Temperature Towers and other tweaks

As mentioned in the last post, I did some temperature towers (using Bob's sneaky gcode) and found better settings for PETG. The standard towers are a little too detailed to work well with the 0.8mm nozzle, but I tried anyway. Here are the Fast Informative Temperature Tower, at 4 different retraction rate settings (default 1.4, 2.0, 2.5, and 3.0):

String galore, but less so at about 220°C and 3mm retraction. I don't think I want to retract more than that, I see several sources saying it can lead to hot plastic getting too far up towards the PTFE tube.

My 0.4mm nozzle was pretty full of PLA. I didn't want to scratch it using any kind of metal tool, instead I heated it up in a candle flame to get the PLA soft enough to run out. I would have preferred a butane lighter, but ours was out of fuel. The PLA didn't run out on its own, but I was able to attach a piece of filament and 

For PETG, I'm using the steel powder bed, as the smooth one just doesn't let go of the print, and for the 0.8mm nozzle with 0.4mm layer height, I have calibrated it to a Z level of 1.150. The layer height does make a difference for that. 

I'm using the stealth mode more. It is really a lot quieter (average 46dB instead of 52, measured with a phone app, the phone lying right in front of the printer) and only minimally slower, but it turns off crash detection, which I definitely want to have on when starting a print and leaving. When I'm watching the print, it's not so bad.

Printing at large

With the new powerful 0.8mm nozzle in place, it's time to print more things!

First thing: Feet for our shoe bench. Its current height is such that the Roomba can almost but not quite get in there, and might well just get stuck. So I designed some simple feet, rounded and flanging out towards the bottom. My goal was to make it slope non-linearly, but I started with the straight slope and did a test with that. Doing a draft print using minimal infill and thin walls, I was still able to not just sit on the corner it was supporting, I could plop myself down with force, and all it would do was creak.

Having confirmed the design, I added a simple function to curve it. The trick to that is to create a function that goes from 0 to 1 but with a slant. Good candidates for this are square and square root, this is what I ended up using:

/** A function that given i from 0 to max starts at 0 and ends at max,

    but slopes inwards in between. */

function inSlope(i, max) = i * (sqrt(max/i));

Note that I give it not a value between 0 and 1, but two values that combine to give that. That makes it easier to plot into an existing design. This gives a pleasant volcano-like shape:

With the 0.8mm nozzle, I can print two of them in 1h18m, using 68g of filament. With a 0.4mm nozzle and the default 0.2mm Speed setting, it would take 3h3m but only use 53g - the difference in filament use probably due to the infill being twice as thick without there being half as much of it.

While printing the second pair, the printer did something amazing that I never expected to see. Part of the edge of the first layer had curled up on itself (probably should tune the Z offset separately for each layer thickness), so when the hotend came back to that spot on the second layer, it hit this bump. At which point, it said "Crash detected", moved the hotend off to one side presumably to wipe it clean, then moved it back and simply continued printing! More simplistic printers would either have torn off the print, leading to a messy blob on the hotend, or at least have pushed the X axis out of whack. I had at most expected it to detect the crash and stop, not to fix it itself!

Here's the final result in action:

Next thing is a hook for our fly zapper, which has otherwise had diverse less-than-optimal homes and which also has a rather strange loop for hanging it, so normal S-hooks tended to fall off when removing it. We happened to have some shelf railing pre-installed in a tiny side room to our kitchen, so hacking up some hooks for that was simple. The prints did end up being larger than would fit, so I had to redo it a few times. I guess the 0.8mm nozzle is too imprecise for this kind of work, but with a few adjustments it worked:

I was also looking at doing some auto-watering boxes, but the size I wanted turned out to make it use more filament than I found reasonable.

For the dresser next to Mickey's comfy chair, I made some simple hooks fitting the top edge, allowing her to hang some bags there:

Weird-eye view of the hooks

And finally, in preparation for getting more filament (the white already ran out), I printed two master spools, so I don't have to contribute to plastic waste with the empty spools. It's enough that I contribute with failed prints. The astute observer will notice a veritable spider's net worth of stringing here. Those can be melted away relatively easily with a flame, but it would be better without them. I did some temperature towers (using Bob's sneaky gcode) and found that a combination of printing at 220°C and 3mm retraction gave much better results.

With default settings

With 3mm retraction, 220°C

Bigger is better!

Last episode: I broke it :(

This episode: I fixed it! :)

Turns out the heat assembly works much better when you assemble it according to the instructions rather than just in whichever order. In particular, the nozzle should go in first, then the heatbreak, not the other way around.

