Ed Nisley's Blog: Shop notes, electronics, firmware, machinery, 3D printing, laser cuttery, and curiosities. Contents: 100% human thinking, 0% AI slop.
For some unknown reason, one of the very rare updates to the Ubuntu 10.04 LTS infrastructure (for LinuxCNC 2.5.3 on my Foxconn D510 box, driving the Sherline mill) stopped supporting the system board’s built-in NIC: networking stopped working. The only symptom was that the NIC didn’t respond and all the usual tricks were unproductive.
After some fruitless searching, I took the easy way out:
NIC added to Foxconn D510 PC
That’s the backside of an ancient NIC using the classic Tulip driver. It used to have a full-size bracket, which I chopped off, bent, and filed to suit, much as with that one in the D525.
Fired it up, the kernel automagically picked the proper driver, and networking Just Worked again.
A few months ago I fired the Thing-O-Matic, only to have it wake up dead. Not exactly dead, but spitting out checksum errors on simple G-Code files sent from Pronterface, which used to work just fine. Trying a bit of this-and-that to no avail, I proposed to The Mighty Thor that I could loan the carcass to Squidwrench, reanimate it with a less bizarre set of hardware and firmware than the much-hacked Makerbot menagerie under the hood, and use it as an exemplar in my 3D Printing classes.
Fortunately, that particular Thing-O-Matic has the most well-documented hardware evah…
Matt suggested an Azteeg X3 controller, because it has thermocouple inputs that match the existing sensor, Thor ordered one, and I tinkered up a first-pass version of Marlin that could read the inputs and twiddle the motors. The firmware is on Github, not that you’ll need it for anything you’re doing; more on that later.
Here’s the Official Doc for the microstepping jumpers hidden under the driver boards:
Azteeg X3 – microstep jumpers
That’s XYZE = 16 16 8 4, respectively, with a spare slot (and spare driver, not installed) for the second extruder it’ll never have.
The extruder’s Type K thermocouple connects to the TC1 port on the shield, exactly reversed from the way you see the test thermocouple there: the red lead is to the left, the yellow lead is to the right. If you get it backwards, the indicated temperature goes down when you touch the bead. The printer’s thermocouple has some backstory.
The 10 kΩ thermistor bead connects to the BED port on the main board and isn’t polarized. The Heated Build Platform has a bit of backstory, too.
The gutted TOM286 carcass with the MBI hardware off to the side:
TOM286 – gutted electronics bay
After a few sessions, it looked pretty cheerful again:
This is what you see when looking down through the acrylic baseplate:
Azteeg X3 – inside TOM286
The blurry silver rectangle off to the left is an aluminum channel glommed to bottom of the acrylic baseplate with silicone snot to eliminate a nasty mechanical resonance.
The thermal cutout circuitry isn’t wired in yet; the ATX power supply has its -Power-On pin hotwired to the adjacent ground pin for now. The X3 gets its power directly from the +12 V supply, so there doesn’t seem to be any way to power the X3 from the +5 V Standby ouput, deliver +12 V to the motors, and switch the supply through the X3’s ATX output pin.
The heaters work fine, the motors turn properly, and the extruder feeds molten plastic; all the motor calibrations seem to be pretty close. The first test object was a total botch, of course, but the printer’s parts seem to work OK again.
A discussion on the Makergear Google Group about a heated enclosure prompted me to run the numbers for cooling stepper motors with water, rather than fans and finned heatsinks.
The general idea comes from my measurements of the air-cooled heatsink stuck to a stepper’s end cap. The metal-to-metal conductivity works surprisingly well and reduces the case temperature to slightly over ambient with decent airflow through the heatsink; epoxying a cold plate to the end cap should work just as well. A NEMA 17 stepper case is 42.3 mm square, so a standard 40 mm square CPU cooling plate will fit almost exactly.
The question then becomes: how much water flow do you need to keep the motors cool?
