All-Analog DIY Segway Project

An anonymous reader writes "One of the zany hacker-makers here at MIT just finished this DIY Segway project (video). Difference from the others: it's all analog. The controller is built without a microprocessor or even digital logic. It does use a gyroscope like the real Segway. The functionality looks fairly basic, but the fact that the controller works at all is amazing. The guy has a ton of other projects on his site too. Definitely worth a read for people who enjoy building things."

Comments

and here we have a hacker

[chronoss2010]

a real one that invents and makes stuff...imagine that , quick arrest the terrorist(/sarcasm)

Re:and here we have a hacker

[jobst]

would you please carefully read the ENTIRE message including the very last word, please!

SEGFAULT

[PatPending]

Best. Name. Ever. for this

Re:SEGFAULT

[camperslo]

The ultimate SEGFAULT was a sad one, the CEO of Segway dying from running his off a cliff by his home.

http://www.huffingtonpost.com/2010/09/27/jimi-heselden-segway-boss_n_739983.html

Controllers were analog before they were digital.

[RoverDaddy]

Digital controllers -emulate- analog behavior (at least many of them do). There's a pantload of research and science behind analog control.

The Land Before CPUs

[Glendale2x]

Why is this amazing "that the controller works at all"? There was a time before microprocessors, you know, and they did fun things like travel in space without them.

Re:The Land Before CPUs

[Anne_Nonymous]

Sure, but name one moon mission that had and XBox aboard.

Re:The Land Before CPUs

[Burdell]

The computers that flew were digital, but the computers that tested them were analog. My father worked on the Saturn V guidance system, and they built one of the earliest "hardware in the loop" simulation setups to test the software and flight-certify the computers that flew. Digital computers of the day were not fast enough to simulate the inputs and monitor the outputs in real time, so the simulation was built with analog computers.

Re:The Land Before CPUs

[skywatcher2501]

Talking about spaceflight, the Russians only recently replaced the analogue Soyuz control system with a fully digital one.

Females?!

[lul_wat]

They are all at MIT?! I'd have studied harder in highschool if they'd only told us.

Re:Females?!

[Colonel Korn]

They are all at MIT?! I'd have studied harder in highschool if they'd only told us.

The MIT recruiting video sent to my high school might have convinced you.

A pair of serious undergrads, one male and one female, are working in a lab. The glassware is very impressive and filled with bubbling food coloring or whatnot. The lights are low to draw attention to the Science. Then the two look at each other knowingly, sweep the contents of a benchtop onto the floor and start making out atop it to the wail of an electric guitar.

403 Forbidden

[noidentity]

I guess the webserver also is running on an all-analog server, which is now in a "halt and catch fire" state.

Re:403 Forbidden

[PatPending]

Likely due to a SEGFAULT.

My thoughts exactly

[betterunixthanunix]

Why is it amazing that an analog control system works?

Re:My thoughts exactly

[0100010001010011]

Kids these days couldn't imagine trying to implement a PID controller without a microprocessor. It does make it easy to watch what is going on.

Watching large governors work is pretty cool. No electronics to speak of. Just a properly tuned set of weights and some geometry.

Re:My thoughts exactly

[seanadams.com]

It's not amazing but it's certainly harder. Probably higher part count, more electrical noise to deal with, harder to debug, harder to implement delays and state machines, more wiring etc. It's just impractical for most purposes.

Re:My thoughts exactly

[Jarik C-Bol]

Actually, they are pretty awesome even at a small scale. I took a small engines class once, where we each where given a lawn mower engine to tear down and rebuild. Lawn mowers have governors in them (dinky little plastic ones usually) butI swear, we spent several hours playing with them once we got down that far into the engines.

Re:My thoughts exactly

[robot256]

Analog is not precise enough (people think).

Speaking as a controls engineer, they have obviously not done much digital controls. You have to worry about things like sampling rate, aliasing, round-off error, and digital noise introduced into the (inescapably) analog parts of the circuit. For a simple system, a properly-designed analog controller is much easier to implement, and has advantages like "infinite" sampling rate, graceful failure modes, white (gaussian) noise as opposed to odd frequencies introduced by sampling and clock frequencies, and no programming bugs or crashes.

Analog controllers for simple linear systems (like telescope mirrors) are in virtually every spacecraft ever launched for precisely those reasons. Only recently has the push for miniaturization driven some simple systems into digital FPGA controllers.

