Why the Prusa XL should use OPTICAL calibration
Vložit
- čas přidán 16. 05. 2024
- In this video we discuss mechanical vs optical calibration for multi extruder 3D printers like dual extruders, IDEX, and tool changers. Specifically, we introduce some pitfalls of mechanical calibration and compare it to the advantages of optical.
Ember Prototypes CXC Calibration Tool:
www.emberprototypes.com/produ...
0:00 Intro
2:08 JLCPCB sponsor
3:15 Why do you need calibration
3:59 The traditional line method
4:47 How mechanical calibration works
5:25 How optical calibration works
7:14 Mechanical calibration - nozzle cleanliness
9:21 Mechanical calibration - concentricity
13:03 Mechanical calibration - precision, misalignment and damage
16:18 Advantages of mechanical calibration
17:47 Conclusion
19:52 Outro
JLC3DP online 3D printing service starts at $0.3, save up to $60 from : jlc3dp.com/?from=EmberPrototypes
Multi Extruder 3D Printer Calibration Simplified (CXC) @ModBotArmy
• Multi Extruder 3D Prin...
Is The Snapmaker J1 The Ultimate IDEX Experience? @ModBotArmy
• Is The Snapmaker J1 Th...
Prusa XL 5 Toolhead Live Build @3DMakerNoob
czcams.com/users/liveyy7NDWRp...
Expensive vs Cheap 3D Printer Nozzles - Let's Look Inside @LostInTech3D
• Expensive Vs Cheap 3d ...
Mini Lathe Project: 3D Printer Nozzle @CNCKitchen
• Mini Lathe Project: 3D...
TITAN FUNDAMENTALS: Edge Finding Work Offsets Manually @TITANSofCNC
• TITAN FUNDAMENTALS: Ed...
CNC automated edge finding / zeroing with touch plate @ChristianKnuell
• CNC automated edge fin...
Jubilee Toolchanger TAMV
jubilee3d.com/index.php?title...
GD&T Concentricity Basics
www.gdandtbasics.com/concentr...
Hot End 3D Model from GrabCAD
grabcad.com/library/mk7-hot-e...
emberprototypes.com/
/ emberprototypes
/ emberprototypes - Věda a technologie
Best 20 min ad, I’m sold.
Thanks 🙂 tried to make it the least "ad-y" as possible...something that's probably impossible because I sell an optical product 😅
After seeing Steven Hawes’ use of cameras for the Lumen PnP’s calibration, i was thinking of this too!
Although to for offsets which makes sense in hindsight.
*I was thinking* to use it for Nozzle Cleanliness, and also *wear*. Granted with a Diamondback Nozzle (or whatever one of the decent non-insert based materials was) this is less of an issue, but I thought it would be interesting to see how abrasive filaments wear down a nozzle and know precisely when to swap.
Although one could have it set it clean the nozzle arbitrarily before every print, using a camera to see if it needs to would potentially save time *and* tell if it actually got completely clean.
Yup, lots you can do with optical inspection! I personally use the CXC mostly in a manual fashion (camera-assisted) and even then, I've been able to inspect and see when my brass nozzles needed changing (the holes were very oblong and uneven wear was very obvious).
Fantastic video! I'm currently planning to build an IDEX printer and this seems to be the best tool for the calibration.
Thanks for watching!!
Excellent video! Covers tradeoffs pretty well.
I think it's a matter of degrees and practicality. All methods have tradeoffs in complexity, accuracy, ease of use, repeatability, etc. Practically it only has to be good enough to, say, 5 or even 10 um to not really pose a problem for 3d printing. A small angle between extruders may not pose a practical problem, for example, depending how it's measured at the nozzle.
Optical certainly has its advantages. It also has issues. For example, additional computing hardware and inspection beyond what a 3d printer can do itself. I agree it's not such a big deal given how cheap single board computers have gotten, but you need that software, and there's still uncertainty in image analysis. And a dirty nozzle isn't a problem only for electrical and mechanical methods, it's also hard to automatically analyze optically.
Full disclosure, I'm a J1 owner and think for robustness, and simplicity, I'm quite satisfied with its contact method of XY calibration. It just works. The steps for success aren't difficult to reproduce, though you have to bear in mind the limitations. For optical, more precision is possible, and ideally wouldn't require any user intervention or computing hardware to analyze and apply, but can certainly be more precise and less ambiguous. Plus being able to get a close look at your nozzle is just useful and interesting.
But for me, I'll stick with "easy and good enough!" 😊 I think we can all agree, the caveman method of test prints is kind of awful.
100% agree here that both have their own tradeoffs. It's true that optical requires additional computing power, and in the case of manually assisted camera calibration - requires an additional computer or laptop nearby to help calibrate (unless you have something like Klipper or a Pi that ties in directly to Mainsail or something).
