Ed Nisley's Blog: Shop notes, electronics, firmware, machinery, 3D printing, laser cuttery, and curiosities. Contents: 100% human thinking, 0% AI slop.
Tag: Improvements
Making the world a better place, one piece at a time
The first step of a good print requires nailing the extrusion to the build platform. The Skeinforge Splodge plugin seems to thicken the first part of each filament on the first layer, which is not helpful. So I turned that off and added a few lines to start.gcode that do a much better job.
I also disabled the Wipe plugin, because you really can’t wipe the nozzle after the first few layers without having some part of the Z stage clobber the object. Rather than enable Wipe for just the first layer, I put a manual wipe in start.gcode, too.
The relevant sections look like this; they fit after the homing sequence at the end of the file:
(--- manual wipe ---)
G0 X54 Y-57.0 Z15 (move above wipe start)
G0 Z8 (down to wipe level)
M6 T0 (wait for temperature settling)
M101 (Extruder on, forward)
G4 P4000 (take up slack, get pressure)
M103 (Extruder off)
G4 P4000 (Wait for filament to stop oozing)
G0 Y-40 (wipe nozzle)
(--- manual splodge)
G0 X-50 Y-55 (to front left corner)
G1 Z0.50 (just over surface)
M108 R2.0 (set stepper extruder speed)
M101 (start extruder)
G4 P2000 (build up a turd)
Depending on a myriad imponderable factors, the manual wipe sequence flips off either a huge tangle or a tiny strand. That’s why I used a 4 second delay: it’s long enough to leave the extruder pressure in a consistent state no matter how it starts.
The manual splodge location depends on your platform layout; I’m thinking of putting it entirely outside the build area. It must be somewhere near the front left corner, because Skeinforge starts each new layer from that direction. Two seconds of extrusion at 2 rev/min forms a blob with a generous contact patch, although the nozzle must plow through the side on its way out.
Note that I leave the extruder running at the end of start.gcode, which means that it’s printing all the way to the outline. That won’t interfere with any part of the object, because (by definition) the first layer of the object lies entirely within the outline.
The Outline plugin puts a single filament around the entire object, allowing me to measure the actual nozzle height and extrusion width on the first layer. More on that later.
The final result looks like this:
Manual Splodge with Companion Cube
Notice that the splodge turd isn’t firmly glued to the platform, but the thread leading to the outline sticks like it was glued and the outline comes out perfectly formed. That’s the whole idea in a nutshell: paste the thread down from a stationary nozzle, then start moving with the turd acting as an anchor.
Trying to start pasting the filament with the nozzle moving doesn’t work well, as witness the left edge of the outline around these test pieces:
ABS coating on aluminum build plate
Admittedly, that was with a DC extruder, but the same principle applies to stepper extruders.
Over the past few weeks I’ve printed the gears and plate from TheRuttmeister’s Coloso-Gear MK5 extruder Thing and flatted the shaft on a moderately husky (but not hyperthyroid) NEMA 17 stepper motor. While tearing the Thing-O-Matic down to add thermal switches to the Extruder Head, I converted the MK5 Filament Drive into a stepper extruder. Much to my astonishment, when I plugged the cable in and fired up ReplicatorG … It Just Worked!
Even more amazing: the first pinout arrangement turned the motor in the correct direction!
Coloso-Gear Stepper Extruder
Some nasty pincushion distortion makes the larger gear look misaligned, but it’s parallel to the mounting plate and correctly engaged with the drive gear.
The motors arrived with short stubs of thin yellow wire on the IDC motor connectors, which I soldered directly to a much longer cable. The Parts Heap disgorged a chubby 8-conductor signal cable; I used pairs of wires for each motor connection, although one conductor would have entirely enough copper. The two cable ties around the motor prevent flexing those delicate wires as the Z stage moves.
