reed relays

Thermals come from the contacts and the connecting parts. Latching relays are no different, just different mechanics.

Do they specify thermal EMF for those latching relays? Why would stupid metrology grade instruments' designers use VERY expensive COTO reed relays with whopping <0.5uV thermal EMF instead of those about zero thermal jewels that cost $1 a bucketful?

Yeah, they just want to sell them as a cheap junk instead of charging at least an order of magnitude more... And they don't know that, e.g., $1 Susumu resistors are much better in real life so they buy oil-filled VHP101 and such from Vishay for $100+ apiece...

Thermocouples are very crude devices, good to +/- 1 degree Celsius or such for the low temperature ones. The high temperature ones are worse. They are not anywhere near "precision".

A good Pt probe easily gets to .01 degree precision. A [limited range] thermistor is even better, .001 degree is achievable. An old Hart/Fluke Tweener 1504, e.g. has .003 degrees accuracy typical with .0001 degrees resolution when used with a good probe. An SPRT sibling, 1502, has .006 degrees accuracy with .001 degree resolution. Those are OLD instruments.

Kaye all-in-one (you only provide power and talk over RS-232 with it) IRTD-400 has a range of -196 to 420 degrees Celsius and has a guaranteed accuracy of .025 degrees over ENTIRE range with .001 degree resolution. Much better farther from extremes.

This is where all those fancy components are used. For a stupid thermocouples with "cold/warm/hot/too hot" measurements [almost] ANY signal relay is OK.

With a latcher there is no continuous coil power dissipation heating everything up. They switch in 2 milliseconds. We have a cute coil driver circuit that multiplexes nicely.

It takes a thermal gradient, a heat source, to make voltage.

Beats me. I'm just an engineer.

SSRs are good about that. But if the figure of merit is Ron*Coff, which is technically a time constant, real relays are vastly better than SSRs.

And DPDT is handier than SPST.

If you have the wit to understand them, they tend to be pretty reliable. Testing the parts is as much about testing your own understanding of the data sheet as it is about testing the parts themselves.

Not a point that John Larkin wants to think about.

You do have to calibrate each one to get that. I've used +/-0.2K Betatherm interchangable thermistors (and Yellow-Springs offer even tighter spec parts at ten times the price).

Platinum resistance thermometers are much more stable, and 10 microdegrees is attainable (with careful multipoint calibration and linearity correction).

Metrology does get written up in peer-reviewed journals like the Review of Scientific Instruments. Their refereeing isn't all that reliable and I've published a few comments there when the refereeing has let them down badly, but the good stuff can be very good indeed.

Larsen N T 1968 Rev. Sci. Instrum. 39 1–12

Sadly, the relay operating coils aren't the only components that dissipate power on a printed circuit board. There's always some kind of thermal gradient across a board.

John Larkin does claim to be an engineer. A lot of what he posts suggests otherwise.

It's called "vacuum welding". Soft metals - like gold - diffuse into one another if left in contact for long enough.

It happens faster if the temperature is higher. We bonded gold-plated electroformed copper grids to gold-flashed ceramic by squeezing them together for a few minutes at 400C. That's called pressure welding.

Sure. However, there are better thermistor PROBES, with very precise calibration and VERY stable. They are NOT interchangeable and have limited measurement range but they are very precise within that range.

I have Amphenol calibrated Fluke 5640-D probe with calibration certificate thicker than an average manual. It has .0015 degrees Celsius accuracy from 0 to 60 degrees and .005 degrees per year drift. Paired with Hart/Fluke Tweener 1504 it is accurate to that .0015 degrees.

Good Thermistor Standard probes are extremely accurate and very stable, with very low drift. Calibrated Tweener 1504 matches that accuracy and very easy to calibrate with a set of calibration resistors made of Vishay VHP101 resistors. It is probably the ultimate precision combination for a limited temperature range. Can be checked with a TPW and Gallium cells at 2 points within their range.

PRT does not allow for such precision because its thermal coefficient is musch lower so it is difficult to measure to the precision that thermistor standard gives.

You'll need a VERY precise/stable current source and very high resolution and low noise ADC to get that precision from a PRT -- the resistance change for 10 microdegrees can be easily less than your noise floor... Much easier with a thermistor with its crazy sensitivity but lower temperature range...

Exactly. A sensitive instrument even senses your breathing from several feet with its pure copper terminals shorted with pure copper. When hunting for single ppm you have to keep your entire setup within closed thermally insulated box. Or at least keep tens of feet distance from it and control it remotely. A slightest draft can set you off by tens ppm. That is copper on copper, CRIMPED pure copper wires.

