Ed Nisley's Blog: Shop notes, electronics, firmware, machinery, 3D printing, laser cuttery, and curiosities. Contents: 100% human thinking, 0% AI slop.
A stray sunflower seed decided that the spot just outside the garden gate was perfect and gave Mary’s garden an attractive marker. It will eventually have a dozen blossoms, each one serving as a buffet for the local bumblebees:
Sunflower with bumblebee
Each bee makes several complete circuits of the florets, draining the nectar and collecting pollen as she goes:
Sunflower with bumblebee – detail
Mary tucks the open gate inside the garden to avoid disturbing the pollinators, as wasps tend to have short fuses and multiple-strike stingers:
Sunflower with wasp
The bumblebee traveled clockwise and the wasp went counterclockwise, but I don’t know if that’s the general rule. I certainly won’t dispute their choices!
In a few weeks, long after the petals fall away, a myriad small birds will harvest the dried seeds…
Walkway Over the Hudson – Sturgeon Moonwalk – 2015-08-28
The view from the middle of the Walkway northward along the Hudson makes a nice panorama:
Walkway Over the Hudson – Sturgeon Moonwalk – North Panorama – 2015-08-28
The black rectangular lump on the left is a steel I-beam that I didn’t notice until too late.
Taken with the Canon SX-230HS, hand-braced on the rail, and stitched with The GIMP’s panorama tool. It’s surprisingly easy to stitch a decent panorama from five low-detail images with plenty of overlap…
Sticking an extruded plastic thread to the platform of a 3D printer requires absolutely accurate alignment and spacing, maybe ±0.05 mm across the entire platform. I’ll leave the topic of automatic alignment measurement & compensation for another day; here’s how to measure the actual platform alignment.
If the wall thickness (in the XY plane) doesn’t come out exactly right, fix that first by verifying the filament diameter setting, then adjusting the Extrusion Multiplier. If the extruder doesn’t produce the same wall thickness that the slicer calls for, you won’t get good results from anything else. In this case, they all have 0.40 mm thick walls, with 0.25 mm layers.
The first layer of all five boxes should be identical:
M2 V4 nozzle – thinwall box first layer
If the platform isn’t absolutely flat and properly aligned, those five first layers won’t be the same thickness. It’s surprisingly easy to spot differences under 0.05 mm, so pay attention to what’s happening.
When they’re done, pop them off and measure their actual height. I measure across adjacent sides, leaving the corners / stray hairs / snot out of the measurement, and figure the eyeballometric average of the values, which usually differ by less than ±0.03 mm. Write the height on the side to eliminate future angst:
Thinwall Box – platform height
The boxes should be 5.00 mm tall, so the leftmost box is short by -0.02 mm and the rightmost by -0.15 mm. The five boxes were 4.98, 4.95, 4.93, 4.92, and 4.85, with a mean of 4.93 mm. The variation across the 200×250 mm platform is 0.13 mm, which is pretty good.
Comparing the bottom layers of those boxes, the first layer of the 4.85 mm box is definitely squashed:
Thinwall Hollow Boxes – first layer bottom view – 4.98 4.85 mm
Once you know what to look for, it’s also obvious from the side (4.98 on the left, 4.85 on the right, bottom layers facing each other):
Thinwall boxes – 4.98 4.85 – bottom layers
Keep in mind that’s a difference of 0.13 mm = 130 µm, just over the ±0.05 mm I usually bandy about. The nominal layers are 0.25 mm = 250 µm.
A bit more magnification shows the nicely rounded first layer of the 4.98 mm box (rightmost thread of leftmost set):
Thinwall Box – 4.98 mm
And the squashed first layer of the 4.85 mm box (likewise):
Thinwall Box – 4.85 mm
Because the Z axis moves upward (on the M2, the platform moves downward) by exactly the layer thickness at the end of each layer, the first layer must absorb the entire difference between the desired thickness and the actual nozzle-to-platform distance. That squashed first layer is 0.10 mm thick, a bit less than half of the nominal 0.25 mm. The second layer of each box looks just like all the higher layers.