In terms of image analysis, there isn't really additional uncertainty *unless* you are implementing TAMV for automation - which in my honest opinion, isn't really necessary for 99% of scenarios. It's cool yes, but you get diminishing returns at that point because manually assisted camera calibration already reduces the calibration time dramatically and improves accuracy for little effort. Dirty nozzles are a problem for optical, but at least it's a bit easier to catch than the alternative methods and can tell you things about your nozzle that you would otherwise not know.
Glad to hear that the Snapmaker XY calibration is working for you! Like I said, I'm pretty sure most of these machines will work, but there may be edge cases that don't and in those scenarios, can probably be traced back to some of the pitfalls talked about in the video.
Thanks for watching and contributing your thoughts! And yes, line patterns are the worst 😂
Here are my two cents:
I have a five-head prusa XL. Speaking from experience, the mechanical calibration on it got things about 95% aligned. However one of the toolheads is off by about 0.15mm or so in one direction. Given the general inaccuracies in 3d printing it's fine, but there is a barely noticeable hairline gap on top surfaces from it.
My go-to nozzle for everything is a DiamondBack. They have an amorphous diamond tip pressed into a brass housing and they have pretty amazing properties and they're also extremely well-made. However, they have their logo embossed on the side of the nozzle which makes the cone-shaped face eccentric. I haven't tested this, but I imagine their logo would cause problems with Prusa's mechanical calibration procedure.
That sounds totally plausible. Optical calibration would probably work perfectly for dialing that in.
What nozzle size are you using? Offset errors are generally less visible the larger the nozzle size as well. I know on my E2, 0.15mm error would definitely be observable.
Is there an implementation for the duet without the need for extra compute? Great video btw.
Thanks for watching! If you want to fully automate it with TAMV, I believe you need a Pi tied to the Duet. But if you want to do it manually (camera-assisted), then all you need is a laptop that you can plug the camera into and run the software with. This is actually how I calibrate all my multi-extruders because its easy and doesn't require a lot of setup.
great vidoe
Does the optical tool have customized opencv setup? Also, is the cross hair overlay builtin, in the calibration routine. What is the min focusing distance from the lens surface, total thickness of the camera housing(with lens), the overall dimensions. It could interesting, if you have the ability to detect angularity of the nozzle(like in a delta 3d printer nozzle). Seems like an interesting tool, esp. with a capable default plugin/code for opencv.❤👍
Yes, there are open source versions of TAMV which use OpenCV to automate calibration.
Doing it manually but camera assisted is also very easy and is how I do it personally.
You can find more info here: www.emberprototypes.com/products/cxc
I use TAMV on my tool changer. As long as the nozzles are clean, it works beautifully, and is much faster than Prusa's method. However, due to the nozzle orifice needing to be *very* clean to get good circle detection, it can be a pain. That's where manual optical calibration can be advantageous. Prusa could easily integrate the entire manual optical calibration process into their touch screen, but I think I think their ultimate goal though is ease of use, and for the machine to have the smarts and not burden the user with it. Definitely some trade-offs with all methods. Great explanation of different methods in this video. Nice work!
Tradeoffs is the name of the game haha! And yes, camera-assisted manual calibration is VERY easy, and I'm sure that if they wanted, they could integrate it into the UI and make it incredibly easy to do. The user would literally just have to "drive" the nozzles into the center of the crosshairs and the compute can do the rest. If there's a permanent or alignable place for the camera then that's even easier. Thanks for watching!
@emberprototypes I was rhinking of buying this for my proforge 4 build and saw this video and is that a proforge 4 at the 16:40 mark? If so, what hotends are those and where can i get them?
Highly recommend for the Proforge! And yes that's a PF4, but a heavily modified one from a secret company that I can't say any more about 😅
I reckon you could get the camera to do Z offsets too just by comparing the size of the primary nozzle orifice then moving the bed up or down on the 2nd tool until the size of the orifice matches.
Yup you totally can. You just need to calibrate for mm per pixels before hand.
However, the only reason I don't personally do this is because I don't think this will be reliable over time with uneven nozzle wear.
You can also try to calibrate Z offsets by looking for the best focus plane, but again, might be susceptible to wear and nozzle type/geometry.
I've heard of some people doing this though - and for brand new nozzles, should be no problem.
I’d have to rewatch/read up, but i *think* this is how the Open Source PnP Machine “Lumen PnP” did so.
Could maybe even adapt it somehow.
@@ericlotze7724 calibration for mm per pixels is a very common way to use cameras to measure/detect things with machine vision. Again, I think it can work, but I do think there are some caveats that could make it a bit finicky. I've heard that while TAMV itself is great, it also doesn't *always* work. I've found that manual camera-assisted is easy and quick enough to not warrant machine vision for most applications. It's not as cool, but the human eye is amazing at detecting things quickly that machine vision has trouble with. That being said, once we set up kTAMV on our Klipper tool changer, I'll probably make another video on it 🙂
Nozzle wear is precisely why this works best. Because you would be comparing the size of the aperture of the nozzle. As the nozzle wears and the surface grinds away from the bed the aperture would reduce in size, so using some clever machine vision you could just raise the bed X amount until the aperture fills the circle.
