Piezoelectric disks for Z contact detect and bed levelling
Posted by leadinglights
|
Re: Piezoelectric disks for Z contact detect and bed levelling October 23, 2017 10:29AM |
I have the final results of the tests at 80°C showing the stability of discs in compression (un-drilled) and in flexing (drilled disc) disc after a week of soaking. The rig was not really satisfactory for the drilled discs as there is more bounce than a small child on a trampoline - see the scope plot with the wild ringing at the point of contact.
Even this stability was only got by reducing the the pressure change from 450 grams to 250 grams. The output voltage for the drilled discs has been multiplied by 1.8 to correct for this.
I think these plots show that even under this considerable mistreatment the piezos should give an adequate life.
Mike
Even this stability was only got by reducing the the pressure change from 450 grams to 250 grams. The output voltage for the drilled discs has been multiplied by 1.8 to correct for this.
I think these plots show that even under this considerable mistreatment the piezos should give an adequate life.
Mike
|
Under-bed Method January 19, 2018 03:27AM |
So I got completely sick of how crap the BL Touch is for both accuracy and repeatability on sensor/bed height detection let alone the fact it will randomly drop the probe and ruin prints.... Considering this crappy sensor uses a bit of heatshrink to set the probe height and length to hall effect sensor, when it is mounted just in front of the hot air coming off a hotend, it is just a crap design all round. This is made worse for me as I use a lot of different filament types, a heated enclosure and change the build surface dependant on the material being printed which means I am constantly wanting to run a calibration before each print. So I have an Ir sensor that is "ok" but this suffers also from changes in the build surface and I end up having to reset the Z-offset dependant on what material is being used.
So I decided to design up a solution for my Tevo Little Monster delta printer to suit the 27mm piezo disc for under-bed detection. The design incorporates some heat shielding/insulation built into the piezo holder as well as keeping the piezo as far away form the heated bed as possible. Initial tests show on the Tevo LM running for 6 hours, with a bed temp of 150 * Celsius and full enclosure with air temp of 60* Celsius, that the piezo is not exceeding 39* Celsius (keeping in mind the ambient room temp is 32* Celsius to begin with), in this situation I have to tape the BL Touch pin up and not dream of running a calibration before hand if the printer has been printing before hand. The piezo disc I bought locally (taken out of a piezo buzzer from Jaycar) seems to have plenty of trigger sensitivity with the springs I am using for bed levelling (springs are out of AA battery holder spring for the negative side of the battery) and showing no sign of being affect by the heat. I have not got a piezo amp board and multi sensors yet, just recording and watching the piezo output on a scope at the moment and I am able to the faintest touch on the bed with just one sensor (I mean a piece of filament touching it faint) without any noise from the motors moving around.
The lower part holding the piezo sensor is attached using existing M5 thread in the bottom plate and is attached from under the plate from outside the printer and holds the piezo inside via a screw in cap. It is printed with only 15% infill with a full layer every 5 layers at 0.8 layer height and does touch the plate with the bulk of the part (except the thread), this ensures maximum insulation from the heat due to creating multiple layers of overlapping air pockets all round the piezo sensor. The piezo sensor is setup to be flexed by a M2 bolt in it's centre with it supported around its edge and clamped in place by the parts screw in lid. It has a slot through the side to allow the cables to come through with the ceramic surface facing the bolt (seems to be the most sensitive way of setting up the sensor).
The top part goes through the top plate which holds the heated bed, it is held in place by a matching "nut" which is sized to fit down the centre of the spring between the bed and the bottom plate in order to hold the spring in place as this replaces the bolt that was once there as I use 3 bolts to hold heated bed in place and set the bed height and level, then using the other 3 left over holes for mounting the piezo bed level sensors (also made a couple of different sized nuts to try different springs).The M2 bolt then runs down through from the top to part that has a M2 thread in it to allow you to set the initial pressures and height of the piezo triggering which of course can also be used to fine tune the trigger sensitivity on top of any adjustment of the sensor board itself. This means you can leave an air gap between M2 bolt and piezo, or have it just touching the sensor or add a little constant pressure which all changes the point the piezo triggers and the intensity of the trigger. The lower part then has a cap on it to allow you to easily change out a busted piezo sensor and the centre of the M5 thread is hollow and large enough (without a thread) to allow the M2 bolt, affixed to the build plate, to move up and down easily but without any sideways movement.
