Sand Table Updates

I built the sand table as a quick project for the Milwaukee MakerFaire, and it showed.  The electronics and cabling were thrown together and everything was held together with zip-ties.  Now that the table is at my house, I wanted to dress it up a little before the final-final upgrades that are being planned.

The electronics enclosure for the sand table was butt-ugly, just as it was when it was on my 3D printer, so I decided to do something about it.  I found a wonderful piece of junk on the hack-rack at the makerspace- an aluminum enclosure, anodized black, that contained two MeanWell 12V 10A power supplies.  I tested them and both worked fine (of course, they're MeanWells).  The box had a line cord jack, fuse holder and power switch ready to go, and there were a lot of ventilation holes in the box.

The first thing I did was test the sand table running from 12V.  There were no problems at all, even running at 500 mm/sec.  Switching the sand table from 24V to 12V operation meant I could get rid of the DC-DC converters that were powering the LED strips, so wiring was simplified and made more reliable.

I pulled one of the power supplies from the box and installed the SmoothieBoard controller in its place, then got to work on wire management.  When I bought the t-slot material used for the frame at the scrap yard, it came with the plastic slot covers, so I routed the wires in the slots and placed the covers over them. The result is much nicer looking, even though it is under the table and mostly hidden from view:

The new electronics enclosure.  No wires hanging all over to temp the cats to chew on them!

New electronics enclosure- all the cables enter the box through a hole near the top.  It's still pretty inconvenient to reach under the table to select a file to run.  The next version will probably have a wifi capable controller.

"A" motor and Y axis limit switch.

"B" motor and X axis limit switch (behind the motor).  Cables are routed in the slots in the table's frame and covered with the plastic strips that are made for the task.

The wires to the LED strips really need to be reworked so they don't exit along the back edge of the table, and I should add a connector so that they can be disconnected easily if the table has to be taken apart to be moved.

Since my original post on the sand table, Jeff Eberl and others have been making huge upgrades to Sandify and it now produces even more interesting patterns than before, and has many added convenience features such as allowing the gcode files to be saved under names you specify.  That makes it a lot easier to combine gcode pattern files into one big file.  It also stores the parameters in each gcode file so you can recreate it if needed.  You can also grab pattern files from several locations on the web, too.

Ms. Kitty enjoying one of the new patterns produced using Sandify.

Final upgrades being planned:

1) Switch to NEMA-17 motors for lower vibration, quieter operation.
2) Install a controller with 256:1 microstepping drivers to reduce noise.
3) Replace the big, ugly limit switches with something smaller and quieter (reed switches, maybe...)
4) Finally, get a nice looking glass-topped table and rebuild the whole mechanism into it so it looks a lot nicer and is more presentable and usable as a piece of furniture.  I'm watching Craig's list for a good deal on either a whole table or a nice glass top.

When I update anything else I'll make another post about it.  When I get NEMA-17 motors installed I'll post some video comparing the noise levels of the NEMA-23 and NEMA-17 motors, and then do it again when I install a higher microstepping controller board.

Comparing Steel Core and Glass Core Belt Stretch in UMMD's Z Axis

A recent debacle led me to replace the belt clamps, and while I was at it, the belts, in UMMD's Z axis.  I took the 10 mm wide steel core belts out and installed 9mm wide, glass core, Gates LL2MR09 belts.

In a previous post I had tested the stretch of the steel core belts under a print load up to 4 kg (plus the mass of the bed and it's support structure, another 3.5 kg).  I found that the steel core belts stretched about 42 um/kg of print load, which translates to worst case stretch in any 0.25 mm layer of 1.2 um- completely inconsequential.

Today I clamped my digital gauge to the printer's frame with the bed about mid way down the Z axis and loaded it up the same way- just stacked a few spools of filament on it.  Photos below show the resulting stretch:

Unloaded and zeroed.

Glass core belt stretch when loaded to about 4 kg.

0.58 mm/4kg = 0 .145 mm/kg which is 3.4 x the stretch I measured with the steel core belts.:

Steel core belt stretch.

The steel core belts stretch came to about 1.2 um maximum in any 0.25 mm thick layer (entire bed covered with a layer of PLA).  That means these glass core belts will stretch a maximum of about 4 um in any 0.25 mm layer.  It seems unlikely to cause any issues in a real print situation, but I'll probably put the steel core belts back on the machine with an improved belt clamp design.

