Low ohm trimpot reliability

There shouldn't be any backlash in a single-turn pot, but there might be some stick-slip energy storage that causes the setpoint to move when shocked. I'm going to try it several ways when my Digikey order comes in.

I expect that vibration can be basically solved with a drop of glyptal, at least for the single-turn case.

Cheers

Phil Hobbs

Bill, you have some time, we need part numbers.

I can't guarantee that that's the part, but it's expensive enough, and the data sheet claims exactly the kind of smooth trimming that Phil Hobbs is after.

Farnell/Element14 doesn't stock it down to 10R, but Newark might - their url tells me I'm not allowed to log onto it - and you can always ring/e-mail Vishay.

Bill Sloman,

Even with single-turn pots, there is usually a spring on the wiper. That'll store energy and also cause a problem with vibration, glyp or not. Pushing pots beyond what they're designed for isn't a good idea, IMO. Pots aren't a good idea.

If a pot-based circuit will work with an appropriate resistive limiter in series with the wiper, and no appreciable temperature rise in the (admittedly marginal-range availability) element, there shouldn't be an issue.

Cermet trimmers in the 10-50 ohm range were common in video/cable transmission analog stuff.

RL

There is also slack that can't be controlled. In the bushings (or drill holes where the shaft runs in) and sometimes in the wiper assembly to shaft interface. Potmeters aren't made like Swiss watches. Well, some are but they have mil-spec price levels.

Quick hack idea to reduce effects of the base spreading resistance: since the effects of Rbb are proportional to the emitter current, couple a voltage proprtional to it from the control current input (resistor junction at base of Q2) through a diode connected MAT-whatever to rhe input of summing amplifier/integrator A2. Then you could use a larger Trimpot (on the order of 1k) to null it.

Multiturn pots have no better settability than singles, but are bigger and cost more.

Hi

Sorry, no hard evidence one way or the other, but I also remember hearing about avoiding trimpots in such places. Unfortunately I can't remember where exactly I've seen it (long ago and I was basically a teenager then), but I vaguely remember that trimpots had problems with DC current being drawn off the wiper. Something forms or degrades at the contact surfaces, making the contact unreliable and the pot "scratchy" much faster than the typical ageing under "normal" use.

The advice was basically either to avoid a DC component on the whole pot altogether or at least AC-couple the wiper so that no DC current flows through the track-to-wiper contact surface.

This may be anecdotal evidence, again, I have no hard facts to back it up. The advice somehow "stuck" in my memory (better safe than sorry).

A parallel array of resistors (R-2R style) with small MOSFETs to switch them might be an acceptable alternative...

Dimitrij

Joerg and others made some suggestions about using shorting links or series/parallel pairs of resistors chosen at test time. Do you have any other ideas that will work down in the low ohms? "Pots aren't a good idea" unless they're actually the best thing for the particular job.

Thanks

Phil Hobbs

The wiper is being fed from two current sources whose sum will never exceed 10 mA, so I think I'm probably pretty safe.

Cheers

Phil Hobbs

It sounds like a trimpot is ideal for your problem.

Lots of people have a "never trimpots" prejudice, which is silly. Sometimes a trimpot is absolutely the best solution. Screwdrivers are cheap and easy to use.

There's a lot of lore left over from the days of cheap, unsealed carbon-track pots that tended to have erratic wiper contacts. Remember Spra-Kleen, the pot fixer spray?

Good sealed cermet trimpots work great and are very reliable. We've used about 80,000 so far, no problems. The downside is that they aren't settable to 10+ bit resolution, and you have to open a box to get at them. Soft calibration is generally better, but a pot is ideal for something like Phil's circuit, or setting the gain cal of a 500 MHz photodiode amp that has no digital interfaces.

The resistor/mosfet array will need lots of parts, logic, nonvolatile memory, a programming interface, and software. As compared to one pot and one screwdriver.

A rheostst is a 2-terminal variable resistor. I still use that terminology.

Bourns 3314G are 82 cents, surface mount, work great.

Thanks. In a differential pair it's a bit more complicated, because it depends on the current splitting ratio. The low impedance is needed for three reasons: simplicity, noise, and (especially) phase shift. A couple of picofarads times 1k is 12 milliradians at 1 MHz, which would be death and destruction in this circuit. At 10 ohms it's only 120 microradians.

One way of generating the compensation voltage is shown in Fig 12. I'm actually doing it slightly differently, with a current mirror on I_signal instead of I_comp, which allows me to put both currents into the base of one side.

The circuit is similar to Fig 3, only with the photodiode bootstrapped very efficiently, (*) allowing a 300- to 500-ohm resistor in the emitter of Q3 (which is actually a BFT92 or maybe an Intersil HFA-series PNP array). That makes it easy to make a copy of I_signal with another PNP, and also fixes problems due to the low Early voltage of the PNP.

I have to buffer the voltage on the anode of D1, of course, and in the process get rid of the buffer's input capacitance, which would destroy the bootstrapping otherwise. It's actually the first real-world application of a negative capacitor that I've ever found.

Cheers

Phil Hobbs

(*) I've been making a fairly deep study of the best way to do bootstraps. I have an inexpensive 3-transistor circuit that gives me gains above 0.999 and output impedances of a fraction of an ohm. That gives a lot of flexibility that I otherwise wouldn't have.

Assuming that the tempco mismatch isn't too bad and that the trim survives shipping and graduate students, I think so too. We shall see!

Thanks

Phil Hobbs

I designed a language lab console once, a big panel full of rotary switches and bus wiring. It had some high-frequency crosstalk and the department head was very picky about acoustics. The fix was negative capacitances to null out the crosstalks, tweaked with pots of course. Worked great.

Fun.

This one is an ADA4817-1 unity gain buffer, with a fairly heavy load (300-500 ohms). It has a pot in series with the output (inside the loop), with the wiper going to the noninverting input via a 4 pF cap.

That works fine at all frequencies I care about (up to 1 MHz at high accuracy) but rolls off soon enough to prevent instability, which is the usual problem with negative caps (the phase lag in the amp turns negative capacitance into negative resistance).

Cheers

Phil Hobbs