TAP-XXX Nozzle Contact Sensor

While I wait for my diamond nozzle, I realized that I had never posted a video of the TAP-XXX piezo nozzle contact sensor. I've been using it regularly for months, and just recently realized a very cheap and quick improvement involving "O" rings. Gerber, STL, and DOCX files are available if anyone is interested.

So here it is:
[www.youtube.com]

A nice demonstration, but for my peace of mind that it is truly consistent, can you try the following.

• Get a map of the surface using at least a 5 by 5 grid and save the readings.
• Move the sensor to two other positions fairly far from, and not symmetrically the same as the first position
• Repeat the mapping of the surface for each new position.
• Compare the grids, subtracting each from the original. Obviously there will be a travel time difference but dynamic (Chladni Plate type) errors should be obvious, often being in the opposite direction to what you would expect.

If it passes this test it should be good to go - there still could be problems in other circumstances (see Boeing 737 Max Angle of Attack sensor) but due diligence will have been done.

Edit:-

Quote
rq3
…………………….. Gerber, STL, and DOCX files are available if anyone is interested……………..

Very interested, where can I find the files?

Mike

Edited 1 time(s). Last edit at 08/22/2021 01:51PM by leadinglights.

Quote
leadinglights
A nice demonstration, but for my peace of mind that it is truly consistent, can you try the following.

• Get a map of the surface using at least a 5 by 5 grid and save the readings.
• Move the sensor to two other positions fairly far from, and not symmetrically the same as the first position
• Repeat the mapping of the surface for each new position.
• Compare the grids, subtracting each from the original. Obviously there will be a travel time difference but dynamic (Chladni Plate type) errors should be obvious, often being in the opposite direction to what you would expect.

If it passes this test it should be good to go - there still could be problems in other circumstances (see Boeing 737 Max Angle of Attack sensor) but due diligence will have been done.

Edit:-

Quote
rq3
…………………….. Gerber, STL, and DOCX files are available if anyone is interested……………..

Very interested, where can I find the files?

Mike

Mike, because of the 800KB limit here, I've posted the Gerbers, the STL for the NEMA17 mount 35mm version (TAP-350), and a descriptive document at:
[www.thingiverse.com]

I know you've had issues with bed resonance and piezo sensors, but this is a whole 'nother animal. I use it for delta calibration and bed leveling (10x10 grid) on a 380mm diameter delta bed, and can reliably print a one layer film 50 microns thick over the entire bed surface, which is a good enough test for me.

Plus, as you can see from the video, the sensor really doesn't care where it is. As described, there were two sensors in that video, one glued under the bed, and the visible one just resting on the bed. I can switch between the two, and if I use the one just resting on the bed, it doesn't matter where I put it, or where the nozzle contacts the bed. Remember, this is desiged to react to jerk, not acceleration, whether that acceleration is Earth normal, motor vibration, or an earthquake. Only a change in its current acceleration state will make it respond. Think of it as constantly re-zeroing its response to its environment, until it receives an impulse too quick to re-zero. It's not looking for deflection, it's looking for a change of state in its current acceleration condition.

If you'd like to give me your mailing address, I'd be happy to send you an actual, complete sensor which you can play with to your heart's content! On me. No charge.

Somewhat related, the MAX737 issue seemed to be a matter of an excessive number of sensors, software that couldn't discriminate between them, all installed in an embedded system that the pilots were unaware of and had received no instruction in how to over-ride. Too much reliance on automation, with no ability for humans to "interfere". If Boeing had said, "Here's this nifty new safety system, and here's how to turn it off", things might have been very different. In fact, at least one MAX737 was piloted by a man who was, somehow, aware of the system, and he managed to save the plane during a similar upset.

The Air France Airbus incident was almost as bad. Side stick controls and fly by wire, with no feed-back between the left side and right side control sticks. The Captain was pushing forward to avoid a stall. The co-pilot was hauling back to maintain a climb. The computer was in the middle, fat, dumb, and confused while the plane plunged into the Atlantic.

Edited 8 time(s). Last edit at 08/22/2021 05:20PM by rq3.

PM with address sent.

I agree about automation having unpleasant surprises for the unwary. My own favorite aircraft was a Piper J3 "VP-YEM" without so much as a starter - the only electrics were the mag switches

Mike

Quote
leadinglights
PM with address sent.