Very noticeable with, e.g. HP419A null voltmeter on the most sensitive 3uV full-scale range with a shorted input. It is ANALOG so it is fascinating to see how the needle moves following your breathing at something like 5 feet from it. When AC starts in the room it pegs the needle :)

Platinium resistance thermoneters are made from very pure platinum, and just platinum, and have a positive temperature coefficient. Thermistors are sintered metal oxides and the useful ones have negative temperature coefficients. That introduces problems.

You don't use an ADC. You use an AC bridge built with ratio transformers

"Coaxial AC Bridges" By B P Kipple and G H Rayner, ISBN 0-85274-389-0

is a useful text on the subject. 10:1 ratio transformers (with eleven taps) are accurate to about 0.1 parts per million, so it is worth stacking up six of them. 2:1 (bifilar wound) parts are accurate to about

The winding techniques get complicated. Rope windings are easy, but you can do better. Strap windings offer a lot advantages but are hard to realise. Ribbon cable could offer some interesting options - not a lot of copper per unit volume, but it is easy to buy in long lengths.

He used an AC-exited bridge with ratio transformers. Like Kibble and Rayner he was a national standards laboratory employee.

Brian Kibble is the "Kibble" in the Kibble balance that now defines the unit of mass.

We did the same to bond gold beam leaded silicon devices to large ceramics with gold paths back in the 70's and 80's. These large hybrids are still used in telecommunication central office and #4 Toll switching centers. Here the process was called thermocompression bonding. It took less than a second using a heated tool and pressure.

That's crazy.

When hunting for

There are shielded versions available, mainly to achieve high density interference free mounting, but the benefit extends to any nearby hazard or victim component.

And those application-specific problems can appear regardless of the manufacturer's advertised low contact EMF.

EMF problems that people are mostly seeing are the thermocouples at the junctions of component leads/posts and the copper traces on the PCBs. This is again symmetrical, but proximal and asymmetrical heat sources, even the relay coil heat itself, can disrupt the cancellation.

It's hard to believe you can have low signal failures with the reeds advertised for instrumentation applications. There are all kinds of things that destroy the performance, and it's mostly bad manufacturing or mishandling: (from AI) Micro-Cracks in the Glass Seal: If the glass-to-metal seal gets a microscopic crack during manufacturing, shipping, or soldering, the inert gas will leak out and oxygen will leak in. Once air is inside, the blades will begin to oxidize, causing the classic low-voltage connection failure. Organic Outgassing Contamination: If the manufacturing facility is not 100% sterile, trace amounts of organic compounds or cleaning solvents can get trapped inside the glass capsule before it is sealed. Over millions of cycles, the physical impact of the blades can cause these trapped organic molecules to break down into a thin polymer film right on the contact point, blocking low-voltage signals. Plating Flaws: If the sputtered rhodium or ruthenium plating peels or blisters off the underlying nickel-iron blade, the raw base metal is exposed. This can cause erratic contact resistance. Mechanical Shock Damage: Dropping a reed relay can slightly bend or misalign the internal blades. If the magnetic field from the coil can no longer pull them together with enough physical force to flatten out the microscopic contact points, contact resistance will spike.

What was the failure? Open contacts? Seems the main problem is the excessive research required to select the most appropriate relay for the application, and to verify the attributes of the received batch before use. You can hold back a representative sample for long-term testing in anticipation of a future recall.

There's nothing crazy about it. The micro-climate inside a closed room reacts to small temperate difference (and your exhaled breath is warmer than room temperature) by creating convection currents which act to reduce the temperature differences. I once had to put draft shields (cylinders of paper) around a sensitive weighing head when I was testing it in the lab.

We got the advice from the Welding Institute in Cambridgeshire UK.

Trifilar windings and quadrifilar windings have the same virtue as bifilar windings - each wire sees an identical environment.

Put seven wires together into a rope and there's a central wire which has a different environment than the others.

If you used it for different job - carrying the current that sets up the magnetic field in the core - the other six wires should see identical magnetic environments and generate identical induced voltages.

A heavier central wire could be wrapped with a single layer of more numerous sensing wires. I've never seen it done. It ought to be obvious to those skilled in the art but there aren't all that many of them.

Presumably you are reacting to John Larkin's original post.

He won't know why he thinks his reed relays failed. He probably managed to try to make them break contact while in series with an inductor, which can weld the contacts together.

The usual failure mode is that the contact resistance gets too high. Each make and break roughens up the contact area and eventually the conducting area present when the contacts are closed is too small. They typically offer 10 million operations

Mercury wetted relays don't have that problem, and last about ten times longer than dry-reed relays, but their contact resistance also rises with repeated cycling.

The users told me so. Thus was a big liquid helium temperature and level measurement system at the JeffersonLabs/CEBAF electron accelerator. The reeds tended to fail open.

Reeds are absurd. Some people seem to have obsessions with them.