You can set the Z-axis offset for a very slight squish, with the maximum nozzle-to-platform distance at 0.25 mm and the minimum set by the other end of the total misalignment, because the plastic won’t adhere to the platform when the nozzle-to-platform distance exceeds the nozzle diameter.
Think of it this way: the plastic emerges from a 0.35 mm nozzle as a (slightly larger than) 0.35 mm cylinder that must squash to become a 0.25 mm high x 0.40 mm wide thread. Given the measurements above, setting the Z-axis home position to make the average box height equal to 5.00 mm would make the tallest box come out at 5.07 mm, which requires a 0.32 mm actual first layer that probably wouldn’t stick well at all.
When your printer can consistently produce five thinwall boxes with the proper wall thickness and height, then you can move on to more complex objects.
Well, it’s actually not “sensing”, but the demo code now sizes the graph to the paper size reported by the plotter, so you can plot on cheap and readily available A-size paper. The Vulcan Nerve Pinch that switches paper size on the fly is Enter+Size; I leave the DIP switches set for B-size sheets, because they’re more impressive and take longer to plot.
A collection of A-size plots:
Superformula Plots – A-size paper
The perspective foreshortening makes the sheets look square and the plots seem circular; they’re not.
The plots lie in rough time sequence from lower left to upper right, showing that I tweaked the n1 parameter to avoid the sort of tiny middle that gnawed a hole right through the center-bottom sheet. I also removed higher m parameter values, because more than 50-ish points doesn’t work well on smaller sheets.
I figured out how to use the Python ternary “operator” and tweaked the print formatting, but basically it’s a hack job through & through.
The Python source code, including the hacked Chiplotle routines that produce the SuperFormula:
from chiplotle import *
from math import *
from datetime import *
import random
def superformula_polar(a, b, m, n1, n2, n3, phi):
''' Computes the position of the point on a
superformula curve.
Superformula has first been proposed by Johan Gielis
and is a generalization of superellipse.
see: http://en.wikipedia.org/wiki/Superformula
Tweaked to return polar coordinates
'''
t1 = cos(m * phi / 4.0) / a
t1 = abs(t1)
t1 = pow(t1, n2)
t2 = sin(m * phi / 4.0) / b
t2 = abs(t2)
t2 = pow(t2, n3)
t3 = -1 / float(n1)
r = pow(t1 + t2, t3)
if abs(r) == 0:
return (0,0)
else:
# return (r * cos(phi), r * sin(phi))
return (r,phi)
def supershape(width, height, m, n1, n2, n3,
point_count=10*1000, percentage=1.0, a=1.0, b=1.0, travel=None):
'''Supershape, generated using the superformula first proposed
by Johan Gielis.
- `points_count` is the total number of points to compute.
- `travel` is the length of the outline drawn in radians.
3.1416 * 2 is a complete cycle.
'''
travel = travel or (10*2*pi)
## compute points...
phis = [i * travel / point_count
for i in range(1 + int(point_count * percentage))]
points = [superformula_polar(a, b, m, n1, n2, n3, x) for x in phis]
