With input shorted, 10Hz-100kHz output is about 1.76mVrms, which indicates the referred input voltage noise density of the EmoeNAP is about 450pV/√Hz (at 1kHz).
And connect a 50-ohms load to input, the 10Hz-100kHz output is about 3.654mVrms.
The theoreticle Johnson-Nyquist noise (thermal noise) of a 50-ohms resistor in 10Hz-100kHz bandwidth is 316nVrms, combined with EmoeNAP's noise floor (1.76mVrms/10000)，the total input voltage noise should be about 361nVrms, and after 80dB voltage gain, 3.61mVrms is very close to 3.654mVrms. This test verified the EmoeNAP's extraordinary ability to measure a 50 ohms resistor's thermal noise precisely.
And I alse tested the LM399 voltage reference's output voltage noise.
According to LM399's datasheet, we can estimate it's output voltage noise in 100Hz-100kHz should about 30uVrms (the best). And the datasheet also gives the voltage noise in 10Hz-10kHz, we can see it varies.
And we can test it now. The output result shows that my LM399 gives about 44.2uVrms output voltage noise in 10Hz-100kHz bandwidth.
More Information (building process and instruction) can be found on external links (on emoe.xyz, in Chinese).
Project files and simple usage docs are hosted on github