An engineer sweeps a wideband receiver and watches the noise floor sink several dB the instant a strong blocker enters the band, then bounce back the moment it leaves. Nothing on the board changed, and chasing the wrong cause can burn an afternoon on a problem that does not exist.
This blog helps RF engineers tell real noise floor behavior apart from measurement artifacts by isolating three mechanisms: Automatic Gain Control (AGC) gain reduction, front-end compression, and wideband measurement artifacts such as reciprocal mixing and FFT/ENBW correction errors, and matching each to a diagnostic signature so the reported noise floor reflects true receiver performance, not a setup or display illusion.
Understanding the Three Major Causes of Apparent Noise Floor Change
AGC action, compression, and measurement artifacts stem from different mechanisms, but frequently occur together. Understanding mechanisms can save hours chasing symptoms.
1. Automatic Gain Control and Referred Noise
AGC reduces analog gain when a strong input arrives.
Displayed power ≈ kTB + NF minus ΔG(t)
Where ΔG(t) is the uncompensated gain reduction. Common causes:
- The AGC reacts within microseconds, faster than the instrument can update the displayed trace
- Trace averaging that blends multiple gain states across a burst
- Detector type, sample versus peak, capturing different points of the transient
Figure 1. The displayed floor drops for the duration of the AGC gain reduction, then recovers.
2. Front End Compression and Desensitization
Compression is a genuine effect, not a display artifact: a strong signal drives the amplifier or mixer toward saturation, reducing gain applied to noise and nearby weak signals.
- Local reduction: gain at the compressed stage genuinely falls.
- Desensitization: sensitivity near the strong tone drops.
Compression gets worse and starts affecting the noise floor sooner under a few specific conditions:
- A strong blocker driving the front end near saturation
- Insufficient input attenuation ahead of the compressed stage
- Third-order products masking the true floor nearby
The effect is not limited to the frequency of the blocker itself. As the front end compresses, the reduced gain applies broadly to everything passing through that stage, so the noise floor and any weak signals nearby are suppressed along with it.
Even 1 dB of compression can bias a reading by several dB.
Figure 2. Gain compression bends the output curve away from the ideal linear response.
3. Measurement Artifacts in Wideband Capture
Reciprocal mixing raises the apparent floor near the carrier as LO phase noise convolves with the tone, opposite the AGC effect above.
N_disp = N_true + 10 log10(ENBW / RBW_nominal)
A Hann window has an ENBW near 1.5 times the bin width; the wrong factor shifts density. The floor distorts when:
- Local oscillator phase noise is high relative to the offset measured
- The ENBW correction does not match the window function
- ADC quantization statistics change near full scale
Figure 3. The measured floor rises symmetrically around the strong tone as the offset narrows.
How These Effects Interact, and Why It Matters for Diagnosis
Treating AGC action, compression, and measurement artifacts as three separate, unrelated problems can lead to chasing the wrong fix. In practice, they frequently occur together and can mask or mimic one another:
- A blocker large enough to trigger AGC often also compresses the front end
- Compression-generated spurs can be confused with a reciprocal mixing skirt
- Correcting for gain state without verifying ENBW can trade one artifact for another
Correct Test Setup: The First Line of Defense Against False Noise Floor Readings
Before troubleshooting a receiver or LO for a noise floor problem, rule out the test setup itself. Confirm before any floor measurement:
- Fixed gain or AGC disabled during characterization
- Detector type and trace averaging matched to intent
- ENBW correction verified against the active window
- Input attenuation set with margin below compression
These settings are not a one-time configuration; you must repeat these settings above before trusting the next measurement:
- Signal burst duty cycle
- Input attenuation
- RBW or window selection
- Local oscillator tuning
A change in any of these can shift the gain state, detector behavior, or bandwidth correction enough to bias the reported noise floor, even if the physical receiver has not changed at all.
Instrument and Setup Considerations
Good practice for measurement setup:
- A local oscillator with adequate phase noise for the offset of interest
- Keeping the strong tone comfortably below ADC full scale
- Calibrated attenuator steps rather than AGC for headroom
- Isolating the strong signal path from the floor measurement path
- Confirming dither functions as intended near full scale
Many floor problems trace back to setup, not the receiver.
Practical Debugging Tips
- A floor that drops only during a burst → check AGC gain state and display correction.
- A floor that rises only near the carrier → check reciprocal mixing and LO phase noise.
- A floor that shifts with RBW or window changes → verify the ENBW factor.
- A floor that shifts with attenuation settings → examine compression margin.
Comparing fixed gain and AGC captures helps isolate the gain loop from the front end.
Final Thoughts
An apparent noise-floor drop does not necessarily indicate improved receiver performance. AGC behavior, compression, reciprocal mixing, and ENBW mismatch can all produce misleading results. Identifying the underlying mechanism can turn hours of troubleshooting into minutes of focused diagnosis.
As bandwidths increase and signals become more dynamic, these effects are increasingly likely to occur together. A trustworthy noise-floor measurement therefore requires understanding both the receiver's operating state and the measurement conditions behind the number.
Read all the blogs in the Transceiver to FPGA series.