## scale and transpose...
path = [ ]
for r, a in points:
x = width * r * cos(a)
y = height * r * sin(a)
path.append(Coordinate(x, y))
return Path(path)
## RUN DEMO CODE
if __name__ == '__main__':
plt=instantiate_plotters()[0]
# plt.write('IN;')
if plt.margins.soft.width < 11000: # A=10365 B=16640
maxplotx = (plt.margins.soft.width / 2) - 100
maxploty = (plt.margins.soft.height / 2) - 150
legendx = maxplotx - 2600
legendy = -(maxploty - 650)
tscale = 0.45
numpens = 4
m_list = [n/10.0 for n in [11, 13, 17, 19, 23]]; # prime/10 = number of spikes
n1_list = [n/100.0 for n in range(55,75,1) + range(80,120,5) + range(120,200,10)] # ring-ness 0.1 to 2.0, higher is larger
else:
maxplotx = plt.margins.soft.width / 2
maxploty = plt.margins.soft.height / 2
legendx = maxplotx - 3000
legendy = -(maxploty - 700)
tscale = 0.45
numpens = 6
m_list = [n/10.0 for n in [11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59]]; # prime/10 = number of spikes
n1_list = [n/100.0 for n in range(15,75,1) + range(80,120,5) + range(120,200,10)] # ring-ness 0.1 to 2.0, higher is larger
print "Max: ({},{})".format(maxplotx,maxploty)
n2_list = [n/100.0 for n in range(10,50,1) + range(55,100,5) + range(110,200,10)] # spike-ness 0.1 to 2.0, lower is spiky
plt.write(chr(27) + '.H200:') # set hardware handshake block size
plt.set_origin_center()
plt.write(hpgl.SI(tscale*0.285,tscale*0.375)) # scale based on B size characters
plt.write(hpgl.VS(10)) # slow speed for those abrupt spikes
plt.select_pen(1) # standard loadout has pen 1 = black
plt.write(hpgl.PA([(legendx,legendy)]))
plt.write(hpgl.LB("Started " + str(datetime.today())))
m = random.choice(m_list)
pen = 1
for n1, n2 in zip(random.sample(n1_list,numpens),random.sample(n2_list,numpens)):
n3 = n2
print "{0} - m: {1:.1f}, n1: {2:.2f}, n2=n3: {3:.2f}".format(pen,m,n1,n2)
plt.select_pen(pen)
plt.write(hpgl.PA([(legendx, legendy - 100*pen)]))
plt.write(hpgl.LB("Pen {0}: m={1:.1f} n1={2:.2f} n2=n3={3:.2f}".format(pen,m,n1,n2)))
e = supershape(maxplotx, maxploty, m, n1, n2, n3)
plt.write(e)
pen = pen + 1 if (pen % numpens) else 1
plt.select_pen(1)
plt.write(hpgl.PA([(legendx, legendy - 100*(numpens + 1))]))
plt.write(hpgl.LB("Ended " + str(datetime.today())))
plt.select_pen(0)
Thinwall open boxes – side detail – 4.98 4.85 measured
Alas, the shutter failed after that image, leaving me with pictures untaken and naught to take them with.
The least-awful alternative seems to be gimmicking up an adapter for a small USB camera from the usual eBay source:
Fashion USB video – case vs camera
The camera’s 640×480 VGA resolution is marginally Good Enough for the purpose, as I can zoom the microscope to completely fill all those pixels. The optics aren’t up to the standard set by the microscope, but we can cope with that for a while.
A bit of doodling & OpenSCAD tinkering produced a suitable adapter:
USB Camera Microscope Mount – solid model
To which Slic3r applied the usual finishing touches:
USB Camera Microscope Mount – Slic3r preview
A bit of silicone tape holds the sloppy focusing thread in place:
USB Camera Microscope Mount – cap with camera
Those are 2-56 screws that will hold the cap onto the tube. I drilled out the clearance holes in the cap and tapped the holes in the eyepiece adapter by hand, grabbing the bits with a pin vise.
Focus the lens at infinity, which in this case meant an old DDJ cover poster on the far wall of the Basement Laboratory, and then it’ll be just as happy with the image coming out of the eyepiece as a human eyeball would be.
I put a few snippets of black electrical tape atop the PCB locating tabs before screwing the tube in place. The tube ID is 1 mm smaller than the PCB OD, in order to hold the PCB perpendicular to the optical axis and clamp it firmly in place. Come to find out that the optical axis of the lens isn’t perfectly perpendicular to the PCB, but it’s close enough for my simple needs.
