Major differences between IR Sensors and FSR Sensors for bed leveling?
-
Doug, will do. If these prototypes work as well as I expect, then we may use some of the remaining PCBs to build a small number of pre-production boards. We'll only supply these to users with lots of Duet + delta experience, but you definitely qualify.
-
Might I also be considered?
-
I would be interested also!
Planning a new delta with Duet Ethernet and Robotdigg 2040 parts. It will be interesting to see how magnetic balls/arms compare with the traxxas versions I have been using.
-
For the OP, one consideration in sensors is tilt of the effector, or anything that alters its movement. Any probe that doesn't use the nozzle will show errors based on any tilt in your system. In that aspect I think the FSR / Piezo type probes that use the nozzle have advantages (tilt still has some impact, but it's largely mitigated).
One thing to do, regardless of the type of probe you use, is to check its trigger height at varying points around your bed to be sure it's triggering the same regardless of the point probed.
-
I'm also on Bord with a test PCB if possible. ~Russ
-
That PCB effector looks novel combining it all into one part. Tolerance junkies could probably machine a little aluminum plate to layer it with.
My next delta will be a large scale one, so I'm definitely looking for a good effector. I have RobotDigg effector and carriage parts currently, but the PCB kit looks great.
I'm new to DIY Deltas, my last built printer was the Rostock Max V2 with a few mods, so I'm really looking forward to finishing one entirely built from non-kit parts. My Rostock was manually calibrated by putting the hot end top against a tiny scale for the three towers and middle - so I'm sure auto-calibration in any form will be easier.
I've done a lot of sculpture printing (from the Nefertiti hack data), so my Rostock is nearing 60 days print time, but I'm moving to parts that need to be more precise, so precision for the bottom layer is coming into play now.
Thanks for all the input, I'm sure I'll post more questions in the next few weeks.
-
I have used basically every probing system developed to date. I have a Rostock printer set up with SeeMeCNC's accelerometer probe, David's IR probe and FSR probing with my new(wish) mount design so I can compare apples-to-apples. Of course I do have to lock the bed from moving to use the accelerometer. I'm currently putting together a piezo probe based on the discussion on the google group to test. I will also test David's new probe when ready.
At this point, I get the best results from correctly setup FSRs and have these on 7 delta printers and two Cartesian printers. My Prusa has his PINDA probe.
Here are my thoughts on probing…
Probing using the nozzle tip as the contact point minimizes effector tilt issues during probing and is arguably a better way to probe. However, I do probe with my hot end hot and many complain about dots of filament on their beds. I simply swipe the bed with the edge of a long thin spatula and probe. The molten filament gets pushed out of the way and I don't believe (although I have no data to back it up other than 1000s of successful probes/auto calibrations) that it interferes with the probe contact offset. Probing with a cold hot end requires that you make sure there is not a little tuft of hardened filament on the nozzle tip - a much more daunting task than you might think. But the deal breaker for me is the ability to probe and auto-calibratebefore every print - I set that up in my g-code header. My "open mesh" fly reel parts require perfection to get 100% yield and after several years of effort, RRF auto-calibration for each print gives me exactly that, 100% perfect parts 100% of the time.
I like the IR probe and the non contact aspect has appeal but it would not work for me. I print on PEI and over the years have patched/repaired minor imperfections with CA glue (I describe how to do this on the SeeMe forum). Unfortunately, CA has different IR reflectance characteristics and I would get odd results. These patches are nearly invisible and a couple of my early PEI plates with 3+ years of continuous use have a lot of patches!
The accelerometer probe is very sensitive to the bed surface (hardness) and even different pieces of PEI resulted in having to completely retune the probe - something I can do in the firmware I developed in the controller used to interface to Duet but not something the average user can do. And it is time consuming and a PIA. I suspect that the 3M tape used to adhere the PEI results in this variability as some of my PEI is held down by 3/4" strips, some by 2" strips, some by a single sheet and others had the adhesive applied at the factory. Each of these behaves differently as do every other bed surface. I finally gave up on accelerometer probing due to these inconsistencies. If you only have 1 bed / surface and it is "hard" you might be Ok with it.