Two tweaks to the MK6 Stepstruder profile in thingomatic.xml produced the right answers:
Set motor_steps = 1456
Set stepspermm = 48.2
Running the motor at 2.0 rpm for 30 sec should produce exactly 1 revolution of the big gear. I marked and counted the teeth on the larger gear as it rotated, and came up with 56 teeth. It’s a 51 tooth gear, so reducing the default 1600 steps/rev by 51/56 produces 1457. A defunct MBI stepper driver board that now only does full steps provides power; I resoldered all the chip pins and the fault isn’t due to external causes like no-lead solder.
Then run it for 60 seconds at 2.0 rpm and it’s under by maybe 1/10 of the tooth-to-tooth spacing. Adjust 1457 x 101.9/102 = 1456. Run it for another minute and it’s spot on.
I measured 60.45 mm for two revolutions of the big gear, so it’s 30.23 for one rev, which requires the aforementioned 1456 steps. Averaging more revolutions would yield more digits, but given the rubbery nature of molten filament, three significant figures seems entirely sufficient. I suspect this depends greatly on how deeply the extruder drive embosses the filament, so it’ll require some fine tuning.
Back of the envelope for the DC extruder at 255 PWM: feed = 45 mm/s, 0.35 mm thickness, w/t = 1.7 = 0.56 mm width gives 6.9 mm3/s. The filament is about 2.9 mm dia = 6.6 mm3, so it passed through the extruder at a bit over 1 mm/sec. There’s some windage involved in all those numbers and the extruding rate obviously depends on the temperature.
The stepper (from the usual eBay seller) is a Minebea 17PM-K150, which doesn’t appear in their catalog listing, so it’s likely one of their many custom motors. The stack length resembles the 17PM-K3xx series, which means roughly 1 A rated current. Setting the driver current to 500 mA (VREF = 1 V) produces enough torque that I cannot pull the filament back hard enough to stop it.
The step rate at 2 rpm is:
48.6 step/s = (2 rev/min) x (51/7) x (1 min/60 s) x (200 step/rev)
At that lethargic pace, the K3xx motors have something like 0.250-0.300 N·m of torque at rated current. At half current, call it 0.100 N·m and multiply by 51/7 to get 0.700 N·m = 100 oz·in.
The effective drive diameter is 30.23/π = 9.6 mm, so the available force on the filament is 0.7 N·m / 0.01 m = 70 N ≈ 7 kgf = 15 lb. Yeah, but that little 7-tooth gear will snap right off …
The reversal plugin cranks the big gear backwards at 35 rpm, which works out to 850.5 step/s. That ought to work, particularly seeing as how it’s not actually pushing anything.
The NEMA 17 steppers I picked up from eBay as part of the stepper extruder upgrade project have round shafts; that’s not surprising, as they came with pressed-on timing gear pulleys. In their new application they’ll sport plastic herringbone gears and those have setscrews.
Herringbone gears with nut inserts
Both nuts have epoxy potting to prevent moving / rotating under duress. Remember to load the screw threads with beeswax and run it all the way through before you pot the nuts, lest the screw become one with the nut. Yes, the left gear fits a NEMA 23 stepper.
(Those are 14-tooth gears. I’ll actually use a 7-tooth gear, but I printed a bunch of gears to get the hang of it.)
Any time you tighten a setscrew on a motor shaft, it’ll raise a burr on the shaft. You can pull a plastic / printed gear off a ruined shaft because the burr will simply carve a gash through the plastic. A metal-hub gear or pulley will jam solid on the burr; you definitely don’t want that to happen.
The solution, which comes standard on many motor shafts, is a flatted section where the screw can raise a burr without causing a problem. In addition, the flat prevents the screw from sliding around the shaft and producing a circular scar that makes the gear impossible to remove.
Adding a flat requires a few minutes of Quality Shop Time, but will save you considerable hassle later on. Just Do It!
Mummify the motor in masking tape to keep grinding grit and metallic dust out of the shaft bearings, then grab the shaft in a smooth- or soft-jaw vise. I grabbed a machinist’s vise in the bench vise, but use what you have.
