Thanks a lot for reminding me the theory. What a Yoke!
How could I have forgot about that!
In any case, it is good to know the theory to exactly know where (and why) are the limits.
I think that the theory is one thing, but the practice (or should we call it "implementation") another totally different.
The SOURCE leads are carrying the current. So for 1mA, we probably don't care about lead resistance, but certainly we want as less as possible when the test current is 10A or 100A. MegaVolt asked for numbers and methodology. Megger provides a nice document related to their DLROs: "For the more rigorous tests, separate test leads are used and the current connections are positioned away from the potential connections by a distance that is 1.5 times the circumference of the sample being measured." This is the zone with uniform current density. (they refer to ASTM B193-65).
The SENSE leads should not carry much current, with typical DVMs the input impedance is at least 10M (usually much higher, like 1 or 10Gohms). Other measurement systems don't have that much input impedance, but this should not create any issue on modern DVM if the system calculates both the SOURCE currents and SENSE parameters and cancels out any imbalance.
But this is not always the case (on some systems) and IEEE Standard Test Code for Resistance Measurement (Std 118-1978) provides a few paragraphs about this:
"Some four-terminal resistance measuring instruments and techniques measure R(x) almost independently of the values of terminal resistances Ra, Rb, Rc, and Rd, the resistance of the leads connected to the terminals, and the contact resistance between the leads and the terminals. Other measuring instruments and techniques only reduce the effect of these resistances by putting them in series with resistances in the measuring circuit whose values are higher than R(x) or by connecting them in other less sensitive parts of the measuring circuit." (page 3)
Later on the Kelvin bridge section (4.5.3) it provides additional info:
"This arrangement permits four-terminal measurement of resistance elements, essentially eliminating the effects of lead and contact resistance errors in the measurement of low resistance (see 2.1)
...
equation 21 is useful, because it shows the necessity to keep the resistance of the yoke (Ry) as small as possible in order to minimize the error caused by lead and contact resistances to the unknown and standard in case of discrepancies between the ratios RA/RB and Ra/Rb. For the highest accuracy, care must be taken to ensure that the connection resistances are balanced, because Ry is not negligible."And with all the above in mind then just you can now realize why the ESI 242D has a test leads balance system with two knobs that are called "Lead ADJ" and "YOKE ADJ".
Related thread here:
https://www.eevblog.com/forum/metrology/4-wire(kelvin-measurement)-with-very-thin-lead-wires/This was gold (as usual from Mr. Edwin G. Pettis):
"DVMs do have limits to compensating 4-wire connections, there is usually a specification called out in the manual for this and whatever effects it may have on the accuracy, DVMs measure and mathematically subtract out lead errors but it is not entirely accurate. It still depends on the DVM's accuracy of measurement which is limited at such low values. The Kelvin bridge was developed to MINIMIZE errors due to interconnecting lead resistance, it does not eliminate the effects. In more modern bridges such as the ESI 242D, there are two adjustments which compensate for lead and yoke resistance up to 0.1 ohms by modifying resistance in the bridge arms to compensate for the extra resistance, these adjustments come very close to eliminating errors within the limits. There is also a 6-wire modification of the Kelvin bridge which further compensates for other errors at very low resistances, the NIST website has information on the various methods of resistance measurement.
No measurement technique is entirely free of error, each one has its own sources of error and knowing just what those errors are can help in making more accurate measurements."