I am no expert, but by process of elimination it seems most of the S21 noise in VNA mode comes from just having an open port near the bottom of the noise floor.
Here are my comparisons:
First, in SA mode, TG= 0 dBm, RG405 4” cable directly from Port 1 to Port2 and the SA displays a very low noise trace showing signal at 0 dBM for 1 MHz RBW, full span. Doing same in VNA mode gives near identical trace.
Next I placed two Mini-circuits 18GHz 20 dB attenuators onto each Port and connected Port 1 to Port 2. I did this in SA mode and VNA mode and the Thru traces look pretty clean and almost the same in both modes
VNA 40dB THRU
I wanted to see the noise floor under very low signal, so I added an additional 10 dB attenuator to each Port (total 60 dB)
The SA Thru trace looks good though for the BW setting, pretty flat, low noise across the full span not quite getting down to -60dBM.
On the other hand, doing the same in VNA yields that very noisy trace (even averaging 100x with 700 points).
This looks like the noisey trace I saw when checking out the reflection on my Eagle RLB:
VNA 60dB THRU 700pts
Next I just removed the cable from Port 1 and left the 60dB of attenuators connected to Port 2 and you can see most of the noise is still there even without a input signal:
[ Specified attachment is not available ]VNA 60dB NO TG
The SVA1000x datasheet says there should be a minimum dynamic range for the SVA 60 dB between 10MHz – 1.5 GHz and 55 dB above 1.5GHz. The Trace noise should be < 0.1 dB rms with 50 averages.
So I am puzzled if this is considered out of spec. Can anyone else with a SVA1000x try to duplicate my observations?