I do also always calibrate with a hot bed. These large delta beds move quite a bit as they heat/cool. I HAVE data on this for Onyx movement and simple borosilicate glass with a Kapton heater movement and it can be as high as 2mm in the center for 90°C bed temps and depending on how you mount the bed/heater.
I've helped literally 100s of people with FSR setup. Those how have had problems can almost always be attributed to one of 2 problems:
-
the bed is too flexible. This was the case with early Mini Kossel users who simple had a borosilicate class bed support in 3 places. This flexes quite a bit. The FSR mount plate I designed and on my blog supports the bed so there is very little flex. That gives me very consistent probe results across the entire bed.
-
the FSR mount system binds in some way. In my earlier "printed FSR plunger mount" design this could be a problem if one did not carefully prepare the mount and plungers and position them.
Once you have these two issues worked out, FSRs probe reliably. I have 2 printers with FSR probing that are >2 years old and have never had a mistrigger or other probing problem. They work and work well.
-
-
I have setup a delta with mhackney's FSR mounting plate and followed his installation guide and it is first layer perfection every time.
-
Here's a link with more info: New FSR PLate mounting system
-
Micheal, thanks so much for that really comprehensive rundown of probing. For deltas this is a critical requirement - I am looking forward to your review of our new method!. #itscomingsoon #spoilers #notsurewhyIamusinghastags
-
so Micheal, good stuff, you went over a lot of the concerns and thoughts i'm having at the moment.
I also use PEI so its good to know about CA glue.
let me ask you a question. cant we just move the 3 point FSR to the endefector. I would think if we did that we can get the same or better results as the plate dezighn.
but this just makes more scene. then i would think the weaker beds would be less effected?
any how. just some thoughts.
also if you can run this test for me so i can add your data that would be greate:
https://www.duet3d.com/forum/thread.php?id=1330
Thanks!
~Russ
-
Russ, adding the FSRs on the end effector is similar to what the guys are doing with peizos on the effector, just that piezos are in some ways more suitable in that application (mainly around pre-load)
-
And being able to drill the piezo means mechanically the system works far more directly, there are FSR hotend designs with dual fsrs one either side of the hotend. If there is an fsr out there with a hole in it then it might be much more possible. But its also the sensitivity at 50g of force for a typical FSR to reliably record a trigger there has to be quite a bit of compliance in the hotend to get this to happen resulting in the wobbly nozzle, which we have eliminated from the piezo sensor modules by pre-loading the much more sensitive piezo disc, also because of the hole the force is transmitted in the most direct way possible to the piezo meaning 10g is more realistic probing force, Mike thinks he can get down to 1g!
The piezo system is much more sensitive to mechanical noise than an FSR though, so probing has to be done with low jerk/accel. I haven't found this a problem, I want my probing to be accurate not fast.
-
I may do the FSR sensors after re-reading this thread.
I wired up all of my motors and sensors last night with the exception of the effector wiring and extruder.
I ended up shortening all of my stock wires and sleeving and crimping them with the included Duet WiFi connectors in addition to labeling all the plugs with Dymo heat shrink. Is there any downside to sleeving with paracord? I had some for a different project and it seems to work almost the same as tech flex - just with a fabric feel.
I have the melamine plate from Trick Laser that I bought with an FSR kit and I may go ahead and use it on the new build instead of upgrading my old Rostock for now.
The Onyx bed is mounted to the melamine and it has to "float" on the three FSR sensors on top of silicone pads that are positioned beneath each tower, correct?
Is there a reason the FSR bed has a single tab that is different from the other two? Does the different tab need a specific orientation to the X, Y, or Z tower?
Also, what are people using to secure the glass? I usually use 3x binder clips - but I would like to find a lower profile method of attaching the glass securely if anyone has recommendations.
-
Could you use the type of clip that they use on picture frames
-
@Russ:
so Micheal, good stuff, you went over a lot of the concerns and thoughts i'm having at the moment.
