Kossel 2020 Build log

I have been logging my build at the Google Groups site, but the lack of ability to edit a post has me looking for an alternate site.

Background:
I have build 2 mini Kossels. The first was cmpletely stock and follwed the directions. The second was to see how cheaply I could build one. It uses printed PLA sliders which I upgraded to Nylon with Brian H. sent me a set to try. I'm not looking to scale it up to a Kossel and use Misumi 2020 extrusions.

Design goals:

1. Modify the current Johann mini Kossel files to use 2020 extrusions - github of parts
2. Use PLA sliders to have an evaluation test matchine
3. Use an X5 controller - Arduino/RAMPS for sale
4. Update my visual calculator for 2020 and the customized parts
5. Have a 9" build surface diameter
6. Use M3 t-nut and machine screws - I already have a cache of M3 screws. Longest screw needed M3x20mm

Progress:

• Version 0.5 of the parts are designed and printed - it took a few iterations and several test parts to get here
• Partial assembly and most of the parts in in hand - missed a few important parts and some didn't work out (5mm OD Bowden tube)
• Photos posted

Using my Visual calculator, and using the most current parts here is my machine specs:

• Horizontal extrusion length: 342.9mm
• Vertical extrusion length: 857.4mm
• DELTA_RADIUS: 170.99mm
• DELTA_DIAGONAL_ROD: 344.888mm
• DELTA vertical length: 299.516mm
• Delta_rod_angle: 60.2785mm when homed
• Build plate radius: 116.645mm (232mm diameter)
• Max Build Height: 377.421mm assuming a narrow tower or cone shaped build.

These are the base pieces


This is my redesigned carriage. Since I'm using printed sliders I wanted to move the pivot to the center of the block. These are as yet untested.


Sowing the top frame assembly


Showing the assemble bottom. Waiting on one more stepper damper to complete the base. I did braid and then sleeve my wires in Techflex (Carbon is the color name).

I'll update this post as I progress if the system allows

-Jay

Edited 2 time(s). Last edit at 04/20/2014 01:02PM by jaydmdigital.

Bill of Materials: (this is subject to change. Always double check if you choose to use my list. I try to be thorough, but I'm human.

I used

• 5 sticks of Misumi 2020 extrusion (yes from the First 150 promotion) each 1200mm long
• 3 - I cut 342.9mm (13.5") off one end. The remainder is the vertical beams. The other 2 I cut into 342.9mm (13.5") horizontal beams for the base
• 1 pack of 100 3mm pre-interstion t-nuts
• 3 Kysan nema 17 motors
• 1 geared Kysan nema 17 motor
• 3 mechanical endstop switches
• 1 rod kit with the ~12" (305mm) rods, including the 4mm setsrews and Traxxas rod ends
• 6 meters of GT2 belt
• 1 set of 3 16T GT2 pulleys
• 1 set of F623ZZ flanged bearings - if you plan to use Johann's extruder, the mini Kossel bearing kit is what you'll want.
• 1 box of 100 M3 x 8mm stainless socket head screws
• 1 box of 100 M3 x 20mm stainless socket head screws Note: I tried to design all of my parts to use this as the maximum length screw needed.
• 1 box of 100 M3 x 16mm stainless socket head screws
• 1 pack of 100 3mm washers
• 1 machined hob
• 9" build plate

Edited 1 time(s). Last edit at 04/20/2014 01:37PM by jaydmdigital.

My 16 tooth pulleys arrived from tridprinting.com (they are fast shippers) so I started assembly on the two towers with the mounted motors.


This image shows the Kossel to the mini Kossel I have with Nylon sliders (7" build plate).

I'm still not sure I'll keep the vertical beams so tall, but we'll see.

Hi Jay - I've followed your build over on GGroups as well, as I am interested in this build. The lack of editing there bothers me, too.

Wanted to ask, though - can you tell me or direct me to an explanation of why the X5 prompts you to sell an existing RAMPS setup? I know the X5 is cheaper, but beyond that I am just curious about it, no other agenda. I see that, as well as others like a Melzi that looks significantly cheaper, but I have no experience or idea about all the consequences of such a choice.

Anyway, got some 2020 from the First150 deal burning a hole in my pocket!

Quote
revwarguy
Hi Jay - I've followed your build over on GGroups as well, as I am interested in this build. The lack of editing there bothers me, too.

