Why an SDR Has a High Noise Floor

[Pavel]

Short answer: A high displayed floor can come from gain, bandwidth and FFT scaling, local interference, computer noise, an active accessory or receiver overload. Compare antenna connected, antenna disconnected and a 50-ohm termination with identical settings to separate external RF from internal display and receiver behavior.

Before you start​

• Split the fault into four layers: device discovery, stable sample streaming, reception of a known signal, and the final decoder or audio path.
• Save the exact error text and a screenshot of the relevant settings after removing usernames, serial numbers and private locations.

Step-by-step method​

1. Step 1: Record sample rate, gain, FFT bandwidth and averaging. Reduce gain in steps; if the floor and many signals fall together, the previous display was dominated by gain rather than a new environmental noise source.
2. Step 2: Replace the antenna with a suitable 50-ohm termination. A large drop points to antenna or local RF pickup; little change suggests receiver, USB, computer coupling or display scaling.
3. Step 3: Run the receiver from another computer or a battery-powered laptop and remove USB hubs, monitors and chargers one at a time. Watch for comb-like noise moving or disappearing.
4. Step 4: Reconnect the antenna with filters, LNA and bias tee removed. Reintroduce one component at a time while keeping manual gain fixed.

Concrete example​

Suppose the floor is -55 dBFS with antenna, -72 dBFS with a 50-ohm load and -70 dBFS with the antenna at a quiet outdoor location. The indoor antenna environment is the main contributor. If all three remain near -55 dBFS until gain is reduced, overload or gain is more likely.

How to judge the result​

The diagnosis is useful when one controlled change explains a repeatable floor change and wanted signals respond consistently. Report relative comparisons because an uncalibrated dBFS display is not automatically absolute RF power.

What to record​

• Repeat each test with the same sample rate and gain so a comparison has only one changed variable.
• A direct motherboard USB port, a verified data cable and a lower sample rate form a useful stability baseline.
• The first layer that fails identifies where to work; later decoder settings cannot repair an unstable USB stream.

Common mistakes​

• Changing several hardware, software or RF variables at once, which removes the controlled comparison needed to identify the cause.
• Treating one autoscaled screenshot or one unusually good result as proof without recording the settings and repeating the test.
• Turning on an LNA to lower the apparent receiver noise figure can make a strong-signal noise floor worse by overloading the LNA or SDR front end.

Final check​

Repeat the original failure case after the proposed fix and after a restart. Record the first failed layer and exact change that repaired it so the result remains diagnostic rather than anecdotal.