Measuring noise with an AM receiver

The problem

Chas, VK3PY, had made Noise Figure measurements of a number of receivers, and was puzzled that some of the receivers returned much better and unbelievable Noise Figures in AM mode. The question was, that if the Noise Figure is predominantly determined in the first stage or stages of a receiver, why is the AM performance so different?

Chas used a Rhode & Schwarz SKTU Noise Figure meter using the ‘twice power’ method. The Noise Generator is connected to the receiver and the source turned OFF. The receiver audio output power is noted, then the source is turned on and the level adjusted until the receiver output power has doubled. The receiver Noise Figure is read directly from the SKTU meter. Note that the indicated Noise Figure is equivalent to the Excess Noise Ratio of the instrument’s noise generator.

The technique is a good one in a lot of ways, but it is exposed to non-linearity of the receiver at low signal levels, such as may be experienced with an AM envelope detector. This is not a manifestation of the classic slew rate problem of envelope detection, but rather the effect of a non-zero detector threshold, and the non-linearity in the region of that knee.

ENR (dB)
Output Y (dB) AM SSB
3.00 2.8 4.8
4.77 5.5 7.4
6.00 7.2 9.1
7.00 8.3 10.3
7.80 9.2 11.3

To explore the problem, Chas made a series of measurements of noise generator output (ENR) for different relative receiver audio output power (Y). The measurements are reported in the table above.


The total equivalent input noise power (K) was calculated for each of the cases, and plotted against the relative audio output power.

The graph above shows receiver relative output power vs total input power. In a linear receiver, this must be a straight line with zero Y intercept. The measured data has a distinct curvature at the lower end, and a least squares linear curve fit gives a Y intercept well away from zero. The graph reveals significant non-linear behaviour, especially around the lower two points which is where the Noise Figure measurement of 2.8dB was made.

If this non-linearity is mainly due to the detector knee threshold and non-linearity, then the slope of the measured line at the higher measurement levels is probably close to correct. The ‘Estimated’ line in the graph is a linear estimate with zero Y intercept based on the slope between the last and third last data points. The Estimated line gives Pout=2.01 for Pin=290K (termination alone), and Pout=7.55 for Pin=3288K Using the Y factor method, the receiver noise figure can be calculated (eg using Noise Figure Y factor method calculator or NFM), and the result is 4.8dB, a lot different to the 2.8dB indicated by the test at lower power level. Note, this is consistent with the Noise Figure measurement in SSB mode reported at 4.8dB.

The solution

Above is a plot of the measured Pout vs Pin response of the receiver in SSB mode, and a least squares curve fit with zero Y intercept. It can be seen that the SSB detector response is much more linear and better suited to such measurements.

SSB detectors do not suffer from two of the distortions (slew rate distortion and threshold distortion) that affect envelope detectors.


Noise Figure measurement of the IC706IIG by the ‘twice power method’ was inaccurate, and apparently due to non-linearity of the receiver in AM mode, non-linearity that was not evident in SSB mode, and presumably caused by the envelope detector.

Note that not all AM receivers use an envelope detector, and for example, synchronous detection does not suffer from some of the linearity problems of an envelope detector.

SSB detectors do not suffer from two of the distortions that affect envelope detectors, and are more suited to noise measurement.

This phenomena has the potential to introduce errors to all noise measurements when using an envelope detector. Celestial noise measurements (eg Sun noise rise) are an example.

© Copyright: Owen Duffy 1995, 2017. All rights reserved. Disclaimer.