Low ohm trimpot reliability

It been very long ago that I needed such low potmeter values. Or potmeters at all (I generally never design with them unless a client wants it). But back then the golden rule was to use wirewound pots for anything below 100ohms. Be prepared for sticker shock.

For 10ohm trimpots this is the only one stocked at Digikey:

With knob, single-turn:

Deluxe edition:

Older folks still call those rheostat.

Having been burned by trimpots from a major manufacturer that were purchased from a legitimate distributor, my opinion is: don't use trimpots. (I wouldn't call it "wisdom," since it's more like "prejudice," but we'll settle for "opinion.")

Not many people use trimpots anymore, and you don't want to be the last.

Also, it's better to build stuff that can be calibrated automatically, without opening the box. Even if it costs a little more.

-- john, KE5FX

Wiper resistance is supposedly serious and noisy, which is why pots are preferred to rheostats. But I can't quantify that, so you should probably measure some and maybe whack them a bit.

If you have enough signal, you could use the pot in pot mode and divide down a lot. Your circuit has an infinite adjustment range, and maybe some other circuit would be less sensitive to pot value. The transistor equivalent resistance should be fairly repeatable.

We were just, 20 minutes ago, discussing how much current we could run through a pot wiper. That is also discouraged in ancient lore.

I've used 20 ohm cermet pots to trim the burden resistance of current transformers, with some amount of current, no problems. I think modern cermet pots with multiple gold wiper fingers are pretty good.

Hey, my new Celestron USB microscope just arrived. Here's a Bourns pot.

The microscope is only fair. Amazon said it was 5MP, and they lied.

I've seen Chinese manufacturers grind a V into the side of an SMD resistor to trim an offset. They applied some kind of dope or conformal coating afterward to seal it. I think it used to be a common practice with thin film designs?

Good pots are reliable. And it's a lot easier to turn a pot than design in a dpot or an mdac and write software.

We use pots that work fine up to 1 GHz. It's hard to tweak gain at 1 GHz any other way.

I use a 100 ohm trim pot to set the gain of a wien bridge osc. I've never noticed any problems with the wiper... but haven't looked to hard either.

George H.

Oh, I should add that when I went looking for 10 ohm pots all I found were carbon composite... (I'm guessing noisy, though I never measured.)

GH

About thirty year ago I got popular with the final test technicians at Camb ridge Instruments by up-grading a regular 19mm multi-turn pot to a relative ly expensive Vishay equivalent with multiple gold wipers. It was much easie r to make tiny incremental adjustments with it than with the cheaper part f or which it was a drop-in replacement.

The catch with wire-wound trim pots is that the wiper tends to step from tu rn to turn, and winds up on one turn before it steps to the next. Multiple wipers span two or three turns and the response isn't as lumpy.

Surface mount resistors are trimmed with an L-cut, done with a pulsed laser or an airbrasive knife. The coarse cut - across the width of the resistor - is followed by a fine cut directed from termination to termination.

The process of making the cut changes the resistance while it's going on, so you cut and try, then cut and try again. It's mostly automated.

• I would guess from the git-go, that they would be unreliable because the resistance of the wiper contact would definitely not be stable. One might be able to prove that on first principles.

Possible improvement by factor of four (max):

o---/\/\/\/\/\--+---o ^--------+

(snipped R. Baer's ASCII art)

Thanks. Doesn't work in this case because I'm using the pot as a current sp litter.

Another approach would be to use a 1-ohm resistor with three or four 2N7002 s parallelling it, but I'd have to temperature-compensate that, and it need s to be pretty linear (about 0.1%). OTOH possibly the FETs would track the silicon resistance that I'm trying to cancel.

Spice is nearly useless for this.

Cheers

Phil Hobbs

Well, if I could get a 500-tap, 10 ohm dpot, I'd probably use it, yeah. ;)

I could make a binary weighted series string of resistors and short them ou t with relays, but I'd have to worry about 8 sets of contacts and the trace resistances, not to mention all the additional strays.

The fall-back position for initial production is coarse and fine pots plus a resistor selected at test time. If we sell more than about 100 units, I c an spend some more time on the test fixture, and maybe select transistors. (That would involve measuring the log conformance of a 40 GHz transistor ac curately without making it oscillate--potentially hard to avoid if it's in a test socket.)

Version 1.0 will hopefully be all-analogue, with a factor 10 wider bandwidt h and 60 dB minimum cancellation compared with the New Focus one.

Version 2.0 is slated to have a whole lot of automatic tweaking (3 terms of the Maclaurin series in j omega for each of 3 branches, plus four DC tweak s). That ought to get me another factor of 10 in both sped and cancellation .

At that point the limiting factor is the variation of propagation delay wit h position on the photodiode. I've found one where that's below 30 ps in a

Cheers

Phil Hobbs

It would have a ton of capacitance, which would vary with the setting.

Cameras are specified in Marketing Megapixels. It takes a group of four monochrome pixels with filters to make one colour pixel. No marketeer worth his cyanide would miss a chance like that, so you need 4x the number of camera pixels to match monitor pixels.

Cheers

Phil Hobbs

I have a bunch on order as we speak, ranging from 75 cents to $4 apiece.

The problem is that if I use too large a resistance, the signal currents get phase shifted by the strays. I care about milliradians. Scope probe type RC compensation only works for one pot setting.

The automatically tweaked version is much more forgiving, at least on paper, but several times more complicated. (It can even fix path delay skews up to 10 cm or so.)

I have this tendency to try to go directly to Version 3.0, which is not always the right answer. With these ones I'm trying to start out simpler, while having a compelling performance advantage.

Most of the ones I see quote 50 or 100 g shock for a 1% resistance change, which means the trim isn't guaranteed to survive dropping the box on the floor.

You can get trimmable through-hole resistors, so I might have a two-step approach: trim with the pot to get the correct value; measure the resistance; set the pot to maximum to make it stable; trim a resistor off-line to the required parallel value; then solder it in. That should take only a few minutes per board.

Cheers

Phil Hobbs

Resistance in series with the wiper isn't a big issue in this circuit--the input comes from a good quality current source (two in parallel, actually).

Cheers

Phil Hobbs

I have done PPM zero trims (in NMR gradient drivers) with a half-pitch DIP switch for the coarse and a pot for the fine adjust.

Well, you of all people should have easy access to lasers of all kinds. So you could place regular raw (high tolerance untrimmed) resistors and do an active laser-trim. The first time I did that was in the 80's and I was instantly sold on that technique. Very high precision, no mess and most of all no potmeters. Ok, this was for a product that run in five-digit qties per year. But the rig-up for this laser step was surprisingly simple.