The XJW-01 bridge - grounding question

[6SN7WGTB]

I have the aluminium bodied, 12V DC external PSU powered version (so no ground path). Kelvin leads.

Was testing a standards box which is a shielded metal case with a series of capacitors, resistors and inductors inside. The unit has a PL259 plug as connection, the shroud of which is firmly connected to the metalwork.

Results were way off on C and R (but not L) until I grounded the system (to mains earth as it was to hand).

Testing components on their own yielded correct results.

What is actually happening with the metal-bodied standards box to give this issue?

So, I need a reliable system ground to be sure I get the right results? (And yes always mindful of any possible L-E shorts although shouldn't be using an LCR bridge on live circuit...)

[Kean]

External interference from other electrical devices like LED lighting.  Without the metal box earthed, it will happily couple the noise into your measurements.

[6SN7WGTB]

Yes I thought that.

What happened initially was that I had about 2 hours of spot on measurements, no ground, and then suddenly it started wandering. I hadn't turned anything on or off on the bench and indeed no nearby LED lighting had changed state.

However, noise it must indeed be.

I suppose as none of the 4-wire connections are grounded, the marge metal box of the standard is acting as an aerial.

But/and what is the downside of my grounding the system? And what do bench 4-wire LCR bridges do? Would I have this issue with one of them - where the unit probably IS grounded and thus adding a new ground woul be undesirable?

[Kean]

I don't know much about the XJW-01 design, but the 12V DC PSU you mention will probably have a Y capacitor for EMI suppression between the AC and DC sides which could introduce noise, apart from general pickup by the metalwork and cables.  Maybe something else was switched on in the house/building that was injecting noise into the mains.  Using a battery as the power source could help isolate the issue.

Having one hard earth connection somewhere is generally best to avoid ground currents.  Alternatively, if the XJW-01 has a guard terminal, you could connect that to the metalwork instead of the earth.  Ideally the guard is via the coax shield around each force & sense connection.

If you were to use a bench LCR meter with an earthed chassis, then that would possibly provide an earth connection as part of the guard (shielding) in 3 or 5 wire measurement configurations (2 or 4 wire plus guard).  Depending on the design and use case, I think the guard will either be driven by the test signal, or bonded to the force low (which in turn may be earthed).  Apart from any coupling affecting measurements, there is also a risk in having the force signals floating and potentially being a hazard when injecting bias voltage or currents from external equipment.

This would normally be covered in the manual supplied with precision equipment.

Some discussion about guarding for low level measurements are under this thread.  Mike Wyatt who started that thread would know a lot more about this topic than me.
https://www.eevblog.com/forum/testgear/measuring-small-capacitance-with-ks34465a-th2830/

You might find additional relevant info in this amazing doc, although it doesn't specifically related to LCR or AC measurements
https://download.tek.com/document/LowLevelHandbook_7Ed.pdf

[6SN7WGTB]

I've added a ground connection which I can connect to mains earth. Seems to work well - the meter can be zero'd and reads a stable 000.03 ±0.01pF on 7.8kHz range, and when measuring results is now stable and repeatable.

It seems quite a competent unit.

[mawyatt]

When using Kelvin Clips we found that measurements were more stable and repeatable when made over a conducting ground plate (per Hioki Instruments recommendations). Our PCBrite baseplate serves as the ground plate when connected to a mains ground terminal from one of the nearby Power Supplies or other instrument. The Hioki IM3536 has a "Guard" capability so this is utilized for the ground connection when the IM3536 is used.

Best

[kloetpatra]

mawyatt is right, you ideally need a ground plate for a shield to be effective

https://youtu.be/Hj-HVve1QC4?t=678