Masked motor in vise
Apply a Dremel grinding stone / cutoff wheel along the shaft to produce a flat about the same width as the tip of the screw. The object of the game is to make the flat wide enough to keep the burr on the flat, but not grind half the shaft away.
Don’t grind the shaft without clamping it, because the vibration will destroy the bearings. Clamp the shaft to stabilize it and isolate the motor, then do the grinding.
Flatted shaft with screw
Here’s the shaft after installing & removing the gear. Notice the burr:
Flatted shaft with screw scar
And a detail of the burr:
Flatted shaft scar – detail
It’s not like I’m over-tightening the screw, either: that’s what a hardened screw does to a soft motor shaft.
Eks forced me to take a pile of crap useful make-froms, including a gooseneck task lamp that was probably bolted onto a machine tool in its former life. It sported a 20 W halogen bulb, but looked to be just about exactly the right size for those LED floodlights, which is why I didn’t put up much of a fuss about taking it off his hands.
The LED lamps are much bigger than the halogen bulb, but they fit neatly into the housing diameter. All they needed was a bit more front-to-back room, which looked a lot like a chunk of PVC pipe. The housing screws together with a 1.5 mm thread that I can’t produce on my inch lathe; I’m still not set up for thread milling. This being a low-stress application with a lamp that ought to outlast me, I figured I’d just make the belly band slip-fit the two threads, glue it in place, and move on.
I sawed off a length of PVC pipe, faced off the ends in the lathe, then CNC milled a recess to clear the male threads on the gooseneck part (I hate precision boring in the lathe). Given the rather tenuous grasp of that 3-jaw chuck, I made two passes around the perimeter: pipe ID 52.1, thread OD 54.5, remove 1.2 mm all around, about 9 mm down.
Milling top recess
On the other end, the female thread ID = 52.2 and the pipe ID = 52.1, so I glued another ring of PVC pipe inside to provide enough meat to turn it down. Once again, saw off a ring, face the ends, then cut out a segment so that the OD circumference of the inner ring is just slightly smaller than the ID circumference of the outer pipe. The result looked like this:
PVC insert sizing
Apply a heat gun to the inner ring until it’s soft enough to stuff into the pipe, clamp it until it hardens, apply PVC cement, and clamp overnight. Contrary to appearances, the ends of the two pipes are flush at the surface. Once again, you cannot have too many clamps:
Clamped PVC insert
Turning down the outside to fit the threads shows just how little meat was left on that pipe:
Skinning down to the insert
While it was chucked up (and despite my dislike of boring) I bored a bevel to accept the LED lamp and adjusted the OD so the lamp fit snugly between the end of the belly band and the lens holder on the front of the housing:
Floodlight in holder
The switch comes from the Parts Heap. A D drill puts a slightly undersized hole that’s just right for the threaded switch; I simply turned it in by hand. A length of zip cord carries the power up the gooseneck, where various ends get soldered to the switch and lamp.
I applied some hot-melt glue to the threads and pushed everything together:
Finished LED Floodlight
The glass lens on the front fits in a molded holder with an annular air gap. The LED lamp housing has all those fancy cooling fins against the inner pipe, so there’s a bit of cooling air flow around the lamp and out through the rear black section. A thermocouple reports the lamp temperature gets up around 75 °C in a 14 °C shop; a 50 °C rise might be a tad warm in the summer, but we’ll see what happens.
The power supply came from the Parts Heap: a 12 V 1 A wall switching power supply in the shape of a wall wart. For now, the zip cord from the lamp terminates in a coaxial power jack that (amazingly enough) fits the wart’s connector, but I’ll eventually put a box in there somewhere.
Clamped the butt end of the gooseneck to the backsplash on the countertop under the mill and It Just Works!