@@emberprototypes
@@Festivejelly I think practically it is not as easy as it sounds. Nozzle wear can also affect the size and shape of the orifice itself which would mess up your results. Not saying it isn't possible, just saying that there are probably subtleties that will give erroneous results unless you have a very sophisticated machine vision algorithm.
Did you tried your optical system on the XL? So far i don't have any issue's with my 5 TH XL (only had to do the mech. calibration once so far at the beginning after my upgrade). Looks very interesting to see how this can be implemented in the Firmware. Thanks for the video.
Thanks for watching! I personally don't have an XL - but I have had a handful of people reach out to me to use it on their machine with success. I presume that for whatever reason, their machines were suffering from one of the issues I talked about in the video.
@@emberprototypes I really wouldnt know how to implement that on my XL. But interesting to hear that some people got it to work.
@@Leo-yh1lj it's pretty simple actually. The optical tool we sell is universal. All you need to use it is a laptop, and the ability to move the gantry manually and override the offsets manually...most printers let you do this already!
@@emberprototypes I see. Not that i have a laptop, but also manually putting in the value's isn't my thing. But who knows, if i got really big issue's in the future that is hard to solve (and related to these off-set settings) i will look into it.
@@Leo-yh1lj yup! If what you have is working then no need to change anything 🙂 the manual stuff isn't that bad, and the plus is that doing it that way works across all printer brands and isn't tied to firmware or locked down systems
I'll also repost my comment here: Would love to see an OctoPrint plugin for us Marlin users! This is very cool, and that convenience for installation and control would be huge.
Do you mean an OctoPrint plugin for using the CXC?
I have a Prusa XL 5 Tools. Any idea how can I make it jog 0.1mm? The default move is 1mm minimum.
Prusa is weird, you have to use Pronterface to move smaller than 1mm - most other machines let you control it to smaller moves.
I put a request in with them to add
@@emberprototypes Figured out how and ordered a set of your Camera calibration tool. I can make 0.1mm moves with OctoPrint connected via USB-C to the back of the Prusa XL (tiny hole, didn't know it was for USB-C until looked closely). OctoPrint Windows or Android works to make these small movements. Yeah, Pronterface works too.
@@user64x that's great! Didn't know octoprint could also work. I'll update our docs 🙂
yo what is that prusa mk3 idex thing
It's from a cool company in Germany called Caribou3D - this is their website:
caribou3d.com/
How would this work with z offset?
You can measure the nozzle orifice size using mm per pixels, or by looking for a common focus plane. I don't recommend this for two reasons though...one is that different nozzles sizes, wear, dirt, etc. can introduce variability, and two is that calibrating Z offsets with a feeler gauge the old school way is already very, very easy and quick to do so you see diminishing returns here IMO.
this is mildly off topic but you could play with your audio EQ since your mic picked up a lot of plosives, it's the popping sound whenever you start saying words that start with p or b. usually just dropping all frequency below 120-160hz already helps a bunch. There's better way to remove it but the quick and dirty method should already help a lot
Thanks for the tip. I noticed this as well and was thinking about getting a sock or something but this advice is useful thank you!
Cool technology, but just don't cheap out on your nozzle, and clean it before calibration
The nice thing about Nozzle calibration for the XL is that it doesn't require much user input. Not to be negative but the majority of the population will mess it up if it requires too much user input imo.
As an engineer in a technical field if a user can mess it up they will. I've seen it over and over again. The more automated you can make the calibration the more consistent it will be in my experience.
In general I agree, except that automation is only helpful if it's reliable. If it's not reliable (for reasons stated in the video) it can be worse than any other method because issues are hidden. In general if implementation is good, these are not common problems. But when they *are* problems, they are not easy to troubleshoot for the average non-engineer. I've had my fair share of work in robotics and automation and have experienced both sides of the coin. For me - I'd rather do a bit of work and know exactly what I'm doing. I'm not saying one method is superior to the other, there are trade-offs like all engineering problems.
@@emberprototypes I agree with you and manual setup is always the best. I think what you've come up with is great and for a person like me it's what I would prefer. Its also good to understand what you have. I just don't think the average person is smart enough.
Now days people can't even do basic Google search for things.
@@bluerider0988 haha yes, I am experiencing your comment on basic Google search more and more these days 😂
I'd like to see automated calibration on more machines in the future like on the XL. Hopefully, in most cases, it's foolproof, but when it isn't optical can be a good/alternative option.
Also - my gut is that cheaper, more accessible multi-head extruder printers are trending upwards. Many which probably will not come with any automated calibration...so in this case, having a firmware agnostic option like the CXC (or any optical solution) is waaaaay better than the ol' stupid line pattern method.
Thanks for watching and for your feedback!
Well there is a reason pick and place machines are using cameras
🙂
Please get a pop filter for your microphone and/or don't talk directly into your microphone