I asked a similar question some time ago and received an excellent answer from Gerald on the post linked below. The method he describes for setting front end gains work well to optimize the s/n ratios for any given antenna and location.
Outside of that, you should not look to the radio itself to improve your noise floor... investigate some low noise receiving antennas or track down the sources of your noise.
Gerald's response is here:
Variable phasers like the MFJ-1026 probably won’t reduce your
noise floor (unless your noise floor is determined by a local noise source),
but yes, they can be quite effective in nulling out locally-generated noise from one direction whatever your location. However they also null out desired signals
from that same direction. Receive
antennas as K9SO mentioned are other (and usually better) options.
They include the Pixel loop, Beverages, 4-squares, SALs, etc. (I use Beverages).
I'll refer you to the comparative receiver testing data published by Sherwood Engineering to answer your question:
Here's what he says about what the noise floor measurements mean:
"Noise floor measures how weak a signal one can hear. Practically it is only of significance on the higher HF bands due to the higher level of band noise on the low bands. This assumes you are listening on your transmit antenna. If you are using a Beverage or a low gain loop, then it could be an issue on any band.
The noise floor is measured with a 500 Hz CW filter bandwidth, assuming the radio has a CW filter. There is a note on the measurement if the radio only had an SSB bandwidth. Older radios (Drake, Collins) had no switchable preamp. Compare them to a modern radio with Preamp ON or Preamp #1 ON. A noise floor of -135 dBm is more than adequate on 15 meters in a quiet rural location. A lower noise floor (-138 dBm) might be useful on 10 meters in a quiet location. Serious 6 meter DXers often use an external lownoise preamp to get the noise floor down to -140 dBm or a few dBm lower. If you are in the city, hardly any of this matters due to all the local noise. (On 15 – 6 meters, hardline would be important to reduce the feedline loss to make the best use of the noise floor.)
Noise floor is quoted in dBm (power). Consider it a similar measurement to Sensitivity on SSB, which I quote in microvolts (uV). Noise floor is a 3 dB S+N/N ratio, usually measured at 500 Hz bandwidth. Sensitivity is a 10 dB S+N/N ratio, usually measured with a 2.4 kHz bandwidth for SSB. Due to the wider bandwidth on SSB, the signal vs. the noise sounds about the same by ear. "
Here are my 20m 6600 numbers:
On USB @2.4 khz BW
-102 dbm with no pre-amp
-118 dbm with 16 db pre-amp
-128 dbm with 32 db pre-amp
On CW @400hz BW
-111 dbm with no pre-amp
-126 dbm with 16 db pre-amp
-136 dbm with 32 db pre-amp
With the antenna (3 element SteppIR (band after dusk):
USB @2.4khz BW
-103 dbm with no pre-amp
--113 dbm with 16 db pre-amp
-115 dbm with 32 db pre-amp
CW @400hz BW
-111 dbm with no pre-amp
--122 dbm with 16 db pre-amp
-124dbm with 32 db pre-amp
The key phrase in your original question is "My preamp/RFgain is not on."
With no antenna the signal strength will drop as preamp gain is increased until the frontend A2D converter is no longer below minimum signal level.
As shown in the two comparison sets, the APPARENT signal level will drop with preamp gain until the antenna noise is within the A2D range, then it remains steady as preamp gain is increased.
As the superhet RX raises the S-Meter with RF Gain reduction, the Flex raises the S-Meter as the preamp gain is lowered. The reading being 0 signal + or - the RF Gain/attenuation equivalent in dbm