The Cobblebot 3D Printer In 2020

Got a Cobblebot kit in your closet? Know someone who does? A good proportion of “news” on the net which includes the name “Cobblebot” is completely free of useful information, although there are some useful tips out there. Here’s my January 2020 brain dump of what’s going on with my Cobblebot. 

On January 1st, 2020, I 3D printed an ABS Bahtinov mask for a Celestron 8” SC Telescope on my Cobblebot 3D Printer. I did it in one try, thanks to my 15 inch square, Marlin-UBL-calibrated heated bed. (Bahtinov process explained at the bottom)


If you’ve heard of the Cobblebot at all, you’ve probably heard the plentiful anti-hype. Either that its design was criticized, or the creators were disingenuous, or whatever. It was originally a Kickstarter project, in the summer of 2014, and later a Indiegogo project. There were several versions of Cobblebots, mine is the original model. The details of how much money was crowd-funded, and some of the aspects of communication by the people behind Cobblebot drew much in the way of criticism, and even accusations of misconduct, but despite all of that, I netted a working 3D printer with a huge build envelope. The costs/benefits to get it built and updated to its current form were:

So I have about $650 and a hundred hours in it, so far. For the price, I have cartesian style 3d printer with a 300mm cube build envelope. It uses Marlin 1.1.9 Unified Bed Leveling to self calibrate. Calibration usually lasts until something torques the z-gantry, but it rights itself in the time it takes to have of a cup of coffee. Despite plentiful uninformed gossip to the contrary, it holds calibrations through numerous prints and does not drift appreciably.

The design was actually very straightforward. It’s a big, cartesian XYZ machine constructed of V-Slot Linear Rail, with a stationary bed, and moveable gantries that carry the hotend in X, Y and Z directions. If you gave a class of college engineering students access to, and an afternoon to design a 3D printer, you’d probably get several designs that looked a lot like Cobblebot, even if none of the students had never seen a Cobblebot.

The bed heats up to 80ºC within one minute, and is always to temperature before the hotend. I don’t have my RAMPS controlling the bed, I just set the temp on the Keenovo control unit.

Lately I’m using a 0.5mm nozzle, which is great for big projects. I usually buy nozzles in assortments from Amazon. 

My print bed rests on coil springs which fit into the v-slot underneath, and has 4 low profile screws which engage t-nuts in the rails, such that the screws push the bed onto the springs and provide a level adjustment. I typically keep my 15” square bed within 0.2mm of Z-variation corner-to-corner, with the bed level compensation turned off. If I tell Cobblebot to probe the bed and it measures more variation, I adjust screws. 

The original Cobblebot design had the Z lead screw nut tabs sharing with the X drive axis. I split that up by buying separate tabs for the X drive at  When the Z and X axes shared tabs, one had to compromise between centering the Z axis, which moved the X belt into the build plate and limited Y travel, or maximizing the build volume and uncentering Z. Once you separate X and Z, the Z tabs can be set at opposite corners, which greatly improves Z repeatability and reduces binding and sag. 

I would say that a majority of my time has been spent devising and applying a bed-leveling scheme. I mucked about with Mesh Bed Leveling in Marlin around version 1.0.2, got my inductive probe and UBL going around Marlin 1.1.5, and just updated to 1.1.9, in which UBL works very well, and is well-exposed in the LCD menus. The combination of the milled aluminum bed, the inductive probe, and UBL really made this printer viable. 

Keeping Your Head In The Right Place

I had a learning experience last year around how head crashes, especially dragging the nozzle across the bed, affect nozzle shape. One goes through various exercises trying to get extrusion right, to get first layer contact right, and etc, but I hadn’t really thought much about nozzle damage until I decided to look at my old nozzle under a microscope. 


Old 0.4mm crashed nozzle New 0.4mm nozzle


The gist is: if you do a nozzle-bed-drag, change the nozzle. Note that there is significant casting cruft around the new (cheap) nozzle, and one could seek to drill it, or something, but I speculate that trying to do anything with sub-millimeter drill bits is not part of a happy lifestyle, so I’ll live with the cruft. The crashed nozzle, on the other hand is probably pretty far from 0.4mm in terms of filament volumetrics, but also appears to want to spew sideways, rather than down. I suggest getting at least a simple microscope, for so many reasons, including this one. 

Future Directions

I’ve kind of got my eye on the Zesty Nimble v2 a remote-direct drive extruder. It puts the drive in the tube, instead of the filament. Another interesting DD is the Bondtech. Either would probably necessitate changing the hotend, so I’m procrastinating, because such projects tend to get prioritized behind “real life” during which time my 3D printer is non-functional. 