I agree about automation having unpleasant surprises for the unwary. My own favorite aircraft was a Piper J3 "VP-YEM" without so much as a starter - the only electrics were the mag switches

Mike

Got it. I'm shipping tomorrow (Tuesday) via USPS international priority mail (I think). Customs will be declared as zero value electronics samples. I'm sending two: the 41mm version (TAP-410) comes apart. The 35mm version (TAP-350) does not. If you want to re-program, you'll have to download the free Picaxe Editor, and buy an AXE027 cable from [picaxe.com]

I have ordered an AXE027 cable and will download the software and drivers. I will run some tests and report back with results.

Mike

Quote
leadinglights
I have ordered an AXE027 cable and will download the software and drivers. I will run some tests and report back with results.

Mike

They are on their way. You should find that they act like an automotive engine knock sensor, insensitive to noise and vibration, but immediately responsive to a very light impulse anywhere on their mounting surface. Originally, I had mine mounted in the top case of my delta, with easy access to the power supply, and it worked fine there, but gaining access to the sensor itself was problematic.

They will run quite happily for at least a day on a 9 volt "transistor" battery. The supply can be anything from 7 to 35 VDC.

The two "go to" variables in the firmware, should you choose to modify it, are the sensitivity, in millivolts, currently set to 24. And the delay, in microseconds, currently set to 80, which is the length of time that the "contact" LED flashes on, and not coincidentally also the pulse length that the printer controller Z axis pin sees as the sensor output goes tri-state so that it will pull up. My controller seemed happiest at 80 microseconds, although anything from 2-200 seemed OK.

Have fun!

I have just received the sensors and hope to work on them early next week.

The plan is to set up some dummy beds - 220 mm square in 2mm and 4mm aluminium on my Cartesian Printer. I will initially try them in the manner that you designed them for and then look for areas where the signal might be compromised. To do this I will take the signal straight from the piezo to one of my own conditioners so that I can extract the data in the time immediately before your conditioner trips. On another channel, I will monitor the contact point using a gold flashed nozzle and contact plate to get the time for the contact event signal to reach the piezo through the bed.

Hopefully, I will be able to find that your method is robust and repeatable, or if not, that some simple "do this" or "don't do that" instructions can be found to guide potential builders.

Mike

Quote
leadinglights
I have just received the sensors and hope to work on them early next week.

The plan is to set up some dummy beds - 220 mm square in 2mm and 4mm aluminium on my Cartesian Printer. I will initially try them in the manner that you designed them for and then look for areas where the signal might be compromised. To do this I will take the signal straight from the piezo to one of my own conditioners so that I can extract the data in the time immediately before your conditioner trips. On another channel, I will monitor the contact point using a gold flashed nozzle and contact plate to get the time for the contact event signal to reach the piezo through the bed.

Hopefully, I will be able to find that your method is robust and repeatable, or if not, that some simple "do this" or "don't do that" instructions can be found to guide potential builders.

Mike

Glad they didn't sit in customs! If they don't rattle when gently shaken, they are good to go. If they DO rattle, it means the jerk mass has come off the piezo. You should be able to power them up on a shaking bench top (I use a large window fan to induce vibration), and not see them trigger until you drop a BB on the bench from a height of about one inch, about 4 feet away from the sensor.

The only "quirks" I'm aware of are Marlin induced. Marlin has an odd "overtravel" issue, where it continues motor stepping for a couple of steps AFTER the processor has received a Zmin trigger. I have my delta bed mounted on 6 viton "O" rings to give the bed a few microns of compliance to work around that. Otherwise, Marlin "averages" the extra steps into its math, and all hell breaks loose, especially on a delta calibration. My delta cal routine uses 16 contact points, 1 at the center, 3 orbiting around the center, and 12 around the extreme edge of the bed. UBL leveling is a 10x10 grid covering the entire reachable bed area. I can print a 50 micron thick, single layer PLA film over the entire bed surface. Also, I have the "Input Interrupt" capability turned on in Marlin, and the nozzle contact speed set to 10mm/second.

I also find that the nozzle and bed have to be surgically clean. The slightest blob of plastic on the nozzle (which varies in thickness as it gets tapped during probing), or a human or cat hair on the bed, can cause probe failure. My go-to delta has 5 micron motion resolution, so this isn't surprising.

Have fun!

Edit: one other quick note. The piezos are connected so that they are polarized correctly when the units are mounted under the bed, with the connectors down. They will work the other way, but you will see a delay in the response as the piezo signal incorrectly goes negative first, and then positive.

