Low ohm trimpot reliability

Thanks, I bookmarked it under "Useful Parts". Although I haven't used trimpots in any designs for the last 35 years unless a client absolutely wanted me to.

Yup, if the bandwidth is narrow enough, or if you're just trying to push an RC rolloff out in frequency a bit. Series peaking is easy to do, and gets you a factor of 1.4 in general. A T-coil can get you 2.8x.

In this case, spice says that the negative cap gets me 20 dB improvement in cancellation at 1 MHz. We'll see.

Cheers

Phil Hobbs

I have seen high-end lab equipment where they didn't skimp on trimpot quality become unreliable. Like unexplainable offset jitter that reacted when tapping on or near a trimpot. My usual quick fix if in a hurry is to mark the slot position with a fine-point Sharpie, turn the pot back and forth several times and then back onto the marker.

The last case was an old analog scope from the "We'll probably never need this anymore but let's keep it just in case" heap. The whole thing had the shakes. We had the schematics but there was no time. So I did the Sharpie & turn routine on just about any potmeter I could reach in there. That fixed it. At least for that day.

An Wild-West method, used by a friend decades ago: He normally would have needed a variable capacitor at one spot on his antenna. But in that area anything mechanical will corrode soon and also it would require a scary climb. So he provided lots shunt wires. The alignment was performed by blowing them away one by one using his rifle.

I think I mis-explained my idea anyway. In differential exponential converters where the tail current is controlled by an op amp current source, it's possible to reduce the effects of Rbb by injecting a current directly proportional to the control current back into the control input summer, taken from the voltage developed between the output of the op amp in the tail and a resistor from the diff pair emitter junction to the op amp output.

Don't know if anything like that is possible here, as the long tail current is controlled by the reference current source.

Technically, yes, but I've seen three-terminal ones in meeting rooms and schools being referred to as rheostats even though they were wired like a potmeter to be able to dim to zero. Probably from the days when power just came out of the next nuclear power station and was considered "free". Newer ones had a core and were essentially variable-tap transformers but they still called those rheostats.

I see--that helps with the effect of beta nonlinearity on the Rbb' compensation.

Unless I still don't understand, it seems like that would only work when Rbb' >> beta * Ree', though, because it doesn't do anything about Ree' For the first go, I'm using magic discrete transistors: Rbb' = 6 ohms, Ree' = 0.2 ohms, beta ~ 600, VAF ~ 1000V, f_T ~ 40 GHz. There the Ree' term is 20 times larger than Rbb'/beta. (I've been doing a bunch of lawyering for the past few days--super boring stuff with a lot at stake--so I'm a bit fried ATM.)

Cheers

Phil Hobbs

Here's the setup; the idea was originally devised to compensate for the Rbb' error in expo converters for analog music synthesizers, which caused the current output from the exponential converter to deviate from the 1V/octave standard and caused errors at high frequency (this schematic is probably from the early 80s, but a similar scheme using modern devices is still used today in commercial products)

Another example schematic with a better explanation:

AoE3 has a table of low-noise bipolars, p 501. Looks like H+H spent a huge amount of time on that (or have cheap scut bunnies.) Rbb's range from 760 ohms (BCX70, my gumdrop standard) to 1.24 ohms (ZTX951). That last one has En of 0.21 nv/rthz. Of course it has 50x the capacitance of my BCX70.

Your magic transistor is not in that list.

Mine's a BFP650, the big brother of the 640. The datasheet spice model isn' t the same as the downloadable one.

If I were Win and Paul, I wouldn't put that one in either--it's prone to os cillate at frequencies far outside the range of most test gear. You probabl y remember that circuit of mine that oscillated at 14 GHz--I measured it by watching its offset voltage vary sinusoidally as I moved my hand towards t he board and away. Seemed to go through a complete cycle in a bit over 1 cm , so around 14 GHz by my "manual wavemeter".

Sure is a nice part if you use the right beads, though.

Cheers

Phil Hobbs

Low ohm pots are a real problem. For one thing, if they're low-ohm, they're probably not being used as pots at all. Use a power rheostat.

That said, I haven't had the need for a pot, except in a throw-away design, in forty years. There's always been a better way. The link idea is much better than a pot. It's probably easier/cheaper to set, as well. I could fairly easily be fit into an ATE environment.

I sure hope it wasn't an AR-15. That would scare too many, here.

As I said in my very first post, I'm using it as a current divider. Three t erminals.

So you're an expert in 1970s pots, very good. So was the guy who told me ab out them. The '70s are a long time ago now, which is why I was asking.

So you're saying that it's easier to tweak something by cutting and resolde ring 8 to 10 traces N times in a binary pattern than it is to turn a pot ba ck and forth? You're cracked.

Cheers

Phil Hobbs

Current through pots is also a real problem. I'd do some serious checking with the manufacturer before thinking about using a pot, here.

They really haven't changed all that much.

No, I'm saying that an ATE system can switch contacts and with a little programming tell you which one to cut. No farting around at all.

One difference is that in the 1970s there was no other affordable choice in mass-market designs, so trimpots were very widely used.

Now they're weird legacy parts. Sort of like those huge "computer-grade" e lectrolytics that used to be everywhere. You can still buy them, but they' re expensive, and manufacturers have had no incentive to improve or even ma intain their quality over the past 20-30 years. Instead, they have every r eason to shave every penny they can. If you've bought any of those huge sc rew-terminal caps in the past 10 years or so, you may have found (as I have ) that it's a good idea to test them before installation, because the manuf acturers apparently aren't. From their point of view, they're doing you a favor just for keeping them in their catalog. They're no longer competing on quality, there's no high-volume business to be done, and as soon as the last guy who knows how to make them retires, that'll be that.

Same deal with trimpots.

-- john, KE5FX

So they're not what the cool kids use, I geddit already. Got a better suggestion for the 10-ohm level?

Cheers

Phil Hobbs

Modern sealed cermet trimpots are cute and cheap and reliable. There are not many alternate ways to cal the gain of a DC-to-500 MHz amp.

JFET and a DAC?

-- john, KE5FX

It's been too many decades since I actually designed anything, but the approach I would take is a place a number of resistors and jumpers surrounding a pot, such that snipping a jumper simultaneously moves the center range of the pot and reduces its sensitivity. Then you could have a cal procedure whereby the position of the pot wiper determines which jumper to cut next. Eventually the system would end up calibrated and any residual errors from pot drift would be minimal. At this point, I'll let the actual designers here comment on how this architecture could actually be implemented.