Marco Reps has an interesting article on low noise amplifiers:

Ultralow Noise Tester: 9V Battery vs. 7805 vs. LTZ1000
https://www.youtube.com/watch?v=XpbDMo8an5w

- uses ADA4523: $3.030 Mouser, Digikey, 4.2 nV/sqrt(Hz), RRIO, 100pA Ib
- Datasheet https://octopart.com/search?q=ada4523

Cross-correlation would improve noise floor. Hard to get down to
Gerhard's level, but maybe doable.

1. Here is a brief explanation of how cross-correlation works:
"Understanding Phase Noise - the Cross Correlation Method"

https://www.youtube.com/watch?v=Sf7qiysPFbQ

2. The benefit of cross-correlation depends on how long you are
willing to wait. From

"TSP #162 Tutorial on Theory, Characterization & Measurement
Techniques of Phase Noise"

https://www.youtube.com/watch?v=SOHjFtw0sgo

each 5db of improvement requires an order of magnitude increase in
the number of correlations.

dB 5 10 15 20 25 30 35 40
N 10 100 1000 1E4 1E5 1E6 1e7 1e8

So going from -180dBc/Hz to 220dBc/Hz would require 1e8 correlations.

3. Nobody would wait that long. But how many correlations do you need?

A commonly accepted figure is your measurement noise floor has to be
at least 10 dB below the noise you are trying to measure. We could
pad that by 5 db, which would require

15 / 5 = 3
1e3 = 1000 correlations. That is not too bad.

4. It is interesting to note the Holsworth HA7062C Phase Noise
Analyzer goes to a maximum of 1024 cross-correlations. 100
correlations (-10dB) takes 7 min, 9s at 10 MHz.

https://holzworth.com/Portals/0/HA7062C_Web_Datasheet.pdf

5. Finally, "Frequency Signal Source's PN (Phase Noise) Measurements
Challenges and Uncertainty", by Ulrich L. Rohde, has some very
important information on cross-correlation errors. This file is at

https://www.mrmonett.com/pdfs/2015-IFCS-Rohde-Oscillator-noise.pdf

--
MRM