Why do spectrum traces have more variation after noise compensation as compared to the spectrum without noise compensation?

To illustrate this with an example, assume that without noise compensation, the measured noise power is 5E–6 ± 0.5E–6 watts. The corresponding power, in dBm, would be –42.23 ± 0.36 dBm. Observe that the variation in power is 0.36 dB. Further, assume that the instrument noise is estimated to be 4E–6 watts. After noise compensation, the reported power would be 1E–6 ± 0.5E–6 watts. On a logarithmic scale, this would translate to –50.62 ± 2.38 dBm. This implies that the variation in power when represented on a logarithmic scale, has increased from 0.36 dB to 2.38 dB.

The following figure illustrates this example graphically.

Power is expressed in Watts on the x-axis, and in dBm on the y-axis. The blue '+' denotes the point (5E–6 watts, –42.23 dBm). The blue vertical guides around this point illustrate a variation of ± 0.5E–6 watts on the x-axis. The blue horizontal guides around the same point illustrate the corresponding variation of ± 0.36 dB.

The red '+' denotes the point (1E–6 watts, –50.62 dBm), obtained after noise compensation, assuming an instrument noise of 4E–6 watts. Analogous to the previous observations, the red vertical guides around this point illustrate a variation of ±0.5E–6 watts on the x-axis, same as the blue vertical guides.

The red horizontal guides around this point illustrate the corresponding variation of ±2.38 dB, illustrating the magnification of power variation after noise compensation, when viewed in logarithmic scale. scale. This emerges from the fact that slope of a logarithmic curve, logx, is inversely proportional to x. Smaller values of x result in large changes in logx and vice-versa.

You can use RFmx SpecAn Spectrum measurement and enable the Average RMS Detectors to reduce trace variation after noise compensation.