Because I've been wanting to try it out, I replaced the old nozzle (which needs cleaning) with a 0.8mm one. That, of course, is an entirely different beast, whose main purpose in life is to squirt out as much hot PLA as it possibly can. Without breaking things, that is. Having little idea how much that actually is, I searched and found this wonderful site crammed with useful information: http://projects.ttlexceeded.com/ (a.k.a Muppet Labs). Several pages there are dedicated to how to work with different nozzles (which he likens to using different lenses on a DSLR, in particular that it's an important feature and if you're not using it, you're not getting the best out of your printer).

The first thing is calibrating volumetric rate. The Maximum Volumetric Speed is MVS = Extrusion Width X Layer Height X Speed, and is supposedly 15mm^3/s. I'm going to follow Bob's instructions on how to calibrate, which requires manual control. I've connected the printer directly with USB using Pronterface. Ah, a familiar face. In some cases you just need full manual control. Pronterface lists the extrusion motor speed in mm/min, the default being 100mm/min. The cross section of 1.75mm filament is 2.405mm.  Then 100mm/min = 2.405 * 100 / 60 mm^3/s =  4 mm^3/s, noticeably lower than the possible amount.

Bob says to use 60mm of filament per test. I'll start on that once I get up near the limits, at first I'm just checking that things work (including that there's no filament oozing out around the heater block). Finalizing the nozzle insertion, I first run a PID Autotune (M303) which has the printer figure out how the heating behaves - a nifty trick I hadn't heard of before. Since I usually run my filament at 215°, I should tune it there. I happened to first just do the default 5 cycles at 150°, then following this page on reprap.org did 8 cycles at 200°, then used the built-in autotune from the menu. The original values, as told by M503, were Kp: 16.13 Ki: 1.16 Kd: 56.23. At 150°, the final values were Kp: 21.45 Ki: 2.20 Kd: 52.33. At 200C, they were Kp: 20.45 Ki: 2.07 Kd: 50.46. So there was definitely a difference from the original.

It's unclear if the time needed to heat up from ambient temperature figures into the PID values, so I let it cool down to 39° before the 215° run, though I don't remember if I did so before the 200° run. The 215C run gave Kp: 20.85 Ki: 2.06 Kd: 52.64. I ran one the next morning from totally cold, which gave Kp:20.46 Ki:2.00 Kd:52.25 It automatically stores it when run this way, so that's easier (as expected). And given how little variation there was between the last runs, I'll just take what the built-in autotune set.

Next step is heating to 285° and giving the nozzle a final slight tightening. I wish I had an electric screwdriver with torque limit showing actual Newton values.

Now for the extrusion speed test! With heat at 215°, starting at 100mm/min, increasing by 100 at each try, I got clicking at 500mm/min, but was able to extrude at 490 without clicking, an impressive 19.6 mm^3/s. However, from 250mm/min and up, it would increasingly curl up on itself after an initial straight line, a sign that it's not able to heat it to full temperature. Trying again at 200°, I accidentally ran it at 3000mm/min, which unsurprisingly it wasn't too happy about. Instant stripping. Whoops. Fortunately, the filament wasn't broken, so I could help it unload and then reload.

Running at 300mm/min shows some curling of filament, at 400mm/min it curled enough to twist back on itself and hit the hotend. I suspect that while extruding, it doesn't apply nearly as much power as when heating, because it just tries to keep the temperature steady, but doesn't take the amount of filament fed through into account. I can see it being difficult to adjust correctly, but also leading to much faster printing if done right. There's probably other problems I haven't considered, lots of smart people have thought a lot about this.

In any case, I was able to start printing again, and at about twice the speed of the 0.4mm nozzle.

Nothing good lasts forever

Hot (or rather, not so hot) on the heels of boasting of my many successful prints comes a tale of woe and sorrow! Woe! Sorrow! Or at least some technical problems.

While printing a faucet filter tool of presumably the right size for ours faucets (I was too lazy to measure), my white filament ran out. Supposedly, the printer should handle this nicely, running an unload procedure immediately. For whatever reason, it didn't, just letting the filament continue through until it was past the drive and thus the point of no return. Apparently, but I didn't know at the time, it's feasible to just feed in more filament immediately, but I tried re-running unload, then pushing the filament piece further down. To no avail, it was stuck.

Opening the side (which hinges up nicely), I found a lot of white debris (it had ground down the end of the filament). After vacuuming most of it up, this was the sight that greeted me:

Taking a closer look with my proper camera, it was clear that the filament sat in a PTFE tube and wasn't a giant blob or anything.