And then it fits just like you’d expect:
USB Camera Microscope Mount – on eyepiece
Actually, that’s the second version. The distance from the camera lens (equivalently: the PCB below the optical block, which I used as the datum plane) to the eyepiece is a critical dimension that determines whether the image fills the entrance pupil. I guesstimated the first version by hand-holding the camera and measuring with a caliper, tried it out, then iteratively whacked 2 mm off the tube until the image lit up properly:
USB Camera Microscope Mount – adjusting tube length
Minus 4 mm made it slightly too short, but then I could measure the correct position, tweak that dimension in the code, and get another adapter, just like the first one (plus a few other minor changes), except that it worked:
USB Camera Microscope Mount – first light
That’s a screen capture from VLC, which plays from /dev/video0 perfectly. Some manual exposure & color balance adjustment may be in order, but it’s pretty good for First Light.
It turns out that removing the eyepiece and holding the bare sensor over the opening also works fine. The real image from the objective fills much more area than the camera’s tiny sensor: the video image covers about one digit in that picture, but gimmicking up a bare-sensor adapter might be useful.
That can also come from a sensor failure, but it takes perfectly good movies. That’s the differential diagnosis for shutter failure, because movies don’t use the shutter.
The shutter still functions, in that peering into the lens shows the shutter closing as it takes a picture, so I suspect it’s gotten a bit sticky and slow over the years. None of the various shutter-priority speeds have any effect, which means that the shutter isn’t responding properly.
A quick read of the service manual shows the Field Replaceable Unit for this situation is the entire lens assembly. Back in the day, a new lens assembly came with its own calibration constants on a floppy disk that you’d install with Casio’s service program (the latest version ran with Windows 98!) using a special USB communication mode triggered by a Vulcan Nerve Pinch on the camera. At this late date, none of that stuff remains available.
While I could take the camera apart and crack the lens capsule open, I doubt that would make it better and, in this case, ending up with a crappy camera doesn’t count for much. Extracting the lens assembly requires dismantling the entire thing, which, frankly, doesn’t seem worth the effort…
That image is number 7915: so it’s taken a bit over two images per day for the last nine years. I can’t swear the counter has never been reset, but that seems about right.
A gallery of SuperFormula plots, resized / contrast stretched / ruthlessly compressed (clicky for more dots):
SuperFormula Plot – 01
SuperFormula Plot – 02
SuperFormula Plot – 03
SuperFormula Plot – 04
SuperFormula Plot – 05
SuperFormula Plot – 06
SuperFormula Plot – 07
SuperFormula Plot – 08
SuperFormula Plot – 09
SuperFormula Plot – 10
SuperFormula Plot – 11
SuperFormula Plot – 12
SuperFormula Plot – 13
SuperFormula Plot – 14
SuperFormula Plot – 15
The gray one at the middle-bottom suffered from that specular reflection; the automagic contrast stretch couldn’t boost the paper with those burned pixels in the way.
Those sheets all have similar plots on the back, some plots used refilled pens that occasionally bled through the paper, others have obviouslybad / dry pens, and you’ll spot abrupt color changes where I swapped out a defunct pen on the fly, but they should give you an idea of the variations.
The more recent plots have a legend in the right bottom corner with coefficients and timestamps:
SuperFormula Plot – legend detail
Limiting the pen speed to 10 cm/s (down from the default 38.1 cm/s = 15.00 inch/s) affects only the outermost segments of the spikes; down near the dense center, the 9600 b/s serial data rate limits the plotting speed. Plotting slowly helps old pens with low flow rates draw reasonably dense lines.
Each plot takes an hour, which should suffice for most dog-and-pony events.
I fill a trio of Python lists with useful coefficient values, then choose random elements for each plot: a single value of m determines the number of points for all six traces, then six pairs of values set n1 and n2=n3. The lists are heavily weighted to produce spiky traces, rather than smooth ovals, so the “random” list selections aren’t uniformly distributed across the full numeric range of the values.
Because the coefficient lists contain fixed values, the program can produce only a finite number of different plots, but I’m not expecting to see any duplicates. You can work out the possibilities by yourself.
