What is a correct way to make Kelvin connection with multiple sense resistors

[prasimix]

I presume that topic is probably already discussed someplace else, but I'd like to ask it again with pictures. Let's take as an example five current sense resistors that is connected in parallel. What is in your opinion better way (i.e. more accurate) to make a Kelvin connection:

a) 4-points are made on each resistor and their signal outputs are concentrated:



b) Only one resistor has 4-points wiring:

[capt bullshot]

Using this layout, current distribution won't be equal for the sense resistors. The middle one will see the highest current.
I suppose to relocate the slots you've made: one from top resistor to the second last on bottom, other side vice versa. This way each resistor has an approximate equal amount of copper connecting it.

For the kelvin connection:  Either I'd use only two sense lines, taken from maybe the middle resistor or from opposite resisors (one sense line from the topmost resistor, the other one from the bottom resistor. Or, if you want to make the best effort average out all errors: use sense lines from each resistor of equal length, or tie them together with small series resistors.

[prasimix]

Thanks, if I understood you correctly you propose something like this:

c) Power wiring on opposite ends, signal wiring in the middle:

[capt bullshot]

Yes, that's my proposal. Rationale is to connect each resistor with equal resistance of the involved copper traces. With this layout, each resistor has aprox. equal copper trace length for the main current. To be correct, it's not only length, but resistance of involved copper trace, a function of length, width and thickness - if you keep thickness and width constant, length is to equalized.

Edit: I believe, this layout above is the best solution, trying to combine sense signals, or tapping the outer resistors on opposite sides will bring more copper into the sense path. Copper has worse TC than your resistors and so your measurement will be worse vs. temperature changes.

[richard.cs]

An alternative is to individually tap every resistor and then resistively sum them through resistors >> than the sense resistor. That then works however poor the current sharing between the sense resistors is. Last time I did this was two paralleled 20 uOhms and used 4x 1kOhm for the summing and proved it was well behaved and gave accurate overall current even with unrealistically poor layouts that made the sense resistors vary 2:1 in current. The contribution of the sense resistor tolerance to the overall error is negligible as it is the product of their error and the sense resistor error.

[AG7CK]

Thanks, if I understood you correctly you propose something like this:

c) Power wiring on opposite ends, signal wiring in the middle:

The "cleanest" solution if full symmetry as shown below. The source and sink traces should be outside and symmetrical to the parallel connection branching area.



https://www.electropages.com/2016/01/10-tips-for-designing-with-current-sense-resistors/

The link also describes the method suggested by richard.cs above:

"If it is not possible to design for equal current distribution, or if true four terminal resistors are to be used, you can connect the sense terminals of multiple resistors together into two common tracks, one for each polarity. To prevent high currents flowing in these sense connections, a ballast resistor of at least 1000 times the value of the current sense resistor should be used in series with each connection."

The general theme of Kelvin-sensing series, parallel, bridges and network parts is treated many places / web links. A search for "kelvin sense parallel resistors" also yields https://www.eevblog.com/forum/beginners/4-wire-kelvin-to-parallel-resistors/