I also use PEI so its good to know about CA glue.
let me ask you a question. cant we just move the 3 point FSR to the endefector. I would think if we did that we can get the same or better results as the plate dezighn.
but this just makes more scene. then i would think the weaker beds would be less effected?
any how. just some thoughts.
also if you can run this test for me so i can add your data that would be greate:
https://www.duet3d.com/forum/thread.php?id=1330
Thanks!
~Russ
Russ, this is a good idea and one I experimented with. This is very much like the electronic probes for milling machines work. It is a little challenging to construct a mechanism with X-Y movement constraints and virtually no constraint in Z when the entire apparatus must move all around! I know other have tried this with varying degrees of success. In the end, I felt a bed mounted solution would be less problematic.
-
I may do the FSR sensors after re-reading this thread.
I wired up all of my motors and sensors last night with the exception of the effector wiring and extruder.
I ended up shortening all of my stock wires and sleeving and crimping them with the included Duet WiFi connectors in addition to labeling all the plugs with Dymo heat shrink. Is there any downside to sleeving with paracord? I had some for a different project and it seems to work almost the same as tech flex - just with a fabric feel.
I have the melamine plate from Trick Laser that I bought with an FSR kit and I may go ahead and use it on the new build instead of upgrading my old Rostock for now.
The Onyx bed is mounted to the melamine and it has to "float" on the three FSR sensors on top of silicone pads that are positioned beneath each tower, correct?
Is there a reason the FSR bed has a single tab that is different from the other two? Does the different tab need a specific orientation to the X, Y, or Z tower?
Also, what are people using to secure the glass? I usually use 3x binder clips - but I would like to find a lower profile method of attaching the glass securely if anyone has recommendations.
saffi, not sure about paranoid sleeving's melting point. That would be the only concern (flammability too).
Yes, the Onyx and bed are mounted to the the melamine plate I designed and this floats on the 3 FSRs on silicone pads. This arrangement is rigid, insulates the FSRs from the bed heater, doesn't over constrain the sensors, and the sensors are positioned as far outside of the bed diameter as possible (at least for the RMax).
Yes, that single tab accommodates the Onyx LED and wiring. This plate was designed for the SeeMeCNC Onyx heater but can be used for other heaters.
I use three of the hold downs supplied with the printer-one at each "ear". These ears are designed so you can use printed hold downs, or even bent aluminum if you prefer. I do use 3X binder clips on my RMax V3 - it has an earlier version of the plate without the ears. They work fine too. I think the lowest profile would be 1/16" (1mm) aluminum strips bent and drilled. I'll try to make some to photograph on my blog this weekend. Maybe I can convince Brian@TrickLaser to include 3 with the plate!
-
Could you use the type of clip that they use on picture frames
Yes absolutely. They may require a little more bending to accommodate the melamine.
-
Micheal, thanks so much for that really comprehensive rundown of probing. For deltas this is a critical requirement - I am looking forward to your review of our new method!. #itscomingsoon #spoilers #notsurewhyIamusinghastags
"Hello, I'm Michael and I'm a probaholic…" I really have spent more time with probing than is reasonable. In the beginning it was simply to make sure I could print stuff like this reliably and repeatably:
That "mesh" requires near perfection for first layer height (and bed adhesion too). It wasn't until I climbed aboard the dc42/Duet train with FSRs and PEI print surface that I was able to print these parts (and even more challenging parts) with 100% yield. A first layer height off by even .03mm is "visually offensive" to the human eye - it looks fat and squished.#ontheroadtoprobingrecovery #lookingforwardtothenewprobetotest
-
Russ, adding the FSRs on the end effector is similar to what the guys are doing with peizos on the effector, just that piezos are in some ways more suitable in that application (mainly around pre-load)
ah yes, FSR might not be the one here, it dose seem that piezoelectric is a better choice all around. That is if you dont want movement in the attachment. But a flat piezoelectric attached to the PCB in the correct configuration might work. just some random thoughts… using something like this: http://www.imagesco.com/piezoelectric/index.html
Russ
mZp-PEcHKindauY22CC1-Xw