Robust rotation measurement (rotation transducer)

I've had very good luck with encoders in the past. If it's a good quality encoder, you're not going too fast for the encoder, and you haven't screwed up your decoding in the electronics or the software, then about the only ways left to have problems are that you're abusing the encoder somehow.

Resolvers are a good alternative to encoders. Good quality pots are not to be sneered at. RVDT's work well over limited stroke lengths.

What sort of rates would an LED/photo-diode not track?

I saw a cute Hall-effect rotary encoder this week. The output was 360-degree, with any sort of encoding desired (mag-dir, quadrature, binary,...).

If I'm not mistaken Allegro Microsystems has those, too.

There is an upper limit that you ignore at your peril. The ones with open-collector outputs have, I think, a more severe speed limit* than ones with totem-pole or CMOS output.

• And, obviously, more sensitivity to your supporting circuitry.

Absolute optical or mechanical encoders, synchro/resolvers, inductosyns, sin/cos pots, maybe some equivalent capacitive thing. RVDTs or pots for modest angles, less than a full rotation.

Incremental encoders are usually pretty reliable. All sorts of printers and things use them. Maybe you have a signal conditioning problem.

There's some sort of cool encoder that uses round PC boards, with inductive coupling between traces.

John

We use Q encoders on fulcrums to detect position that does at times move very quickly.. The problem i've seen is the electronics doing the decoding.. If you don't account for chatter pulse from vibration where it would cause a pulse to render it self off/on on channel and not the other and depending on your coding scheme, you can see extra and fewer pulses than expected..

Using a RS D flip flop to latch signals on each transition does remove this problem. It seems many ready made process controllers for Q operation that I've seen do exhibit these issues how ever, in most cases the applied application does not lend itself to cause problems.

It would be nice to know how you are doing the actual decoding of the A,B,Z signals?

Jamie.

Is this about absolute angular displacement? There are Gray-code wheel types of sensors that can give good performance in the missing-codes sense. Quadrature is a crude 2-bit Gray code, but the sensitivity to error gets less as you go to higher bit counts, with more 'known-illegal' transitions between states.

Even if you miss step 12, it's comforting to know that THIS one is step 13. Unless you're superstitious.

With 6-bit Gray code, the position only gets ambiguous if you miss dozens of steps.

the

I'm doing the decoding using a state transition table using a microcontroller that is polling the A and B channels with a sampling rate to be much higher than the possible state transitions rate on the encoder. I'm not sure what a 'Z' signal refers to.

The datasheet inidicates that the encoders may have some extra logic in them as well but it does not specify any behavior explicitly. The encoder is an AEDS-964x.

Not familiar with Alegro's product. These are ICs. You add your own mechanics on top, or bottom.

Saw some neat motion sensors this week too (all the disties were through with their toy vendors in tow).

Obviously there is an upper limit, but I wouldn't have thought it would be in the range of encoder mechanics.

Melexis also have them.

-ek

That's the Inductosyn. Electrically it's just a resolver with lots and lots of poles, although there are enough detail differences that the 'just' needs to be taken with a grain of salt.

Synchros and resolvers are cool. Analog Devices has s/d converter chips for about $12.

John

It's not that difficult to measure the outputs directly and do the math in a processor, either.

We've been thinking about that lately, or at least doing it in an FPGA. The most popular tracking algorithm is very elegant; it's available in a number of places on the web. The problem with a synchro is that, twice a line cycle, the signals all go to zero and you are blind for a while. The algotithm essentially has inertia and cruises through the line nulls nicely, so the ability to track speed and acceleration is amazing.

Page 19+

John

Perhaps one of the more exotic things that i have ever seen is a widget called a resolver tracking bridge, it did the first 12 bits or so and passed the residue (tan theta) to a 16 bit SAR ADC. All of the RTB was done with transformers and relays and stuff. It was just some channels out of hundreds feeding the ADC aat 50 ksamples/second. Late