MakerGear Forum

It's been a pain in the kaboose trying to perfect this but I've successfully figured out how to do it.

Callibration
--------------------
We want the ABS to lay down in extremely flat layers in order to adhere well to the glass bed, so get the extruder as close to the bed as possible. It's critical to get it /extremely/ flat because ABS does not want to stick, and it's far more temperamental than PLA in this regard. I do this by lifting the bed until it's /just/ touching the extruder (With the red light just a slight turn away from z end-stop engagement), then back off one small turn so there's a razor thin gap between it and the extruder (I clean off the extruder with a knife to expose the tip.). I do this until all four points on the "cross" have a razor thin gap with the bed. I find the "business card" or "paper" method doesn't really reliably get the kind of razor-thin gap ABS needs.

1) Wash down the bed with acetone/water.
2) Print the first layer really slowly. I personally do 10mm/s (I have a special program that reduces the first layer wall/perimeter speeds to 3 mm/s). The reason is that the corners/edges of the ABS lift up really easily when going faster, likely due to the 'tug' of the extruder at the corner/edges, so going slowly helps minimize lifting corners and getting a nice flat layer. If the line being laid down looks "round" /at all/ (Not even half-round works), you're likely going to have problems - It needs to be flat like a tapeworm!
3) Keep the 50mm fan off. The cooling causes the object to pop off the bed after the object reaches around 3-7 mm in height, due to the cumulative tension of the ABS wanting to shrink.

I find that 230C on the filament and 110C on the bed works well.

Welcome to the world of no-mess and perfectly shiny and flat ABS prints!

(Btw, the "cross" is the callibration technique I use. Front-middle, then back-middle (To level the Y axis first), then middle-right and middle left (To level the X-axis second). Works well for me 100% of the time.)

I...what...huh?

Really? This works?

Interesting! (Should dub it "the Tapeworm method".) Does it hold for really tall prints (over 4") or do you still have to fight the cracking?

Absolutely no problems here! Seems to work perfectly with hatchbox ABS. This other cheap chinese ABS I got off of ebay, though, does seem to have a problem with delamination so I guess it depends on the ABS. I'm currently printing a large 6" tall part and it's showing no problems, and none have shown problems yet. It seemed like turning off the fan was the biggest improvement for preventing the part from popping off the bed, and it seemed like getting the extruder extremely close to the bed and laying an extremely flat bead was necessary for greatly improving ABS-glass adhesiveness. Otherwise, it seems the bottom layer easily peels off with the extruder going back and forth.

6" tall ABS part just finished, all looks good.

It's actually quite sticky when the glass bed is 110C, and pulling the part off the bed while hot can warp the bottom layers appearance (Make it bulge a little). I've tried water cooling, which quickens the cooling process, but it seems to cause warping of the part (The entire part bows upward). Letting the part air cool until removal seems to be the only cooling method I know of that leaves a perfectly flat bottom and leaves the part non-warpy. It'd be nice to figure out a good way to quicken cooling without warping the part. (It seems like the part easily detaches below 60-70C)

I suspect one could probably print with a glass bed at 80C with this method, given how hard it is to remove the object at 80C. That might quicken the cooling/removal process. It seems the flatter the bottom layer, the greater it sticks to the glass bed.

Got a new glass bed, tried this method at .28mm layer height. Not good, seemed to be extruding too much causing the edges to lift in parts. However, it also seemed to be lifting way too easily so I washed it with acetone then wiped it with a water-soaked terry cloth to get rid of the streaks (The spots and streaks transfer to the bottom layer, making it look ugly). I'm thinking the new glass might've had something on it, probably something that came with the cutting/manufacturing process - perhaps cutting oil or coolant. Anyway, I'm trying again! Let's see how it works. This method worked fine at a layer height of .2mm and .15mm - maybe it's finicky with .28mm layer heights? Let's experiment and learn.

Edit: Ok, cleaning it didn't help. The nozzle is probably too far away from the bed. Let's try to get it even closer, lol. The cool thing is, once it's callibrated, it's good to go for some time before another callibration is needed.

Turns out that it can't be "too close", but it also can't be "too far". There's a pretty tiny window where it's the perfect distance from the bed where it stays flat and produces no filament that's hanging on the side of the line on each pass. I'll try to find a method to find this "perfect window" (Tuning the bed in-print seems to work, but requires constant attention during the first layer. But, once it's tuned, it's tuned.)

For .28mm layer height and 110% first layer height with Blue 1.75mm Hatchbox ABS and extrusion multiplier of 1.00 and extrusion width of .40mm with a .35mm nozzle (All of which applies to me), it seems the following works for calibrating.

