Hall rotary encoder tempco--wisdom?

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Always with the optimism, eh?

Best regards, Spehro Pefhany

Grin, (I hope it's OK if a laugh at your pain a bit.) Everything magnetic h as a tempco. This has nothing to do with your question, but I dunked a fer rite bead into LN2 today, and measured the resistance from one end to the o ther. I was looking for a phase transition (which I didn't see), but the r esistance went from 35 ohms to 50k.

George H.

I'm guessing these encoders are being interpolated? Why not just use an optical encoder with sufficient resolution so you don't have to interpolate an analog signal? You can get encoders with 5000 counts/rev fairly easily, for just a little extra, you can go to

Jon

Well, it would be a semiconductor!

That would be MnZn, no? I wonder if NiZn freezes out more, or completely. Hmm, I doubt ferrites are pure enough to be controlled for dopant level impurities, and defects (they're polycrystalline besides). There might be residual resistance, at least until very low temperatures.

No idea if anyone's tried doping ferrites as semiconductors. Mixed oxidation state materials tend to be very difficult to control (e.g., ZnO, which I believe is almost always P-type due to oxygen deficiency), so magnetic transistors (magistors? spinistors?) are unlikely, at least by direct equivalence (electron/hole junctions and so on).

Tim

-500ppr-2ch/dp/1654865

nes on it to make it work. I've tried to take a look at the data-sheet, but can't get it to download.

eady hurting a bit at 4096, since the actual measurement range is only 8 de grees.

If you are going to have to make your own code-wheel anyway, and since you are only working over 8 degrees, it would be easy enough - in theory - to w ork with a 16 degree segment out of a larger "code-wheel" and treat the seg ment you look at as a 2048 segment out of an up to 46,080 counts per rev th eoretical code wheel. The Avago screed talks about a "code strip" and bodgi ng your mechanical set-up so that your 8 degree range moves the code-strip a useful distance isn't impossible, though I can see all sorts of arguments for not doing it.

The best - as you point out below - is that you don't actually have to.

old spray is _completely_unable_ to move the indicated angle--the same puls e width variation that made a 7-degree change yesterday gives less than 1 c ount (0.09 degrees) today. So it looks like measuring the right thing fixe d the problem more or less completely.

It often does. I'm very glad to hear it.

roses. Amazing the buried treasure you can find in other people's softw are. (Sometimes mine too, of course.)

There's nothing quite like asking the right question.

Well, it just says 0.5 degrees at room temperature and 0.9 degrees over

-40 to +125. That can be read several ways.

In any case, it looks like it's pretty stable with temperature now.

Cheers

Phil Hobbs

It's okay, I'm used to it by now.

The tempco of the magnet is taken out to leading order by combining the four Hall elements--when you compute ratios, the field ideally cancels out.

Cheers

Phil Hobbs

There are a couple of options for sensing shaft rotation. As someone else already mentioned the optical encoder I will confine my response to suggesting that you could use a Synchro/Resolver to monitor the shaft. These can be quite accurate over fairly wide temperature regimes. The decoding of resolver signals (Sin/Cos) is not that difficult and there are chips from Analog Devices that will provide a digital output from the signals. You can even get rate of rotation informtion from them.

On a sunny day (Wed, 08 Jan 2014 16:40:06 -0500) it happened Phil Hobbs wrote in :

Right, large gear box, lot of play.

No gear box?

Stepper motor, worm drive. constructon with springs that eliminates any play. Would take a re-design. You need help from a mechanical wizzard.

On a sunny day (Thu, 09 Jan 2014 00:52:38 -0500) it happened Phil Hobbs wrote in :

Phil, just an idea, how about a small [air?] rotating capacitor, like for tuning radios, and measuring pF (or frequency). That shifts the problem elsewhere....

John already solved it--the encoder frequency isn't guaranteed stable, and the software was measuring pulse width instead of duty cycle.

Cool gadgets.

Cheers, James Arthur

The most negative way is usually the intended one. When you're writing a spec you have to make your product look better than the competition, and they're doing it too.

That's great! I hate things that "almost work".

By the way, getting a real environmental chamber might not be a bad idea now that you've got the room for it (and 3-phase power?). You can discover a lot of interesting stuff by logging gigs of data and running over the whole temperature range- even on supposedly mature designs.

I was recently debugging someone else's design and discovered an interesting temperature-dependent instability at would never have been seen otherwise. It went bonkers repeatably at two distinct temperatures (with the temp moving in either direction). Can't discuss the cause or cure, but that was a strange one.

Best regards, Spehro Pefhany

Hi Tim, Yeah I was going to plot it up, log (R) vs 1/T.. but I've got a new 'puter at work and I've yet to get my old copy of Origin installed... But maybe la ter today/

Well I don't know exactly what it is. Pieces that I inhereted from elsewhe re. There's a nice article "Low Temperature Transformation in Ferrites", L. Bic kford, Rev. mod. Physics, January 1953 pg75. Where he looks at magnetite a nd see's a nice phase transition at ~120 K. (change in R, magnetization an d peak in specific heat.) Kinda showy.

OK I don't know about the semi conductor nature. I looked at some other be ads and found that the conduction wasn't a bulk effect. But looked more li ke percolation... resistance was not proportinal to the length.

George H.

AH HA!

I was wondering how in hell you could have a 5 degree variation over 100 K -- that would have you falling off the lines of a real encoder.

Now I know...

I would go looking for rotary LVDTs. When that failed, I'd be mightily tempted to roll my own. Yes, you'd need some signal processing somewhere to decode the LVDT, but if you've got enough capacity on a microcontroller it can all be done with good accuracy in software.

With the right magnetic design, you could probably even make a restricted- range LVDT with high sensitivity over your 8-degree span, at the you- don't-care cost of crappy performance outside of that.

Failing that, if you have room for a long moment arm consider a couple of magnets on sticks attached to your shaft, and a couple of linear hall effect sensors on a board nominally centered on the magnets. Then read the outputs, do some figgerin' and have an answer.

Interestingly the software fix seems to have got rid of the issue, hurrah.

Cheers

Phil Hobbs

My actual lab isn't that large--only about 200 square feet, and I'm pretty crowded in there now. (A lot has changed since I took those lab tour photos.)

Interesting. Cryogenic or ambient?

Cheers

Phil Hobbs

with more than one magnet that could quickly get complicated, permanent magnets are ~2000ppm/C

-Lasse

That might work fine, but it would be one more science project that I'd like to avoid at the moment!

Thanks

Phil Hobbs

Ambient.