The stock Zzipper fairing handlebar mount consists of an aluminum bar with a plate welded to each end at more-or-less the correct angle to match the fairing curve. The plate has a 1/4 inch hole in one end, wherein a 1/4-20 nylon machine screw clamps the fairing to the plate, with a nylon washer distributing the stress. That doesn’t cope well with the vibrations caused by riding around here, let alone our summer vacation trips on crushed-stone rail trails, and the fairings tend to stress-crack at the holes.
These 3D printed plates are just the latest in a long series of attempts to distribute the stress over a larger area. The outside view:
Fairing mount – outside
The open hole gets another screw to hold the plates in position. The bump on the far side is an Oozebane turd, about which more later.
The view from inside the fairing:
Fairing mount – inside
You can’t see the layer of black foam rubber salvaged from a mouse pad between each plate and the fairing. That should prevent any local stress concentration at the screw and ease the transition to the tapered plate edges.
The solid model looks about like you’d expect:
Fairing Mount Plates – Upper
The hole position depends on the fairing position, as the fairings have three holes. The pictures show the fairing on my bike; it’s in the lowest position, with the screw in the topmost hole. The OpenSCAD file has an option to put the holes where you need them.
The plates are only 8 layers thick, printed with 4 solid layers top and bottom to eliminate any fill. You could do the same by setting the fill to 100%, I suppose. Using 4 outer shells (3 additional) makes the flanged edge nice and flat and uniform.
The layer height is 0.33 mm, with w/t=1.7 for a width of 0.56 mm. Feed rate = 43 mm/s and flow rate = 255. DC Extruder, alas.
Running the first layer at feed = 0.5 and flow = 0.75 produces some fluffing in the fill, but there’s no way to get a lower flow from the DC extruder motor. Flow = 0.75 corresponds to PWM=191; anything lower sometimes fails to start the motor. If it starts, it’ll run, but … that’s not dependable.
I printed them on an aluminum plate for a nice flat bottom surface.
The OpenSCAD source code:
// Clamp plates for Zzipper fairing on Tour Easy recumbents
// Ed Nisley - KE4ZNU - Mar 2011
// Build with...
// extrusion parameters matching the values below
// 4 outer shells
// 4 solid surfaces at top + bottom
// slow feeds to ensure hole perimeters stick to fill
include </home/ed/Thing-O-Matic/lib/MCAD/boxes.scad>
include </home/ed/Thing-O-Matic/lib/MCAD/units.scad>
// Select hole layout
// The if statement seems to work only for CSG object trees
// Fortunately, I need only two different layouts...
HoleSelect = 1; // 0 = his, 1 = hers
HolesTop = (0 == HoleSelect) ? [0,1,1] : [1,0,1];
HolesBottom = (0 == HoleSelect) ? [0,1,1] : [1,0,1];
// Set these to match the extrusion parameters for successful building
ThreadZ = 0.33; // extrusion thickness
ThreadWidth = 0.57; // extrusion width = ThreadZ x w/t
HoleWindage = ThreadWidth; // enlarge hole dia by extrusion width
// Plate dimensions
HoleDia = 0.25 * inch; // these are 1/4-20 bolt holes
HoleSpace = (1) * inch; // center-to-center spacing
// usually 1 inch, but 15/16 on one bike
CornerR = 5.0; // corner rounding
Layer1X = 90; // against fairing surface
Layer1Y = 32;
Layer1Z = 2*ThreadZ;
Layer2Margin = 1.5; // uncovered edge
Layer2X = Layer1X - 2*Layer2Margin;
Layer2Y = Layer1Y - 2*Layer2Margin;
Layer2Z = 3*ThreadZ;
MountX = 46.3 + HoleWindage; // handlebar mounting bracket end plate
