Decommissioning my workhorse of four years: The Mendel90!

My Mendel90 is on its last legs – these are the things about it that kept me using it for as long as i did as well as what needs to change with its iteration.

Infos on the Mendel90

ATX supplies
Plum couplings

Decommissioning my workhorse of four years: The Mendel90!

What’s up everyone, Tom here, and sometimes, you just have to let it go. And that’s exactly what i’m going to do today with my old, tired workhorse of a 3D printer, the Mendel90.
And to pay the last respects to this trusty machine, i kinda want to give you a rundown of how i built it, what kind of decisions went into it, and which parts worked out well and which not so well. It’s been my most used 3D printer, but also the most-modified one, so it’s in a really bad shape right now.

Let’s start out with the basics:

This was the third 3D printer i built for myself after using a stock Prusa Mendel i2 for a while and then cheaply converting it to a standard Mendel90 to see if that concept worked out at all. And it turned out the Mendel90 is actually a pretty darn nice machine, so in mid 2012, i set out to build this one. And the awesome thing about this 3D printer is, that because Nophead actually designed it completely in OpenScad, you could just change the parameters to larger or smaller rods, different belts, a different print volume, and it would generate a complete set of printable parts plus cutting templates for the frame parts and the bed undercarriage.

So, obviously, i wouldn’t just build another “standard” sized 3D printer, it had to be large. Double-size, in fact, and using 18mm plywood instead of 10mm MDF. I basically designed it around having two heater PCBs on there, which basically had it ending up with a build volume at a bit over 400mm wide and 200mm deep and tall. The other thing i wanted to eventually do with it was to use it as a super-light-duty CNC router with a Dremel, but even though i used 12mm steel rails, when they’re spanning over 500mm, they’re everything but solid. Actually, at that length, they perform exactly like 8mm rails in a regular-sized printer. Also, i simply grabbed ground stainless steel bar stock for these, which was a mistake, as the LM12UU linear ball bearings drive ridges into that material really easily.

So the other thing i did to make the motion system a bit stiffer was to use these 15mm wide HTD belts. They are the same family as the common GT-series belts, but are made for lower loads, and as such, cheaper, and these in particular are a 3mm pitch. I used some fairly small aluminum pulleys to give the maximum stiffness and torque from the motor. I had started out with printed pulleys, but they just never ran true and ended up breaking, too. The Y-axis on this Mendel90 actually uses two belts and stepper motors, each with their own driver, because the bed assembly is actually really heavy.

Of course, the heater PCBs and the glass I had on there originally does add a bunch of weight, but the undercarriage is much too heavy, too. It’s made from 6mm thick FR4, so the same material as PCBs are made out of, and it performs basically like carbon fiber, costs 50% less, but also adds 50% weight for the same dimensions. Now, the bed is incredibly stiff, that’s not going anywhere, but the weight did run me into a few issues. The toughest one being resonances, and I’ve been sensitized for them ever since. Basically, when you’re printing an area with short back and forth strokes for infill, you are eventually going to hit just the right frequency to get the entire axis assembly resonating. And at that point, your stepper motors are going to skip. So as a consequence, the y-axis runs on really low accelerations and does slow down the entire print. And the part that i figured this out on were the links of these drag chains. There’s one below the Y-axis and there used to be one for the X-axis as well, but i took that one off since i was getting some weird artifacts and irregularities that i thought originated from that drag chain having… drag. Turned out it was the filament actually pulling up on the extruder, so after i added that reverse bowden tube, it started printing a whole lot better.

Now, on the topic of extruder and hotend – right now, it doesn’t have one. It started out with a Tom’s Greg’s Wade’s extruder, but almost every single hotend or extruder i’ve tested over the years was mounted in this very printer and as a result, this carriage area is now a complete and utter mess. It was never intended to be as heavily modified and tinkered with as it was, so there are no connectors in here, it’s not modular at all and as you can see, right now, it’s even still got a boost converter dangling in there from, i believe, the Dyze-End-X review. If i end up rebuilding this, it definitely needs a complete redesign.

So as we move around the printer, you can see the general idea behind all this paneling around the frame. Now, the Mendel90’s frame is awesome in that regard, because it makes up an almost perfect box shape, you can lay it down flat onto either side, and i decided to close all side off with some more plywood. The left and right sides are mostly closed with some vent holes for the power supply and the electronics, the back is home to a filament rack, the top has this cutout for the filament to enter the printer as well as reflective surface for the RGB and white leds that illuminate the printer from the inside and provide a bit of status information, and the front was going to be a plexiglass panel, but i never got around to properly attaching that.

I wanted to eventually build it with a hinge assembly that would, like, swing up, instead of just a simple hinge on the top. Most of the time, though, i used it with the left and rear panel installed and everything else off since i regularly needed to get to the electronics in the rear right column.