Another interesting wrinkle is laser cutting, which is well-documented around the web and on YouTube. Since the Cobblebot has an empty second extruder position, I think I could put a laser on there for cutting, without disconnecting the FDM hotend. I would want to make an enclosure before I start firing lasers, so that would also be a big project, but looks like it could be done within a couple hundred US$. 

Printing The Bahtinov Mask

In order to avoid making a post that turns up in searches for “Bahtinov mask” and not providing what the searcher probably seeks — 

With UBL and bed heat, it was not difficult to print this 10-layer mask. It took about 2 hours. It was a great test of 1st layer precision, which was pretty good, there were places where it was pretty thin, but no place where it was loose. It’s not a particularly good anti-model warping test, since there are no stresses from cooling upper layers. 

The result was a bit delicate to get dislodged from the bed, but with my trusty scraper and some positive attitude, it came off undistorted. I let the bed cool to touch-able before I pried it up, but I didn’t let it return to room temp. 

There are several b-masks on Thingiverse, including an OpenSCAD model that’s adjustable. The others are fixed, and either printable in two parts, or the wrong size. The OpenSCAD one doesn’t produce a manifold model. I ran the output through several fixers, looked at cleanup in Blender (which allows you to select all non-manifold vertices), but it was going to be a long slog.

In the end, I took an SVG file produced by the Bahtinov generator at: 


Imported it into Blender, converted it to a mesh, and extruded it upwards 3 mm  to make a solid shape. Due to weather, we have yet to test the mask on a star, but this will probably be remedied soon. I think that my aforementioned interest in laser cutting will probably produce a simpler, faster way…

Cobblebot build: 28 wheel assemblies

The first procedure in the Cobblebot build is about putting the wheels together. I watched Mike Kopack’s video where he used pliers, to good effect, and then in the month since, have wondered a few times whether one couldn’t just tighten the nut and push the bearings in that way. I came up with an alternate way to get the bearings in, using a 5 mm t-nut. The bearings have a sticky grease/rust inhibitor on them, so you may want to get a piece of paper or cardboard to work on. Probably best not to rub your eyes or eat finger food while you’re working with the bearings.


Pic above shows everything you need, minus the hex 3mm hex wrench, which I forgot to put in the pic.


I was able to get one of the two bearings in by just pushing it in with my fingers.



Put the wheel with the one bearing onto the 5mm screw.




Don’t forget the washer between the bearings.



Place the second bearing on the screw.



Spin the t-nut onto the screw, flat side down, until it is snug against the top of the bearing.





Seat the head of the screw onto the end of a hex wrench.


Use a 10 mm open-end wrench (or pliers, or a Crescent wrench) to turn the t-nut, which will push the bearing into the wheel.


An example grip to hold the whole configuration while you’re tightening the t-nut. Also, the tightened t-nut and the seated bearing.


Remove the t-nut (use it again for the others). Finger tighten the lock nut onto the screw, place with the others.

Repeat 27 more times.

I had some trouble getting the pictured hex wrench into the heads on 3 or 4 screws. The hole in the head which accepts the hex wrench was really tight. When I tried a different hex driver, a cast-formed bit for my electric driver, it was much easier to get into the disagreeable screws.

(I think that the folding hex key set as pictured is made by stamping into a die, and the wire that they put in the die needs to have just enough material to fill the die, without being too much, which is a hard balance to strike, especially for screws like these M5 guys, which have really close tolerences on the drive hole, which is good. The casting process used on the driver bits (I have this set)  puts the right material into a mold, so it’s a less finicky process.


Cobblebot, I Unbox You (part 2)

The remainder of Cobblebot came on Friday, right before the 3-day weekend, even though USPS tracking predicted Tuesday after Memorial Day. I did a thorough inventory, comparing to the parts list in the instructions, and found that it is all there, AFAIK, with a few extras not listed in the currently-released build instructions.

On Saturday, I spent time with the electronics, not to build them, but to test them and make a few videos about how they work together. Unfortunately, I’ll need to upload those on Tuesday. I started to upload them from out here at the Lake Cabin, but the Internet connection becomes unusable for other things, and it would take about 12 hours, anyways.

Some unboxing pictures:

IMAG1184-obf IMAG1189 IMAG1190



I did a complete count, there were a few extra nuts and what-not, but nothing missing…