Keith's Electronics Blog

While trying to print a replacement cap for a spray bottle (which I ultimately want to do in PLA, but that’s what broke my extruder and I’m working my way back up to it), sizing the cap properly for the bottle was complicated by the cap shrinking unevenly while the build was still in progress. I knew I needed to turn down the build platform temperature as soon as the first layer had adhered; but the cap was so thin that the platform didn’t lose heat fast enough to make much difference.

I connected a spare DC fan to my bench power supply, set it just outside the CupCake’s build chamber, and dialed it down until it barely spun. Holy schmoly! The left cap (you’ll need to click the image for the larger version), made with no cooling, looks like an art project woven out of twigs; the next one is extremely smooth on the side that was facing the fan — better than anything I’ve printed before!

The “seam” was on the side away from the fan and doesn’t look so as great as the fan side — but if one fan is good, two are surely better, right?

I designed and printed clips to press-fit onto a couple of 1.6″ DC fans I had sitting around (the clip is good enough to use but I have some dimensional tweaking to do before posting on Thingiverse) and positioned them at opposite corners of the build chamber, blowing around opposite sides of the object on the build platform. (Hey, I’m from Kansas. Vortices come naturally to us. Go buy a Vornado.)

The third and fourth prints above are with two fans running, trying to find optimal airspeed to cool the deposited filament but not cool the nozzle so much that the ABS solidifies before exiting. (I filed down the seam on the fourth; it didn’t really print quite that well.) I think the ultimate combination may be an enclosed build chamber to raise the overall temperature and fans to cool the layer just printed.

Many CupCake, RepRap, and Hydraraptor operators have added cooling fans, but I don’t recall seeing dramatic before-and-after pictures. Based on what I’ve observed during mesmerizing hours of watching the CupCake print, it’s obvious why it helps — without active cooling, the previous layer is still molten enough that the soft extrusion being deposited by the nozzle still applies enough force to squish the previous layer out of the way in some random direction. It may have been deposited in the right place, but it’s not in the right place any more after the next layer moves it. With cooling, it stays put.

Z-Axis Wobble and Ooze

The CupCake carries the extruder up and down on a platform supported by four threaded rods. It’s a very economical design, but the rods are less precisely straight than Acme threaded rod and aren’t a perfect fit for the ID of the pulleys at the top. The combination of these factors results in the extruder platform swaying from side to side as it moves up and down. The effect is very visible during high-speed movement but still present during low-speed movement.

My printed objects are finally at a high-enough quality that I can see the Z-axis wobble represented in the perimeter of the object. The faint ripples visible toward the left edge — about four and a half filaments high — match the thread pitch of the threaded Z-axis rods and are caused by the extruder circling through the X-Y plane as it raises during the print.

This effect is well-known and is a solved problem. Thingiverse user “twotimes” has designed a wobble arrestor that adds rigid, smooth rods to the Z axis and new bushings attached to the extruder stage to follow them in a perfect vertical motion. It’s time!

The ripple is visible at the left end of the elevation view of this fan clip (and more obvious in person than in the photograph), but it is other imperfections that catch my eye. From the front, the bottom section looks very good, but the tower at the top (which once installed is the lower part of the extruder stage clip) has considerably more variation in filament position.

From the back, note the relatively high quality up to the top of the rearward extension for the clip — until which point the entire object has a single, contiguous perimeter. Above that level, the quality of the rear tower is much worse and the wall facing it is somewhat worse than below. The quality of the forward tower drops precipitously after the top of the main clip body.

Some of this may be due to the smaller cross-section leaving less time for the fan to cool the filament, but I place most of the responsibility on the filament drive in my extruder. It’s still the DC gearmotor design; the geartrain has a lot of backlash; and it’s not good at fast reversals. I’m optimistic that a stepper design will control the oozing, already improving the print quality, as well as making it practical to enable Skeinforge’s Cool plugin to pause printing between layers while the fan cools the just-extruded filament.

I had previously noted the shrinking of the first few layers of a build on my CupCake and attributed it to the ABS shrinking too rapidly after extrusion because the room and the build chamber weren’t terribly warm. Although I’d leaned things against all of the CupCake’s windows so the heated build platform warmed the whole chamber, I thought that too much heat was leaking out when I prepped the print and removed the pre-print test extrusion and that it took a number of layers to heat the interior back up to non-shrinking temperature.

I figured I could retain more heat inside if I didn’t have to move the front “curtain” to remove the test extrusion, so I went to The Yard and picked up a big pair of Kelly forceps. Since I wanted to use them as giant tweezers, I Dremeled off the locking mechanism, leaving myself with a big pair of plain ol’ forceps.

I heated the build platform and the now-more-enclosed chamber for half an hour (longer than the time it takes to print past the warped area), then sneaked the forceps in and snaked out the test extrusion. Whaddya know — the build warped in just the same way as before.

Hm.

I don’t yet have my build platform’s thermistor connected, so I’ve been running the heater open-loop using a lab power supply to adjust the temperature via current regulation. I remembered a while back when I had it so hot that the bottom half inch of my objects stayed melty-squishy while being built … and although I no longer run it that hot, I’ve been sloppy lately about turning down the current once the first layer is adhered to the platform.