Another Interesting 3D printer Failure- How NOT to Design a Belt Clamp

UMMD has a belt driven Z axis using steel core polyurethane GT2 belts from China.  The belts are held in 3D printed clamps and use short pieces of the same belt to lock the belts into the clamps.

This is a test piece, but the clamps used in UMMD's Z axis use this technique to clamp the belt.

Several months ago I noticed the Z axis belts were flopping around a little, as if they had lost tension.  Initially I thought it was due to the top pulley mounts shifting under the constant pull of the belt tension, and later, after readjusting them, I thought maybe it was caused by broken steel wires inside the belts.

I recently retightened the belts and within a couple days, found they were flopping around again.  I decided it was time to take them off and see what was going on.

Here's what I found:

The polyurethane portion of the belt was sliding and stretching over the steel cables at its core.

The tension on the belt caused the end of the belt to stretch over the steel wires that run through it.  That's not good.  The same thing could happen to any belt held in a clamp the way I have designed the clamps in this machine.  I think it's better to use clamps that fold the belt over on itself to lock it- that way the belt is less likely to slide and stretch on the core.

Here's how the XY stage belts are clamped in UMMD.  The belt folds back over on its own teeth to lock it in place.  This seems to be a better design... it's been working trouble free for >2 years.

I replaced the belts with some glass core, 9mm wide Gates GT2 belts (the same type used in the XY stage), and then redesigned and printed new clamps that fold the ends over on themselves so that this won't become a problem again.

The new design, which is very similar to the design used in SoM's X axis, which has been working for about 5 years:

SoM's X axis belt clamp has been working for >5 years.

I used this printed gauge to check the slot widths for the belt material I had on-hand,

Gauge used to check slot widths needed to clamp different belts.  1.0-3.0 mm in 0.1 mm steps.

then designed the new belt clamp to fit the bed lifting brackets and the belt:

The new belt clamp for UMMD's Z axis.  The space around the posts is just wide enough for the belt to fit, and the entry and exit slots are just wide enough for the folded belt to fit with the teeth interlocked.

Note: the belt clamp file linked above is not the exact dimensions I used in UMMD- I was unable to locate the original file so I recreated an approximation of it that you can easily customize to fit your printer and belt.

A New Self-Locking Belt Clamp Design

Here's a quick, simple one, in case you're tired of my long-winded posts.

I am rebuilding the Y axis in SoM for belt drive (long-winded blog post will appear soon) after finally giving up on getting rid of the resonance in the ball screw drive.  To that end I needed to make a belt clamp that will mount on the carriage plate (SoM is a bed-flinger).

I decided to try a printed belt clamp because they seem to work well in SoM's X axis and in all three of UMMD's axes.

The drive pulley on the motor and the end pulley, on an aluminum mount I pulled from a piece of surplus junk, are at different heights (about 1 mm).  The belt has to be parallel to the guide rail on the side of the loop that attaches to the carriage plate.  I could either figure out a way to put the pulleys at the same height, or I could adjust the belt clamp design to hold the belt at one level above the carriage plate on the motor side and at a slightly higher level on the end-pulley side.  That was the easier way to go.

I played with a couple single piece designs, but in the end I decided to make two different but almost identical parts and screw them down on the carriage plate.

Here's what I came up with:

The printable belt clamps, one holds the belt 1mm higher than the other.  The belt enters the slot on the other side, folds over the steel pin, then goes back into the slot.  The slot is just high enough to fit the belt with the teeth interlocked, so the belt can't come loose.

A sectional view of one of the clamps.  Each clamp is held down with a single flat head screw.

The belt folds over in the slot and its teeth interlock.  I used a 3mm diameter steel pin (left over from the sand table project) to stop the belt from pulling back through the slot, though I suspect it would be OK to make the whole thing printed plastic.  Even if the pin broke off the plastic, it won't fit through the slot so it would still do its job.

Y axis belt clamps in place.  There's plenty of clearance for the belt above the clamps.

I pulled the upper part of the belt away to make the clamps more visible.

This design depends on the strength of the plastic on either side of the belt - if the slot is too tight the belt might produce enough force to split the plastic layers.  We'll see how long it lasts.  SoM's printed X axis belt clamp has been working fine for about 5 years...

Fusion360 design file is here.

Interesting 3D Printer Failures

I recently experienced a couple failures and almost failures that might be interesting to people who build 3D printers.

