i know machined blocks are great, it be nice to hit those nano numbers, but the interesting bit here after seeing the simulation is that, it may not really need a huge block of copper to get those numbers (maybe more layer? i dont know), and it could be done just by accurate PCB fab. and correction could be done by shifting the leads on the solder pad (if allowance is there)
i think problem of thermal EMF already have many ways to mitigate its effects, mainly by thermal insulation to force parts to reach equilibrium quickly, so i wont worry about that. what is out of reach of many hobbyists is ability to access milling equipment and do precision milling, which in this case is mitigated by just using good PCB layout. it is really killing 2 stones with 1 tetra (a short and a zero ohm). and 1 PCB is about 13 grams, but of course, nobody have anything in tetra PCB format to say it works yet
updateI can't see any advantage between round or triangular.
i was itching to find out too.
here are 2 similar layout. in both cases, i shifted 1 of the nodes off by 0.02 inch to see how much of the field lines are affected.
im not sure how much of this is real in actual, but it seems by suffering a shift off a main axis, the symmetry will suffer badly. but in my case of the simulation scale (=1uV full scale) badly = 20-30nV or so it seems (99 lines, change of about 2-3 lines). so if we look at zlymex's experiment, and compare with this, with a "fabrication" accuracy better than 0.02 inch, anyone should be able to attain very good accuracy effortlessly in theory by placing the circuit entry points smack on the dot.
so i guess this in theory shows the amount of deviation if the circuit pads are in-accurately placed.
A1-2, vertical shift (WRT +ve)
A3-4, vertical shift(WRT +ve)
A5 and 6 horizontal shift
added a7 = crude representation of normal shorting plug. sense side is out of symmetry by 9 -10 lines (90-100nV per uV)