Wanted to ask, though - can you tell me or direct me to an explanation of why the X5 prompts you to sell an existing RAMPS setup? I know the X5 is cheaper, but beyond that I am just curious about it, no other agenda.

Well, it was more of a statement of "I'm moving on". The X5 is very smooth. The RAMPS/Arduino I have running my first mini Kossel is rock solid but I love the compact nature of the X5.

Quote

I see that, as well as others like a Melzi that looks significantly cheaper, but I have no experience or idea about all the consequences of such a choice.

Anyway, got some 2020 from the First150 deal burning a hole in my pocket!

I got a lot of the printer assembled tonight. Should be calibrating tomorrow. I'll post photos tomorrow

Quote
revwarguy
Wanted to ask, though - can you tell me or direct me to an explanation of why the X5 prompts you to sell an existing RAMPS setup? I know the X5 is cheaper, but beyond that I am just curious about it, no other agenda. I see that, as well as others like a Melzi that looks significantly cheaper, but I have no experience or idea about all the consequences of such a choice.

My Megatronics 2.0 with GLCD (single-board with same electronics as Mega2560/RAMPS) seems to be at the limit of CPU performance driving a Delta. Even printing from the on-board SD card, it will pause or slow down momentarily every minute or so like it is catching its breath. Forget about any hopes of doing 1/32 stepping. The Melzi uses a A1284P CPU, which is also an 8-bit processor and most likely also operating at the redline on a Delta. I'm with Jay in needing to move on from the current generation of controllers. I had put my eggs in the RAMPS-FD (Due-based CPU) basket, but with the departure of bobc from this forum and uncertain future of RAMPS-FD, I gave up on waiting and ordered an X5 over the weekend as my next step.....

Thanks for the replies - I guess I will look into the Azteegs.

Just out of curiosity, though, what would ever be the need for 1/32 microstepping? From my CNC work, most of the steppers we use for that or for our 3D printers (including the Kysans) are 1.8 degree +-5%. (I know there are others that have 3% tolerance or a .9 degree step, etc, but these are not common at all.) This means that there is a 10% tolerance band for the positioning error for any given step to the motor (although it is a cyclic error, so it averages out to zero over a full revolution.) This error is the native ability of the motor, has nothing to do with the drivers. So, this also says that anything more than 1/10 microstepping can have no real effect on accuracy, it just increases empty resolution.

There is an argument about it running with a smoother motion under 2-3 rpm but in my experience, you cannot really observe much difference between running a driver at 1/10. 1/16, or 1/32 stepper in terms of "smoothness." I suppose one way to measure it would be to measure the sound with a decimal meter while running at the various steppings, but I never saw much difference past 1/10. Plus, your stepper wiring has to be more robust to support the higher frequency pulsing at those rates, let alone employ a processor fast enough to keep up times 4 motors.

Not trying to be argumentative, but if there is a need for 1/32, I don't quite see it.*

(* actually, implementing a closed loop stepper system with an encoder might be a good reason for it, but I don't think putting a $150 encoder on a $20 stepper is likely on these inexpensive 3D printers.)

Edited 2 time(s). Last edit at 04/24/2014 03:57PM by revwarguy.

My personal thoughts on 1/32..."measure it with a micrometer, mark it with a crayon, cut it with a chainsaw." However, as Jay said, it is time to move on with newer technology. I am glad that he is posting stuff here in the Delta section, since my own feeling is that this section needs more technical discussion and less support threads for the 3DR.

I've only been tinkering with RepRap stuff for 4 months now, but have observed that a large percentage of the community seems to be stuck on cloning yesterday's work and not pushing the envelope to develop tomorrow's innovations. My own self-designed Delta has changed so many times that I don't even bother fine-tuning the calibration any more, and I am tinkering with ideas like using heated glass as used in airplane windshields and store glass freezer doors to make a more uniform heatbed. The J-head, Mk2 PCB hotbed, and Marlin/Repetier firmware are all somewhat serviceable items, but it is time to move on.

You may also want to take a look at that silly-named "T-Bone" adapter being discussed in the electronics section. If I remember correctly, they are breaking new ground with motor controller chips, and I thought that I read something about closed-loop on X and Y in one of their posts.....

Quote
vreihen
However, as Jay said, it is time to move on with newer technology.

Yeah, didn't mean to hijack your thread, Jay.

Quote
vreihen
You may also want to take a look at that silly-named "T-Bone" adapter being discussed in the electronics section..

Thanks for the pointer.