MountHoleSpace = 13.0; // end to hole center
MountY = 16.3 + HoleWindage;
MountZ = 4*ThreadZ; // recess depth
MountCap = 3.0; // endcap arc height
MountR = (pow(MountCap,2) + 0.25*pow(MountY,2)) / (2*MountCap); // ... radius
Layer3Margin = 1.5;
Layer3X = Layer2X - 2*Layer3Margin;
Layer3Y = max((Layer2Y - 2*Layer3Margin),(MountY + 8*ThreadWidth));
Layer3Z = 3*ThreadZ;
PlateZ = Layer1Z + Layer2Z + Layer3Z;
// Convenience settings
BuildOffset = 3.0 + Layer1Y/2; // build Y spacing between top & bottom plates
Protrusion = 0.1; // extend holes beyond surfaces for visibility
//---------------
// Create plate with selectable holes
module Plate(hs) {
difference() {
union() {
translate([0,0,Layer1Z/2])
roundedBox([Layer1X,Layer1Y,Layer1Z],CornerR,true);
translate([0,0,Layer1Z + Layer2Z/2])
roundedBox([Layer2X,Layer2Y,Layer2Z],CornerR,true);
translate([0,0,Layer1Z + Layer2Z + Layer3Z/2])
roundedBox([Layer3X,Layer3Y,Layer3Z],CornerR,true);
}
if (0 != hs[0]) {
translate([-HoleSpace,0,PlateZ/2])
cylinder(r=(HoleDia + HoleWindage)/2,
h=(PlateZ + 2*Protrusion),
center=true,$fn=10);
}
if (0 != hs[1]) {
translate([0,0,PlateZ/2])
cylinder(r=(HoleDia + HoleWindage)/2,
h=(PlateZ + 2*Protrusion),
center=true,$fn=10);
}
if (0 != hs[2]) {
translate([HoleSpace,0,PlateZ/2])
cylinder(r=(HoleDia + HoleWindage)/2,
h=(PlateZ + 2*Protrusion),
center=true,$fn=10);
}
}
}
//---------------
//-- Build the things...
translate([0,BuildOffset,0]) Plate(HolesTop);
translate([0,-BuildOffset,0])
difference() {
Plate(HolesBottom);
translate([-(HoleSpace + MountHoleSpace - MountX/2),0,PlateZ - MountZ/2 + Protrusion/2])
intersection() {
cube([MountX,MountY,(MountZ + Protrusion)],center=true);
union() {
cube([(MountX - 2*MountCap),MountY,(MountZ + Protrusion)],center=true);
translate([ (MountX/2 - MountR),0,0])
cylinder(r=MountR,h=(MountZ + Protrusion),center=true);
translate([-(MountX/2 - MountR),0,0])
cylinder(r=MountR,h=(MountZ + Protrusion),center=true);
}
}
}
I loves me my Thing-O-Matic, despite its annoyances…
[Update: Stepper extruder parameters and a tweak to make the mount plate track the hole position correctly.]
This is a proof-of-concept lashup of a circuit to shut off the Thing-O-Matic’s power should the Thermal Core overheat. It vaguely resembles those doodles, but with the thermal switch cases grounded and an indicator for the main thermal switch.
[Update: You should read the rant at the bottom of that post to understand why this isn’t a firmware mod and doesn’t contain a microcontroller.]
Operation is straightforward:
The black NO (Normally Open) momentary switch energizes the DPDT relay, one NO pole of which then holds the relay power on.
The red NC (Normally Closed) momentary switch interrupts that circuit and releases the relay.
An NC thermal switch detects an overheated Thermal Core, opens that circuit, and releases the relay.
The other NO relay pole connects / disconnects the ATX power supply’s -Power On line from the Thing-O-Matic Motherboard. That connection requires a circuit-board cut to splice the relay into the Motherboard.
The LEDs:
Lower Green = ATX AC power on (from +5VSB power)
Upper Green = +Power On signal active
Red = Test / Fault (on = relay inactive)
Yellow = low over-temperature alarm
Orange atop box = high over-temp switch active
I included a second NO thermal switch that activates at a lower temperature, mostly because I had one, but that’s certainly not required. The multitude of LEDs makes for a happy-looking box; labels would be a nice touch, I agree.
