Precise voltages from imprecise parts..

On a sunny day (Sun, 19 Apr 2015 21:56:17 -0700) it happened Robert Baer wrote in :

I did some calibration at Tek, Tek used one turn trimpots. Drop one from 10 cm (4 inches for you 'merricans) on the table. then recalibrate again.

I worked in a big scientific setup, 10 turns was the standard everywhere.

I don't know who pease is, and not sure it is worth knowing. Pipo the clown said a red nose is better.

Is your nose red, Jan?

Perhaps he meant that if you need the resolution of a 25-turn trimmer, you probably designed it wrong.

Pease is always worth reading, even if one doesn't always agree with him. His beef with multiturn trimpots was stability under vibration and temperature cycling, iirc.

A drop of glyptal will keep a single-turn pot stable, but multiturn ones have backlash and stick-slip energy storage that can't be entirely controlled by holding the adjustment screw still. How important that is in a given situation will vary with the quality of the pot, the environment, and the required stability. Even lab gear usually has to be shipped someplace, and shipping can be brutal.

The other issue (which I don't recall if Pease raised) is the mismatch between the tempco of the pot and the other resistors, which frequently makes the apparent precision advantage of a multiturn adjustment mostly illusory.

Cheers

Phil Hobbs

Baer

t Baer

nominal

A to

n?

V if all

l value.

are

t of 25V

math..

s and

he way.

in

I don't recall ever designing in coarse and fine trimpots, but there was an occasions when I up-rated a trimpot from a 25-turn Bournes to a 25-turn Vi shay part, and made the final test technicians a lot happier - the Vishay p art had a multi-element brush, and with it they could get the resistance ex actly where they wanted it.

The Cambridge Instruments EBMF 10.5 electron-beam microfabricator reflected more than a decade of continuous development, so it wasn't so much designe d as evolved. When we designed what would have been it's successor - if it had ever made to production - we didn't need that much precision in the bea m steering electronics, but we certainly need it in the 10.5 .

The very first commercial project I was involved with was based on my boss reading about a "window comparator" configuration in a magazine. He thought these were great and should be used everywhere. So for this pneumatic valve sequencer thing there were about 8 window comparators with 16 of those trimmers sticking out of the side. We used thousands of them. The valves in fact needed to be in a strict sequence for the thing to work but the window comparators let you put the on/off times anywhere you wanted. So it was very hard to get things setup right. If you wanted to adjust the "test time", say, you had to adjust the "test time end" trimmer a couple of turns clockwise (or anticlockwise, you could never remember which because it was backwards). Then you had to adjust 5 or 6 of the others too, by the same amount, to keep things in the right sequence. The instructions read like "turn the test time end trimmer 30 turns clockwise then 10 turns anticlockwise. Turn the test time start trimmer..."

We supplied one of those little yellow plastic trimmer tools with each unit. They were double-ended with an open screwdriver blade in the unused end. We used to machine off that end otherwise you got a hole drilled into the palm of your hand from all the "trimming".

So anyway I now have a long-standing aversion to trimmers...

Of course he should have done it the right way, with a chain of one-shots, and saved half the trimmers.

;)

Cheers

Phil Hobbs

Valid for 3/8" multiturn trimpots, less so for 19mm trimpots, where the brush travels in a straight line along the fixed resistance.

There's more than one sort of multi-turn trim-pot. It's unwise to generalise until you've specified which one you are talking about.

A ten-turn precision potentiometer can be seen as a multi-turn trim-pot, but you wouldn't open it up to lock it with a drop of glyptal.

Trimming 15ppm/C 0.1% resistor with a 100ppm/C trimpot usually works pretty well, because the trimming element is typically 0.3% of the total resistance, and doesn't contribute much to the over-all temperature coefficient.

You can get into more trouble if you are silly enough to use the trimpot to fill in between the E96 grid - E96 values are about 2% apart, and trimpots are typically 100R/200R/500R,etc but a parallel or serial padding resistor can rescue most situations.

That was a later version! Ye Olde 4538. A slightly less insane design, with the important times brought out to some nice 10-turn dial pots. Then I revolutionized the product line with a state-of-the-art

Trimpots are the signature of a novice designer. My designs typically have no trimpots at all. However, I do use lots of settings registers and calibration files. ;-) I guess I'll always be a novice, but I've got better at hiding it...

Jeroen Belleman

Me too, for the last ~20 years.

You can do a lot with software correction too, software auto-zero, auto scale and so forth. I have had a single precision resistor that ends up getting switched around calibrating everything else.

HP3458A has a single precision resistor and voltage reference. All the voltage, current and resistance ranges get auto-calibrated to these two built-in standards, to within 1 ppm IIRC.

Dpots and suchlike are good for comparatively slow stuff. There are situations, especially in high-Z and high speed work, where a tiny trimpot is just the right medicine.

Cheers

Phil Hobbs

Sure thing. But if you're only trimming 3000 ppm, you probably don't need 25 turns (125 ppm per turn!)

Cheers

Phil Hobbs

I think he argued that a multiturn trimpot doesn't even have better setting resolution than a single.

A DAC or dpot or a software cal factor is a better way to trim. If you must use manual trimming, one good way to do it is to use a dipswitch or hex rotary switch for the coarse trim (switch some really good resistors) and use the pot for the final PPMs.

Well, maybe that's my beef with them instead of his. ;) I'll have to go back and check.

I agree in general, for sufficiently slow stuff. Dpots and MDACs are very disappointing above a few hundred kilohertz, though, due to the nasty tradeoff between the ON resistance and OFF capacitance of the mux.

They're great for out-of-band adjustments such as tweaking a power supply, nulling the offset of an inverting stage, or controlling VCAs and so on.

As I said to Bill, if you keep the adjustment range small enough that the pot's tempco doesn't bite you, you probably don't need 25 turns.

Cheers

Phil

Indeed. Hewing strictly to rules of thumb when it makes no sense in the circumstances is a mark of a novice designer.

--sp

I you used a trimpot you probably designed it wrong.

These are good well above 1 GHz...

but you need a special tiny ceramic screwdriver to set them. The resistance path is short and direct:

I figure that a small 1-turn pot has a usable resolution around 0.2%. A little wider makes it easier to set in production.

Using it as a pot, not a rheostat, makes the tempco much better.