The first one was discovered when I started to rebuild the Y axis in Son of MegaMax (blog post will be made when the work is done), my bed flinger printer that lives at the Milwaukee Makerspace.  I took the bed plate off because I was going to make a new bed plate and convert the ball screw drive to belt drive.

Here's the bed plate about a year after it was put into use on the machine.  It has a self-adhesive 450W kapton heater.  I don't know what type of adhesive it had on it.

And this is what it looks like today, after 5 or 6 years of temperature cycling:

Notice the brown spots- there are air bubbles that formed between the heater and the bed plate under them.  Air is a great thermal insulator, so the aluminum bed can't take the heat away from the heater where there is a bubble and the result is hot spots.  It's probably safe to assume that almost any self-adhesive material is going to eventually let go this way, and the heater will eventually burn itself up.

The Keenovo silicone/fiberglass heaters, and probably a lot of others, come with 3M 468MP adhesive transfer sheet on them.  UMMD has had such a heater on it for a little over 2 years.  I recently noticed that the heater was starting to peel off the underside of the print bed.  Right now I have a piece of silicone foam wedged under it to keep the heater pressed against the bed plate, but sooner or later (probably sooner) the rest of it is going to start peeling off.  This is why it's a good idea to mount TCOs on the heater, as I should have done, instead of on the bed plate, as I did.  If the heater comes off the plate, having the TCO on the plate won't keep the heater from burning itself up, and maybe other things too.

Silicone foam used to keep the heater pressed against the bed when the adhesive started letting go, a little over 2 years after it was installed.

Finally, again on UMMD, the PEI print surface was mounted on the bed using 2" wide tape labeled 3M 200MP.  The standard stuff people use these days is 3M 468MP adhesive transfer tape, and if you look closely at the label on it, it says "200MP Adhesive" on it, so the two seem to be the same thing.

468MP adhesive transfer sheet commonly used to hold PEI and heaters on aluminum bed plates.

The 468MP transfer sheet that uses 200MP adhesive was letting go of the heater on the underside of UMMD's bed plate.  Then another odd thing happened just a couple days ago.  I started a largish ABS print on UMMD and went away and after an hour or so, the print failed.  It looked strange.  Like the edges of the print lifted, but closer inspection revealed that the PEI lifted up off the bed plate- the ABS was still stuck to the PEI.

Here's the print that failed.  Look at the edges of the bed plate and PEI...

Here it is from a lower angle.  The tape that was used to stick the PEI to the plate remains stuck to the plate, and is no longer sticky on the top side.  

The central area of the PEI was still stuck to the bed plate.  I removed the print from the PEI and found that the PEI laid back down flat on the bed surface and you'd never know anything was wrong just by looking at it.

A couple years ago, when I noticed the PEI starting to lift at the edges of the bed on SoM, I removed it and retaped it to the bed plate.  One of several things that recently prompted me to start a rework of SoM's Y axis was that the PEI was starting to lift at the edges for the second time in about 5 years.

I think there are a few things to learn from this:

1. 468MP adhesive transfer tape using 200MP adhesive has a limited life span when it is heat cycled regularly.  I've been getting about 2 years of use, but I print a lot of ABS.  If you print PLA or other lower temperature materials, you might get more than 2 years from it.  
2. It's a good idea to inspect the bed frequently and pay special attention to the heater to avoid disasters.  When you're inspecting it, try lifting the edges of the PEI and the heater away from the plate to verify that the adhesive is still working.
3. When you're assembling parts with 468MP/200MP, follow directions on 3M's web site to get maximum bond strength and lifetime.

An Update to the Sand Table Mechanism

After the MakerFaire I looked at the sand table mechanism and found one problem.  The magnet carriage was tilted badly because the spring under the magnet was putting downward force on one side of the carriage.  The magnet carriage, made entirely of printed ABS, was wearing out quickly from sliding on the 16 mm square, powder coated X axis tube.  Apparently, when ABS slides on powder coat, ABS loses.

I decided I needed to do something about it or the whole thing would not be very reliable.  The UHMW bearings on the Y axis are holding up well, so I decided that I need to figure out a way to put UHMW bearings on the X axis, too.

After much thought and many CAD models, I came up with what seems to be a pretty good idea.  I used a piece of 1" square aluminum tube that fits over the 16 mm square X axis tube.  I added UHMW bearings to fill up the space between the two.  I milled some slots in the tube to allow easy attachment of the belts.  The belts pull on the aluminum, not printed plastic, so there are no worries about plastic breaking due to the stress.

