Kelvin SMD connection

Thanks, I've seen quite a few of those myself. ;) Copper is amazingly better as a conductor than most metals, and especially alloys. However, achieving 1% accuracy in a 100-milliohm shunt doesn't happen by accident. It requires either a dedicated 4-wire sense resistor, or else some actual thought, measurements, and math (perish the thought).

Cheers

Phil Hobbs

So the solder is ~50 um thick? I wonder what sort of variation in thickness one sees. (Putting in numbers.. I get a very low solder resistance... ~10-9 ohms. (Area ~= .5mm x 1mm, length = 50 um)

George H.

"Snap!" Scratch that... ~ 10^-5 ohms (what's a few orders of magnitude between friends?)

George H.

Darn, you beat me to it.

Interesting. My newsreader Thunderbird converted the "^-5" to a superscript when I read the message. Very nice.

The adhesion of solder is due to its propensity to dissolve copper, and maybe nickel, so you can add half an order of magnitude because the interface layer isn't pure solder. Alloys have higher resistivity, generally.

You might also want to allow (if the whole process isn't handled in a clean room) a bit of dust: 100 microns is not an impossible dustmote size, which would another half order of magnitude in the event the dust prevents the component from lying flat.

I have Kelvin shunts marked 150A 50 mV, which comes to about 6 e-3 ohms. So, if I were to use that kind of shunt in surface mount, it'd have to be a 4-terminal SMD package OR I'd want to calibrate after assembly.

One should use a lead free solder. Preferrably the gold alloy or even better still the Indium stuff.

Baloney. Molten solders would expel any such lighter mass object, and your supposed interface resistance difference is so near nil as to be negligible.

And yet you somehow think that a "dust mote" is going to sway your capacity to accurately monitor anything once it is up and has been characterized. That characterization is a key part of using any Kelvin Connection supposition circuit. One cannot simply mass produce it. They should ALL be individually characterized. Maybe build it right in to the operational firmware of the monitoring circuit.

That or it is inconsequential and should be ignored for all intents and purposes here.

I vote number two.

It is difficult to mass produce lab instrument level accuracy, so one should never give one's circuit or product any such moniker without actual instrumentational level calibration of said circuit or device.

Being surface mount, I presume the monitoring circuitry to be quite local to this shunt device, so it would not be hard to devote some real estate to a simple PIC setup to assist in a post solder calibration (and maintainence) routine.

Well worth another $6 in parts... ten even.

I suspect the pad resistance doesn't amount to much, the main thing is not to include the trace resistance in one's measurement.

(This is power supply stuff, not critical at all.)

I'll make some measurements if I get the time later in the week.

Cheers, James

motto: "Why wonder when you can measure?"

Sure. But if the end cap is included in the rating and you flip it, the installed resistance will be lowered by 2x the end cap resistance.

It's probably not much.

John's idea of copper pour under such a part is an interesting heat-sinking wrinkle.

I was surprised to see the parts you linked only go down to 50 milliohms.

Cheers, James Arthur

Nice parts.

James

What size trace do you use to handle 150 A of current?

Umm, about .33 e-3 ohms?

Every now and then I lay out a 4-layer PCB with a collection of miscellaneous circuits. The next one will have

Two versions of a pulse generator output driver

A 500 MHz instant-start triggered oscillator

A bunch of 0603 0805 and 1206 resistors with various amounts of copper heat sinking, so I can measure temperatures and thetas. Maybe I could add some Kelvin thingies to test.

Some PCB inter-layer transformers and coupled transmission lines

A few microstrip/diode/phemt thingies, for testing parts

Maybe a TDR circuit

Maybe an RJ45 passthrough with SMA snoop connectors

Any other suggestions?

Yep, that's right. I must have mistyped onto the calculator (the stack still shows the wrong result, right above the one that agrees with you). Drat! In the old days, with a slide rule, that mistake wouldn't happen.

Yeah, and your "dust mote" would change that .33 e-3 to .3300005 e-3

Wow.

...

...

Looks like an opportunity to make TDR test standards; some transmission-line items with artful blemishes applied, suitable for calibration and application of serial numbers...

Any lug or busbar that can take a 3/8 inch bolt can be easily connected... at 240 grams (about half a pound), it's not likely to attach nicely in a reflow oven with a standard preheat

So why are you talking about surface mounting these things?

People do make and sell TDR test thingies. If I try to make a 50 ohm trace on a cheap FR4 board, I might expect to get 45 to 55 ohms, maybe occasionally worse.

A TDR demo board might be interesting, like a trace with various widths, corners, and some vias.