SNR measurement using a spectrum analyzer

Many signals have a well defined bandwidth. Examples are OFDM, SC-FDMA, and to some extent CDMA/WCDMA.
Signals with a nearly rectangular spectrum tend to have good bandwidth efficiency, because only a small fraction of the available channel bandwidth is left unused. They are mostly found in modern communication systems, because the higher efficiency comes at the cost of more required processing power both to generate and to detect the signal.

The spectrum analyzer sweeps a tuneable bandpass filter and plots the power within the filter bandwidth, which is also called "resolution bandwidth" (RBW).
By setting a narrower or wider RBW filter, the absolute power levels on the display will change. That is of no concern for SNR measurement, since only the ratio of signal to noise is required.

The pictures show the same signal using two different RBW settings:

OFDM signal on spectrum analyzer using 1 kHz resolution bandwidth

the same signal in 100 kHz RBW

The levels differs by 20 dB - $10\log_{10}(\frac{1}{100})$ : Both signal and noise are flat over frequency, and the 100 kHz filter will capture 100 times more noise than the 1 kHz filter.
Using a smaller RBW gives more accurate results, but in this example only because the measurement takes longer, leading to more averaging.

The RBW setting should not affect the SNR!

For signals with a rectangular (or nearly rectangular) spectrum, SNR is simply the ratio of signal and noise power densities.
The plot shows two sweeps on top of each other:

Signal and noise
Noise only

That said, it is often a good assumption that the "fuzz" seen outside of the signal bandwidth continues through the signal, setting the noise floor.

This method gives only meaningful results if the signal has a reasonably flat spectrum.

For comparison, the example below (a GSM EDGE signal) is not “reasonably flat”.

Using the method with sine waves

It is common to see illustrations, where the signal-to-noise ratio is shown in the above manner, between the noise floor and a sine wave.
However, the situation is now much more complicated: a sine wave does not have a “flat” power spectral density.
Comparing two SNR readings obtained this way is safe, because there is a scaling factor that cancels out.
However, obtaining an absolute result is more difficult, because one needs to take the noise bandwidth of the filter (which is not equal to the resolution bandwidth) into account.
It can be easily seen, that by changing the RBW, one can obtain almost arbitrary SNR values, and that alone may give indication that something is wrong:

“Measuring” an absolute SNR by comparing a sine wave to the noise floor is dangerous:
One may not compare an absolute power with a spectral power density

As said, I can get meaningful results, if I can explain how the power density of the noise and the absolute power of the sine wave relate to each other through the spectrum analyzer's filter.
However, this isn't nearly as straightforward as comparing two constant power spectral densities.

© Markus Nentwig 2007-2008
The content of this page is provided without any warranty and may not be reproduced without permission.

Comments? Questions?

Please send me a mail! mnentwig@elisanet.fi