Saturday, July 29, 2017

Ezevor and other robot arms

I stumbled across the Evezor Kickstarter page recently and was fascinated to see the range of things this SCARA arm can do. It can 3d print, carve or mill like a CNC machine, laser-engrave, pick and place electronic parts, paint or draw, and more. It also appears on Thingiverse.

There are Youtube videos of many of these operations. The resolution is good. The effective build volume is enormous. The speed is decent. As I watched more videos, I got a little bit obsessed with the whole thing and decided to blog about it.

The Kickstarter campaign was regrettably unsuccessful. You can still pre-order one for $2650, but I can't justify an expense like that currently, and I suspect the lead time would be substantial.

In the midst of my obsession, I started poking around the interwebs to see what other similar projects might exist, and might be more affordable as well as open source, and came across quite a few.

Several arms work like a Luxo lamp, using parallelograms along the arm segments to keep the end effector in a consistently vertical orientation.
These are all appealing for cost reasons, but I worry about the resolution. They'd be fine for picking up small objects, and maybe some drawing, but I'd be surprised if they could do 3d printing or CNC milling with a Dremel tool.

There are a couple other SCARA arms. These are more commercial and expensive, and don't appear to be open source. But for that price they probably have the resolution. I found a couple more that were neither Luxo-style nor SCARA. So if you get obsessed with the original idea of a robot arm that can do a wide range of precision tasks with a potentially heavy payload (like a router for doing CNC milling), what should you do? My own inclination would be to print parts for an Ezevor. It would be interesting to see whether either the SparkFun arm or the Tindie arm could be a serviceable 3d printer. If not, I'd probably hunt around for a larger 3d printer than the one I have at home.

Saturday, January 09, 2016

Graceful shutdown for the Raspberry Pi

The Raspberry Pi is a great little board, but because it runs Linux, you risk corrupting the file system on the SD card if you simply switch off the power. The standard advice is to type "sudo shutdown -h now" and wait until the shutdown process has finished, or has at least unmounted all the file systems, before switching off the power.

What if the RPi is part of some embedded widget with no keyboard and screen? How then to initiate a shutdown, and how to know when it's finished? It was while contemplating the creation of just such an embedded widget that I hit upon this piece of brilliance. First, build this circuit.

How does this work? When you plug in the wall wart, we don't want to draw power from the 9 volt battery yet, so we want Q1 to be off. This is accomplished by using Q4 to keep the Q2/Q3 Darlington turned off. The wall wart power flows thru the three diodes in parallel and operates the switching voltage regulator (I picked up a bunch of these from some Chinese outfit on eBay). Meanwhile C1 charges to around 9 volts.

What happens when you yank the wall wart? Why, magic of course. Magic happens. Specifically, Q4 turns off, and C1 begins to discharge thru R1 and R6, turning on the Darlington pair, which turns on Q1 so that now the switching regulator is running on battery power. The battery power remains on for some multiple of the RC time constant (100 uF * 500K = 50 secs), and you end up with about two minutes to get the Raspberry Pi to shutdown.

I've started a Github repository for this, and it includes the two scripts that make graceful shutdown a service that starts automatically when the RPi boots.

Sunday, November 15, 2015

How does the Tooba's keyboard work?

The question I was most frequently asked when exhibiting the Tooba was about how the keyboard works. I was asked this by people who were obviously intelligent, and I have a deep understanding of it myself, so you would think there would be no problem in communicating it.

The short answer is, when you touch the copper contact, its effective capacitance increases, causing a change in an RC time constant which can be quickly measured with very simple electronics. The problem is that nearly zero otherwise intelligent people in our society understand capacitance.

I tried several different ways to explain this to people. Usually I said, "it works the same way as the touch screen on your phone", and that's true but it doesn't explain anything. It only gives people a known reference point to indicate that it's not physically impossible.

Capacitance is actually a pretty straightforward idea. You have two parallel metal plates. A battery or some other force pulls electrons out of one plate and pushes them into the other, so the first plate becomes positively charged and the second plate becomes negatively charged.