When you turn on the ATX power supply, the Lower Green and Red LEDs turn on: the “Test” part of the “Test / Fault” indicator. Push the black button, the Red LED goes off, the Upper Green LED goes on, and the Thing-O-Matic is up & running. Push the red button, the TOM shuts down, and you’re back to the starting condition.
The Yellow LED goes on when the lower temperature switch goes on.
Shortly thereafter, presumably, the higher temperature switch opens, the Orange LED goes on, the TOM shuts down, and you’re left with the Lower Green, Yellow, and Orange LEDs: zowie! When the high temp switch cools off a bit, the Orange LED goes off and the Red LED goes on. After a while, the Yellow LED will go off, and you’re back to Square One again.
What’s not yet done: mounting the thermal switches to the Thermal Core in a way that’s mechanically solid, electrically isolated, and thermally dependable. I just got a bag of 100 °C NC switches, which make more sense than the 65 °C NC switches I’d been fooling with.
The wiring uses 4P4C and 6P6C modular phone connectors and cables, which are cheap & readily available, if not exactly proof against high temperatures. In normal use, failures tend to be open-circuit that will shut off the heater power. Take care not to position the cables so they melt first; they’re not intended as thermal switches.
Achtung: modular cable color codes are not standardized, particularly on the jack side, so pay more attention to the pin numbers than the colors. If I ever meet the guys who rearranged the jack colors, There. Will. Be. Gibbage.
A back view of the box shows a nice rectangular hole that’s obviously a manual CNC job on the Sherline, with no corner filing whatsoever. Hot melt glue holds the connectors in place, so I’m not showing off the inside:
Thermal lockout box – rear
The -Power On connection to the Motherboard requires the single cut shown in yellow:
Motherboard PCB Modification
It looks like this in real life, with the wire soldered to the Arduino header pin. Another dab of my Shop Assistant’s orange nail polish seals the PCB wound:
Motherboard -Power On modification
The remaining wires attach to the ATX power connector pins on the bottom of the board. The yellow wire passes through an unused mounting hole on its way to the top side, as above. Use a cable tie to tie the cable to the board, through a pair of otherwise unused RS-485 connector mounting holes.
Motherboard Connections – Bottom
While you’re chopping away at the Motherboard, add that isolating diode to keep +5 V USB power from turning the ATX fan with the power off.
The overall schematic (clicky for more dots):
Thermal Lockout Schematic
There is no corresponding PCB layout, because the circuitry forms a point-to-point hairball inside the box. If you were doing this for real, you’d want a PCB with a bazillion connections, but …
For example, here’s the FET driver for the Orange (it just looks Red) high temperature LED before a liberal application of heat stink shrink tubing:
Overtemperature LED driver hairball
You can test the thermal switches using a butane lighter.
Although that collet pusher works fine, the locking pin holder often teleported itself inside the vacuum cleaner. It recently reappeared on the far end of the main workbench, a good 15 feet away from the Sherline as the swarf flies. This, to misquote Churchill, is an impertinence up with which I shall not put.
Herewith, a replacement offering several advantages:
Won’t fit up the vacuum’s snout
Easy to grip
Perfect pin alignment
3D printing FTW!
It’s a flat block resting on the flat top of the pulley, with a nice arc matching the pusher’s OD. A small hole for the pin at exactly the right altitude makes the whole thing rock-solid stable: it slides firmly into position.
The 3D model looks like you’d expect:
Pin holder – OpenSCAD model
The finger grips were just for pretty, as you don’t need that much traction to extract the thing.
A similar view of the real object with the bottom surface up and some flash around the edges:
Locking pin holder – spindle end view
The as-printed block put the pin about 0.2 mm above the spindle hole, so I rubbed it on Mr Belt Sander (with the power off) until it fit. I printed the block on the aluminum plate platform; the Z height home setting evidently needs a tweak. However, the hole was exactly the correct distance from the top surface: flipping the block over fit perfectly.
The Smell of Molten Projects in the Morning 