Once in an article... replacement of trimmers

Such as using a dpot?

Dpots are nice for slow things, but even the very fastest ones (AD8403 @

OTOH, a nice small trimmer can work up to above 1 GHz.

Cheers

Phil Hobbs

That's where I use PIN diodes :-)

So far in my whole career I have not designed any trim pots or trim caps into production-ready circuitry. Ever. Not one.

Sonnich, the main reason is reliability. Talk to a seasoned service tech and they'll likely tell you that trim pots are in the top ten when it comes to field failures. Right up there with switches, connectors and such. Because they are basically mechanical devices and thus subject to being compromised by corrosion, shock and so on.

Never ever made transistor audio amp with bias trimpot ? :-) The valve-era or early transistor-era electronics used plenty of trim caps and adjustable inductors.

Manual adjustment operations are expensive for production, too.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

For RF attenuators, sure. For tweaking CMR in a diff amp? Don't think so.

Horses for courses. My stuff sometimes lives or dies based on its AC CMR in weird situations, so the occasional tiny trim pot is just the thing.

All true, but you still use power switches and input connectors. Why? Because they're necessary, and so are trimpots, sometimes.

Cheers

Phil Hobbs

Nope. I did design linear amplifiers for higher frequencies and all of them were afforded electronic bias adjust. If that's possible at megahertzes I assume it should be a piece of cake at audio. Bias trim pots are so 70's. It's like points on cars, the last vehicle I had with mechanical points was a 1969 model Citroen 2CV :-)

At my first employer they called that "rubber engineering".

Yup. You also need personnel with higher training levels, meaning even more cost. One VP of Production summed it up like this: "George, ideally we'd like a test box from you guys that has only two big lamps for output. A green one for "good" and a red one for "bad".

Assuming the tester is still working properly. ;)

I'm not a big fan of having folks with only burger-flipping experience doing production test on my gizmos, and anyway, it isn't much harder to "turn pot R1 until the LED goes out" than to plug in the power.

Cheers

Phil Hobbs

Anything that can be adjusted, can (and will) be misadjusted.

-- Cheers, James Arthur

Many ways to do that. If the voltage swing at that node is miniscule a tiny FET or JFET is suitable. I like to servo them so the tempco in the adjustment goes away. Unfortunately many monolithic FET arrays suitable for DC to super-high speed have gone towards boutique status. So not too much servoing lately.

The other option is to use fast DACs, the ones where you can directly tap their R2R network. With a simple diplexer you can let that "glide over" to a PIN diode (or varicap if the last cent counts) if the DAC doesn't jibe with higher frequencies too well. It's like pulling in the flaps of an aircraft once enough speed is achieved :-)

I couldn't sleep well if I knew that some trim-pot equipped electronics would be thundering into the sky with force excursions up to 60g and tons of shaking, and the trim pot failing could result in a fireball. Or flying on an aircraft that just had a not so wrinkle-free landing attempt and does a go-around with its pilot drenched in sweat.

If it's not mass produced and used in a not too hostile environment they can be ok. I'd still try to avoid them.

Power switches and connectors are required "evils" because the user requires them. I don't like them but can't do anything about it. One can do something about trim pots though. So far I have never encountered a situation where I couldn't.

The most striking example was a board with four ganged ADCs to do a conversion at an effective 800MHz or so (I think, it's been almost two decades now). Back in the days when this was really pushing the envelope. Extremely critical with respect to noise, offset, gain deviations, phase skew. Their old board had something like 30 trim pots and caps. It was a nightmare in production and also responsible for the lion's share of field returns. So I designed a new one for this client, zero trim pots and caps. A few years later their field service manager said to me "George, you made us the Maytag men" (remember the ad back then?). He actually had to let a guy go, no kidding. Kind of made me feel bad but he did find a nice job in production and his family liked it that he came home every evening.

Yes, except I jokingly got a complaint that the virtual pass-fail lights should also read "pasar" and "defectuoso" :-)

It did have a keyboard, screen, statistics, and such, for when the foreman (was actually a woman) wanted to see how things are going. Also, the name of the operator had to be entered at shift start for regulatory reasons. Results were spooled onto a server via a LAN connection and backed up. One could even see a "streaming" view from computers outside the clean room.

Me neither but this changes majorly when you design something that rolls off the conveyor belt at several thousand units a month in China, for the next decade or two.

Sure, I've done that too. But it isn't always suitable in a front end, especially when it needs to float.

All that stuff has a lot of capacitance, and getting decent linearity is very difficult. A small trimpot is way under 1 pF, and as linear as you like.

