Mesh bed compensation with E3D Kraken



  • Can I Use Mesh bed compensation with my E3D Kraken or would the other 3 Nozels rip my print of while the active one follows the Grid?



  • Depends how much compensation is being applied, but you're right to be concerned about it.



  • The bed should be mostly Flat I´m using a machined Aluminium Plate with 4 Leadscrews (Yes, I know it´s over constraint but i didn´t when I ordered it) and the true Z alignment enabled (https://duet3d.dozuki.com/Wiki/Bed_levelling_using_multiple_independent_Z_motors)



  • You'd need to perform a detailed mesh compensation routine to see how flat the bed surface is. There's also gantry sag or skew that can show up.



  • @Phaedrux Ok I will let it generate a detailed Grid map from time to time to check if everything is all right


  • administrators

    I found the key with the kraken is for everything to be totally level across the build space, or as close as possible to that. its a very difficult print head to use for this reason.

    I wrote about this when e3d released it:
    http://blog.think3dprint3d.com/2014/01/4-extruder-printing-duet-kraken.html

    to summarise:

    • Set one of the 4 nozzles as your primary and have that slightly lower than the other 3 to start with. get everything working as well as possible with this one.
    • its important that this nozzle is the same distance from the bed across the whole build space (mesh wont help you when you bring the other 3 nozzles into play).
    • once you have good results with one nozzle then lower the print head to the bed so its just touching, loosen the other three grub screws and drop the other three hotends to the same level. tighten. job done, in theory.
    • this depends on your mechanical system being level across the build-space - no appreciable axis sag or else you will get nozzles ploughing into the print


  • @T3P3Tony I think I've read your article about the Kraken when I first looked into buying one.
    So no mesh bed compensation for me then.
    Thanks again for the help.



  • I respectfully disagree (a little bit). I run a Chimera so I am half the size as the Kraken but the concept is the same. The key is to not have any significant mesh bed compensation happening in the 20 mm between nozzles. As long as you don't have significant variation in that magic 20 mm, you are more or less safe.
    You are obviously going to get a bit of touching and sometimes it catches but overall it works.
    Yes, you get colours intermixing, possibly a bit more because of the touching ....
    These issues are not specific to the Kraken though ... anytime you have more than one nozzle (short of a nozzle lift setup) you will get this happening.
    I am a bit disillusioned with two nozzle printing but the issue at hand is only a tiny part of that. I would not want to try printing with 4 nozzles on the same model ...
    In the meantime I am patiently waiting for lifting nozzles to become the norm.



  • @jens55 said in Mesh bed compensation with E3D Kraken:

    I am patiently waiting for lifting nozzles to become the norm.

    While you wait for that tool changing seems likely to supercede it.



  • Tool changing is another interesting concept but I am hesitant to jump on the bandwagon this early, especially because the substantial cost jump.
    Has anybody ever published hard figures about x and y registration accuracy for different tool heads ? How about z registration between different heads ? Is there support for individually setting up each head's z offset? Something that is automated and doesn't require messing about with a paper shims 'feeling' the amount of drag or baby stepping ?

    I tool changing printer is on my wish list but until it is well established that this system works, I am reluctant to pay the price.
    Two nozzles was supposed to work and had a lot of positive reviews but of course it really doesn't work well.



  • @jens55 Yes it's still early days with tool changing. I think between some clever engineering and smart electronics like the Duet most of the issues should be solvable. Though it is a very expensive way of doing multicolor printing, the costs become more reasonable when doing multi material printing and when (if?) more tools come along the benefits will make it seem cheap perhaps.

    I fell for the early hype of the single nozzle multicolor/material promise of the Mozaic Palette. It never really materialized. The limitations and gotchas were just too much to make it functional. They've since released a paid upgrade kit that replaces some of the hardware with slightly better refinements, but it still doesn't solve the biggest issue of having constant buffer underruns with no way to mitigate them.I have yet to successfully get a 4 color print that wasn't a keychain.


  • administrators

    @jens55 said in Mesh bed compensation with E3D Kraken:

    Has anybody ever published hard figures about x and y registration accuracy for different tool heads ?

    not that i am aware of but repeat-ability on the kinematic couplings is reported by e3d as better than single digit microns.



  • Unless the figure is properly published, it's just marketing speak. I would also be concerned with the fact of it being just the coupling rather than the overall system. It seems that only the coupling should cause errors but I do not know what other errors are added by the time we get to the tool tip.



  • @jens55 said in Mesh bed compensation with E3D Kraken:

    what other errors are added

    It still strikes me as odd that they chose sensorless homing for the X and Y endstops.



