motion control

Sony make a system which depends on a magnetically coded rod, which works better under really messy conditions, and you can go to a Zygo laser interferometer if you want real precision, but a Heidenhain Moire-fringe optical scale is most likely to be the right way to go.

Potentiometers were used commercially in cheap systems when I was young

- and back then I worked up a capacitive position sensor which was cheaper and better, but sadly bulkier - but today they'd be strictly for very cheap and nasty gear.

--
Bill Sloman, Nijmegen

Or encoders for longer strokes and (possibly) less precision.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

google LVDT for stuff like this

martin

Heh. "Bullshit detector". I like that. The closest I've ever come to that was a 68HC11 in an industrial universal battery charger: 24V, 40A, with SCR phase control, and I sampled the output 16 times per mains cycle, and averaged the voltage and current readings to decide on the timing for the next cycle. (it did both half-cycles symmetrically - I tried to do it a half-cycle at a time, but couldn't get them balanced), but if the current ADC had one reading that was out of range, (indicating a short), it immediately shut down and refused to trigger the SCRs again until the short had been cleared and the processor was reset.

It's kind of a good feeling, when you've got a 24V supply pumping out

40A, and when you short it, the box goes, "pup" and quits. Then starts up again when the short is gone. ;-)

Cheers! Rich

I had to troubleshoot an interesting problem once. It was an automated sawmill, that used a fancy travelling-salesman type algorithm (8 dual pentium boxes) to optimise cuts based on current market prices. very fancy laser/optical doodads for measuring knots in logs etc, monstrous rams (it chopped up 25yo pine trees), a 12' circular saw blade on a ram driven by a 600kW machine etc.

problem was, whenever any of the drives turned on, it went apeshit. rams slammed up and down, sawblades came alive, that sort of thing. All up, very bad. plant commissioning did *not* go well at all :)

they were using some fancy position measuring transducers, that were basically a piston inside a tube. The piston had a magnet on the end, and an RF signal was pumped into one end (IIRC). some of these were 20m long, and absolute accuracy was a fraction of a mm. a most excellent transducer, but they did *not* like stray H fields, especially huge HF ones.

ultimately the problem was unshielded motor cables. Well actually they were shielded, but an instrumentation guy insisted the shield be connected at one end only. big mistake. fixing that fixed the problem.

Of course if the s/w guys had run a bullshit detector past the transducer data, it would have been far less problematic. the resultant tantrums were quite impressive though, but scary if standing nearby.

Cheers Terry

I'm looking for something linear, cheap, possibly nasty, extremely reliable and only need an accuracy of about 1mm. Any ideas?

--
Dirk

The Consensus:-
The political party for the new millenium
http://www.theconsensus.org

And you ignored the LVDT why?

--
JosephKK

Depends how long the stroke was, but my capacitative position sensor compared two capacitances between two lumps of printed crcuit board. On obe side I had a common "receiving" electrode (over the top of a solid ground plane) and on the other side I had two transmitting plates being driven in anti-phase. One saw 100% of the receiving plate all the time, and was driven with a controllable amplitude square wave, and the other was screened from the receiving plate by a moving grounded lump of metal (a ground plane inside a flexy in my case) and driven by a constant amplitude square wave.

The controllable voltage was adjusted to keep the net demodulated output from the receiving plate as close as possible to zero - which reverses what we were actually doing. In fact we adjusted the amplitude of the controllable squre wave, and used a PID loop to drive an electric motor to drive the grounded in and out to screen more or less of the constant amplitude square wave transmitter.

All the switching and demodulation was done in CMOS. The only cute bit of the circuit was the high impedance amplifier on the receiving plate which was FET-bipolar pair whose AC gain was set by a 0.1pF capacitance between the broken ends of a printed track (we cut the track on the first prototype, but had neatly round ends on the next layout).

E-mail me if you want more detail (at bill dot sloman skip this anti-spam insert at ieee dot org).

--
Bill Sloman, Nijmegen

There are a couple of possible explanations. For one thing, an LVDT isn't cheap. For another, the range over which it is linear (or even useful) is a lot shorter than the transformer doing the measuring, or the moving core whose position is being measured.

And he was asking for suggestions, and you've phrased your suggestion in a pretty unfriendly fashion.

--
Bill Sloman, Nijmegen

OK. Here's what I want as a preliminary spec. Reliable non contact. The movement could be up to 5Hz over a max 30cm stroke for weeks ie tens or hundreds of millions of measurements before failure.

Cheap -

Would this work over a 300mm stroke? See my other post for a few more details.

--
Dirk

The Consensus:-
The political party for the new millenium
http://www.theconsensus.org

for

My capacitative sensor could fill the bill. The original was accurate to about 0.1% - measured with a 6-decidmal digit DVM and a milling machine bed.The flexy that ran through the position sensing part of the sensor also ran through the compensating part, which helped, but my system wasn't non-contact because the flexy wasn't that stiff. A system devused for you application could be reliably non-contact, but lot depends on what your moving electrode can look like, and how you can support the sensing and driving electrodes

The price may be attainable - but that depends on what you'd be paying for the double-sided printed circuit boards for sensing and driving electrodes.

--
Bill Sloman, Nijmegen

and

Slidepot with a boot?

The thing from an old dot-matrix printer?

I saw a linear system with toothed belt and whatever you call those toothed wheels that the belt rides on, with a 10-turn precision pot as one of the wheel axles.

Good Luck! Rich

and

My example worked fine over 100mm. The interesting questions about getting a 300mm stroke would be all mechanical, and would depend on the details of your mechanism - the system should scale up nicely.

--
Bill Sloman, Nijmegen

for

What's that thing for?

robert

for

Have you looked at those capacitive linear sensors- like the working part of electronic vernier calipers? Simple, robust, cheap, I can buy a

300mm scale for £25 in ones, so 1000s should be no problem. Typically 0.1mm resolution. Downside: slow, stiffish (probably doesn't have to be), fairly bulky in the commercially available ones. But some variation on the theme might do it, even if you have to design it yourself and up the excitation rate to get the speed. Buy one and take it apart- it's just PCB, the electronics is pretty simple.

Paul Burke

Not seen one of those. Can you post a URL please?

--
Dirk

The Consensus:-
The political party for the new millenium
http://www.theconsensus.org

and

It would wear out in a matter of hours, if not minutes.

Anything mech complex would not be reliable enough over 100m repetitions

--
Dirk

The Consensus:-
The political party for the new millenium
http://www.theconsensus.org

Thanks. No indication of how it works however. What I am trying to avoid is anything that requires accurate mechanics or magnets. The device will be working in a fairly hostile environment from a magnetic POV.

--
Dirk

The Consensus:-
The political party for the new millenium
http://www.theconsensus.org