Kelvin SMD connection

I'd be concerned that the resistor rating includes the end-termination resistance. Flipping changes that (reduces), a possible calibration error.

solder .--------.----------------------// fillet | | \ /| | sense resistor / '--------'.------------------------// .--------'------------| '---------------------'

On Friday, February 13, 2015 at 3:40:53 PM UTC-8, Jim Thompson wrote: [about Kelvin connection to surface-mount resistor]

These shunt resistors are precision types, trimmed; the manufacturer ought to test their variance after soldering. So, there must be a recommended pad layout, and maybe solder prescription, if you know where to look.

Several millimeters of NiCr, a silver-plated endzone, and a probe jig ought to be repeatably different from the as-soldered component. The prospect of a tiny grain of grit causing a few mils of SAC solder in series with the precision shunt is ... very annoying.

There are surface-mount 4-lead shunts.

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Specifically, what resistor type and value are you attempting to turn kelvin?

It might not be worth the trouble........

RL

Exactly. I have never felt this layout was definitive.

Very good if you mount the resistor upside down. Not really good if you don't because you don't get the sense points close to the R-film/terminal boundary.

Yep. It will have 4 contacts. But we won't want to pay for That! lol

"Rightside up" does not get to film-terminal boundary.

Terminal RES Terminal \ FILM / .----==========----. |XXXXXXXXXXXXXXXXXX| '---- ----' Cu--------+. .+--------Cu \ / K1 K2

"Bottomside up" does get closer to the film-terminal boundary. It still may not be "perfect" for a given pad geometry and the accuracy of placement. The pads should be longitudinally "too long," especially on the "inside." Terminal Terminal \ / .---- ----. |XXXXXXXXXXXXXXXXXX| '----==========----' Cu--------+. RES .+--------Cu \ FILM / K1 K2

So get the Manufacturer to load your reels upside down.

1-oz copper has a sheet resistance of 0.5 milliohm per square at 25C. That puts a bound on how bad it can realistically be.

Zero-ohm jumpers have typical end-to-end resistances of very roughly 20 milliohms. See e.g.

.

I suspect that variations in the solder fillets at the ends of the resistor will make more difference overall than the sense trace placement.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

These are nice.

Nice fat (equipotential?) contacts. Huge peak power capacity. See sheet 3 of the PDF for their opinion about how to connect to them.

----'

-------+. RES

Yeah, I am not so sure the "problem" is the Cu pads. As you mention, solde r would seem to matter. The terminal wrap itself would seem to count, but maybe that too is very low. So my idea is that getting the copper sense po int closer to the actual "film" boundary would help on those counts.

.

Not sure how much of that is terminal and how much is "film." Is there a p resumption that the inevitable solder is de-embedded in those numbers? Any way, it seems obvious that when the solder and/or terminal resistance are g etting anywhere near the resistance of the resistor, then substantial error can creep in if one is not careful.

.

Hard to avoid the solder, and maybe control it (stencil height variance, ap erture variance, for example; and even part placement). An implication to what you are writing might be taken as "'accurate' current sensing with lo w value two-terminal SMD parts is not possible, regardless of any reasonabl e Kelvin taps." yes/no?

The current sense resistors that John Larkin linked are interesting. Not a ll sub-tenth ohm R's are made that way. Many look "standard": wrap around terminals with "stuff" on top.

I sound like I am contradicting myself. lol

It's no trouble--reasonable layout improves the accuracy, for free. It's well worth it.

The board at hand has 2512 and 1206, .010 to .025 ohms. Three of the 0.010 are paralleled to yield 3.3 milliohms.

Cheers, James Arthur

If the resistance is rated as installed, flipping the part impairs the calibration. If it's thin-film, as in your drawing, that also places the heat-generating surface in contact with the PCB instead of the air.

Cheers, James Arthur

Fun thread James, The one advantage I see to the original 2 pads per side, is that it gets rid of the solder resistance. I have no experience but getting down to milli-ohms that might make a difference.

George H.

Not sure I agree with those contentions, but maybe. Sometimes SMD R's are deliberately "single sided":

I'd expect heat removal to be better if the resistor is mounted upside down. Especially if one puts a bit of copper on the PCB topside. But it's academic, since most heat is conducted out of the end caps.

Here's one way to work around that problem:

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

If your layout tool won't let you use different widths for different parts of a trace, you are spending too much on your tool. I have used FreePCB (both in terms of beer and speech) and it does that quite easily.

I suppose. So does using a Kelvin resistor and without the complications.

Really? I guess if you are designing a $10 consumer device it matters. But then I won't be having a problem with my free tools either.

--

Rick

Consider the 3-D nature of the solder joint and I think you will realize it is negligible. It is thin with significant dimensions in the other two directions. The filet that is considered important is lagniappe for the actual resistance of this connection.

_________ | --------- /| / | / | / |_________--------- ----===============--- ^^^^^^^^^^ This is the region with very low solder resistance. The fillet helps with the mechanical attachment and is an indicator of the not visible part of the solder joint being good.

--

Rick

The resistivity of solder is about 10x that of copper, but the fillet is much much thicker than the pad. Also the fillet is far more variable from unit to unit than the pad is. So ISTM one would have to actually do the math to find out. (What a concept!)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

To find out what exactly? My point is that the thin and large solder area under the terminal has a geometry that provides a lower resistance than the tall and relatively smaller area of the fillet. It doesn't require calculations to compare the two because it is purely matter of geometry. Unless you feel the relative dimensions of the solder is not as I have described.

--

Rick

The variability in the calibration of the pseudo-Kelvin connection, of course. High precision milliohm resistance measurements require the application of either thought or money (or possibly both).

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net