Testing ideas

Hello All,

After the failure of EasyMaker on KS, I got myself to thinking about a raft of problems in various spaces.  I also had some fantasies.  My biggest fantasy was around ginger-bread shaped men with a “dogbone outline” shape.  Imagine the outline of a gingerbread man, then imagine it made from dogbone shaped extruded HDPE.  Imagine it all being 0.01mm in all tolerances.  Imagine it having a bearing quality surface finish.  Imagine being able to embed electronic structures into it.

Then get mad.  It can’t be made at home.  Heck, I’m not sure it can be made at all.  You might be able to get it made at a ridiculous cost.  It can’t be 3d printed ( HDPE does in fact print, but is very dimensionally unstable, and loses its key surface properties when run through an extruder. )  The dogbone shape can be milled in 2D, but not in 3D.  You could make a dogbone ribbon or sheet and bend it, maybe, with HDPE.  When you exit thermoplastics and into things like fiberglass — hoo!

Get madder.  It can’t even be drawn in OpenSCAD.  The ginerbread shape is easy enough.  So is the dogbone.  Put them together — hoo!

And it’s beautiful.  It would open up a whole new world of design.  Phones with integrated bumpers that look natural.  Abstract bookshelves with integrated lips.  All this from being able to manipulate and manufacture, with ease, a slightly complex shape.( and as shape complexity goes up, this problem explodes.  Adding 1 small hole in 1 extra face of a cube goes from a simple problem to a maybe impossible one. )

When I did my analysis of LumenLab and why they failed as a business — it’s precisely this reason.  They had shape, material, and tolerance requirements beyond what can routinely be achieved  by a “broken machinist”.  Yes, some machinists could do it — but those machinist are rare and expensive.  The “25%” best machinists.  All the rest of us, myself included, couldn’t achieve it.  I am a “broken machinist” — I have design dreams that I can’t make.  That almost no home user can.  Heck, we can’t even afford the machines needed to try.  No MakerBot can make it.  No ShapeOko or Zen mill.  I’m not sure a Roland MDX-450( a 20,000 dollar machine ) could do it.  No Stratasys device.  EOS and laser sintering *might* be able to make the shape out of nylon( the wrong material ) — maybe, but not with .01mm tolerances and definitely not a bearing grade finish on all surfaces.

Get even madder.  There’s no CAM solutions that can make it, either.  The 3d printer CAM packages all think in 2.5D, and couldn’t generate toolpaths needed for the tolerance requirements, even on a perfect machine.  The Milling packages at the home level can’t generate toolpaths to all the interior surfaces.  Professional CAM tools can do this — if you’ve got one of the mythical 8-axis machines and 20-30K for software, you might be able to generate the toolpaths needed to do it.

So, even if you’re a brilliant machinist, unless you have an amazing mill and amazing CAM software, this very simple shape, is impossible.

That’s what I set out to solve.  I don’t want to restrict my designs to what the technology of today can do.  I’ve already begun thinking in design of what the technology of tomorrow can do.  Surprisingly, this is more natural.  When I started out, I started with thinking like the technology of tomorrow.  Then I ran into all the limits of the technology of today.  I modified my thinking to what we can do today — but why?

So, to solve these problem, I’ve been running experiments.  Experiments in motion control.  Experiments in bearing loading.  Experiments in error detection.  None of these experiments have worked so far.  But the failures have been fascinating.  Each time, those failures have brought out new ideas.  Those new ideas are changing my thinking and guiding me down a very different path than I had originally envisioned.  They’re also challenging my business models.  I spend so much time with other nerdy people — I’ve come to realize that few of them get one idea ahead of me, and none of them 2.  There’s something about 3d printing that “freezes” you in the wrong spot.  I think it’s the intersection of, “Hey, I can make that better!”;  “Wow, that’s cool!”; and “OMG, this F*ng sucks!”.  These three thoughts seem to lock us all in place…  It’s hard to learn how to emotionally detach from them and realize, “Wait. Wait. Wait.  If this did work that way, then people could make X.  X could make Y more affordable.”

Hall Effect Switch Tutorial

Hi All,

I recently picked up some cheap hall effect switches. Under the microscope, they say 44E 938. I think these are allegro hall effect sensors, though I’m usure as to the model number. I’ve been trying to figure out which pins are which on this switch, and which was is “left” on these switches, as well as their normal state. They’re normally “on” switches, as opposed to the normally “off” type I thought they were.

It took me a little while to figure out what they were and some characteristics about them, so I thought I’d document my findings for others to use should they also want to use a hall effect switch as an endstop. The switches themselves are cheap — around 20 cents or so each. They work with small magnets, with the field strength determining distance to activation. They need a 10K pull up resistor.