Some numbers:
Water’s heat capacity is 4.2 J/g·K
1 J = 1 W·s, 1 W = 1 J/s
NEMA 17 motors dissipate about 5 W (13 W if you’re abusing them)
We’ll cool all four motors in parallel, for a total of 20 W
Allow a 5 K = 5 °C temperature rise in each cold plate
Rub them all together:
(20 J/s) / (5 K * (4.2 J/g·K)) = 0.95 g/s
For water, 1 g = 1 cc, so the total flow is 1 cc/s = 3600 cc/h = 3.6 liter/h, which, here in the US, works out to a scant 1 gallon/hour. It’s tough getting a pump that small and cheap flowmeters run around 0.5 liter/m…
If you don’t want a pump. put an aquarium up on a (sturdy) shelf and drain it through the cold plates. A cubic foot of water, all eight gallons and sixty-some-odd pounds of it, will last 8 hours, which should be enough for most printing projects.
If you want reliability, drain the coolers into a sump with a float switch (high = on), put another float switch (high = off) on the aquarium, and have the pump top up the aquarium. If the pump fails, your steppers stay cool for the next 8 hours. Heating the water about 5 °C during 8 hours won’t require active cooling.
Now, managing the hoses leading to the X axis stepper may be challenging, but a cable drag chain would control the rest of the wiring, too.
Browning HP Mag Blocks – stainless and plastic – side
It’s actually bronze-infused stainless steel powder, so it’s not exactly solid steel. The parts spend a day rattling around in a vibratory polisher that slightly rounds off their edges and smooths the surface, but (as with all 3D printed objects) you must learn to love the results; it’s certainly more photogenic than the black plastic version from my M2.
The bottom view shows the hole I added to reduce the metallic volume; they charge a bit under $0.01/mm3, which encourages airy design:
Browning HP Mag Blocks – stainless and plastic – bottom
A cross-section view of the solid model shows the interior structure:
The vent pipes are somewhat larger than in the plastic version and, obviously, I didn’t include the yellow support structures in the model I sent to Shapeways.
Their specs give a minimum wall thickness of 3.0 mm, which I’m definitely pushing on some of the internal features. The pipes came out perfectly, as nearly as I can tell, although some polishing media did get wedged in the smaller hole. Air passes freely across the top, which is the important part.
Although the specs list a ±2 mm (!) tolerance, a comment in a Shapeways forum said that applies to larger objects, with 0.2 mm being typical for smaller objects. The steel and plastic parts match within 0.2 mm of the nominal model dimensions, so that lower tolerance seems about right; I have no idea how consistent it is.
Another comment recommended carbide tools for secondary operations and that’s definitely true; I wrecked a perfectly good HSS tap trying to thread the central hole. Fortunately, I made the block slightly smaller outside and slightly larger inside, specifically to avoid having a deep thread; I intend to ram a standard M3x0.5 SHCS into that hole and epoxy it in place without worrying about thread damage.
That image has desaturated red to suppress the camera’s red burnout. It looks better in the realm of pure math:
Planetary Gear Bearing – Kurled – solid model
Reducing the tolerance parameter to 0.4 produced a surprisingly rigid, yet freely turning, bearing that required no cleanup: it popped off the plate ready to roll!
The heavy lifting in the OpenSCAD source code remains emmitt’s work. I replaced the outer cylinder with a knurl and simplified his monogram to stand out better amid the diamonds. This is the affected section:
Turns out that it’s 0.1340 inches, determined by bracketing the sliver above that 0.1300 block with feeler gauges. I don’t believe that last zero, either, as the Basement Shop was about 10 °F below the block’s 68 °F calibration temperature. [grin]
The actual size of that gap makes absolutely no difference whatsoever, but fooling around with the gauge blocks gave me an excuse to renew my acquaintance with them and, en passant, massage some oil over their long-neglected bodies:
Gauge block set
I used La Perle Clock Oil, which isn’t Official Gauge Block Oil, but doesn’t go bad on the shelf. Verily, this bottle may be the last of its kind, as it’s no longer available from any of the usual sources; it appears I bought it back in 2000.
The blocks are in good shape, probably because they don’t often see the light. FWIW, I have experimentally determined that my body oil doesn’t etch fingerprints into steel.
