reed relays

Do the math.

What's the Seebeck coefficient of a copper-copper junction?

Unfortunately it is not the MATH that deflects the meter's needle. Nobody's perfect, you know... Your copper is not 100% copper, not monocrystalline, there are oxides/sulfides and other impurities on both contacting surfaces and there is something BEHIND those binding posts somehow connected to them...

When you measure resistance (or whatever) with even venerable 3458A that took all known measures to compensate for all ill effects, you don't just hook the DUT to it and read what it shows. That is if you want accuracy better than that from a $20 multimeter from Amazon or eBay. You choose a range MANUALLY, use a manual trigger after letting everything to get into thermal equilibrium for at least several minutes, throw away the first measurement and take the second. You almost always need to cover the input terminals and DUT terminals or somehow thermally insulate them. The probes' wiring is also critical -- there is leakage that can steal precious ppms by shunting your DUT, triboelectric effects, EMI noise and much more.

Sure enough you don't have to bother with all that crazy trickery if 0.1% accuracy is more than sufficient for you. The math will work just fine.

Just try it yourself. Connect, say, a decent standard resistor to 3458A, put it to auto-trigger and let it run for, say, half-hour observing it from far away or remotely over GPIB. Then come closer, say within 5 feet, and see the last 2-3 digits change.

Some reed relays have a wiping action.

That’s a gradient effect.

You can get offsets in the tens to low hundreds of nanovolts from different copper alloys, but surface films are as near isothermal as makes no odds, so you won’t see them.

Cheers

Phil Hobbs

How does the solder bridging that junction come into play?

How pure is the copper on either side of the junction? Apparently even stress differences across the junction can give you a Seebeck coefficient (if not a big one).

It's hard to do the math when you don't know what numbers to plug in (and you don't seem to).

So John didn't know. Reeds failing open is odd (though a sufficiently high contact resistance might get registered as a open failure).

Anybody serious about the problem would have pulled the failed reeds and found out what had actually gone wrong with them.

My guess would be that the operating coils hadn't got the current they needed to make the relays close. Big liquid helium set-ups can be big - when I was working in the Nijmegen University science workshop I had to clamber around its super-conducting magnet set up some monitoring gear, filling in for colleague who had been missing a bit of his leg since he was kid in WW2.

Remote test gear can be stuck at the ends of very long cables.

But big enough to keep national telephone systems working for a couple of decades.

They have specific virtues and vices. For some applications they can be very handy. John Larkin doesn't know much and seems reluctant to learn anything new.

John Larkin isn't aware that reed relays seal the switching elements inside a vacuum tight glass capsule (reed). Helium and hydrogen can still diffuse in, but can't insulate the switching contacts.

That's one of the attractions of mercury wetted reeds. The mercury film soaks up some of the kinetic energy of the moving contact. It sloshes around the end of the moving reed, but not for milliseconds.

Zilch. You can have N extra metals in series and still get a good measurement from a single pair of junctions, provided that both ends of each extra metal are at the same temperature.

A normal temperature solder joint between two otherwise isolated wires is going to be very very nearly isothermal, so you’d never see it. Terminal blocks, PC boards, and such are a different matter.

Cheers

Phil Hobbs