Quote
revwarguy
Not trying to be argumentative, but if there is a need for 1/32, I don't quite see it.*

(* actually, implementing a closed loop stepper system with an encoder might be a good reason for it, but I don't think putting a $150 encoder on a $20 stepper is likely on these inexpensive 3D printers.)

For me, 1/32 stepping is "quieter" than 1/16 which is much quieter than 1/8 stepping. Something like the Gecko 251 (I use it on my CNC router) would be great as it automatically moved from 1/10th to lower microsteps depending on the speed. However, A $61 251 versus a $13 DRV8825 doesn't work out

There is so much headroom on the X5 and Smoothieboard (runs at 120MHz) that the choice between 1/16th and 1/32 doesn't require sacrificing top travel speeds. Also, the X5 is smaller than the arduino board by a small amount and there is no stacking of a RAMPS (a la CRAMPS).

As for an update on the build:

Build notes:
Some build notes:

• My belts ended up being 70" long. The Z_max height is 390mm.
• Interesting my calculator gave me 170.99mm for the delta radius. The actual value is 172.5mm.
• I build my diagonal rods with the 12" rods from tridprinting, not cutting. Diagonal rod length measure (with a steel rule but already mounted on the printer) at 340mm.
  I do need to find some sort of springs or bands to put on the rods...they rattle a bit due to the stick-slip of the 2020 and PLA.
• I am using stepper dampers on the motors and mount tape on the bed/glass tabs. This guy is super quiet. except for the rods :/
• I did not realize that I could adjust the delta radius with the M665 command so I was updating the config then resetting the X5. This was as painful as calibrating in Marlin (compared to Repetier FW and Host) as Repetier would bitch since the com port disappeared and so I'd have to kill Repetier then unplug the X5 (it's USB powered right now) then replug it, then restart Repetier host. Smoothier Gcodes are here
• After calibrating and getting good flat travel I tried printing a test piece (a 100mm square 3 layers thick). I thought it was my hotend expanding a bit when hot because I had to adjust my endstop offsets by over 0.6mm each. Turns out, a PLA mount on this hotend with no fan (never needed a fan on the MG V3) is not a good idea. I will cut a net mount from Birch ply like I have on my original mK.
• I printed a final calibration part (4mm square with a 2mm square cutout) before bed last night. A 2020 extrusion is tight but does fit. Outer dimensions are 40.14mm x 40.10mm and inner is 20.06mm x 20.12mm and height is perfect at 1mm. I plan to play with the diagonal rod a bit but for a first round that's pretty good IMO.

Progress photos:
I completed the assembly and most of the calibration. The printer is in a "not decided" state so sorry for the homely look of it.

I bought 5' of 6mm/4mm ID tubing for a 3mm filament Bowden tube. I'm not cutting it until I decide if I am going to use 3mm or not. So it's draped over the machine and I think is the cause of some serious ooze.


The X5 is screwed to a mounting plate. Happily, the plate which I designed to slide into the slots of 1515 extrusion are able to "hook" in the slot and let the mount hang at an angle. This is temporary until I decide what I want to do. Right now I am running all motors (yes extruder too) at 1/32 microstepping. Yes overkill I think, and I plan to experiment with other setting so I can compare.


This is my old MakerGear V3 0.5mm nozzle from my Prusa i2. It's not liking the Bowden and oozes some very fine stringers all the time it's not printing. Also, I have to print at a low 180C with this one.


This is to show my endstops. The wires run in the inner slot of the 2020, I may change this on the sliders design to be the back slot instead.


This is the Wades extruder I designed for mounting to 1515 and 2020 abstracted from Greg's and other's enhancements. I liek that it's compact and quiet. I was not able to use an 8mm bolt because all of mine, the head was off center. So I used some 8mm threaded rod and 4 M8 nuts (I have a lot of both from my previous Sells and Prusa Mendels).

-Jay

Out of curiosity, what retraction speed and distance are you running? I am at about 700mm with my 1.75mm filament Bowden tube, and running 6mm retraction @ 100mm/second to stop the E3D from oozing. These settings seemed a little bit excessive based on slicer help tips and other people's posted experiences, but they cleaned the ooze problem right up.....

In Cura I use 3.5mm at 20mm/s. Since I have ordered a new E3D to replace this MG hotend, I'm not going to worry about calibrating the retraction yet..

