ultra potentiometer

[coppercone2]

So I was thinking about my current source, with ten turn wirewound potentiometer on it, the range switches etc



If you were to take a better potentiometer, like a metal foil one, possibly get a better contact for it (precision ground or better) and remake the actuator mechanism with something like a precision geared low backlash micrometer controlled thing, or even piezo actuators (like a hard drive head)... what would it be capable of?

Helipot was better. But I mean if you made this thing really good.


It might have a linearity problem and difficult thermal design. But how far can it be dialed in for resolution? I don't mean in that old current source, it just inspired this thought.

As good as a LVDT?

[Doctorandus_P]

Potentiometers do have a (theoretical) infinite resolution.

I don't know about "metal foil" potentiometers. The better quality versions generally "conductive plastic".

A few days ago an KTR-10mm arrived from Aliexpress. It's a linear potentiometer which quite resembles a pneumatic cylinder. The label says it has a linearity of 0.05%, but it's from Ali, so who knows what it really is. With a 12 bit ADC I'm hoping to get a better then 1um resolution out of this, but to test this, I would have to make a pretty good test stand first. Linear potentiometers like these can be bought in lots of different sizes, even more then a meter.

A few years ago I also bought a WDD35. These are rotary potentiometers and intended to be used as position sensors on motor shafts. They do not have end stops (but they do have a small dead band). They do have ball bearings and a quite elaborate wiper made out of 10 or more wire contacts, and they should survive millions of revolutions. The WDD35 comes in different linearity classes. They are calibrated by grinding away a part of the conductive plastic. WDD35 also has a quite large (35mm) outer diameter, and this increases usable resolution too of course.

For final resolution, mechanical stability is probably the limiting factor. If you rotate the knob a tiny bit, but instead of moving the wiper around the center of the axle, it's possible that the wiper stays in place but the play in the axle moves to "the other side".

LVDT has some inherent advantages. There are no touching parts, so no wear and less to no mechanical friction. Depending of course on seals and how the parts are kept together.

[TimFox]

Wirewound 10-turn potentiometers do have a finite resolution, due to the wiper moving in "bumps" from one turn to the next.
More expensive pots with conductive plastic over the wirewound element avoid this, and have less noise when turning the pot.
Bourns spec sheets for real wirewound pots (both panel and trimpots) give this resolution explicitly.

[Conrad Hoffman]

You want resolution? Just make it really big. See the Leeds & Northrup K-2 slide wire potentiometer.

[coppercone2]

you would think if you ground down the surface of a wire potentiometer like a variac, it might be somwhat infinite, but with discrete jumps

like peacewise infinite

[TimFox]

Discrete jumps are why the resolution is finite.
A continuous track, such as carbon, cermet, or conductive plastic, achieves infinite resolution (in principle).

[TimFox]

Here is a Bourns datasheet for 10-turn panel-mount pots, both wirewound and plastic coated:  https://www.bourns.com/docs/product-datasheets/3500.pdf?sfvrsn=8cb7c7ec_4
The plastic parts state "essentially infinite" resolution.
Resolution for several normal wirewounds are shown in the lower left corner.
Lower resistance pots use fatter wire, so their resolution is worse.
Similarly, a Bourns datasheet for wirewound trimpots:  https://www.bourns.com/docs/product-datasheets/3057.pdf?sfvrsn=d650d9ef_9
The resolution specs in the lower right go from 2.4% resolution at 10 ohms resistance to 0.24% at 50 k\$\Omega\$.

[ArdWar]

You might be able to eke out a little more resolution and precision from tightening the mechanical design, but imo the way solid-solid sliding contact work don't exactly well suited to form continuous, monotonic, hysteresis-free operation.

[donlisms]

I generally don’t expect a good wirewound pot to be realistically useful at better than 100 parts per turn.  This includes a typical Helipot, Spectrol, or whatever. Some are better than others, of course.

I  reached this conclusion by fiddling with several different pots for hours on end, connected straight to the 34401, and trying to to hit specific values - you know, like you do in the Real World, trying to hit specific values.

The resolution was apparently not limited by the stepped effect of discrete windings, which I believe to be much finer than 100 increments per turn.  What I found in the testing is what I find in Real Life while twisting calibration pots and such.  The value goes high, you back up it goes low, and back and forth, jumping across the target - but not between the same values.  They are essentially randomized. Approaching carefully from one direction helps.

All this made me think the mechanicals are much more important than I had first guessed.  They’re sloppy, with backlash and wiggle, and an important part of the big picture.

So the game is… if you want 0.375 of the total resistance, how realisticly can get there in a reasonable amount of time (not so much fiddling!), and to what effective resolution?  And my design number is 100 parts per turn, 1 part in a thousand for a ten-turn pot.  Realistically, as a resolution I can hit while using the thing.

To get better resolution over all, I’d use the pot as a smaller part of the total resistance.  With a 99x resistor stuck on the end, or split over both ends, I’d get 10ppm from a Helipot in a narrow range.  Realistically.  Maybe slightly better under good conditions. (And with a TC improvement as a bonus.)

I have some HUGE old Helipots with the ball bearings and all; I haven’t opened them, but I imagine there is nice construction it there.  They’re more than three times better.  And will be very difficult to fit behind a panel!

This does give me encouragement that it would be possible to build a really, really good pot, really, really expensively.  But I think it would require becoming a really, really good mechanical engineer and a really, really good machinist first.  I doubt the improvement would be better than ten-fold, and maybe you could get to a point where the winding steps were visible.  Or… something smoother than wire, with all the other requirements met. 

[TimFox]

100 steps per turn is comparable to a linearity spec of 0.1%, which is better than the normal spec for 10-turn pots, but found in premium parts.
It's hard to find the huge old 40-turn pots anymore.
To achieve a better result than that, I have connected two 100 ohm ganged 10-turn pots with a 10 k 10-turn pot in between, as fine and coarse settings (although one can't easily read the result directly) to adjust a voltage reference.
Again, read the "resolution" specs in the Bourns datasheet I cited.

[Doctorandus_P]

100 steps per turn is comparable to a linearity spec of 0.1%, which is better than the normal spec for 10-turn pots,

Now you are mixing up linearity with resolution. Those are two different things.

[TimFox]

100 steps per turn is comparable to a linearity spec of 0.1%, which is better than the normal spec for 10-turn pots,

Now you are mixing up linearity with resolution. Those are two different things.

I wrote “comparable”, not equivalent.
A larger helipot dial has 100 lines per turn, also comparable to 0.1% of 10 turns.