# Typical resistance of an SMD solder joint

Do you have an idea of the typical resistance between a PCB track and a SMD resistor pad? That is, the resistance of the solder on, for example, a 0402 resistor.

The idea behind my question is that I'm currently designing some shunt resistors (to measure currents), and I would like to know below which resistance I have to start to use "tricks" to really measure voltage across the resistor and not across resistor + two solder joints. (By tricks I mean taking the measurement track in the "inside" of the footprint, or even go for 4-terminal resistors.)

Currently I have 4, 5 and 120 mΩ shunts. My guess is that for the 4 and 5 mΩ resistor I need 4-terminal resistors, and that for the 120 mΩ taking measurement on the inside side of the footprint should be enough.

• It's not so much the typical resistance of the solder, but its variability depending on how much paste gets put into the joint, or whether the resistor and its pads are under or over-sized, or whether it's sitting flat on the board after soldering. That's the point of a 4 terminal resistor, those details can vary without affecting the measurement. How much of the spec of the original resistor you want to preserve makes a difference as well. I'm sure you can look up the volume resistivity of solder, and do some worst case estimates for the resistor sitting 0.x mm off the board. May 17, 2023 at 13:26
• The typical stencil thickness for 0402 parts is 4 thou (100 microns); most solder paste contains about 55% metal (the rest is flux). Depending on the current being measured, I try to use 4 terminal devices but there are other ways of doing it. See analog.com/en/analog-dialogue/articles/… May 17, 2023 at 13:40
• CMS seems to mean the same as surface mount device (SMD.) What do the letters "CMS" stand for?
– JRE
May 17, 2023 at 14:29
• @JRE There's only one country in the world always inventing their own abbreviations... fr.wikipedia.org/wiki/Composant_mont%C3%A9_en_surface Supposedly they got hubris after the big success of SI :) May 17, 2023 at 14:34
• @Lundin: Thanks. I searched for it but couldn't find.
– JRE
May 17, 2023 at 15:04

Per Peter Smith's excellent comment - I thought of the same Analog Devices "report from the tranches": Optimize High-Current Sensing Accuracy by Improving Pad Layout of Low-Value Shunt Resistors by Marcus O'Sullivan (here available also as a pdf).

In short: you will need to experiment. The results will also depend on the copper weight of the PCB, the tempco of the resistor, the amount of solder paste used, the type of reflow process used (IR vs vapor phase), etc.

In my limited experience, I had excellent Kelvin behavior with solder paste coverage near the minimums and solder-phase reflow. But it's a trade-off. The measurement accuracy of such SMT-mounted devices is often at odds with longevity, i.e. the soldering/reflow settings for good accuracy may have low yields and/or low durability.

The ultimate answer depends on what do you need: what are the actual specs? Temperature range? Allowable system-level tempco? Current range? The volume you have available? Etc.

For very high currents I've used crystal cans as current shunts and quartz crystals as highly accurate temperature sensors. But that's by no means some ultimate solution: it worked in the narrow niche I needed it for.

For resistances in the <0.1Ω range, PCB traces offer a lot of flexibility in tuning the Kelvin behavior, especially since they can be very well modeled using basic finite-element methods for 2D resistive sheets. As long as you can measure their temperature well, copper isn't too shabby a material. Having very thin temperature sense traces interwoven with the sense traces may work. It really all depends on how much calibration and characterization you can afford.

If this is a low-volume product or a one-off that doesn't have extreme volume constaints, I highly suggest using actual Kelvin sense resistors that have four dedicated terminals - whether SMD or through-hole.

There are some truly excellent Kelvin sense parts both in through hole and SMD, and you get what you pay for in terms of long-term parameter drift, tempco, voltage coefficient, etc.