Now, electronics and the mainboard. So in here is an Arduino Mega + a Ramps, and if you’ve been watching my videos, you’ll know that i can’t really recommend this setup to anyone. But it’s still ok that I use it, because it has very little to do with what the Ramps board typically looks like. I’m not using the on-board Mosfets for the heated bed, or heated beds at the time, i’ve got the second y-axis motor driver wired in separately and there are a few changes to the power delivery on top of that. Overall, it doesn’t have all that much in common with your $6 Ramps board from Aliexpress anymore. Next to it, there’s just barely enough space for Raspberry Pi, as well as a fan to cool the stepper drivers.

Now, the actual high-current magic is happening in the other pillar. There’s an 850W 80+ silver ATX supply in here, which did cost 80€ at the time, but i’m sure the better efficiency more than made up for the premium up front over the lifetime of the printer. And i didn’t want to just chop up the wiring, so i made an adapter PCB that the PCI Express and ATX connectors plug into, and that PCB is also home to the two Mosfets for the heated beds, so that i wouldn’t have to route 30A from the left column to the right one and then almost all the way back.

Did i mention there’s a 3D printed cable channel in the rear of the printer and a webcam integrated in the front?
Alright, so that should be a good overview of what this machine is all about. Now, what would i change when rebuilding this? Quite a bit.

So starting with the most drastic part, the double heated bed. Now, these two are completely independent. They have their own control Mosfet, and thermistor, they are controlled independently in firmware, and when i wanted to save some power for a smaller print, i was actually able to heat up just one and leave the other one cold. This gave me a 12 to 15cm wide strip on the edge of the glass that was mostly the same temperature.

However, i had to make some severe changes to Marlin to actually be able to control two beds – usually you have one or more hotends, but not individual bed zones. So i ended up never updating Marlin and was stuck with a version from, like 2013. A lot changed since then. Eventually, i swapped in this 150mm aluminum bed that was intended for the Cerbris, and not only was it a whole lot faster, but is also turned out that that size was plenty since i wasn’t really used to having a usable large printing envelope anyways. So i’m not sure if i’d pursue a split-zone heater again, maybe with a lot of thought going into it, but having a large, aluminum one, maybe insulated on the bottom, is probably the smarter way to go.

Next up, linear motion. 15mm belts and 12mm rail – pretty sweet, i’d do that again any day. But only with proper hardened rails, in fact, the z-axis has already been upgraded to chromed and hardened rails. The z-axis couplers, commonly called “plum couplers”, link in the video description, turned out to be pretty awesome, they have zero longitudinal play and are actually stiff enough to guide your threaded rod or whatever you use instead of letting it wobble around, and as a result, this printer has nearly zero z-wobble. The thick frame works perfectly, since it’s primed and painted, it never warped and provides plenty of strength for the printer.

What i will definitely need to improve, though, is the wiring, as if it weren’t obvious enough, but as it stands, it has no flexibility, as soon as i tried to change a single thing, i’d have to dig through layers of wires and find which one did what, and there also aren’t any proper connectors on them, at most soldered pin headers, which are anything but ideal connectors. Though I have to say, the printed drag chains are pretty neat, even if did question them along the way. When i rebuild this, i’m definitely going to put them all back into place.

But by far, what got me to use this printer a whole lot was the Rapsberry Pi + Octoprint in here. Having the pi integrated and always on, supplied by the 5V standby rail of the ATX supply, made it incredibly comfortable to quickly print something. I knew this printer was completely reliable, so for most prints, I didn’t have to come down here at all until the print was finished and ready to be picked up. The mainboard would turn on the rest of the power supply and the LEDs, so i could watch the entire process in a browser window, and when it was done, it turned the power supply off again to save power, but it was always ready to receive the next print order.

So as i said in the breakdown of the CerbrisReborn, I’m not sure if i’m going to rebuild it immediately, but something definitely needs to be done, as right now, it’s in no shape to be useful. It does still print if i’d put an extruder and hotend in here, but everything about it is just not in good shape.

Alright, so that’s the quick version of what my Mendel90 is all about – and the first and last time i’m showing it on camera in this state, in this sort of detail. So if you’ve been wondering why you only ever saw detail shots of it – yeah, that’s why. I hope you found this video helpful or at least interesting, if you did, i’d appreciate a thumbs up, if you didn’t, a thumbs down. If you haven’t done so already, get subscribed or consider dropping a dollar or two on Patreon. And that’s it for today, see you in the next one.

Prusa i2 image by prusajr

Music is Jahzzar – Pause, licensed CC-BY-SA


You can support me without spending a single penny!