Great thread 
@Whosa_whatsis how long is a short bowden tube? I’m currently setting up a bowden for my alu plate prusa i3 and wonder wether to mount the extruder on top or bottom of the alu plate. A bit worried that motor vibrations will make uneven layers if it is mounted on top.
@Tom_Oyvind_Hogstad the bowden tube on the Bukito is only 300mm long. For most machines, I would tell your to add about that much to however long you think it should be, but the Bukito is arranged to keep the bowden tube as short as possible.
Given the 1.75mm vs 3mm horse has been well and truly flogged above. Care to talk about the bots design a bit @Thomas_Sanladerer ?
@Tim_Rastall i started drawing it up two days ago. What i’ve settled on and started modelling for now is this:
- A compact and precise dual-head printer. I already have one that does large prints and now that i have four Massdrop motors (and a couple spares) sitting around, i wanted to put them to good use. I want to get this one making engineering-worthy parts of arbitrary shape, so PVA supports are a must.
- Two E3D bowden hotends, possibly with a fairly small nozzle for the main hotend.
- Cheap, where possible - i’m scared, but i’ll use the 40€ Geeetech Melzi board (plus a small addon board for the second extruder). Generally, i’ll try to use materials that are either cheaply available from China or common mailorder articles. For example, 2020 is 4.80€ per meter (including cutting) from a German ebay seller.
- Total cost should end up just below 300€. Electronics have come way down in price, and not having belts helps as well.
- Build volume (xyz) 150x150x200(ish)mm. 150mm square heated beds seem to be a common “compact” size and i’ve found that 95% of my prints fit on that area anyways. Also, no mechanical bed adjustment - that’s so 2013.
- 2020 aluminum frame, enclosable. Probably going to use rack brushes to get the top covered up.
- Few and no-nonsense printed parts. Some will end up simple enough to be made with a drill press and a block of UHMW. Combined functionality, where possible. Also, as few screws as possible - i’m still suffering from the aftershock from that Sells Mendel.
- Ultimaker-style cantilever bed.
- Overhead CoreXY-based motion using spectra line. I had to pull a couple tricks to fit the CoreXY inside an aluminum frame.
- Top-mounted electronics and motors, where possible.
- If i feel like it, some way to shove finished prints off the bed. Using glue stick, my prints stick perfectly during printing, but are completely loose once the bed has cooled.
@Thomas_Sanladerer , what ‘small addon board for the second extruder’ are you using?
I have a Melzi setup and would love to add a second extruder for supports, as you have planned (multiple colours is just for fun, but 2 extruders would be a GODSEND)
I only know of RepRapPro’s approach which is a second Melzi board for an extra 2-3 extruders…
@plexus_eye maybe your study was limited to ONE type of extruder, it sounds like that may have influenced your decision because I have not had 1 failure with my RepRapPro extruder / hotend / bowden (450-500mm) setup. I really use 1.75 EVERY day and have NEVER had a buckle in filament or extruder problem due to 1.75, I am sure your findings are accurate, but I don’t know if they precisely diagnose the FILAMENT as the problem… Since I can vouch for the filament never being a problem with the extruder I have I would say the problem is your extruder, notably; a bigger gap between the extruder drive and the ‘filament guide’ (whether this is a bowden tube or simply the orifice in the extruder body) is BAD when using 1.75 filament.
My extruder (RepRapPro stock extruder) is designed for the 1.75mm and handles it perfectly. I even hand-weld filament ends together (5m left on the old spool = weld it to the end of the new spool) and that feed right on down the tube. I can’t agree that 1.75 “causes buckles” - only that it “may cause buckles if the extruder isn’t designed for it”.
@Jarred_Baines addon board as in a Mosfet, three resistors, some female pin headers and a handful screw connectors on a 5x7cm proto board (one of these: http://www.ebay.de/itm/10-Pcs-Prototype-Paper-PCB-Universal-Experiment-Matrix-Circuit-Board-5x7cm-S9-/390701072620?pt=LH_DefaultDomain_0&hash=item5af79940ec ), which will exactly use up the four remaining pins on the Melzi. I just took a quick look and saw four pins labeled A… and deemed it doable, i’ll have to check if those pins are really what i think they are.
How do they hook the Melzis together?
hang on, 4 pins?
Do you have an LCD as well?
This is the RepRapPro wiki page related to their tricolor Melzi setup:
Ok, i had a second look. The additional extruder will definitely work and there’s even enough pins for another one since the SPI, I²C and serial interfaces are all GPIOs as well. The four mentioned pins are in fact four ADC pins.
