The printer arrived in a box of subassemblies and a bag of M3 bolts, and the first thing I learned is that "some assembly required" is doing a lot of work in that sentence. Three hours later I had a frame that was nearly square, a heated bed that was nearly level, and an evening's worth of optimism that the machine was about to drain completely.
I'd wanted one for ages. The excuse, when it finally came, was a broken clip on a cable raceway that I couldn't buy a replacement for. The honest reason is that I wanted to make things at home with software, and a printer is the most literal version of that I could think of. You write a description of an object, the slicer turns it into instructions, and a motor and a hot nozzle turn the instructions into a thing you can hold. It is deeply satisfying when it works, and I should be clear up front: in week one, it mostly didn't.
the calibration cube humbles you
Everyone starts with the calibration cube. A 20mm cube, no detail, the printing equivalent of "hello world". Print one, measure it with calipers, and if it comes out 20mm on every axis your machine is roughly honest.
My first cube did not print. The first layer didn't stick, the nozzle dragged a sad bird's nest of PLA around the bed for a minute, and then it was extruding into thin air a centimetre above where the plastic should have been. I cancelled it, scraped the mess off, and started reading.
The problem was the thing everyone tells you is the problem and you don't believe until it happens to you: bed levelling and first-layer height. "Level" is a misnomer; what you're actually setting is the gap between the nozzle and the bed at the four corners, and it wants to be about the thickness of a sheet of paper. Too far and the plastic doesn't adhere. Too close and the nozzle scrapes the bed and clogs. I'd set mine by eye, which is to say wrongly.
I levelled it properly with a folded piece of paper at each corner, adjusting until the paper just caught with a little drag. Then I set the Z offset in the firmware so the first layer squished slightly rather than floated. The next cube stuck. It measured 20.1 x 20.0 x 19.8mm, which for a first go I was thrilled with. The 0.2mm out on Z I've since learned is a separate calibration rabbit hole, and I have decided to live with it for now.
the failures, catalogued
It's worth writing these down while they're fresh, because the forums are full of people who have forgotten how baffling the early failures are.
- Nothing sticks. Bed too far from nozzle, or a cold/dirty bed. The fix was levelling, plus wiping the glass with isopropyl alcohol to get the greasy fingerprints off. Clean glass and the right gap solved nine out of ten of my adhesion problems.
- The print detaches halfway up. Warping. The corners of a larger flat part curl up off the bed as the plastic cools and contracts, and once a corner lifts the nozzle knocks the whole thing loose. A bit of bed heat (60°C for PLA) and a brim around the part held it down.
- Stringing everywhere. Fine wisps of plastic strung between separate parts of the print, like the model had been left out in a spider's web. That's the nozzle oozing while it travels. Retraction settings in the slicer pull the filament back slightly during travel moves; tuning that cleaned it up.
- A horrible grinding sound and no plastic. The extruder gear chewing a flat spot into the filament instead of pushing it. That was a partial clog: a leftover bit of cold plastic narrowing the nozzle. A cold pull cleared it.
None of these are interesting problems. That's the point I keep coming back to. I expected the hard part of 3D printing to be the modelling, the geometry, the clever software. It isn't, at least not yet. The hard part is that this is a physical machine doing a physical thing, and most of the failures are mechanical, thermal, or just dirt. The software did its job correctly every single time. The plastic and the metal needed persuading.
the first real thing
On day five I printed the cable raceway clip that started all this. I measured the old one with calipers, drew it in OpenSCAD because describing geometry in code is more my idea of fun than pushing it around in a GUI, and printed it.
It didn't fit. The clip was 0.4mm too wide on the inside, which is exactly the kind of tolerance error that the calibration cube was trying to warn me about. I adjusted the dimension, reprinted, and the second one snapped onto the raceway with a click that was genuinely one of the more satisfying noises of my week.
// raceway clip, second attempt
inner_w = 11.6; // was 12.0, too loose
wall = 1.6;
height = 8;
difference() {
cube([inner_w + 2*wall, height, 6]);
translate([wall, -1, 1.5])
cube([inner_w, height + 2, 6]);
}
That clip cost me about five days of fighting the machine and twenty minutes of actual design, and it replaced a part I couldn't otherwise buy. By any rational accounting it was a terrible use of a week. I don't care even slightly. The thing in my head is now a thing on the wall, made of plastic I melted in my own house, and the next one will take ten minutes because the machine is calibrated now.
The lesson, if there is one, is that the printer is a tool that demands you respect the physical world before it'll do anything clever. Get the bed level, get the first layer right, keep the glass clean, and most of the dramatic failures simply stop. Everything I read told me that. I just had to fail at it personally a dozen times before I believed it.