Re:My thoughts exactly

[udippel]

And I have mod points, but don't give you any. 'Troll' is awfully harsh, I agree. I'd rather give you some 'un-informed'.
The higher part count is surely on the side of the digital controller. Just look at the diagrams offered: analog means direct processing of signals, no A/D. Just some op-amps, pwm, done.
Harder to debug? Nonsense. You debug with a voltmeter instead of a logic analyzer.
You are right with respect to advanced controlling, though, like counting, timing, delays. But none is needed here, some filters are just enough, and filters are implemented easier with some RCs around an op-amp. Also, you need a bridge. A bridge is much more simple if build in an analog manner. So your 'just impractical' is a good reason to not give you any mod points. It might be your opinion, and you sure may have one, but to me, an EE with some experience in developing controllers, it doesn't hold water in the case of a gyroscope.

Re:My thoughts exactly

[Jane Q. Public]

Yes, but the problem is: this ain't it.

His motor driver chips for example (International Rectifier IRS21184), take standard CMOS digital input signals. Digital input, or digital-compatible input, makes no difference. Somewhere along the line you still need to do what amounts to A-D Conversion. Which brings back most of the problems you mentioned.

Unless I misunderstood, and from the spec sheet I don't think so (the schematic shows Schmitt-trigger inputs, which convert analog input to square wave with hysteresis), then no matter what he says, this is an analog-digital hybrid system, not "pure analog" at all.

Re:My thoughts exactly

[Arlet]

The higher part count is surely on the side of the digital controller. Just look at the diagrams offered: analog means direct processing of signals, no A/D. Just some op-amps, pwm, done.

How does adding some op-amps and discrete PWM reduce part count ? Have you counted all the resistors and capacitors for the filtering ?

With a digital controller, you can implement the A/D, filtering, control loops, and the motor PWM all inside the same device. You can even avoid some of the A/D stuff by using a accelerometers and gyroscopes with a digital interface.

and filters are implemented easier with some RCs around an op-amp

An RC filter can be implemented in a single line of code on a controller.

Re:My thoughts exactly

[Prune]

I built an analog PID temperature controller for my espresso machine (as every coffee geek knows, grouphead temperature variation over about a degree C during the ~25 s extraction noticeably affects the taste). It's one of the rare cases I use opamps. The analog part of all my audio projects always uses transistors or tubes, as chip amps have the problem of thermal variation in the latter stages affecting the input stages which are in the same thermal package. This doesn't show up in a steady signal harmonic distortion measurement but there are specific tests for it. Plus, you can't really experiment with new topologies when you use an IC. Anyway, as a software developer in my day job, I find it a great thing to do analog in my hobby, as it adds a sort of Zen balance. I do now some digital in audio as well, but more out of necessity due to my dissatisfaction with the performance of current digital-analog converters. There is an interesting interplay between the two, though, as well as between electronics in general and the neuropsychology of human hearing which makes perception of distortion poorly correlated with typical engineering metrics such as THD in blind listening tests. 2nd harmonic is inaudible to as much as 1% for most people, for example, while some higher than 3rd harmonics and other distortions suchas crossover (class B or AB amps) and amplitude-to-frequency modulation effects which though tiny occur in many amplifiers, are audible at as little as several ppm (a healthy ear has a 120 dB dynamic range)

The best part...

[farnsaw]

The best part is the shopping cart in the lab holding a jumble of electronics.

Re:The best part...

[Chris Tucker]

As seen at MIT:

A Shopping Cart Full Of Roombas.

Re:The Tilting Handlebar

[Sulphur]

Two MIT profs were arguing which was smarter.

One pointed at a 14" disk drive and said "I can make that walk across the room."

He keyed in something and after a few seeks it lurched onto two legs and walked across the room.

The other one said "That's nothing; I can make it turn around and go back."

Re:Controllers were analog before they were digita

[Jane Q. Public]

MIT?? This is kid's stuff. The only difference between your broomstick controller and a $20 DIY backyard sun-tracking sundial is that the broom balancer is 2-axis, and has to be faster. Big deal.

Re:I would like to join the chorus

[Jane Q. Public]

Heisenberg and Planck might disagree with you on that.

Re:Controllers were analog before they were digita

[pz]

MIT?? This is kid's stuff. The only difference between your broomstick controller and a $20 DIY backyard sun-tracking sundial is that the broom balancer is 2-axis, and has to be faster. Big deal.

Um, no. Clearly you have not studied this problem which is a (perhaps *the*) classic PID exercise. A simple P term (proportional) will fail very, very quickly. Add the D term (differential) and you get stability, but drift. Finally, add the I term (integral) and you eliminate the drift and turn the meta-stable system into a stable one. If you want stability to external perturbation, or generalization to a broad range of loads, then you need more analysis and more terms.

Designing one of those from scratch, based only on the mathematical modeling, and building it from individual components, while worthy of no more than an undergraduate exercise at MIT, is non-trivial. Designing a full Segway-like system is a generalization of this problem and also non-trivial.

If you think it's so easy, then please build one yourself -- demonstrating all of the calculations -- and post the video.