06 - Bowden System and Multiple Hot Ends (7/26/14)

Post date: Jul 26, 2014 5:28:29 PM

I frequently switch nozzle diameters for different print applications. For example to do the miniatures I want to use the smaller .25 mm (would like to go smaller if can find one), for larger prints and quicker prototyping, I use the .4 nozzle with thicker layers. Now add the choice between ABS and PLA plastics and there are the 4 combinations. I've read about problems people have had switching between plastics in the same hot end due to their different characteristics like melt temps etc. Stories like material buildup in the nozzle and clogging. Switching hot-ends is a minor inconvenience really, but it means a bunch of unscrewing, disconnecting and then after re-assembly there is the re-calibration of Z. While there are several options to deal with my desired workflows, each with pros and cons, I have had it in my head that I want 4 hot ends that I can switch among easily.

The options:

1) Switch the actual nozzle fitting.

  • pros: fast, simple, no need to disconnect or loosen other components

  • cons: wear and tear (threads, and if it's brass it's soft), might break important seals (leaking), hot end design needs to accommodate it

2) Switch the entire X carriage

  • pros: fast, simple

  • cons:more expensive if using a direct feed extruder then likely multiple motors and multiple extruder mechanisms

3) Stay with one middle sized hot end

  • pros: no work at all to switch

  • cons: compromising on flexibility, still likely have issue switching plastics

4) Multiple hot ends mounted simultaneously

  • pros: get to design some fun new stuff and if done right, simple fast switch-overs, and can move to multiple simultaneous extruders more readily for multiple colors and/or multiple plastics on the same print, single auto Z homing with no need for re-calibration.

  • cons: need to change to Bowden style feed system. I see this as a con because of the "elastic hysteresis" I have seen in these systems, I have gotten the retract pretty well figured out in the direct feed system. From what I read, retract needs to be kept to an absolute minimum with Bowden because of the compression and relaxation hysteresis in the filament tube. The only way I was able to effectively eliminate the "webbing" or stringy filaments in non-print moves, was to use retract.

5) Keep going the way I have been.

6) I know there was at least one more but I have forgotten.

The most common draw-back to a non-permanent setup is the re-calibration of Z even with the auto leveling system (once I add it).

I chose option 4. Doing this will be another learning experience, I just hope I can get as good of results with the revised filament feed as I had been with the direct feed.

The first thing I did was model the hot ends that I have, they are all consistent sizes so that was easy enough, just made one and copied it. Next was to see how compactly the 4 hot ends could be arranged without interfering with each other. This was so I would lose the minimum amount of X and Y print area.

I had a concern about the Z position of 4 hot-ends, they all had to be at the exact same tip position or there could be interference with what was deposited by the "active" hot end. The hot ends I purchase can vary slightly in overall length, so this meant the design I came up with would require adjust-ability. I started playing with clamping ideas and after trying out a few designs, I settled on one that seemed sufficient for prototyping. Another concern arose about initial filament feed into the top of the hot end if a large enough gap were created by adjusting it downward. I designed an adapter with a conical inner wall to guide the filament to the center.

Here is a tinkercad image of the design with a cut-away showing the clamping, adapter, hot end arrangement etc.

After a quick coarse print of a single clamp and some test fitting, I was ready to try the entire "cluster". I increased the quality a little and in about 2 hours had this:

It accommodated all 4 hot ends just fine but looked kinda "Frankensteiny"

I confirmed clearances of the assembly between the X axis guide rods and eye-balled that things looked ok. There was no mounting of this to the X plate I currently had because there wasn't room in the openings for all the hot end goodness. So back to the modeling software to design the X plate. I loaded up the model for the original, thinking I could just make some tweaks on the openings, but when I placed "Franken-holder" onto the plate, I went "Oh, ok, there's collision with the bearing brackets." ( I realized I should have done this sooner but this is about quick prototyping and innovation, what's a little wasted plastic? Also, I didn't dis-assemble anything on the printer so a true "test fit" of the assembly onto the x plate was not do-able).

I also realized I would need to go to a 4 bearing x plate to make room, and I needed to scrunch the hot ends a couple mm closer together in the X. So I printed gen 2 of the Franken-holder.

While that was printing I designed a crude X plate (nothing fancy and a lot of wasted plastic (no optimizing - it was getting late). I started the X plate printing and once satisfied it was going well, got some shut-eye.

There is always that moment in the morning after a night print that feels like Christmas day. I walk down the stairs and peek around the corner with anticipation. The difference is, as a child I don't recall ever being disappointed. Sadly, with the 3D printer, I have been. A tangle of plastic barely resembling what you told it to make lies on and around the print bed - sigh.

But that was not the case today!

I mounted the same hot end used to print the new parts onto the new setup to test things out and connected everything as shown below (it's not visible but the fan is now in the back lower down so it blows across the heat sinks on the hot ends):

And I did a rough mounting of the extruder with existing hardware (I plan to re-orient this)

Here is a wider shot of everything:

Time to do a test print. After confirming everything heated ok and there was no melting, burning etc. I connected the X belt, homed, calibrated and loaded up an already sliced model I regularly use as a test print. (1.75 mm PLA, 185c, 0.4 mm nozzle, .25 mm layer height, .3 mm width, 40 mm/s)

Two things are immediately noticeable; the stringy whiskers and the "serif" on the tips of the letters. Those are not in the model nor are they in (most) prints done with the direct feed. Any pausing causes blobbing because of the continue flow of extrudate well after the motor stops pushing. Time to start experimenting with slicer settings now. From what I've been reading, at least for slic3r, it's best with Bowden system to:

  • avoid crossing perimeters (Print Settings): on

  • only retract when crossing perimeters (Print Settings): on

  • retract on layer change (Printer Settings): off

  • wipe (Printer Settings): on

And in general:

  • reduce the need for thin wall gap fills

    • very fast non-print moves

    • lower layer heights and slower print speeds (this means slower extrusion therefore less hysteresis)

    • still researching...