  • when they (e3d) pushed initial story about the system there were numbers published (repeatability was awesome) but than after they start selling (and they changed some stuff, for e.g. carbon fiber plate is gone) I don't see any numbers any more but I assume they will vary from machine to machine as you do need to do manually a lot of stuff yourself... I'm these days ordering one so..

    as for mesh compensation with multihead, I used some self designed (not very good) dual extruder and try-struder (3 heads) many years ago and the only way I found it work acceptable is to print a big raft out of soluble material (hips or pva or.. I even tried with PLA as support) as the main reason for dual nozzle was main and support and for try-struder main, support, soft (flex or pp for hinges) so I always had one support nozzle available 😄 ... that worked good, you disolve that raft later easily and after raft you have ideal surface to print on so mesh compensation required (not that any of the firmware's supported mesh compensation back in the day 😄 ) ... the only downside is the waste of support material for the raft, but as modern slicers can use soluble support only for the contact layer (I was not smart enough to do that myself back in the day) the waste would not be that big



  • I was not aware that they chose that method and yes, it is a bit odd.
    I think this is one of those little details that will be worked out with time but it certainly makes you think the system isn't quite right for prime time yet.
    In general, I think it is unrealistic to expect a new device to arrive without the need for some tweaking. My thought/hope is that by this time next year the initial teething issues might be mostly sorted. I just hope that with cost being so high, it won't drag the timeline for sorting initial issues for longer than a year.


  • administrators

    @jens55 regarding high cost, you could look at @poofjunior 's Jubilee project. I think @Danal has made one.
    https://github.com/machineagency/jubilee



  • Yes, I've built a Jubilee. It's a CoreXY with up to four tools. Could easily be extended for more. It is E3D tool compatible. Jubilee prints with no purge blocks.

    E3D used metal plates and metal pins for their kinematic coupling and locks, which makes wear a long term proposition; repeatability across a few days is quite good (I don't have any hard numbers). Jubilee uses metal kinematics, but printed plastic locks, and has some good engineering that makes that work well even though it wears.

    Cost? About the same as any other well built corexy, plus a small increment for the tool lock, plus the cost of each tool. Not "expensive" by any stretch, for what it does.

    At this moment, mine is producing excellent single tool prints, and OK-ish multi tool prints. I need further tweaking and tuning to get to excellent multi tool.

    One landmark recently crossed is/was getting very good reliability on changes themselves. I have now completed several prints that contain two to three thousand tool changes, with absolutely no faults. I couldn't have said that a few weeks ago.

    Feel free to ask any questions.



  • @Danal said in Mesh bed compensation with E3D Kraken:

    At this moment, mine is producing excellent single tool prints, and OK-ish multi tool prints. I need further tweaking and tuning to get to excellent multi tool.

    What part of multi tool prints causes issues for you? Is it mechanical or operator/software issues ?

    One landmark recently crossed is/was getting very good reliability on changes themselves. I have now completed several prints that contain two to three thousand tool changes, with absolutely no faults. I couldn't have said that a few weeks ago.

    What were the issues that needed improving ?

    Thanks !



  • What part of multi tool prints causes issues for you? Is it mechanical or operator/software issues ?

    Getting to excellent multi tool is mostly a matter of aligning the tools. So sort of both.

    At the moment, there is no automation to measure the tools themselves. Z is done for each tool the old fashioned paper way, and the Z offset is placed in a G10 command in the configuration (actually, I put it in an external file so I can be sure it is consistent). This actually works well enough; doesn't take that long.

    X and Y, so far, I've been printing a vernier. www.thingiverse.com/thing:2847643 Then make an adjustment to the G10 values. I'm not finding this very satisfactory.

    There's been lots of discussion about automating probing. Heat and wipe nozzle, then touch it to metal plate for Z, then drop it in square hole and probe for X and Y. OR... USB microscope. OR... piezo with a square sticking up, etc, etc. So far, this is all talk, nobody's implemented anything. I MIGHT, if I get some free time (ha ha) give a shot to one of these methods, implemented on the Pi on a Duet 3.

    Summary: Multi Tool Alignment.

    What were the issues that needed improving [to get reliable toolchanges]?

    • My personal first use of RR3 firmware. Learning curve. I seemed to have a very hard time with RR3 concepts. Perhaps been using 2 too long. Anyway, once it "clicked" I do believe the new naming of pins and gathering things up to form a tool is MUCH better than the old way. I'm just thickheaded about some things.

    • Not that many weeks back, RR3 documentation was a lot more sparse than it is today. It has really improved a lot in a short time.

    • A bug or two in the beta. Again, lots of improvement in a short time.

    • My first toolchanger. Learning curve.

    • A couple of mistakes I made reading the instructions and/or Bill-Of-Materials that resulted in the tool-lock ALMOST working. The original designer, @poofjunior is on a discord and is very responsive. Once I had a few "Aha!" moments, and changed it to what the plans/instructions actually said, it started working much better.

    • My first laser cuts (to make certain delrin parts). I'm really loving that, what a great new capability in the arsenal.

    • I hand cut a bed on a bandsaw from an expensive piece of aluminum plate... with a PDF pattern glued on... a pattern that had printed just a FEW percent too small... SH*T!!. Very disheartening. Second attempt: CNC cut with many passes on a CNC router. Worked GREAT. Again, a new learning and a new capability. What fun building these things.

    By the way, the delrin parts and the machined bed are now available from a couple of suppliers.

    Summary: Learning curve.

    • And, what I'd call "normal" tuning. Getting the XY of the parking stalls really truly matching physical reality.

Log in to reply