4-Wire Kelvin test leads vs correction in software

[igor78]

I've built myself a milliohm meter. Ratiometric measuring, nothing fancy - two INA219 in series, one measures the current over known 1% 0.1Ohm resistor, the other measures the voltage drop over the leads+DUT resistor. Current switches back and forth to negate some of the measurement errors.
Now the resistance of the leads - this is where people tend to use 4-Wire Kelvin test leads. I get why.
Instead, I short my two-wire test leads, do the measurement, and remember it in software as "offset" or "error" which I later subtract from the actual DUT.
What is my fallacy? Why use 4-Wire Kelvin if I can short the leads? Convenience? The inconsistent shorting point?

Q2:
Most of the 4-Wire Kelvin use shielded cable, where does it go inside the measurement equipment? Does it connect to the ground? If yes, digital ground? Analog ground? The star-point where all grounds connect? To the metal chassis of the enclosure? How does it work?

[Kleinstein]

Just shorting the leads is not as reproducible as a 4 wire measurement. The contact resistance for the short and connection to the DUT can vary a little. The resistance of the leads can also change with temperature (e.g. while you want too see how some resistor (e.g. PT100) changes).

For where to connect the shield depends on meter. Some meters connect the ground to one of the 4 terminals for the ohms measurement (which in may vary) and some have a separate terminal for the shield. The shield is mainly relevant for higher resistors, not so  much for the < 100 Ohms ones, where you really want 4 wires.

For the DIY circuit the right point would likely be the metal chassis or maybe analog ground - mainly for RF shielding, not so much for DC leakage.
It it is about DC leakage (e.g. R > 100 K), than it depends on the circuit details and may need different connection for the cables.

[Doctorandus_P]

Contact resistances change when you move the connection clips or when you create a "short" on some other location.

You can eliminate (most of) the resistance of the leads by shorting them, but this does not work for the contact resistance between the leads. If you want an accuracy of up to 100mOhm, then just shorting the leads to null the offset is probably fine, but if you want more accuracy then 4-wire measurement becomes important for correct results.

[Neilm]

As far as Q2 is concerned, I would not have a separate digital and analogue 0 V plane as that asks for trouble. I have seen too many bits of equipment where poorly split 0 V planes causes problems with measurements either directly or if there any EMI around.

[magic]

Now the resistance of the leads - this is where people tend to use 4-Wire Kelvin test leads. I get why.
Instead, I short my two-wire test leads, do the measurement, and remember it in software as "offset" or "error" which I later subtract from the actual DUT.

Actually, you shouldn't short the test leads together but connect them next to each other to the same terminal of the resistor under test.
That way the contact resistance doubles, which is exactly what occurs in a real measurement.

Look up Kelvin clips, these are neat.