Never done one of those myself. Probably it isn't your daily grind either. ;)

I don't use them often either, but 'never' is too dogmatic.

Trims can often be avoided, especially in low-impedance situations like that. CMR trims in front ends can be harder, even though they don't have all the thundering buffalo.

Cheers

Phil Hobbs

A few of mine had to float, by law. Or by UL60601, to be precise. The

Until the unit takes a hard fall, then it's non-linear for a split second and after that stuff don't work right no more :-)

A BAR50 is about 0.2pF, has a nice long carrier lifetime around a microsecond and costs 5-10c. Tough to beat.

The strange thing is, even for European ones they often charge more in Europe than over here. Then again, that's the same with Volkswagens.

Lately it has migrated into that direction a bit.

The problem with trimmers is the customers fiddle with them.

"We wound every trimmer from end to end and it still doesn't work"

For one older board I - carelessly - gave a LCD contrast trimmer too much range. About 50% of the field returns were due to a "blank display".

I mean float at signal frequencies. A few pF to chassis wouldn't matter much even with a defibrillator.

Okay, so some of my gear might not survive Dave Jones throwing it off the top of a dam, or driving over it in his car. So what? You break it, you pay for it. Making a 300-milligram trimpot break loose purely from inertia would require quite the shock--enough to put a good sized ding in the case, at very least.

Nice enough part, but you have to hang junk on the ends of it, which have more capacitance than that, and it isn't DC coupled--in fact, it won't work properly below 10 MHz. Again, an RF-only solution. A nice little trimpot is good medicine for wideband DC coupled things sometimes.

Cheers

Phil Hobbs

le,

ked

or

oo

d,

you

Or

We use them, production trimming is done automatically with a motor controlled setting

In our case we need to have extreme low cost and combined with sub uA current consumption, you really have no other choice

The products are not space stuff, even though the trimmers are selected for high G performance.

Regards

Klaus

Red glyptal is good for that--it doesn't stretch, so if you tweak the pot, the varnish cracks in a nice visually obvious way.

Ouch. I've done that sort of thing too, back in the old days--a trim cap that would keep the oscillator from starting if not adjusted right.

That's the main problem with trimmers IMO--people use them as a substitute for thought. I haven't used one in a customer design in some years, but I can think of a couple coming up that might need one or two. One is a pHEMT FET probe for a DNA-sequencing experiment, and the other is a glorified laser noise canceller. (Neither will be a consumer item.)

Cheers

Phil Hobbs

That'll be a bit more of a challenge. Can be done though, for example driving with an optocoupler that has two photodiodes, using ine on the drive side feedback. You can transfer pretty acurate DC levels that way, without any fancy modulation. I usually prefer magnetic signal transfer, via carrier.

What I really liked was active laser trim but I guess that has somewhat fallen from grace with the demise of alumina subtrate hybrid production.

Trim pots aren't much of a concern for scientific equipment. In my line of work "You break it - you pay for it" generally isn't working. Customers will call my clients and tell them "I dun nothin' wrong, honestly".

The junk is essentially just a current source. Or voltage source plus resistor. It's not intended to work below a few tens of MHz. That's why I mentioned the combo, if DC to almost infrared is required one can use a DAC or digital pot for DC to xx MHz and diplex over to the PIN above.

But yeah, there is some complexity to this kind of solution.

It might be worth looking into a laser trim machine. I've designed stuff around them extensively in the late 80's but since hardly anyone does alumina hybrids anymore these days there must be a glut of those machines on the used equipment market.

It was amazing. Load the alignment program into the connected PC, load a rack full of modules onto the hopper mechanism ... phsssss ... bzzzt ... click ... hissssssssss ... bzzzt ... module moves into the finished goods hopper on the right while the next moves into the machine. At one large factory I stayed on a little after their engineers went home, together with the lead engineer who said "You've got to see this". The only other guy I could see was the guard and his dog, even the lights were almost out to conserve energy. And all those machines kept hissing and humming.

"I've tightened'em all real good, Sir, and that dang thang still don't work"

Probably also good as a seal against dust and moisture. But then the usual will happen, they'll steadfastly deny to ever have touched it. "Must have been someone at the factory, or maybe an earthquake cracked it".

For the price of a dpot, you can put in a tamper-evident shock sensor, e.g. a third-overtone HC-18 clock crystal on the MCU. ;)

No, you can't, not if you need a few picofarads or below, or if you want a clean step response. Plus it has shot noise. Laser trimming would be nice, lots of times, but I haven't worked anywhere that was possible since the 1980s. You can get trimpots down to 2.8x2.5x1.5 mm, with

Cheers

Phil Hobbs