The new magnet carriage design.  Belts will attach at the vertical slots in the aluminum body of the carriage, and the magnet holder will attach using zip-ties.

It turns out the belt attachments are almost exactly where they need to be to keep the belts parallel to the X axis guide tube.

Making the UHMW bearings was the hardest part of all this.  UHMW is a little tricky to machine because it's very soft.  When I first tried to fit it all together, I couldn't get the X axis guide tube to fit inside the magnet carriage tube with the bearings in place.  A few minutes with some very coarse sand paper fixed that.

Here's the first test video with the new magnet carriage in place:

New magnet carriage design for the sand table. from Mark Rehorst on Vimeo.

I intend to sand-blast the powder coating off the X axis guide tube which should reduce the friction.

Here's the new carriage with the magnet and spring in place.  When the sand box is on top of the mechanism, only a few mm of the magnet protrudes above the blue magnet box.  The white bits at the center are the UHMW bearings.

UHMW bearings ride on all four surfaces of the X axis tube.

The spring fits in a circular hole in the magnet box, and the magnet is free to move up and down to follow the bottom surface of the sandbox.

Future updates may include switching to a Duet controller board and NEMA-17 motors to try to quiet things down a bit.  I think it will be a lot quieter if I run the motors at a lower speed, too, but then it won't be nearly as fun to watch it work.

3D Model Extracted From CT Scan

More stuff moved over from my old web site.... Just in time for Halloween!


In 2007, when I was in dental school, I had a cone beam CT scan done.  I was going to have braces put on my teeth and I had a dental implant at #3 and I think they wanted to check the implant status before attaching brackets to my teeth.

After the scan was done, the radiology tech showed me the visualization capabilities of the software.  It was very impressive.  I had been reading about 3D printers for years and decided that I wanted a copy of the CT scan data because I planned to be able to print it some day, so the tech burned the data to a CD (remember those?) for me.

Six years later, and I had built MegaMax, my first 3D printer.  I located the CD with the CT scan data and went to work on finding free software that would allow me to extract data and produce a printable model.  It was a real project at the time- there were only a few free packages that could do the job.  I think I spent a total of about 40 hours hunting for software, then learning how to use it, and finally extracting the model and cleaning it up for printing.

I used software called DeVide, which no longer works in Windows, to do the initial job of isolating the tissues for a 3D printable model.  I also made multiple still images by sweeping the tissue density, then assembled them into an animated .gif file:

These days there are other software packages that can do the image processing, and it doesn't matter if the data is from a CT scan (x-ray based) or an MRI (magnetic, images water in tissues).  If you have a Mac, look up Osirix.  For windows there's InVesalius.  3DSlicer works in Windows, Mac, and Linux.  

YouTube has how-tos for extracting models from DiCom data files.  Here's one for InVesalius:

Human skulls have cartilage, lots of soft tissue, bones, teeth, and air spaces.  When you extract a model from a CT scan using tissue density to select teeth, you lose a lot of low density bone (eye sockets, sinuses, etc.).  When you reduce the tissue density to try to get the low density bone into the model, you also get the cartilage.  That sort of thing leads to a lot of clean-up to make the model printable.  If I remember right, I ended up using 4 different software packages to finally produce a printable model from the CT scan data.

Pencil and toothpick cups made from CT scan.

There's an STL file of the cups here.

I also made keyfobs from the same file.  You can find the models here.

One of the difficulties with trying to do this sort of thing is the difficulty of obtaining detailed CT or MRI scan data.  That stuff is all protected by patient confidentiality laws- even if you strip the files of patient name, etc., the file data contains a likeness of the patient's face so it can be linked to that person.  Unless you use your own data, you'll have a hard time finding data that you can use to extract a human skull.  There are several data bases on line that contain data of human and animal scan data, but they are often limited for academic use only.  

I have put my own CT scan data in the form of a Dicom multifile on line.  You can download it here.

If you ever have a scan done, you might want to make arrangements in advance to get a copy of the complete data.  If you wait until months later and request a copy, there's a good chance that only part of the data will be available.  Even though it is your data, you may have to pay a small copying fee to obtain it.

Fractal Hand

In early 2005 I was visiting family in Milwaukee and saw a handbill taped to a light pole on North Ave advertising a psychedelic band that was going to be playing at one of the local bars.  The handbill was a bad photocopy with a hand drawn in ink that had hands at each of its fingertips.  I really liked that image.