Clean the tip of the nozzle exposing a fresh tip.
Put a 65lb piece of cardstock between the nozzle and printbed.
Get it to the point where the cardstock becomes difficult to move, and increase the nozzle height /just/ until the cardstock easily moves around. This should be done /right/ before the z endstop's red light comes on.

Seems like this reliably gets the right height for the above settings.

Not too close, not too far away: Just right.

Edit: Well, this is awfully queer. It looks like the extruder simply completely reversed the filament out of itself. That's never happened before. Is this an S3D quirk? It looks like it started going off in a random direction, laying down filament at random and crossing everything in its path. What the hell happened here?

It appears that it started traveling perpendicularly to the previous line it traveled in, which could only be possible if it was traveling at a 45 degree angle and then the x and y-travel code got switched during transmission causing it to travel perpendicular to the direction it was supposed to travel. I'm willing to bet that's what it was, now why would the x and y-codes get switched? Or, why would the x and y motors switch? Or, wait, I guess all that's necessary is for the x-motor to get a negative value instead of a positive value, switching the x-motor's direction - or, maybe the x motor's direction never got reversed for some reason and it just continued on at the "turning point"? Man, this is weird.

Oh well, it reinforced the importance of wiping with a damp floursack cloth after an isopropyl alchohol wipe (Doesn't seem like acetone is necessary to promote adhesion). There were miniature streaks imprinted on the glassy bottom layer because I didn't wipe it with a damp floursack cloth afterwards.

Edit: It's doing it again. Reversed the filament and it's in zombie mode, at the exact same spot as last time. That means: There's something wrong with the G-code! Which is weird because it's not showing up in the preview window, whatever the extruder thinks it's doing now.

Hmmmm... the log window says "Firmware is unresponsive." repeatedly. That might have something to do with it. But, why's it suddenly unresponsive at the exact same spot in the print?

Okay, restarted S3D, created a new 'workspace' and put the same pieces down and closed down all other programs that might want to use the port. Also, reconnected the USB. *crosses fingers*

Okay, logs as of 1 minute into the print show no sign of firmware unresponsiveness. This is a good sign.

WTF! It reversed the filament and went into "zombie mode" at nearly the exact same spot into the print, despite starting a new file and recreating it entirely from scratch, guaranteeing a new set of G-codes. The logs also show "firmware unresponsive". WTF IS GOING ON HERE?!!! The utter repeatability of this suggests that it's S3D that's at fault but... how?? This is utterly mind blowing and f*cking ridiculous!!

Screw this, I'm running the G-codes from Cura. I'd hate to believe that the G-code is really at fault, but running from Cura will help prove that it's S3D, not the g-codes, that's at fault. Or vice versa, depending on if it does it again.

Edit: Examined the evidence more closely. Doesn't reoccur at the /exact/ same spot, just the general area. Indicates probable issue of S3D as a g-code host, not necessarily a problem of the g-codes. Executing the G-codes off of cura will prove whether it's an S3D issue - if it reoccurs, then it's possible it's a subtle hardware/software issue of some sort that I'm not likely to solve. Or, if it is a g-code issue, it doesn't seem to have exact repeatability. Anyway, I'm closely watching it now.... something's "shrilling" as the extruder goes back and forth. it's getting really close to the 'crime scene', so 'it' could happen any moment now.

Well, cura just executed the gcode perfectly indicating there's a problem with S3D as a g-code host. I'd almost suspect the developer is playing some sort of trick on me through the internet connection (The software is always connected via internet to its licensing server) given that I would doubt that S3D is really /that/ crappy as a g-code host since it's been working perfectly until /just/ now, tonight, and it's been malfunctioning almost the same, but not identically, between different g-code files. Of course, maybe there's a hacker screwing up my prints. Who knows. I guess that's a possibility since it seems my computer's become part of a botnet given how google keeps asking me for captchas, despite knowing I've never used google in a high volume way to deserve such constant captchas from google. So, I guess 'foul play' is the most probable possibility, even though I guess S3D just being 'bad G-code host software' is plausible. Still, reversing the filament and then going into 'zombie mode' seems absolutely bizarre, even for 'bad software'. That takes particular G codes that can't be 'accidentally' done by a typical software bug.

Cura doesn't have constant internet connectivity, unlike S3D, so that may be its saving grace. Okay, well, I've determined that I'm executing my gcodes from cura from now on a separate, dedicated print server. That absolutely settles that question.

swbluto wrote:... why's it suddenly unresponsive at the exact same spot in the print? ...
... The utter repeatability of this suggests that it's S3D that's at fault but ...
... it's been malfunctioning almost the same, but not identically, between different g-code ... files

Unless I've missed something, you're doing all of your printing from Windows through USB, which is a known source of mysterious printing problems that defy solution.

since it seems my computer's become part of a botnet

And that, children, is why friends don't let friends use Windows.

