Using a Spectrum Analyzer Noise Marker Without Confusing Power and Density

A spectrum analyzer’s ordinary marker reports power associated with its measurement bandwidth, while a noise marker can normalize a broadband result to one hertz. Those readings answer different questions. Check which quantity is displayed before comparing noise measurements made with different resolution bandwidth settings.

Identify what the marker computes

The Rohde and Schwarz FSP operating manual describes noise marker corrections for detector behavior, equivalent noise bandwidth, and normalization. The appropriate correction depends on the analyzer implementation and measurement mode.

Record whether the result is dBm, dBm per hertz, or relative noise density. Keep the carrier reference explicit if a relative unit is used.

Use bandwidth scaling only where appropriate

For approximately flat broadband noise, integrated power increases with bandwidth. A hypothetical density of minus 150 dBm per hertz corresponds to minus 110 dBm in an ideal 10 kHz noise bandwidth, because ten times the logarithm of 10,000 is 40 dB.

Use equivalent noise bandwidth where required, not automatically the filter’s nominal three decibel bandwidth. A discrete tone does not scale like flat random noise and should not be interpreted using the same normalization.

Choose a clean observation region

Place the noise marker away from discrete spurs and steep spectral edges unless the method specifically accounts for them. Averaging nearby points can include a tone and bias the result.

Check the analyzer’s own noise contribution and input attenuation. If the source noise is close to the measurement floor, the displayed density includes a significant instrument contribution. A lower trace after changing attenuation may describe the analyzer setup rather than the device.

Verify consistency across settings

Repeat with a suitable different resolution bandwidth and compare normalized density in a region expected to be flat. Keep detector, averaging, and source conditions documented. Allow the measurement to settle after configuration changes.

When the application needs total noise in a band, integrate using a method appropriate to the measured spectrum. Multiplying one density value by bandwidth is only justified when that value represents the band adequately.

If the density varies across the band, retain the trace data and integrate frequency bins with their correct power weighting instead of selecting the quietest visible marker position.

Save marker type, bandwidth definition, detector, averaging, input conditions, and the reported quantity. A useful noise result distinguishes spectral density from integrated power and discrete interference. An ordinary marker that moves when resolution bandwidth changes is not necessarily evidence that the device noise changed; it may simply be measuring a different amount of broadband energy.