Understanding basic spectrum analyzer operation

Reference level

Reference level is the top edge of display and represents the maximum expected power at the spectrum analyzer input. In most cases, the reference level is adjusted so the highest level of the signal is slightly below this level.

Setting the level either too low or too high needs to be avoided. Setting the reference level too high decreases the dynamic range and ability to see small changes in amplitude. If reference level is set too low, the trace pokes above the top of the screen. Setting the reference level too low can also affect the measurement results.

Behind the RF input, some of the first sections of the spectrum analyzer include active components like mixers and amplifiers. If the input level is too high, these devices can go into compression, which creates distortion and negatively effects the measurement results, sometimes very severely. To prevent this from happening, a variable input attenuator is placed between the RF input and these sensitive components. When the reference level is set, this value is used by the spectrum analyzer to adjust the input attenuation and/or the IF amplifier gain in order to avoid overloading the instrument.

Resolution bandwidth

For basic spectrum measurements, resolution bandwidth is, by far, the most important setting. Most spectrum analyzers use heterodyne based analyzers to measure spectrum by sweeping across a span. The trace showing power versus frequency is drawn from left to right, usually repeatedly.

One way to help understanding resolution bandwidth is to think of it as a window that moves across the span, measuring the level as it goes. Anyhow the resolution bandwidth filter or window isn’t square but has a Gaussian or similar shape. The window also doesn’t move, the spectrum is slid past the window instead. The result is the same, and many RF engineers do think of resolution bandwidth as a moving window or filter that crosses a span.

Resolution bandwidth affects is the ability to separate or resolve closely spaced signals. Two narrow signals can only be separated, if the resolution bandwidth is smaller than the distance between these two signals. If a wider resolution bandwidth is used, both signals are covered by the filter as it sweeps past, and they appear as a single signal in the trace.

Average noise level

Another aspect of resolution bandwidth is the effect it has is on noise. More specifically, resolution bandwidth affects the noise floor, also referred to as displayed average noise level, or DANL. The noise floor rises or falls depending on the chosen resolution bandwidth.

What happens to the noise floor when the resolution bandwidth is decreased? As an example, a simple CW signal and a rather large span of 2 GHz is used.

• With a resolution bandwidth of 3 MHz, the average value of the noise floor is approximately -73 dBm
• Narrowing the resolution bandwidth to 300 kHz, drops the noise floor to – 84 dBm
• At an RBW of 30 kHz, the noise floor falls again to -93 dBm
• At RBW equals 3 kHz, the noise floor has an average value of -104 dBm.

Decreasing the resolution bandwidth by a factor of 10 reduces the noise floor by about 10 dB. As a practical matter, to see signals close to the noise floor, a narrower resolution bandwidth should be used.

Resolution bandwidth and sweep time

Lowering the resolution bandwidth provides better signal separation and lower noise, so why not always use the lowest possible resolution bandwidth? Resolution bandwidth is essentially a filter, and narrow filters take a longer time to settle, or get a stable result, compared to wider filters. This means sweeping slows down when using smaller resolution bandwidths in order to get accurate results. Sweeping too quickly leads to both amplitude and frequency errors.

The main factor determining the sweep time of a spectrum analyzer is the resolution bandwidth. What’s the right sweep time? Most analyzers automatically compute sweep time based on resolution bandwidth and span. This setting can be overridden but decreasing the automatically calculated sweep time is usually not a good idea.

The optimal resolution bandwidth is almost entirely a function of the signal being measured, and often must be determined by experimentation. There is a trade-off between speed and selectivity / noise. On most spectrum analyzers, not any arbitrary value for resolution bandwidth can be chosen, but can be selected in certain steps, e.g. 1 kHz, 3 kHz, 10 kHz, 30 kHz.

Video bandwidth

The last basic parameter is video bandwidth. To understand video bandwidth, the term video signal must be explained. Traces are essentially an envelope of the power at individual frequencies, and this envelope is called the video signal. It’s named video since, in the old days, this signal was applied to the vertical deflection of a cathode ray tube in order to draw a video trace on the screen. In modern spectrum analyzers, video bandwidth is a filter used to average or smooth out the displayed trace.

Unlike resolution bandwidth, video bandwidth only affects how the signal is displayed, not the way it is measured or acquired.

Lowering video bandwidth at a video bandwidth of 200 kHz a fair amount of noise can be seen on the signal. This noise is reduced, when the video bandwidth is lowered to 20 kHz, and decreases even further when video bandwidth is lowered to only 2 kHz. Lowering video bandwidth only reduces noise on the trace, it does not drop the noise floor like resolution bandwidth does. It also doesn’t improve the ability to resolve or separate closely based signals.

Choosing video bandwidth

Video bandwidth only changes what the trace looks like, so to a certain extent the correct video bandwidth setting depends on the application. Most modern spectrum analyzers will automatically configure, and update video bandwidth based on other parameters like resolution bandwidth. In many cases a smaller or narrower video bandwidth seems desirable since it reduces noise on the trace. But just like resolution bandwidth, video bandwidth effects sweep time – the smaller or narrower the video bandwidth, the longer the sweep time.