I am trying to measure noise floor levels down to 10nV/rt-hz on a piece of equipment with a spectrum analyzer.
I'm looking at signals < 100kHz. I am using a 40dB low noise amplifier (LNA) , wired onto the DUT board and measure it's output with the analyzer.
This helps but, I am still getting significant broadband noise due to the GND connection between the DUT and the analyzer.

A few notes:
1. the LNA is GNDed on the DUT and is powered by 9V batteries (+-9V) to remove bench supply GND loops from the equation.
2. LNA is built using low noise op-amp, En-rti is about 3nV/rt-hz over 200Hz-100kHz

I can reduce the GND loop noise slightly by strapping the DUT to the analyzer GND with heavy braid but, not good enough.
I will next try "floating" the analyzer but, this is not usually a great idea.

Any other good ideas/techniques I could try?

 

I am trying to measure noise floor levels down to 10nV/rt-hz on a piece of equipment with a spectrum analyzer.
I'm looking at signals < 100kHz. I am using a 40dB low noise amplifier (LNA) , wired onto the DUT board and measure it's output with the analyzer.
This helps but, I am still getting significant broadband noise due to the GND connection between the DUT and the analyzer.

A few notes:
1. the LNA is GNDed on the DUT and is powered by 9V batteries (+-9V) to remove bench supply GND loops from the equation.
2. LNA is built using low noise op-amp, En-rti is about 3nV/rt-hz over 200Hz-100kHz

I can reduce the GND loop noise slightly by strapping the DUT to the analyzer GND with heavy braid but, not good enough.
I will next try "floating" the analyzer but, this is not usually a great idea.

Any other good ideas/techniques I could try?

You are pushing the limits of the measurement. The lower your signal level is, the more noise will affect the results. Once you get rid of the effects of ac line pick-up and radiated noise from fluorescent lights and digital equipment, etc., you run into a whole new set of sources of electrical noise which are much more difficult to minimize.
https://www.allaboutcircuits.com/te...ise-what-causes-noise-in-electrical-circuits/

 

I am trying to measure noise floor levels down to 10nV/rt-hz on a piece of equipment with a spectrum analyzer.
I'm looking at signals < 100kHz. I am using a 40dB low noise amplifier (LNA) , wired onto the DUT board and measure it's output with the analyzer.
This helps but, I am still getting significant broadband noise due to the GND connection between the DUT and the analyzer.

A few notes:
1. the LNA is GNDed on the DUT and is powered by 9V batteries (+-9V) to remove bench supply GND loops from the equation.
2. LNA is built using low noise op-amp, En-rti is about 3nV/rt-hz over 200Hz-100kHz

I can reduce the GND loop noise slightly by strapping the DUT to the analyzer GND with heavy braid but, not good enough.
I will next try "floating" the analyzer but, this is not usually a great idea.

Any other good ideas/techniques I could try?

Build yourself a battery powered low noise differential amplifier, for example using the LT1028 (https://www.analog.com/media/en/technical-documentation/data-sheets/1028fd.pdf) on P19:


If you powered this with a pair of 9V batteries, and connect the OUTPUT to the inner of the coax to the spectrum analyser and the GND (0V, base of 10k, and base of bypass caps) to the shield. It is also advisable to connect the spectrum analyser GND to the DUT GND with a resistor (e.g. 10R), and check that the DUT GND is close to the SA GND before connecting the op amps. It would be best to do this with a separate conductor, not simply using the outer of the signal coax, in order to minimise ground currents in the outer of the coax.

 

Attached is a schematic of the LNA circuit, using the AD8429 diff amp. It seems to be a nice part.
The circuit is powered with two 9V batteries.
The two 46.4K input resistors provide a necessary DC bias to the amplifier inputs. This introduces a conduction path (23.2K)
for common mode current to flow to the reference GND. I will increase the resistance further, to see how much that helps.

I will also try next, to add a parallel GND connection between the DUT and the Analyzer.
Stay tuned ....

 

I've arrived at a good solution .... I did 2 things.

1. increased the shunt resistors in my ckt from 46.4K to 511K each - this increased common-mode impedance an order of magnitude

2. I ensured both the DUT and the Spectrum analyzer had power cords plugged into the same power strip -
this helps reduce potential differences between GND systems of the DUT and the Spectrum Analyzer

Now, I see no change in noise floor as I connect/disconnect the amp to the DUT GND.
Not bad for a Monday! Full steam ahead.
: )