You should choose one of:

• Print from the SD card
• Switch to Linux

Think of it as using the right hammer for the job. [grin]

There's something wrong with the G-code! Which is weird because it's not showing up in the preview window

You should cross-check your G-Code with http://gcode.ws/, because that will give you a command-by-command look at the file. As a rule of thumb, however, if the G-Code looks good, then the G-Code does not cause erratic behavior. That's not always true, but it's a very very very good first approximation.

Feeding G-Code to the printer requires that the PC pay constant attention to the USB hardware while running with fairly precise timing, neither of which Windows does very well under the best of circumstances. Before blaming firmware / S3D / G-Code / hackers / alcohol, eliminate the single common factor that's the main source of printer connectivity problems.

Please?

Well, it's settled. S3D is a bad G-Code host because of its lack of robustness in dealing with Window's USB connectivity issues. Pronterface would freeze when it came across these issues, but I guess S3D's behavior of 'always trying to reconnect and continuing the print' isn't nearly as good as Cura's.

I've found a better calibration technique.

For .28mm layer height and 110% first layer height with Blue 1.75mm Hatchbox ABS and extrusion multiplier of 1.00 and extrusion width of .40mm with a .35mm nozzle (All of which applies to me), it seems the following works for calibrating.

Clean the tip of the nozzle exposing a fresh tip.
Put a TCBY Frozen Yogurt membership card between the nozzle and printbed. (It's the thinnest solid plastic card I have, lol. It's thicker than the 65 lb cardstock I have.)
Get it to the point where the cardstock becomes difficult to move, and increase the nozzle height /just/ until the card easily moves around. This should be done /right/ before the z endstop's red light comes on.

Seems like this method is producing great prints for me. Not too far away so that the ABS doesn't stick, but not too close such that it creates excess on the edges encouraging lifting of the outline on each pass.

An easy way to tell if it's too close is if the surface of the bottom layer's infill looks 'hairy' or looks like it has hairballs; if it's producing visible excess on the edge as it goes back and forth, definitely too close. It should look like it's well filled in and perfectly even and uniform.

Using the above method, it seems that...

TCBY card - .73 mm thickness. Produces "round stuff" that sticks, but sometimes the initial line being laid down comes up (Dragged by the extruder; destroys the beginning of an outline on the bottom layer, but doesn't seem to destroy the part and bottom layer appearance actually looks 'okay' despite this in the final product.) and it also produces visible lines in the bottom layer. It, however, doesn't seem to lift the edges as the extruder makes its passes. Otherwise, it works. Seems to easily pop off the bed without warpage around 90C. (Haven't tested large objects with this 'thickness'. Might warp off the bed with large objects.)

65 LB cardstock - .17mm thickness. Produces nice flat layers that seems to cause no lift and adheres well, with a tiny bit of excess on the edge and no visible lift on the edges as it goes between each pass - the top surface of the bottom layer seems to be a tiny bit 'hairy'; the bottom has /barely/ visible lines and feels perfectly smooth. Going below this level of thickness is likely to cause lifting issues due to excess buildup on the edges with each pass. Easily releases around 80C with no warpage. 10mm part measures 9.86mm thickness.

To test - 110 LB cardstock - .33mm thickness. Initial appearances - seems to be laying down a pretty perfect bead. No excess at all on each pass, yet, seems to be filling everything in. No hairballs or "hairiness" at all on the top surface of the bottom layer. It doesn't seem like the edge is at risk of being lifted and it seems like the beginning of the line doesn't have a tendency to lift. Finished: Bottom is reflective, lines are visible but perfectly even. Feels smooth when rubbing along the grain, but can feel a faintly detectable 'grainy' feel going against it. Easily popped off at 84C with no bottom layer warpage. 10mm part measures 10.09 mm thickness. Might retry with something like "First layer height: 75%" instead of the current 110%.

110 LB cardstock(.33mm thickness) with 75% first layer height instead of 110% - To test. Same result as above except bottom feels slightly less grainy. The part measures a true 10.00mm in thickness.

Will try 50%. I have a feeling anything below 40% will just recreate the 65LB result.

110 LB cardstock(.33mm thickness) with 50% first layer height instead of 110% - glassy smooth bottom, top of bottom surface layer didn't have any hairiness or hair balls, looked perfectly even. Part measures 10.00mm, oddly enough. Part released at 80C and lines on the bottom are barely visible. Looks like we found a winning calibration technique. Start with 110LB cardstock and gradually reduce layer height on the specific printer to find the perfect layer height %.

Anyway, it seems the "sweet spot" is somwhere between .17mm and .73mm. I'm thinking the 110LB cardstock will produce the best result - great adhesiveness, no excess on each pass, no lifting, completely flat bottom with little to no lines visible, and will easily detach around 85C.