With the stainless bolt I am using I am not getting more than a few of degrees of heat coming down into the piezo chamber from the 60 degree air temp in the enclosed print chamber so no issues with heat transfer there. By all accounts I should be able to set the sensitivity of the board pretty darn high then just adjust the M2 bolt to set when the sensor/board is set off reducing the impact pressure on the build plate, I may even be able to remove the springs altogether although they are helpful for getting the bed perfectly square to the uprights on my delta.
I am looking forward to being able to run a decent calibration test before each print and have no concern with changing out the build surface when ever I feel like it, now to get over the $50 cost to get the piezo board and sensors sent over from the UK.....
So I decided to design up a solution for my Tevo Little Monster delta printer to suit the 27mm piezo disc for under-bed detection. The design incorporates some heat shielding/insulation built into the piezo holder as well as keeping the piezo as far away form the heated bed as possible. Initial tests show on the Tevo LM running for 6 hours, with a bed temp of 150 * Celsius and full enclosure with air temp of 60* Celsius, that the piezo is not exceeding 39* Celsius (keeping in mind the ambient room temp is 32* Celsius to begin with), in this situation I have to tape the BL Touch pin up and not dream of running a calibration before hand if the printer has been printing before hand. The piezo disc I bought locally (taken out of a piezo buzzer from Jaycar) seems to have plenty of trigger sensitivity with the springs I am using for bed levelling (springs are out of AA battery holder spring for the negative side of the battery) and showing no sign of being affect by the heat. I have not got a piezo amp board and multi sensors yet, just recording and watching the piezo output on a scope at the moment and I am able to the faintest touch on the bed with just one sensor (I mean a piece of filament touching it faint) without any noise from the motors moving around.
The lower part holding the piezo sensor is attached using existing M5 thread in the bottom plate and is attached from under the plate from outside the printer and holds the piezo inside via a screw in cap. It is printed with only 15% infill with a full layer every 5 layers at 0.8 layer height and does touch the plate with the bulk of the part (except the thread), this ensures maximum insulation from the heat due to creating multiple layers of overlapping air pockets all round the piezo sensor. The piezo sensor is setup to be flexed by a M2 bolt in it's centre with it supported around its edge and clamped in place by the parts screw in lid. It has a slot through the side to allow the cables to come through with the ceramic surface facing the bolt (seems to be the most sensitive way of setting up the sensor).
The top part goes through the top plate which holds the heated bed, it is held in place by a matching "nut" which is sized to fit down the centre of the spring between the bed and the bottom plate in order to hold the spring in place as this replaces the bolt that was once there as I use 3 bolts to hold heated bed in place and set the bed height and level, then using the other 3 left over holes for mounting the piezo bed level sensors (also made a couple of different sized nuts to try different springs).The M2 bolt then runs down through from the top to part that has a M2 thread in it to allow you to set the initial pressures and height of the piezo triggering which of course can also be used to fine tune the trigger sensitivity on top of any adjustment of the sensor board itself. This means you can leave an air gap between M2 bolt and piezo, or have it just touching the sensor or add a little constant pressure which all changes the point the piezo triggers and the intensity of the trigger. The lower part then has a cap on it to allow you to easily change out a busted piezo sensor and the centre of the M5 thread is hollow and large enough (without a thread) to allow the M2 bolt, affixed to the build plate, to move up and down easily but without any sideways movement.
With the stainless bolt I am using I am not getting more than a few of degrees of heat coming down into the piezo chamber from the 60 degree air temp in the enclosed print chamber so no issues with heat transfer there. By all accounts I should be able to set the sensitivity of the board pretty darn high then just adjust the M2 bolt to set when the sensor/board is set off reducing the impact pressure on the build plate, I may even be able to remove the springs altogether although they are helpful for getting the bed perfectly square to the uprights on my delta.
I am looking forward to being able to run a decent calibration test before each print and have no concern with changing out the build surface when ever I feel like it, now to get over the $50 cost to get the piezo board and sensors sent over from the UK.....
|
Re: Piezoelectric disks for Z contact detect and bed levelling January 19, 2018 02:32PM |
Your design should work quite well but I think that having the build stage sitting on three quite long 2mm diameter legs is a little worrying as the stage can be pulled about in the XY plane by the nozzle. Another possible problem is that vibration may cause it to shake.