I will try to update with results of various trials.

There is a list of other improvements to do as well:

  • Make cut-outs in the X plate for wiring (notches on each side)

  • Optimize the X plate mass (remove and thin not-needed material)

  • The hot end heat sink is just a little uncomfortably close to the guide rods and in the one case the belt in the front so I need to make mods to the design to lower the hot ends at least 10mm. For now I will make sure a PLA hot end is mounted there because of the lower temp.

  • Add pre-aligned holes to bolt the 2 assembly parts together (I had to actually use a drill on my printed parts!)

  • Add nut holder cut-outs to the clamps in the hot end holders

  • Rotate the extruder 90 degrees around X so the input is rear and output is front.

  • Design something so the Bowden tube is as short as possible (so far all I can think of is a kind of stilt mount where it rides above the X carriage and keeps the tube distance to an absolute minimum).

  • Replace the hose clamps on the inputs with quick release 4mm ID fittings (I have some similar to these on order)

Update 7/27/14

After making some setting changes, the test print improved a fair amount but there is still work to do (the settings used below are those listed above and when it does retract, instead of 2.5 mm it's now 6 mm):

I jumped right to a more advanced test and it didn't go so well:

compared to how it was working with direct-feed (it is different plastic so I will want to try the red for the cube too):

Another factor to consider is the fan is located differently and the print surface is heated since that red cube was printed. I think playing with print temps is also called for.

Update 2:

After taking care of much needed domestic duties including cutting grass (my back yard was starting to look like a meadow restoration project which I think is cool, but the neighbors and city hall - not so much), I played more with the new setup.

All four hot-ends are installed and aligned

I "quick switched" (Franken-switched?) to the .25 mm end (front right) intended for PLA (so far I've only printed ABS with a .25 nozzle)

(1.75 mm PLA, 185c, .25 mm nozzle, .1 mm layer height, .3 mm width, 15 mm/s speed, 6mm retract only when crossing perimeters)

(It was unusual to see another nozzle going over the part)

The result above is decent enough, It appears a layer width of .3 mm for PLA from a .25 mm nozzle is too wide, it's not getting squished that wide so there are gaps. I'm pretty sure the ABS works fine at .3 mm width. Also noticeable is the "weld" at the bottom of the letters. The .1mm layers are blending nicely though.

Once again I thought I would boldly jump to the wire frame cube

Argh!

Time to try Cura and/or other slicers and do more searching/reading.

Update 3 (7/28)

It seems like an opportune time to change slicers since most the settings for the existing one no longer apply.

After slicing and printing with Cura about a dozen times, the results were in some ways better, others worse. Here is the best one of the bunch

There are a couple improvements, 1) width on each pass is making for a tighter surface and 2) the width of the letters are more like the model

One thing with slic3r to try is not override the width but let it choose by entering zero in the field.

1 other weird things that happened above: a strange external side-wall curve as though X kept moving slightly on each layer. This may have been a fluke, it hasn't happened since. There could have been a strange one-time angular gravity vortex, or perhaps a floating point issue in the slicer...

The wire-frame cube fared much worse, it's not even worth the pictures, It couldn't even get the bottom few layers down, it was generating some wild paths that kept lifting, trying a brim and raft helped it stick but the path for the edges of the cube were terrible, it tried to make the horizontal edges by moving back and forth perpendicular to the direction of the edge (if that makes any sense).

It's only been a couple hours of time getting familiar with Cura so there is much to learn, plus it's is in active development which is a big plus.

While not ready to give up on the direction I've been headed (others have had good success with the remote feeder system prone to elastic hysteresis ;), ruminating on the issues have led to refinement ideas and other choices:

  1. re-position the extruder assembly to where it has a shorter tube to the nozzle assembly

    • half way up the side

    • another carriage mounted directly above (on it's own rod(s). This could be permanently fixed above or freely slide side-to-side.

    • mount the extruder on standoffs a couple inches above on the X carriage (center of gravity seems an issue)

  2. re-design the multi hot-end assembly so the extruder mechanism can simply be directly re-positioned over the desired hot-end (release, lift, turn, drop, connect).

  3. design a different mounting system for the direct-feed such that it is still a single hot end on the X carriage and the hot-end assembly can "pop in". This assembly would contain a permanently mounted and calibrated Z homing servo/switch, this way the only thing to do when changing hot ends (besides the wiring) would be setting the offset between the hot end and the switch.

The mass of the X carriage has not been an issue for print speeds, for me the limiting factor on speed is currently the mass of the print surface (Y axis). So moving the extruder mechanism off of the X carriage was for the convenience of the 'Hydra" design. For this reason, if I can't get the hang of overcoming the hysteresis problem, choices 2 and 3 are looking pretty good.

Update 4 (7/29)

After thinking about idea 2 above, it seemed the potential exists for too much shift of the center of gravity causing front/back tilt. The motor and gear assembly needs to stay centered (in Y) over the guide rails. Also, why crowd things and add the extra weight when only one hot-end will be used at a time? This led to a very rough sketch of the pop-in assembly idea. The holder is still based on the existing clamp, but some ideas are brewing about using the existing notch in the PTFE portion of the hot-end so a different clamp is in the works. The servo needs to be far enough away from the heater, and there needs to be clearance for it under the carriage etc. still many aspects to work out.