Now, i’ve never seen much of a need to have a display and rotary encoder on my printer (thanks to @OctoPrint ). Those addons would eat up the pins needed for more extruders anyways, as each extruder (including hotend) requires one ADC pin and three “dumb” digital GPIO pins.
@Ashley_Webster beds usually have a way to mechanically level them, but on this printer, I’ll assume that the bed will be reasonably level without any adjustments and let Marlin’s auto bed leveling (and a bed probe) do the rest.
Yeah, how’s that work? does it compensate for ACTUAL bed leveling by changing Z?
For example, if the back left corner is .5mm higher than the front, it will raise Z .5mm when printing in that region?
@Jarred_Baines that’s exactly what it does. Instead of skewing the print, it’ll actually tilt it. Here’s a demo of what it can do: http://youtube.com/watch?v=x8eqSQNAyro
yeah, I’ve been suggesting doing that at work in the CNC mills for AGES…
plonk a part down on the table, probe the part in Y direction, move along (say 100mm) in X and probe Y again.
Repeat with X direction.
You don’t need to set the part up straight along the axes anymore, you can just orientate the part in the CAD/CAM package and it’s ready to rock!
And in the newer machines you can have a probe automatically touch the part to determine its ‘angle’ orientation, and, that’s it! it’s ready to machine!
Of course my boss isn’t a fan of using technology, so we continue to screw around getting parts perfectly straight before machining (eyeroll)
Don’t know if I like the idea for a 3D printer though, or in Z for a mill either, it means that the ‘ironing’ or ‘flattening’ effect of the nozzle changes depending on direction.
For .1 or small amounts it would do fine, but my mendel bed can be up to 3 or 5mm out, depending on how the adjusting screws are tightened.
Those 3D probes are functionally the same thing as our bed probes - i’ve seen them working on some mills and they really are awesome little things.
From what i’ve read and seen, it seem that the “ironing” effect isn’t that much of an issue on 3D printers. Though on some fairly steep prints, i was definitely able to tell which infill lines were printed in which direction.
Still, i think it’s reasonable to assume that an unleveled bed isn’t going to be out of whack by more than one or two degrees, and for that it’ll be perfectly fine.
My main concern about using these features is the introduction of moire effects. For a deltabot the effect would be negligible because it’s native resolution lines are curved, and not on planes parallel to any of the cartesian axes, so any interference pattern introduced by changing the coordinate system should be lost in the noise (though @ThantiK 's experience suggests that this may not be the case, at lest for some “auto-leveling” algorithms". On a cartesian bot, however, your the coordinate transformation will results in lines that are almost, but not quite, parallel to the native axes of the machine, which is where moire effects (due both to rounding error in the calculations and to the limits of stepper resolution) are strongest.
Plus making your Z axis participate in moves within a layer means that you have to worry about backlash on that axis, where normally you don’t (even with Z lift enabled, Z axis will always be hitting from the same direction when it starts printing).
Good point @Whosa_whatsis . Now, i’d think the moire pattern would only amplify z-axis inaccuracies which would otherwise manifest in the usual z-banding at suboptimal layer heights, so the moire would only be barely visible lines along the part’s corrected z-axis. With a native 5µ resolution along Z (or should i use M3 threaded rod instead of M6?), i suspect it will only be visible at very low layer heights, if at all.
I can’t estimate how much backlash that cantilevered z-axis is going to have, in any case i’ll try and use the Prusa/Mendel90 double-nut-spring solution, which has worked great on my bots so far. All of them had a Mendel-style axis setup, though.
So for the design and testing phase i’ll just stick with the all-electronic solution and make that decision part of an empirical process. If it doesn’t work out, a bed leveling mechanism is trivial to add and i can still use the probe as a z-endstop.
I haven’t done all the math, but it seems like the optimum layer height would be highly dependent on the angle that the coordinate transformation is compensating for, which of course depends on what direction the extruder is moving relative to the slope of the platform.
It seems like the optimum layer height should probably be the uncompensated optimum layer height multiplied by the cosine of the slope angle, but for moves close to going straight up/down the slope of the platform, I would expect to see bunching where the extrusion radius during the Z step overlaps the extrusion radius of the corresponding step on the previous layer. Since the cosine values will not be numbers that can be stored without rounding error in a floating point variable, and because Moire patterns are worst when you are close to, but not exactly on the the optimum values, it may actually be best to choose a layer height that is, for example, the closes prime number to your desired value to keep the interference as random as possible, since you’re never going to eliminate it. It would be interesting to write an “auto-leveling” algorithm that would fudge the values within a certain tolerance to output values that would result in fewer rounding errors.
Personally, I’m more interested in the idea of using some of the various Z probe designs for assisted tramming (or even automated mechanical tramming) than the platform compensation algorithms that we’ve been seeing recently.