On my delta printer I used 6mm diameter sleeves snugly fitted through 2mm thick PTFE washers to prevent XY movement. You can see this (maybe not clearly thought ) on the picture below - three sensors at 120 degree intervals and three of the sleeves I mentioned between them.
Let us know how well it works.
Mike
p.s., double posting is regarded as not good in this forum
Edited 1 time(s). Last edit at 01/19/2018 05:30PM by leadinglights.
On my delta printer I used 6mm diameter sleeves snugly fitted through 2mm thick PTFE washers to prevent XY movement. You can see this (maybe not clearly thought ) on the picture below - three sensors at 120 degree intervals and three of the sleeves I mentioned between them.
Let us know how well it works.
Mike
p.s., double posting is regarded as not good in this forum
Edited 1 time(s). Last edit at 01/19/2018 05:30PM by leadinglights.
|
Re: Piezoelectric disks for Z contact detect and bed levelling January 19, 2018 11:42PM |
I might of been clear, the build plate on the Tevo LM is held on by 6 M5 bolts that use a set of spaces to hold the build plate away from the base plate (hexagon build plate). I have removed 3 of these M5 bolts and replaced the spacers with springs to allow me to adjust the bed perfectly square with the towers but adjusting the remaining 3 M5 bolts. The holes let over from removing 3 of the M5 bolts have got the new piezo sensors holders mounted into them using the existing M5 threads in the base plate. The build plate does not sit on the M2 bolts going onto the piezo sensor, the build plate sits on and is held up by 6 springs, these spring are compressed by 3 M5 bolts to adjust bed height/angle. The M2 bolts are attached to the build plate inside a printed mount for them (affixed top and bottom to build plate) with the ability to adjust their height with fine detail due to the fine thread pitch which allows you to adjust if/when/how much they are touching the piezo sensors. In other words, I can adjust if the sensors are in constant contact with the piezo sensor and how much "preload" that contact is, or I can adjust them to sit away from the plate so the plate has to move down a set distance in order to have the M2 bolts touch the peizo sensor. This allows me to tune the M2 bolts till they are sitting a poofteenth away from the piezo sensor so not movement from the head and motors cause it to create a signal, only a set amount of movement down will, or alternatively I can put some "preload" onto the sensor with the M2 bolts which means it takes more force to get a reaction out of the sensor, or I can set it to anything in between.
This takes the design you picture to a new level as there is no connection between the plate height adjustment mechanism and the probe hitting the piezo sensor not is there any flex or leverage issues to deal with as well as insulating the piezo away from the effects of the heated bed and heated enclosure on both the piezo and the bed levelling arm/adjuster.
P.S. Last time I checked, one was a post asking if freebies were still available in another thread with a basic outline of the design and the other is a detailed design about my specific piezo design in the thread dedicated to piezo bed level design, but thanks for letting me know "self-imposed moderator", I wasn't aware that everyone read every single thread in this forum and was such a big issue but I will keep this in mind in future when I look to share design info[[/size]/s] 
Edited 1 time(s). Last edit at 01/19/2018 11:51PM by Redemptioner.
This takes the design you picture to a new level as there is no connection between the plate height adjustment mechanism and the probe hitting the piezo sensor not is there any flex or leverage issues to deal with as well as insulating the piezo away from the effects of the heated bed and heated enclosure on both the piezo and the bed levelling arm/adjuster.
Edited 1 time(s). Last edit at 01/19/2018 11:51PM by Redemptioner.
|
Re: Piezoelectric disks for Z contact detect and bed levelling January 20, 2018 01:09PM |
Having not read Redemptioner's post fully and having misunderstood a couple of details I though it was about time I expiated some of my sins and clarified details of my underbed sensor that may not have been truly clear in past posts.
Firstly, the piezo replaces the three (or four) bed leveling screws on a Cartesian printer or some Delta printers, it is not fitted in addition to the mountings. The piezo is also not used as a microphone but acts as a true pressure sensor in series with a capacitor. The parallel mechanism and pressure pad can be replaced by a simple bar but that would introduce some compliance and have more pressure on some parts of the piezo than others. Having the piezo actuated by a fall in pressure allows the build stage some overtravel if the nozzle moves too far down - see the diagram. The preload is used mostly to correct for errors and differences in printing so that the piezo is under a little pressure even if there is an overtravel condition - the piezo will remain held and won't fall out.