Opposite charges attract, so there is a force pulling the plates together, but they are mechanically held apart, usually by a layer of insulating material. The attractive force therefore acts upon the charges themselves, the excess electrons in the negative plate and the "holes" where electrons are missing in the positive plate. The negative and positive charges want to stay where they are, a bit like inertia, and this manifests as a measurable voltage difference between the plates. The only way to change that inertia is a current, a flow of electrons, into one plate and out of the other. A flow of 6.24x1018 electrons per second (or one coulomb per second) is one ampere.

Capacitors can be big or small. Capacitance is the size C measured in farads. Great, what the hell does that mean? Remember that the inertia of charge is measurable as a voltage V. The amount of charge is Q, the number of coulombs originally pumped into the plates by the battery. Then the magic formula is Q = C x V, or V = Q / C. How this pertains to the Tooba is that if Q is the same and C gets bigger, then V gets smaller.

Here is the touch-sensor circuit in the Tooba. The input on the left and the output on the right are I/O pins of the microcontroller. The left one is an output that drives electrons into and out of the capacitor, and the right one is an input to measure the voltage on the capacitor. It's a crude measurement, just a comparison to a threshold voltage that gives a zero or one.

BRIEF DIVERSION: In electronics, current is actually the direction opposite to how electrons are moving. This backwards convention is (no kidding) because Benjamin Franklin didn't have know whether the carriers of charge were positive or negative; the observable behavior was the same to him. He guessed, with a fifty-fifty chance of getting it right, and got it wrong. His mistake got codified into all of electronics, forever and ever, amen.

The output on the left is set to logical zero for a little while. The diode quickly discharges the capacitor to the same low output voltage representing a logical zero. Then the output is set to logical one, a higher voltage. The current flows to the right, against the direction of the one-way diode, so all the current flows through the resistor which slows down the movement of electrons. Here's the tricky part: if the capacitor is small, the voltage rises quickly, and if the capacitor is large, the voltage rises slowly. After a certain fixed amount of time, the voltage will just cross the zero-one threshold if your finger was not touching the copper, and the input will read a one. But if you touched the copper, the voltage will rise too slowly and the input will read a zero.

There is one more tricky bit to this. You may have been told in school that you always need to complete a circuit. Pushing current into the top of the capacitor would require that the bottom of the capacitor is connected to the ground in the Tooba. It turns out that capacitors are the one exception to this "complete the circuit" rule. You can (briefly) push current into a single plate and it will behave like a capacitor even without a second plate. That's why you can play the Tooba (or use the touch screen of your phone) without having a wire attached to you to complete the circuit.

Addendum - a friend read this post and wrote:
There's a missing piece in your explanation of capacitance: yes, the negative and positive charges do attract one another, but the charges in each plate also more strongly repel the same-sign charges in that same plate. That creates a restoring force that tries to push current back through the circuit to discharge the capacitor. 
The greater the capacitance, the smaller this net restoring force is. (You can think of a capacitor of higher capacitance as giving more room for the charges to spread out so they're not so crowded, or putting the plates closer together so that the "inertial" attraction across the gap is greater, or both.) 
If you touch one plate, a teeny bit of the charge can flow into your finger, or charges can become polarized in your finger so that some of the electric restoring force is cancelled out. Either way, it increases the capacitance by relieving some of the electric back-pressure through the circuit. And that's why you don't have to complete a circuit yourself.

Friday, July 24, 2015

The Tooba

Way back when, one of my friends in high school was a guy named Dave Wilson whose father got him into Moog-style analog electronic music synthesizers. Dave created a museum of historical synthesizers in his home in Nashua, New Hampshire. Through my high school and college years, we exchanged ideas and circuits and bits of lore for various synthesizer hacks. These synthesizer modules were implementing mathematical functions that could be implemented in digital electronics. So we both at various points and in various contexts wrote code to do that.

The Tooba is going to be some of that built into a short length of 3" PVC tubing. There will be a two or three-octave touch sensitive keyboard and some linear slide pots for various synth parameters. Sound will be generated in the same Teensy 3.1 controller board that scans the keyboard. Unless the poor little CPU gets overwhelmed, then maybe I'll put in a second Teensy for sound generation.