Maybe, I thinks, maybe the first few layers are too warm and pliable and get compressed as the layers above them cool and shrink. I uncover all the CupCake’s windows, start a print, and dial the heated platform current down from 2.7A to 1.8A after the first layer sticks down. I get perfectly straight walls, so I’m finally on the right track and it’s less heat that I need, not more –

but whatever temperature 1.8A delivers isn’t enough to keep the object stuck to the kaptan tape and the bracket I’m building pops loose when it’s almost finished, naturally when I’m in another room, and the last half mile of extrusion doesn’t adhere particularly well to the bracket when the bracket is no longer on the platform.

2.7A for first-layer adhesion, 2.2A after first layer, slight shrinkage, no popping loose from the platform. Now to the point of tweaking for diminishing returns. I should maybe add a fan pointed at the platform like other folks have done, although that has its own dangers (easy to cool the platform too much).

I’m eager to get the thermistor wired up (heater PCB v1.1 will have less restrictive connector spacing); translate these currents into temperatures; and connect the heater and thermistor to the CupCake for closed-loop, PID control. I will not abide the clickety-clack of relays, so I’ll find some big FETs.

I’m doing fairly well printing on my CupCake now that I have (1) my heated build platform (2) levelable with (3) a bearing-supported axle on my filament drive motor. Still need (4) a stepper filament drive and (5) roller-bearing X-Y guides.

My workshop temperature has been pleasantly cool for humans lately (currently about 67°F). The heated build platform keeps the first layer from shrinking and pulling up off the platform; but at 67°F ambient, upper layers shrink also and deform the build.

I’ve been combating upper-layer shrinkage by leaning things against the openings in the CupCake walls while printing; it traps the heat of the build platform and significantly reduces the shrinkage.

Here you can see shrink in early layers from the front of the CupCake being open while I was removing the test extrusion before the print (long recovery time for the build platform to reheat the chamber) and dramatically less shrink in the upper layers (because eventually it recovered after I blocked the front with my laptop screen). It doesn’t take very high temperatures to reduce shrinkage.

(The bottom of the object is shiny from the heated build platform, doodled with a marker for revisions, and holding a screw as an experiment with acetone and mounting boss thread durability.)

I’m (still) thinking of cutting acrylic pieces to cover the CupCake’s windows. The challenges are

• How best to attach the acrylic for easy removal for service? Hinges? Magnets?
• How to route the heated build platform cables out the back window so they don’t snag? Maybe I should rotate it 90° CW and bundle them with the Y drive and X-Y endstop cables?
• How to remove the test extrusion before printing? An auto-scrubber would be lovely, but in the short term I might get a loooong tweezer and leave an access hole in the front window, biohazard gloves sandblaster style.

Why Does It Do That?

The ABS all cools to room temperature eventually (okay, we could talk asymptotes, but I’d rather not), but it appears that only rapid cooling makes it shrinkalot. Interesting, n’est-ce pas?

Also

Devil bunny needs a ham.

The CupCake plastic-extruding 3D printer can have its printing instructions sent to it live from the controlling computer or stored onboard on a mini-SD card and printed from there. We’ve all seen suggestions to print from the card for improved print quality, but it didn’t make sense to me that USB communications were slow enough (bandwidth) or unpredictable enough (jitter) to make a difference. How wrong, how wrong I was.

I’m intending to print several different versions of the Wade filament drive to test and upgrade my Plastruder MK3, starting with Wade’s original. In the model of the idler block, each corner has a mounting bolt hole all the way through; but the CupCake slows down so much while circling the holes (without the filament drive slowing down at all, thereby depositing extra plastic) that halfway through the holes were closed over and by the end they were blobbed up above the deposition plane. Besides making an unusable object, this snags the extruder nozzle every time the mound comes by.

The bearing is supposed to spin freely in the slot and I couldn’t press it all the way in.

I’ve had tremendous luck with Skeinforge’s Unpause module before, which is supposed to mitigate the time the CupCake’s relatively slow Arduino CPU needs to calculate the toolpath for every segment of a curve, but here it didn’t help a bit. Circling the holes seemed just as slow and the result was almost identical.

I was able to press the bearing all the way in, but it was a tight fit.

The Stretch module widens curves and corners to take into account the inner edge of the filament following a smaller-radius path than the center. It looked promising on the first couple of layers, but soon I had so much blobbing that the Y-axis stepper lost steps.

The bearing fit easily into the intact underside of the slot.

Scrounged up an SD card, copied the code to it, and printed from the card. No Unpause, no Stretch, no tricks. More than an order of magnitude better than any holes I’ve printed before.

The bearing fits perfectly except where I impatiently squeezed the block too hard with pliers trying to pop it loose from the build platform too soon.

“Printing” the code from the MacBook to the card took 15 minutes. Printing from the card took 18 minutes. The similarity of those two times demonstrates that USB-serial transmission speed to the CupCake (and reception thereat) is much slower than I realized. It’s easy to see how the transmission becomes the bottleneck when sending many small steps around tight circles.

Given the widespread knowledge in the larger RepRap community about the advantages of designing (particularly) small holes as low-edge-count polygons instead of circles, I am genuinely surprised that Skeinforge doesn’t have a module to reduce tight curves, with specifications for things like maximum number of segments, minimum degrees of arc per segment, and minimum segment length.

Yes, I’ve had a couple of significant breakthroughs in my CupCake usability this weekend. I’ll cover them when I have a few moments.

Huge Blob

And a word to the wise: If you’re still printing directly from your computer, don’t walk away when your laptop battery is about to die.