Incidently, after nearly two years and many hundreds of prints, only one out of 9 piezo discs (3 printers) has lost any of its mojo - and that one is down from 8V to 7V so is not ready to retire.
Mike
Firstly, the piezo replaces the three (or four) bed leveling screws on a Cartesian printer or some Delta printers, it is not fitted in addition to the mountings. The piezo is also not used as a microphone but acts as a true pressure sensor in series with a capacitor. The parallel mechanism and pressure pad can be replaced by a simple bar but that would introduce some compliance and have more pressure on some parts of the piezo than others. Having the piezo actuated by a fall in pressure allows the build stage some overtravel if the nozzle moves too far down - see the diagram. The preload is used mostly to correct for errors and differences in printing so that the piezo is under a little pressure even if there is an overtravel condition - the piezo will remain held and won't fall out.
Incidently, after nearly two years and many hundreds of prints, only one out of 9 piezo discs (3 printers) has lost any of its mojo - and that one is down from 8V to 7V so is not ready to retire.
Mike
|
Re: Piezoelectric disks for Z contact detect and bed levelling April 13, 2019 12:18PM |
For some time now I have been getting a backlog of projects and have been trying to get some of them finished. An improved way of extracting a good trigger has been a principle objective for at least the last year and I am beginning to get some useful results.
Instead of trying to filter the incoming data I have tried to use a few checks to determine if a nozzle contact is good, looking for a clean background signal followed by an increase at a linear rate between two limits until a trigger voltage is achieved. There is a secondary trigger to signal that nozzle contact had occured but could not be relied on for calibration. This arrangement allows software in an external controller to probe again or signal that the nozzle contact was poor for whatever reason.
Early tests show that plastic on the nozzle is readily detected and mechanical noise that has not been filtered out will again cause a bad contact signal - again before any damage is done to the build stage. Although it wasn't the intention, the filtering method used does seem to be good at cancelling mechanical noise
Th plots below are mightily kludged, but only because there was no practical way to extract all of the data from the 8 pin PIC MCU. However, the blue line (ADC data) is genuine and has been fed back as a data file into the Simulator in Microchip MPLab X to produce the mean of 8 point data and the slope data. Trigger delay is from an oscilloscope.
In the plots, the green line is the running mean of the last 8 points from the ADC (Blue line) The slope is derived from the last 9 points - the middle point ofr pivot point is discarded. Not shown in these plots in the baseline which is the running mean of the last 40 values. The nozzle was advanced at 1mm/sec and the trigger was reached 30 ms after initial contact which occured after 30 microns of movement.
When I have got this completely working I will put the program and circuit diagrams on an open source repository like Sourceforge or GitLab in the event anybody wants to develop it further. Meanwhile, ther is a little more information on [reprap.org] and the piezo mechanism is shown on [reprap.org]
Mike
Instead of trying to filter the incoming data I have tried to use a few checks to determine if a nozzle contact is good, looking for a clean background signal followed by an increase at a linear rate between two limits until a trigger voltage is achieved. There is a secondary trigger to signal that nozzle contact had occured but could not be relied on for calibration. This arrangement allows software in an external controller to probe again or signal that the nozzle contact was poor for whatever reason.
Early tests show that plastic on the nozzle is readily detected and mechanical noise that has not been filtered out will again cause a bad contact signal - again before any damage is done to the build stage. Although it wasn't the intention, the filtering method used does seem to be good at cancelling mechanical noise
Th plots below are mightily kludged, but only because there was no practical way to extract all of the data from the 8 pin PIC MCU. However, the blue line (ADC data) is genuine and has been fed back as a data file into the Simulator in Microchip MPLab X to produce the mean of 8 point data and the slope data. Trigger delay is from an oscilloscope.
In the plots, the green line is the running mean of the last 8 points from the ADC (Blue line) The slope is derived from the last 9 points - the middle point ofr pivot point is discarded. Not shown in these plots in the baseline which is the running mean of the last 40 values. The nozzle was advanced at 1mm/sec and the trigger was reached 30 ms after initial contact which occured after 30 microns of movement.
When I have got this completely working I will put the program and circuit diagrams on an open source repository like Sourceforge or GitLab in the event anybody wants to develop it further. Meanwhile, ther is a little more information on [reprap.org] and the piezo mechanism is shown on [reprap.org]
Mike
Sorry, only registered users may post in this forum.