Here is a prototype, mostly testing the touch-sensitive keyboard. For this version the sound was generated by Timidity running on a Raspberry Pi, and the form factor is obviously different. The keyboard misbehaved in this video because of a sagging under-powered 12-volt supply.

UPDATE: Wrong! The keyboard misbehaved because of a programming error identified much later, now corrected.

I'll add more to this post as the thing gets closer to completion.

I had some confusion about the date for the Rhode Island Mini Maker Faire (last year it was a month after the NYC Maker Faire) so I had to hustle when I got the date correct. But I managed to pull it off and have got the thing working.

I also had to figure out how to mount the copper hexagons on the PVC tubing. Eventually my technique got good enough, but by then I had a mix of well-soldered and poorly-soldered hexagons, and the latter periodically pop off. So in the video below you'll see some missing hexagons that I need to replace before I head down to Providence.

So that's the gadget. There is a lot of clean-up to do on the mechanical design. Now I feel it's stable enough to justify laying out a printed circuit board. And the inside wires should be replaced by ribbon cable. But hey, it's working.

Thursday, June 11, 2015

Eating without grains

Recently a friend had me watch a talk by William Davis, a cardiologist who wrote a book called Wheat Belly. If you have an hour, here are his thoughts on wheat.

If you don't have an hour, here is the gist. Humans originally consumed grains in general and wheat in particular to avoid starvation. They saw aurochs (predecessors to the modern cow) around them eating grasses, and lacking other food sources at the time, also tried eating grasses. They found that the only part of the grass that was digestible was the seeds, and they cultivated these grasses into the various grains we know today.

A few problems. First, the first humans to eat grasses did not have the evolutionary adaptations that the aurochs and the cow have: special teeth, special multiple stomachs including one grinding stomach, special digestive enzymes. We still don't have those adaptations, and we probably wouldn't acquire them for hundreds of thousands or millions of years. Grasses have only been on the human menu for about 10,000 years, and our genotype doesn't change very quickly.

Second, the seed of the grass is the one part that, from an evolutionary perspective, wants NOT to be digested. It has defensive proteins and enzymes intended to discourage (read "poison") any animal trying to eat it. Gluten allergies are human reactions to one of these poisons.

Third, what has been changed in recent decades. This is largely the work of Norman Borlaug, undertaken to improve crop yield. Borlaug received a Nobel prize for his work in creating a dwarf form of wheat with a vastly increased yield. Farmers can no longer prosper growing the earlier tall wheat and so now all the wheat you eat in any form is Borlaug's dwarf wheat. This is true on a global scale, not just here in the United States.

The problem is that this wheat has proven to be mismatched even more poorly to human nutritional needs, and that's why so many people are gluten intolerant, a condition that did not previously exist. It is looking likely that the modern epidemics of obesity and diabetes are also tied to dwarf wheat. Many have been mystified that these epidemics have spread outside the U.S. and this explains why that would happen.

Gliadin, one of the proteins in wheat, is an opiate that stimulates appetite, contributing to obesity. This video by Joe Rignola goes into more detail of the damage that gliadin does to a person's intestines. This damage is not confined to the intestines. T cells (the part of the immune system that attack one's own body, responsible for inflammation) respond to gliadin by attacking both it and something that belongs in your gut called transglutaminase. It turns out that transglutaminase is produced throughout the entire body and all areas come under attack as the T cells attempt to respond to the original gliadin injury. And so now you have a problem of broad spectrum inflammation.

As if this all weren't enough, it turns out that wheat raises blood glucose substantially. Two slices of whole wheat bread raise blood glucose higher than six teaspoons of ordinary sugar. In the hour-long video, Davis goes on to enumerate more components of wheat that cause additional health problems. Elsewhere, he identifies health issues with all other grains as well. So the thing one wants to do is to adopt a grain-free diet. Here are a few resources that may be helpful if you are considering this.
The important thing is to identify foods that you can be certain are grain-free, for example meats, cheeses, fruits, vegetables, and nuts. Recall that grains did not enter the human diet until the very recent evolutionary past. Grains are not, as might be believed, necessary for good health. They are in fact detrimental to it.

