to indicate.
I made a jig with a heater block to heat the disk up, a spring to preload the disk and a solenoid to release the pressure. The pressure pad and parallel mechanism from the present design were adopted but machined from PEEK to allow for temperatures up to150°C.

At 20°C the output from the disks was 25V

but this fell to 10V at 40°C

and only 2.5V at 60°C


There was no usable output by 70°C

Although the low output at 60°C is still quite usable, using this kind of piezoelectric disk in a heated chamber printer would probably not be as good.
Since the data sheets for PZT type P7 indicate that is should be quite usable at well over 100°C I am presently at a loss to know what is happening here - possibly the cheap disks use Barium Titanate instead of PZT.

Mike