Months later, I was trying to learn to use Adobe Photoshop and remembered that handbill and decided to try to make a photographic version of the image.  I put my hand on a scanner, saved the image, then went to work on it.  I don't remember the whole process, but it involved making the background black, scaling and rotating multiple copies of  the image on multiple layers, and using the rubberstamp tool to obliterate the seams where the wrists joined the fingers, and finally flattening the whole thing to one layer.  IRIC, I spent about 2 hours working on it, mostly because I didn't know what I was doing and had to look up each step as I performed it.

This is the result:

Fractal Hand, an L-system fractal

Blogger shrank the image.  The original, is much bigger and more detailed.

Over the years I've had several requests to use the image in books, etc., and I've had a few tee-shirts made that look really great.  I wear some of them myself and have given a very few as gifts to special people.  I get many complements on the shirt whenever I wear one of them.

Someday I'll create a 3D printable version of this thing.  It will probably have to be printed on an SLA or SLS machine to capture the tiniest fingers.

Insect Photos

When I lived in Dallas and later, Missouri, there were a LOT of interesting bugs.  I outfitted my 3 MP camera with a closeup lens and flash bracket and took a few pictures of some of them.  The pictures below were originally posted to my web site and now moved over here...

Mark's Bug Pix

I've always liked optical toys, especially cameras and microscopes, and I've always liked creepy-crawlies (except chiggers, which I hate).  Shortly after I got my first decent digital camera I started shooting pictures of insects and bugs.  I lived in Dallas when I got my Fuji Finepix S602z 3 megapixel camera, and I quickly bought a color corrected 5X closeup lens.  I modified an external flash bracket and use a relatively low powered auto exposure flash unit for close-up work.

I set the camera for minimal sharpening, and usually use manual focus, then move the camera back and forth to focus the image.  Sometimes I get what I want after only four or five photos, other times I shoot twenty and can't get the image I want.  Digital photography is perfect for macro work like this- the film is really cheap!  Of course, the bugs don't always stand still either.  Whenever I can I try to capture them flying or on the run or jump.  It isn't easy but when it works it really works well.
My lens is often just a few inches from the bug.  By getting in close I maximize the bugs size in the view finder and get as many pixels as possible into the image.  In the photos where the background is bright, I used natural light, or set the flash to auto mode.  Using the flash at higher power setting makes the background, which is sometimes distracting, go black or nearly black.  It looks unnatural but focuses attention on the subject.  Yes, those photos are taken during the day, often in bright sunlight.  The flash is so close to the subject so a very short shutter speed can be used even with the lens stopped down to f11 for maximum depth of field.  The daylight illuminated portion of the scene is grossly underexposed making it go all but black, and the portion lit by the flash is perfectly exposed.

3 megapixels is getting to be a little low-res these days.  I hope to upgrade to a higher resolution camera sometime in the not too distant future.

Here are some of my favorite photos.

A pair of metallic green bees.

A leaf butterfly.

This is called a mantis fly.

A bee fly on a thistle.

A Synchlora moth.  This one is about 1/2" long.

I managed to catch this Syrphid fly hovering in front of the camera lens.  Maybe he was looking at his reflection. 

Syrphid fly at rest.

This is a real moth I found on my porch in Missouri.  It was about 1/2" long.

A bright green grasshopper.

Here's a small gray butterfly (Gray Hairstreak?) on a milk week plant near Dallas.  Can you find the aphid in this picture?

A classic "bug".

A small (about 3 mm  long) metallic green fly.

Honey bee surveying a thistle in Dallas.

Another honey bee, landing on a flower.

Another honey bee looking for a meal.  Honey bees are great subjects because they hover a lot and don't seem to mind if you get close to them.

Another honey bee.

Last one, I promise.

This is an Acorn Weevil, (Curculio sp.) about 1/4" long.  More info here.

Another shot of the acorn weevil.  Yes, he's pretty small.

A cicada emerging from its shell.

The cicada, about 15 minutes later.

This dragon fly lifted its head and smiled for the camera.

An Imperial moth.

One of my favorites from Missouri- a male Luna moth.

Another Luna moth.  

A Polyphemus moth

A thistle that's releasing seeds.

A Tiger Swallowtail caterpillar.

This was a surprise!  The Tiger Swallowtail caterpillar has some foul smelling scent glands that pop out of its body just behind the head when it's disturbed.