Monday, March 02, 2015

An overview of the 2014 printer

People have been asking for an overview of the 3D printer that I showed at Makerfaire NYC 2014. I designed it myself, borrowing a couple of ideas from other printers. My printer builds objects from a liquid resin that solidifies under ultraviolet light. Your dentist may use similar stuff to fill cavities. I get the ultraviolet light from an unmodified conference room projector that cost me about $350, the single biggest expense of the whole project. The most important design principle was that a person of very low craftsmanship like myself should be able to build the thing. It pretty much does not require precision at any point in the construction.

The idea to use a projector rather than steering a laser came from the B9 Creator printer. The Peachy printer gave me the idea of floating resin on salt water and projecting the light onto the surface of the liquid. The idea to raise and lower the build platform using three threaded rods driven by a bicycle chain was my own, and it turned out not to be such a great idea.

Soon I plan to post some of my plans for a 2015 printer. Right off the top of my head I can think of four worthwhile goals.

  • Bring it to both the Bay Area Makerfaire and the NYC Makerfaire.
  • Replace the bicycle chain and those three threaded rods with something simpler and more reliable.
  • Clean up the electronics and software so I don't need a laptop to control it.
  • Replace the orange bucket with something with transparent sides, like an aquarium, so that people can watch the printing process.

With the bicycle chain not yet in place, the printer looks like this. Those gray sprockets engage the bicycle chain, which goes around in a sort of diamond shape. The orange bucket is one of the three-dollar buckets from Home Depot. The plywood, nuts and bolts, and threaded rods also came from Home Depot. At this particular point in the work, I thought I would suspend a mirror over the top of the bucket at 45 degrees, which is why you see a piece of wood in the upper right of the photo. But I didn't know about first surface mirrors then, and I lost enough UV going through the glass twice that I couldn't get the resin to solidify, so I then positioned the projector directly over the bucket, pointing downward.

The sprockets were designed using OpenSCAD and initially the teeth were too pointy - correct in theory but too sticky for real-world bicycle chain, so you can see where I cut off the points, and later revised the design. If you order sprockets from my Shapeways store, they should now work fine.

Here is the printer fully assembled. The bicycle chain is driven by a stepper motor. Each revolution of the stepper motor (200 steps) raises or lowers the build platform by 1/20th of an inch because the thread on the threaded rods is 1/4-20.

When it's printing, it looks like this. Only where the light is the strongest is the resin solidified. The resin happily ignores ambient diffuse daylight in the room where I'm printing, so I don't need to use dark room lighting.

Pictured below are some of the objects I've printed with the thing. I started out with a bottle of green resin and when that started getting low, I added a bottle of clear, so my things tend to vary between green and clear.

Tuesday, September 16, 2014

A couple of recent 3D printing successes

With a few last-minute improvements I've been able to substantially improve the performance of the printer. I slowed the stepper motor to reduce vibration, and I allowed a settling time after each motor movement before exposing the next layer of resin. I cleaned up the build platform, it is now a sheet of aluminum epoxied to the plywood. (And of course, within a couple of prints, it has gotten covered with a sheet of cured resin. Best laid plans...) I had been trying exposure times that were too short, so I went back to 60 seconds per layer.

If you'd like to see these prints and others, and the printer that made them, come to Maker Faire NYC this weekend at the New York Hall of Science in Queens, NY.

Here is a chess rook. It has an interior spiral staircase. The windows are a bit misshapen and the bottom flat surface is covered with a big glob of cured resin. I don't know why those things happened, but the detail on the parts that came out well isn't too bad.
This is a dodecahedron, one of the five Platonic solids. In the days of ancient Greece, this shape was the cause of some controversy because it could be used to prove the existence of irrational numbers, which ticked off Pythagoras something fierce. This was posted on Thingiverse, as was the rook.

More shapes to come soon, if all goes well.