So I tried pushing some more, at higher temperatures, but it wasn't budging. Eventually, I had to take the hotend assembly out, and since I'd probably need to take out the nozzle anyway (one page suggested removing the nozzle and then heating), I would swap to my 0.8mm MegaNozzle shown below (banana for scale):

After getting a set of inverted hex bits (very useful!), it was easy enough to unscrew the nozzle, but that didn't help with the blockage. Fortunately for me, they had designed the whole thing so that taking out the hotend is relatively easy, and I was able to unscrew the heatsink and heatblock. They advice against unscrewing the heatblock, but I didn't see how to avoid it, since it's not really accessible. Anyway, with that out, I was able to heat it in the oven and carefully puuuull out the filament until it went "plop" and popped out. This was essentially the classic cold pull, though possibly a bit too cold due to not having the right tools right at hand at first. The PTFE tube needed replacement, but fortunately my kit came with two spares - how prescient!

So I reassemble the whole thing and run a test without filament - after all, I need a nozzle in there first, which requires 285C. Alas, the heating is bad - it heats up, but very slowly, at 100C about 1C per second. I didn't let it run for long enough to get up to full temperature, it took over a minute to get to 135C. To check that it wasn't the thermistor misreading it, I poked an old PLA print at the heat block, but it only deformed slightly, so it definitely wasn't doing runaway heating. That's something at least, overheating could be quite dangerous. But I'm still stuck with bad heating. I double-checked the wiring in the board, and the resistances and voltage were correct. So either there's a loose wiring in the heatblock end, or... something else.

Many good prints done

Using PETG from Beta2Shape on the steel dust surface, I calibrated it to -1.09mm. Prints still fell off unless I used 3dlac. When I tried the flat surface, I was hardly able to get the print off at all. There is also a lot of very fine stringing when I use the default Prusament PETG settings. I'll try some of the tips in https://all3dp.com/2/petg-stringing-3-easy-ways-to-prevent-it/ to improve that.

For PLA, I couldn't use the steel dust surface, instead I use the flat one calibrated to -1.0mm. The PLA from DasFilament has done good by me. Curiously, after switching to the powdered surface and back, the calibration changed to -0.8mm. Clearly recalibration is a good idea after switching surface.

The printer has been quite busy. I joined makervsvirus.org and printed 50+ face shield holders:

I look super derpy in this shot, but it shows how the face mask works, just add a sheet of clear plastic.

50 of them delivered to our local hub.

I printed tools for my coworker Ilham for taking off the faucet filter (in German a "Percolatorschlüssel"):

Generic version that works on many faucets 

Version for a specific size of inner teeth - I didn't even know there were inner teeth! This was the size that Ilham needed, but difficult to get a good grip on

A version with better grip, but too large for our faucets. 

Some other things against Coronavirus:

My wife is doing a bunch of masks, but ran out of bias tape - as did all the suppliers. So I printed this nifty bias tape folding device, it works quite well.

Another thing for masks, this holds the straps instead of your eats getting pulled at. Unfortunately it slides downwards, and so does the mask.

And another couple of useful things around the house:

I got a used bike trailer of a kind I've previously used for the Monachium war chariot, but this one came with an attachment that tried to fill the same space as my disk brakes. Since physics has some stern words to say about doing that, I made an adapter for the other kind of attachment, printed in PETG (which is less brittle and more heat-resistant) and somewhat over-dimensioned. It works nicely so far.

For working from home, I found my WiFi connection utterly inadequate, so I decided to lay Ethernet into my room. These are Cat6-compatible clips that can be stuck onto the wall. I now have a pretty well routed wire, firmly attached. 

The new PrusaSlicer 2.2.0 has some nice details to it, like when an SD card is detected, it will add a button to write to the SD card and after that another button to eject. Now that's a UI that helps smooth the common path. Too many engineers in particular forget to do that, presumably thinking that it's enough that the functionality is available, there's no need to duplicate it. Kinda like the mathematician reducing a fire to "a known problem."

Perfect prints produced by Prusa printer

It printeth! Verily, it printeth! My new Prusa i3 mk3s printeth! And wow, does it do a good job of it. Here's the obligatory Benchy McBenchface, in a quality I haven't even gotten close to on my old printer:

None of the sloppiness around the brow, no extra lines around edges, the print on the bottom totally clear, and it's almost possible to read the tiny writing on the aft. Impressive!

I got this as a kit (of course!), and putting it together was pretty easy, though not trivial. There's quite a tome of a manual, with each step illustrated with photos and text. No soldering is involved, and all the necessary tools were provided. On the down side, some texts were ambiguous ("put the screw in the left hole" - but I can turn this thing any which way!) and the photos were not very good, especially since they had to show details of black on black. I could certainly take better shots (and did a couple of times, just as a sample). I worked together with my friend Christina, and we only made one mistake where I had to go back and undo some of the assembly. Inserting the nuts was sometimes tricky, in most cases we were able to pull them in using a screw from the other side, but not always.