With the numbers facing you on the table, ( So writing side up ), and the leads on the bottom, like in the picture, the leftmost pin is VCC ( +5V ), the middle pin is GND ( 0V ), and the right pin is signal. You need to attach a 10K resistor between the left pin and the right pin( aka a Pullup resistor between VCC and Signal ). The center pin goes to your MCU ground, the left pin goes to your +5V logic, and the right pin goes to your MCU’s input pin. So, if you’re using an Arduino UNO as your MCU, wire it up like this tutorial: I used pin 2 as my input pin.


Magnetic characteristics:

The output of the sensor is 1 when there is no magnet nearby ( HIGH ), and LOW(0) when there is a magnet nearby. The strength of the field depends on how far away the magnet can be. The more powerful the magnet, the farther away the switch will activate. The switch needs a surprisingly strong magnet — I used a very small neodymium magnet, about the size of a medium grain of rice, cubed( 3mm cubed )and got a reading at about 1/2 an index finger’s worth of distance. The switch also seems to be pole sensitive — I don’t know if it reads north or south — but the switch reads only one pole of the magnet.


Here’s the Arduino code to make it work.

int val = 0; // This holds the read value.
void setup()
pinMode(2, INPUT); // Read the input on pin 2
pinMode(13, OUTPUT); // I used pin 13 since it has an LED on the UNO built-in.
Serial.begin(9600); // I also wanted to confirm the value I read.

void loop()

val = digitalRead( 2 ); // Go read the pin.
Serial.println(val); // just to see on the serial monitor what I read.

if ( val == HIGH )
digitalWrite ( 13, HIGH ); // Turn on the LED when the value is high.
digitalWrite ( 13, LOW ); // Turn off the LED when the value is low.



Openrail arrives

My openrail order arrived! I am impressed! They’ve really improved stiffness over the early prototypes I saw! This stuff looks like it can handle real CNC loads! I can’t wait to build something with it.

Hmmm. My new vertical mount Extruder would work great in a Rostock design. Openbeam + openrail = Rostock? Or EasyMaker with dual Z shaft for Easy alignment? So many ideas!


Why Small manufacturing doesn’t work.

Hi All,

I recently went to MFG.com and tried to get a small run of parts made.  I found 2 different suppliers — 1 in the US, and one in China, and had them try and do the work.  The work was CNC milling of some HDPE plastic blocks — hardly rocket science.

Here’s the Skinny — the cheapest US shop came in with a bid of $191 for the job, with $7,000 as the most expensive.  The modal bid was around $500.  In China, the cheapest shop came in at $2.00 for the job, the most expensive at $3,000, and the modal bid was around $50.  I was doing a small run — 1 set of samples, then no more than 100 final parts.

First problem — hardly anyone knew how to read the 2d drawings.  These are old drawings that have been used to produce parts for about a decade.  They were professionally drawn, not something I just whipped together, a longtime ago.  I was trying to build a few copies of an old machine that I happen to like, and found the original blueprints for.  These were standard isometric drawings, with a top, side, front, and back view of each part.  In almost every case, both the American and Chinese shops could not read them.  They kept thinking the views were front views of individual parts — even though each view was properly labeled and numbered!  An example would be “Part 001, Front view”.  Almost no-one provided quotes for the parts.  I had to explain that these were 2d blueprint drawings, not 3d models.

Second problem — everyone wanted 3d models.  This is an old machine, and 3d drawings didn’t exist back then ( they did, but just barely ).  No-one knew how to quote from drawings, and everyone wanted 3d models so they could put it into their CAM packages and get time estimates.  In fact, in order to quote the part, I’m sure many of these people took the drawings and made cheap/dirty 3d models.  Since almost all of the people doing bids on the job were likely sales people, and not engineers, none of these estimates were right.

Third problem — no one delivered.  At all.  The US folks didn’t deliver.  The Chinese didn’t deliver.  Finding manufacturing partners that can do the job and deliver on time is a difficult task.  If you look at Kickstarter, this is perhaps the most common problem why “widget” type projects are late.  The manufacturing partners can’t deliver on their commitments.

This experience has taught me something —  I understand why manufacturing APIs work now.  The humans are doing terrible jobs at basic tasks like reading blueprints — a computer can’t do it any worse.  There are only a few companies I have worked with that I trust to do the job, get it done on time, and in budget — the ones that have hardly any humans in the process at all.  Job intake is by computer, job quoting is by computer( and instant ), job scheduling is by computer, job is done by an automation, and the human is only loading/unloading the blanks.  The few shops that do this are rare and expensive, but worth it.

