Wires and connectors are checked:
Blue 63,8 mΩ ± 200 μΩ
Green 64,4 mΩ ± 200 μΩ
Black 32,5 mΩ ± 200 μΩ
White 32,1 mΩ ± 200 μΩ
(Black and white are thicker)
Also wiggling them around had never an effect on the problem.
My script tests for these currents, with corresponding shunt voltages:
10 mA 5,0710 V
8,2 mA 4,1582 V
6,8 mA 3,4483 V
5,6 mA 2,8398 V
4,7 mA 2,3834 V
3,9 mA 1,9777 V
3,3 mA 1,6734 V
2,7 mA 1,3692 V
2,2 mA 1,1156 V
1,8 mA 912,8 mV
1,5 mA 760,7 mV
1,2 mA 608,5 mV
1 mA 507,1 mV
820 μA 415,8 mV
680 μA 344,8 mV
560 μA 284,0 mV
470 μA 238,3 mV
390 μA 197,8 mV
330 μA 167,3 mV
270 μA 136,9 mV
220 μA 111,6 mV
180 μA 91,3 mV
150 μA 76,1 mV
120 μA 60,9 mV
100 μA 50,7 mV
By tuning the awg to a certain shunt voltage, the actual measurement begins. But only after that shunt voltage is stable.
It seems to really depend on the current measured whether it becomes jumpy (And this is not always the same target current).
Attached is now the 6,8 mA - 3,4483 V target.
The 10 mA-5,0710 V and 8,2 mA - 4,1582 V targets have then successfully passed. (Those failed previously)
At the beginning there's tuning. Which was now heavily osccillating due to the problem. After 30 s, it became stable during 10 reads so the measurement enters the final stage. In this stage no more tuning, just waiting for the measurement to have a low running std dev. Which will not happen..