As for updates, I printed two 1515 mini Kossel top frame pieces. I printed them at 15 x 15mm for the extrusion cutout (no extra) and it fit perfectly both times, so this one is as well-calibrated as my original.

Quote
vreihen
Out of curiosity, what retraction speed and distance are you running? I am at about 700mm with my 1.75mm filament Bowden tube, and running 6mm retraction @ 100mm/second to stop the E3D from oozing. These settings seemed a little bit excessive based on slicer help tips and other people's posted experiences, but they cleaned the ooze problem right up.....

I did try to play with retraction settings yesterday and I discovered that the heating element isn't able to get to 180 an hold that temp while printing. So, I'll wait.

I also added a rubber band to the Z tower rods to remove some rattle. It wast a bit long so I stretched if from one lower pivot to the upper pivot.

Edited 1 time(s). Last edit at 04/30/2014 08:24AM by jaydmdigital.

I dumped the MakerGear V3 and ordered an E3D Bowden from filastruder.com Time was very accommodating and let me order it with a 0.6mm nozzle instead of the stock 0.4mm.

I modified the effector.scad file to open the center mount to make room for th etop of the heatsink to push through, and then the rescess so the Bowden connector would fit. I used the Birch ply mount as before. I originally tried to use Kapton to mount the thermistor (Semitec) that comes from E3D, but the temps seemed unreliable so I took a note from semecnc.com and use Permatex Ultra Copper RTV to mount a new thermistor (with Teflon insulation) when one of the legs broke as I tried to remove the Kapton. I got the thermistop from tridprinting.com.







Now that I have the hotend I need to reprint the Wades to use the E3D Bowden connector, but I can still go ahead and calibrate the printer. I plan to film the process I use which will apply to Smoothie and Marlin users since I will be using M665 and M666 gcodes.

Edited 1 time(s). Last edit at 05/04/2014 11:35PM by jaydmdigital.

It's alive! [www.youtube.com]

I did make some changes:

• I replaced the gold effector with a silver one after I redesign the center opening. The gold one was to deep and did not let the coupler pull the hoten tight so it was floppy.
• I reprinted the Greg Wade body for use with a bigger pushfit for the .25" Bowden tube that came with the E3D.
• I modified the design of the Wades extruder so I could use a machined hob instead f trying to hob the 8mm threaded rod (or bolt) myself. I got the hob from Makers Tool Works
• Since the video I have moved the X5 under the build plate, oh and I increased the build plate to 10" and moved the 9" one to my other mini Kossel

To do:

1. Calibrate - and film/narrate the process
2. Clean up the wiring - now that I have decided where the X5 will live
3. Print something

Cool build thread. I'm thinking of trying this myself once my Kossel Mini is fully operational and can print the parts.

Going to use your video to tune my mini and X5

Please provide any feedback you have so I can make it better for everyone.

I had sheared the brass nozzle off my E3D last week and got the replacement parts from Filastruder this week. I when with the new V6 heaterblock, thermistor, and nozzle. Here is a comparison of the same part printed at 0.2mm layer height and 0.1mm.



After the installation I had to adjust my max Z height and some of the tower endstop offsets. It turns out the 10" cake pan glass plate is actually 9.5" and not totally flat. I did make some adjustments to the delta radius and have it printing very flat where the plate itself is flat.

Hello, F6 and I get this error:
Rendering Polygon Mesh using CGAL...
ERROR: CGAL error in CGALUtils::applyBinaryOperator union: CGAL ERROR: assertion violation! Expr: target(s1).vertex()==target(s2).vertex() File: /usr/include/CGAL/Nef_3/SNC_FM_decorator.h Line: 218

Thanks

What version of openscad are you using?

V. 2015.03-1

I just tried it and get no errors. It looks like you are on a Linux system so I'm not going to be much help. Do other openscad files work, can you comment out some of the modules to see if it works with out any of them? If it does, let me know which one. it may be that it doesn't like the way I use the rail STL and chop it to length. I don't know.

Thanks for the replies, windows error (V. 2015.03):
Rendering Polygon Mesh using CGAL...
ERROR: CGAL error in CGALUtils::applyBinaryOperator union: CGAL ERROR: assertion violation! Expr: target(s1).vertex()==target(s2).vertex() File: /data/OpenSCAD/libraries-mingw32-master/mxe/usr/i686-w64-mingw32.static/include/CGAL/Nef_3/SNC_FM_decorator.h Line: 218

other openscad files work OK..