In the mean time, here’s some tricks that may help you if you try to manufacture things:

1. Have 3d models.  STEP files for CNC, not STL.  Most CNC shops can’t handle STL files. STL files only work in additive shops ( laser sinter/fused filament shops ).

2. Have a high budget.  A small run of a part will cost 10-100x the unit price of a production run.

3. Pick 3 manufacturers, all at the modal price.  The lowest bidder is highly unlikely to deliver.  Even a modal bidder is unlikely — but they’re a bit more serious.


Personally, I think I’ll manufacture all parts myself from here on in, or from the few trusted suppliers that have delivered on time and in budget.  If I ever start a job shop, this will be my number 1 concern — all jobs on time and in budget( and as few humans as possible in the process to ensure this is so. )



Happy New Year!

Hi All,

Happy 2013 to everyone!  I managed to get some time over the christmas break, and have made some progress on new toolheads for the next robot.  I now have a new extruder designed, meant for vertical mounting, a MakerBot Mk7 drive gear, and a standard groove mount for J-HEad/MakerGear hot ends.  I’ve got some fit/finish changes to make, then off to print it and test it.  A vertical extruder has the motor on top of the hot end, instead of to the side, and mounts to a vertical plate instead of a horizontal one.  This particular one is designed to mount to a 2020 metric face of an extrusion.  It is compatible with EasyMaker, ShapeOko, and any other design that uses vertical plates.

Benefits of a vertical mount extruder:

  1. 100% z efficient on certain mill/printer designs.  ( EasyMaker/ShapeOko/most CNC mills ).
  2. Power — full size motors and gearing allows large filament pushing power  ( useful for unheated garage printing in Winter time ).
  3. Maintenance — Mk7 gear can be cleaned “live” — disassembly should never be needed to clean in this design.

It’s 100% z efficient on an EasyMaker/Shapeoko.  That is, it can print all the Z height.  I felt this important, since it allows CNC mills to function as printers without modification.  Most Horizontal extruders take about 40mm ( 1.5 inches ) of Z height.  If you’re using a CNC mill, which may have only 3″ to begin with, you can lose a lot of printable height with a horizontal mount extruder.  With this extruder, you get it all.  This efficiency does have a cost — this extruder requires support to print, and the cheapest service provider I could find would charge $100 to make one.  It would be at least 3 setups to mill — and the geometry is pretty complex, so molding is out.  That means I have to produce each one, since it isn’t affordable to make it any other way.

I hope to have pictures/videos of the new extruder in action sometime this month or so.


Hi All,

It’s been over a week since the last update, and I wanted to put out a little post about what we’ve been up to.  It’s Thanksgiving season here in the states, so for the most part, thanksgiving break has been taking a lot of time.  I’ve also printed some proof of concept parts for the “secret feature”, and found some issues that bother me around how to strengthen and control the motion.  We’ve printed out another iteration of new linear motion system, but haven’t really tested it yet.  Already, I see issues in this iteration that should make it difficult to use, which will have to be addressed.  One interesting thing — U-Groove wheels — I’ve never seen a linear motion system based on them — yet I see plenty of V Groove designs.  I wonder if U Groove linear motion systems have something inherently wrong with them?  I haven’t found any data at all on the strengths/weaknesses of U Groove systems.  This is important, since my hybrid motion system uses a U groove wheel and 2 Igus plain bearings.  I wonder if I could do it all with just the U Groove bearings?

Cold extrusion in the garage is turning into a pain, as temperatures are now dropping below freezing, and the garage drops below 10C, making printing near impossible with the current extruders. I’ve been mulling over a new, “vertical mount” extruder design that merges elements of Wade’s and Ron’s extruders, and has a very narrow footprint.  This would allow stacking of groove mount nozzles close together in a dual extruder setup.  However, I’m loathe to commit to the design, as again, I’m not pursuing EasyMaker, and only it and the ShapeOko mill could mount such an extruder.  In Theory, a vertical mount extruder would also be good for a Rostock or other platform with a large Z footprint,and would be far more powerful/easy to maintain than current extruders by a large margin.  Finals are coming up, so I’ll likely go silent over the next few weeks, as my time/attention will be spent there.  After finals is Christmas/Travel season, so I don’t think we’ll do much there either.  I don’t expect to be back to any real level of design work until January.  When I get the time, I’ll have to decide if I should design an extruder or not.

Anyway, I hope you all enjoy your holidays.  See you in a while!