The basic design is the same as my current one, with the bed moving in the Y axis, the extruder running along the X axis, which is in turn lifted by threaded rods on the Z axis. The details have improved a lot, though. Gone are all the endstops, since the motors are able to sense when they skip, which is enough to detect the end. The extruder has a built-in filament sensor, so it can stop printing if running out of filament. There's also a power panic system in case of power outages. The bed is with a separable bendable steel plate, held on by so many magnets that there are warnings not to use it if you have a pacemaker.

The operations are also very easy. There is a panel in front with a little dial, which allows controlling the printer in many ways. Starting a print from an SD card is three clicks away, and it auto-calibrates the bed height before each print. The self-test and larger calibration was quite effective, pointed out an issue I had missed.

For Z axis calibration, there is a live adjustment option, so while the print is happening I can adjust - micrometer by micrometer - how close the the bed the first layer should be. To get the right height, I printed the Prusa logo model that came on the SD card, adjusting the height by 0.05 mm every two centimeters or so. I stopped printing after the first layer, so I could see the quality:

Top side of print. Left part is with the nozzle closer to the bed

Bottom side of print. Left side is nozzle closer to the bed. 

It was clear when the nozzle was way too low or way too high. In between, at about -0.95 to -1.1, the printing was just about perfect. This is a way better calibration than the single-strip built-in method.

The main problem since I got printing going has been to make the prints stick during printing. At first, I just used the recommended isopropyl alcohol swap to clean the plate. That didn't help much. Increasing the bed and extrusion temperature made it possible for a large flat piece to stick. Washing thoroughly with soap made things somewhat better, but printing this little hedgie proved too much for that. Next step was my trusty hair spray, which didn't help either. The trick from this video of putting a bit of gluestick on and spreading it with alcohol made no difference whatsoever. Only when I smeared the printing area with gluestick did it actually stick, but then that also adds more thickness. After reading various discussions about this problem, I ordered a flat sheet instead. It's supposedly less durable, but being able to print without smearing the plate is important.

I also ordered some filament - this is a 1.75mm extruder where my old printer is a 2.85mm - and some different-sized nozzles. I even got one of the experimental 0.15mm nozzles, since I want to try printing miniatures for gaming. I understand even using the 0.25mm one is tricky, so I don't expect instant success. I grabbed a 0.8mm nozzle for the old printer, it can be relegated to the easier tasks.

All in all, this is a really nice printer. The printing world has moved a lot since I got my Mendel90 half-kit, and it shows. I expect (hope) to be able to spend less time repairing the printer and more time designing.

The new wonder! What shall I call it?

An upgrade

The last post showed the broken foot pedal on our bio garbage can. I managed to epoxy it back on, but because it was so cold where I did it, the epoxy didn't fully set (and ran over the support so I had to cut through it), and within a month of further stepping, it broke off again. No saving it this time.

My printer has mostly been standing still, partly due to my own carelessness, partly due to parts failure. I neglected to level properly before a big print and ended up with so much backpush that the filament stripped, then the rest of the filament in the hotend oozed out. This resulted in stuckness that couldn't be resolved with a cold pull, because there just wasn't enough heating up going on. In the end I took the cooling part of the hotend off and placed it on a toaster (flatbed style, known as "Deathtrap" to non-Danes) until the filament became soft enough to pull. This bought me a bit more printing time before the aforementioned cracks reappeared, and I'm not going to try to re-glue that piece. Fortunately, I had printed another extruder block, which I'm now cleaning of excessive elephant's foot before mounting. I also tried, for extra reinforcement, to apply some contact cement to the side of the block. I imagine it can seep in through tiny cracks and bind it together a bit. Would be interesting to compare epoxy and various glues not for bonding two parts, but for reinforcing one.

In the meanwhile, I have turned 50 (who'd have thunk?), and in that connection my wife arranged a Christmas and birthday gift conspiracy to get me a new printer. Enough money came in to get me a Prusa i3 mk3s, and maybe also the MMU. It should just barely fit in my workbench. While the Mendel90 has been good, it has also suffered from some build problems and limited upgradability. The rest of the world has moved on, and getting proper bed levelling, motor feedback, a large print volume, a better plate, and many more improvements will allow me to focus more on the design part than on maintaining the printer. I will probably still keep the old one around (somewhere) for when I need more printing done quickly, but my day-to-day printing should be easier with the Prusa.