Hi Kleinstein,
If the case parts are metal, they would have to be connected to PE (or maybe floating), but they should not be connected to the ADC part circuit ground. A good shield would need to be connected to the ADC ground = low side of the voltage measurement and well insulated from the case and high side of the input.
you are absolutely correct. Chassis is connected to PE, and the ADC part is isolated except for the ribbon connection to the PSU board.
With the typical cheap generators with SMPS supply and no PE connection the signal has a good chance to have quite some common mode voltage to ground (e.g. about half of mains via the class y capacitors at the supply). With just a short at the input one would not notice that. Chances are the hum comes in with common mode voltage.
For a test one could check what happens of the DMM low side / generator is also connected to PE (e.g. meter case).
I did a test run connecting one function generator's BNC connector's shield to the Prema's metal back-plate via alligator clip and that helped a lot. The meter is pretty stable now in the 20 VAC range and only reads about 40-50 mV low (assuming the function generator is precise). So I guess additional shielding is not needed the noise came in via the Uni-T's power supply.
Low amplitude AC measurements - Prema 5000 / UTG962E common earthI'll use a power bank to run the Uni-T from battery and avoid PSU related noise issues. I'll probably also make a banana-to barrel cable so I can hook it up to my linear bench PSU which should be a lot better than the 5V adapter that came with the function generator.
I also noticed an interesting issue where I'm getting a declining reading upon range switch between high Z and non-high Z ranges. Looks almost like some capacitance related artifact. A short demo is available
here.
While the meter was open I also measured the big 4700µF 16V cap at the bottom left of the PSU board, just because it was easy to access and does not require removal of the board. It measures a bit high ~5500 µF but ESR is >0.1 Ohm (in circuit). I heard that older caps tended to rather read high then low - so I guess its still good.
Also the lithium cell must still be good (enough) because according to the manual the Prema runs 3 self tests (Contr. 1-3) at startup. Contr 2 "calculates a check sum over the calibration factor data stored in the buffered RAM and compares those with a checksum". If the battery was to weak I think I'd just get an "Err. 8: Error during self-test 2: Calculated checksum and control checksum do not match (data-loss in battery buffered RAM).".
I guess now I need to come up with a plan to get the meter calibrated. Resistance range is good and pretty close to the Keysight. The <=2VAC range looks also good.
Also there is an interesting safety remark in the manual chapter. 8.1. translation:
"It must be considered that the maximum permissible values of 50 VDC or peak voltage between the "LO" input and guard (see section shielding) and 200 VDC or peak-to-peak AC between guard and chassis are not exceeded. For high-voltage devices without mains isolation this must be considered during polarity selection."
So if I get that right they are saying: If you measure high voltage DC, don't connect DC+ to the black input and also if you measure high voltage AC make sure you don't connect any AC signal (or anything floating 200V above chassis/PE) to guard, correct?