The modified Chiplotle demo code bears little resemblance to the original:
from chiplotle import *
from math import *
from datetime import *
import random
def superformula_polar(a, b, m, n1, n2, n3, phi):
''' Computes the position of the point on a
superformula curve.
Superformula has first been proposed by Johan Gielis
and is a generalization of superellipse.
see: http://en.wikipedia.org/wiki/Superformula
Tweaked to return polar coordinates
'''
t1 = cos(m * phi / 4.0) / a
t1 = abs(t1)
t1 = pow(t1, n2)
t2 = sin(m * phi / 4.0) / b
t2 = abs(t2)
t2 = pow(t2, n3)
t3 = -1 / float(n1)
r = pow(t1 + t2, t3)
if abs(r) == 0:
return (0,0)
else:
# return (r * cos(phi), r * sin(phi))
return (r,phi)
def supershape(width, height, m, n1, n2, n3,
point_count=10*1000, percentage=1.0, a=1.0, b=1.0, travel=None):
'''Supershape, generated using the superformula first proposed
by Johan Gielis.
- `points_count` is the total number of points to compute.
- `travel` is the length of the outline drawn in radians.
3.1416 * 2 is a complete cycle.
'''
travel = travel or (10*2*pi)
## compute points...
phis = [i * travel / point_count
for i in range(1 + int(point_count * percentage))]
points = [superformula_polar(a, b, m, n1, n2, n3, x) for x in phis]
## scale and transpose...
path = [ ]
for r, a in points:
x = width * r * cos(a)
y = height * r * sin(a)
path.append(Coordinate(x, y))
return Path(path)
## RUN DEMO CODE
if __name__ == '__main__':
paperx = 8000
papery = 5000
tscale = 0.45
numpens = 6
m_list = [n/10.0 for n in [11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59]]; # prime/10 = number of spikes
n1_list = [n/100.0 for n in range(15,75,1) + range(80,120,5) + range(120,200,10)] # ring-ness 0.1 to 2.0, higher is larger diameter
n2_list = [n/100.0 for n in range(10,50,1) + range(55,100,5) + range(110,200,10)] # spike-ness 0.1 to 2.0, lower means spiky points
paramlist = [[n1,n2] for n1 in random.sample(n1_list,numpens) for n2 in random.sample(n2_list,numpens)]
if not False:
plt=instantiate_plotters()[0]
plt.write('IN;')
# plt.write(chr(27) + '.H200:') # set hardware handshake block size
plt.set_origin_center()
plt.write(hpgl.SI(tscale*0.285,tscale*0.375)) # scale based on B size characters
plt.write(hpgl.VS(10)) # slow speed for those abrupt spikes
pen = 1
plt.select_pen(pen)
plt.write(hpgl.PA([(paperx - 3000,-(papery - 600))]))
plt.write(hpgl.LB("Started " + str(datetime.today())))
m = random.choice(m_list)
for n1, n2 in zip(random.sample(n1_list,numpens),random.sample(n2_list,numpens)):
n3 = n2
print "m: ", m, " n1: ", n1, " n2=n3: ", n2
plt.write(hpgl.PA([(paperx - 3000,-(papery - 500 + 100*(pen - 1)))]))
plt.select_pen(pen)
plt.write(hpgl.LB("Pen " + str(pen) + ": m=" + str(m) + " n1=" + str(n1) + " n2=n3=" + str(n2)))
e = supershape(paperx, papery, m, n1, n2, n3)
plt.write(e)
if pen < numpens:
pen += 1
else:
pen = 1
pen = 1
plt.select_pen(pen)
plt.write(hpgl.PA([(paperx - 3000,-(papery - 500 + 100*numpens))]))
plt.write(hpgl.LB("Ended " + str(datetime.today())))
plt.select_pen(0)
else:
e = supershape(paperx, papery, 1.9, 0.8, 3, 3)
io.view(e)
The Smell of Molten Projects in the Morning 