Edited 1 time(s). Last edit at 09/03/2021 12:30PM by rq3.

I had been intending to do a bit more research with the TAP sensors that rq3 sent me, but life got in the way and I have been delayed in this project – and a few other projects as well.

While I haven’t got any rigorous results yet, I thought that I would note down the results that I got more than a month ago.

I built a dummy bed (shown below) to mount above the existing bed and allow space for the TAP sensor to be mounted under the bed. The bed is made from 3mm float glass and has adjustable supports mounted on each corner and three positions for mounting the sensor, all of which are glued to the glass plate. The positions for the sensor are one at dead center, one on a diagonal line between two opposite corner supports at a distance midway between the center sensor and a support, and the last one at a point midway along the Y direction and 80% along the X-direction. These were chosen as they looked like areas where sensitivity would be at a minimum.



In the foreground of the picture is a programmer for the PICAXE on the TAP. The cylinder lying on the glass is an 8.000mm diameter gauge to set the nozzle height

The intention was to apply real-world probing movements with a nozzle and map the response at many positions and nozzle speeds, firstly with the intact sensor and then with taking the signal from the mass-loaded piezo disk and sending it through a charge amplifier to see what the generated signal can tell us.

I only got as far as building the rig and adjusting the trigger threshold to above the noise level when as I mentioned before, life got in the way. However, I did some initial “drop an 8mm ball from 4mm above the table” type tests which showed no dead spots. There does seem to be a bit of a problem if the nozzle is dirty: I did a Q & D simulation by putting some 50 micron polythene sheet between the ball and the bed which gave erratic and unpredictable signals.

Hopefully, I will be able to continue with my many projects – and a couple of others that I have thought about, in the next couple of months.

Mike

Mike, it's very good to see you back in action!

This looks really interesting!

Too bad you don't sell a kit for us lazy people with too many projects already

I will have to dig out some piezos and do some experimenting. I like the possibility to level with heated bed and nozzle, I am currently rebuilding my kossel mini that originally had FSR bed leveling. That wasn't all that reliable and was removed while fitting a heated bed. The major problem I see is making sure the nozzle is clean.

I am very interested in this sensor. Do you have the STL and Gerber files available for the under bed version?

Quote
labraticmp3
I am very interested in this sensor. Do you have the STL and Gerber files available for the under bed version?

Certainly:
[www.thingiverse.com]

I've been using it for over a year now, and it's still reliable and consistent. Have fun, and let us know what you think.

Edited 1 time(s). Last edit at 02/07/2022 05:02PM by rq3.

Quote
rq3

Quote
labraticmp3
I am very interested in this sensor. Do you have the STL and Gerber files available for the under bed version?

Certainly:
[www.thingiverse.com]

I've been using it for over a year now, and it's still reliable and consistent. Have fun, and let us know what you think.

Thanks for the quick reply, but I only find the 35mm_Nema17_Inserts.stl at the link you provided. Is the 41mm_Knock_Gerber.zip file for the underbed version?

Quote
labraticmp3

Quote
rq3

Quote
labraticmp3
I am very interested in this sensor. Do you have the STL and Gerber files available for the under bed version?

Certainly:
[www.thingiverse.com]

I've been using it for over a year now, and it's still reliable and consistent. Have fun, and let us know what you think.

Thanks for the quick reply, but I only find the 35mm_Nema17_Inserts.stl at the link you provided. Is the 41mm_Knock_Gerber.zip file for the underbed version?

The circuit boards are electrically and layout identical, except for the shape of the board perimeter. The 35mm version is cut into a circular disc from the 41mm rectangular board. The 35mm version is, of course, smaller than the 41mm, and operates identically. Either version should be mounted under the printer bed.

rq3,

Thanks for making your design available. I have built one and installed it on my Anet A8 printer. I am very satisfied with the performance so far. I don’t think I am getting the performance that you get, but I think it is because of the lack of rigidity of the A8. It is much better than the other probes I have tried.

Quote
labraticmp3
rq3,

Thanks for making your design available. I have built one and installed it on my Anet A8 printer. I am very satisfied with the performance so far. I don’t think I am getting the performance that you get, but I think it is because of the lack of rigidity of the A8. It is much better than the other probes I have tried.

Glad it's working for you. If you are using Marlin, makes sure you enable "Endstop_Interrupts" in configuration.h.
It can make a huge difference if your board supports it, and most modern 32 bit boards do.