Low ohm trimpot reliability

The obvious question, then, becomes, "how did monitor marketeers miss that opportunity?". ;-)

My highest power laser is about a watt, so I'd have to buy something to do that. Not necessarily awful--I could put the board on a translation stage and line it up at low power.

The main issue is that the optical test needs its own jig, and combining them would be a huge pain and take up a lot of room that I don't have. So a two-step process using a pot to find the correct set point, followed by a trimmed resistor would work until and unless we get some real volume going. (Or so I think at the moment.)

This morning I used a Chinese diamond needle file to trim a 1.0 ohm RN60C resistor to 1.500 ohms to within 0.1%. Took a couple of tries. It's sitting on my bench waiting for the epoxy overcoat to cure, at which point I'll torture test it.

Cheers

Phil Hobbs

I'd forget the 75c ones.

In RF module manufacturing they place wire snippets at strategic places, standing up. When it's all on the network analyzer these get bent until it all fits. Similar to adjusting carburetors during our youthful days.

That's one of the many problems. The other is creepage over time and in a few years wiper contact issues.

Or just get a limited selection of 0.1% resistors, provide two resistor pads in parallel or series, set up a spreadsheet where you key in the measurement and out comes the combination of resistors to solder in. No trimming, no dust, no sealing.

Colour TV standards, I expect. One shadow mask hole = one pixel.

Of course back then marketeers were supposed to drink hemlock rather than cyanide. ;)

Cheers

Phil Hobbs

The wiper is being driven by the collectors of two transistors, so wiper resistance only matters if it gets to be more than 20 ohms or so. The transistors are running with fairly small V_CE, and there are a few pF of capacitance on the net.

It's more the setpoint stability that's the issue here.

Cheers

Phil Hobbs

Interesting idea. Dunno if it would work at these resistance levels, though--I expect that the contact resistance variations are as much as

Cheers

Phil Hobbs

Well, they're Bourns 3361s, which are generally decent.

That would be a possibility too, especially if I use three or four resistors. The Susumu 25 ppm ones are pretty nice, and reasonably cheap.

The other wild card is how R_ee'+R_bb`/beta actually varies with temperature and bias current. It's obviously a function of carrier density at some level. I'll definitely have to do some measurements.

For this version I'm bailing on the matched transistors and using a pair of discrete SiGe ones. If I have to, I can keep delta-T under control pretty well by adjusting V_CE of the two transistors to keep their dissipation matched. (The Early voltage is effectively infinite, so this shouldn't screw up the cancellation.)

Fun.

Cheers

Phil Hobbs

The pot is splitting the current between the 1-ohm compensation resistor and ground. It'll normally be near half-scale.

The compensation is for log conformance, so it's necessarily DC coupled. (The logging behaviour, specifically having the fluctuations split in the same ratio as the DC, is what makes the cancellation work.)

The tempco of the doped silicon is going to be pretty large, so it may be that a few 2N7002s is a better match anyway. We shal see!

Thanks

Phil Hobbs

If you do an actual mtbf survey, you may find that a trimpot is significant - depends on what else is in there.

This is 'noise' compensation, but it looks like the bandwidth goes down to DC in your app, so cap coupling probably not an option. You've already rejected buffering.......

Do you need the full adjustment range? Passing the majority current through fixed components would allow the pot wiper to handle only the adjustment range.

RL

Here is another idea: Place several resistors in parallel, R2R style and with short but nicely exposed sacrificial "arrow patches" (the ones used on ancient computer boards). Measure, have a spreadsheet tell you which ones to disable, then cut the corresponding arrow patches. In the old days that was done with a knife, a sharp screw driver blade or a drill, nowadays a wee touch with a Dremel will do.

The nice thing about the arrow patches was that technically they were "parts", showed up in the schematic and in contrast to masked traces they could be soldered shut in case of an "oops, that was the wrong one".

Of course, a real cheapskate would make a low resistance "PCB resistor" (glorified thin meander traces) with shorting bridges at the turns and then use a small thin Dremel grinding disc to zing away bridges until things match. 10 ohms is a bit much for that technique but it looks like you could go lower. Ok for a few boards but gets to be a smelly affair if you have to produce a lot.

Thanks, good ideas.

I'm going to do a small test board to hold those hot SiGe BJTs and some decoupling components, then stick it on my transistor testing jig to see how the log conformity changes with temperature. It could be a big number--doped silicon's tempco is one or two tenths of a percent per degree, iirc.

So those copper resistors might be just the ticket. ;)

Cheers

Phil Hobbs

a scalpel will do to cut a trace no need to pull out the big dusty guns

-Lasse

It does but often you have to make multiple scrapes, scalpels can snap, and I've seen one very bloody accident where someone cut traces.

The Dremel is faster. I did that on 10 boards a few weeks ago. Laid them all in a row ... tsssing ... tsssing ......... tsssing ... done.

Then you don't have a problem! It's only bad when the "CRV" (contact resistance variation" is an issue. Using a pot as, well, a "potential" device - i.e. the wiper current is "small" - obviates this, whatever the end-to-end resistance. Setpoint variability, while limits (see the old Pease article) does not vary much with end-to-end resistance.

Until a guy performs a "union toss" as a shipping company. I have personally seen some.

Should say "at a shipping company".

Conservatively, a surface mount 4mm pot, single turn, can be set to, and will stay, to within 0.2% of rotation. 0.1% gets a little touchy.

They could, instead, have counted phosphor dots.

I guess cyanide makes them much more creative.

I'm surprised the backlash, alone, isn't worse than that. That's about 1/3 degree.