The block set, which is similar to a current box o’ blocks from Enco, claims “Workshop Grade”, but the ±0.00050 inch = 1.27 μm tolerance shown in the top row of the labels is much worse than even grade B’s sub-micron tolerance. That newer box claims “Economy” accuracy with the same spec, so I suppose somebody kvetched about mis-using the terms.
Ah, well, they’re far better than any measurements I’ve needed in a while and entirely suitable for verifying my other instruments.
Turns out that I managed to crunch it, exactly as I expected: I’d added a block to the Z-axis stage that poked the home switch just slightly before the anti-backlash nut unscrewed from the top of the leadscrew, but the stage could continue moving another few millimeters.
You can see the gap just above the brass anti-backlash nut:
Sherline Z-axis leadscrew nut – top end
At that point, the nut has barely a single micro-smidgen of thread engaged; that last 0.1340 inch of travel (yeah, I measured it) isn’t usable.
Rather than put a collar around the end of the leadscrew, I opted for a brute-force block atop the Z-axis saddle nut that will slam into the bottom of the stepper motor mount just before the anti-backlash nut disengages:
Sherline Z-axis Overrun Block – rear view
A strip of tapeless sticky (double-sided tape, minus the tape) holds the block in place on the saddle nut. It’s not subject to any particular stress: as long as it doesn’t fall off, it’s all good.
I ran the stage upward until it stalled, then epoxied a new switch (with the old fluorescent tape) in place. This shows the result after backing the stage down a few millimeters:
Sherline Z-axis Overrun Block – side view
The solid model shows off the bevel that provides a bit more room for anti-backlash nut adjustment, not that I ever adjust it that much:
Sherline Z-Axis Overrun Prevention Block – solid model
Obviously, it doesn’t print in that position, but it’s easier to design it in the natural orientation and flip it around for printing.
The OpenSCAD source code:
// Sherline Z-axis Overrun Prevention Block
// Ed Nisley KE4ZNU December 2013
Layout = "Show"; // Show Build
//- Extrusion parameters must match reality!
// Print with 2 shells and 3 solid layers
ThreadThick = 0.25;
ThreadWidth = 0.40;
HoleWindage = 0.2;
Protrusion = 0.1; // make holes end cleanly
//----------------------
// Dimensions
BlockZ = 30.0; // overall height
ZLimit = 17.0; // Z travel limit
TongueX = 9.0; // beside Z axis dovetail
TongueY = 10.0;
StubX = 6.0; // behind Z axis pillar
StubY = 3.0;
BlockX = TongueX + StubX; // overall X
TabY = 3.0; // behind brass bracket
TabX = BlockX - sqrt(2)*TabY;
TabZ = BlockZ - ZLimit;
BlockY = TongueY + StubY + TabY; // overall Y
//----------------------
// Useful routines
module ShowPegGrid(Space = 10.0,Size = 1.0) {
Range = floor(50 / Space);
for (x=[-Range:Range])
for (y=[-Range:Range])
translate([x*Space,y*Space,Size/2])
%cube(Size,center=true);
}
//- The Block
module Block() {
difference() {
cube([BlockX,BlockY,BlockZ]);
translate([-Protrusion,-Protrusion,-Protrusion]) // remove column
cube([(StubX + Protrusion),(TongueY + Protrusion),2*BlockZ]);
translate([-BlockX/2,-Protrusion,-Protrusion]) // form tab
cube([2*BlockX,(TongueY + StubY),(TabZ + Protrusion)]);
translate([0,BlockY,(BlockZ/2 - 0*Protrusion)])
rotate(45)
cube([3*StubY,2*StubY,(BlockZ + 2*Protrusion)],center=true);
translate([0,0,-Protrusion])
cube([sqrt(2)*TabY,2*BlockY,(TabZ + Protrusion)]);
}
}
//-------------------
// Build it...
ShowPegGrid();
if (Layout == "Show")
Block();
if (Layout == "Build")
translate([-BlockZ/2,-BlockY/2,BlockX])
rotate([0,90,0])
Block();
