Spectrum Analysis Fundamentals, Part 4: Video Filtering and Trace Averaging
2020-08-31 | 7 min read
Welcome back to the Spectrum Analysis Basics blog series. In part 3, I went over detector types and how spectrum analyzers gather and display data, including average detection. In this blog, I’ll go over averaging processes. Averaging processes help to smooth the variations in envelope-detected amplitudes. This blog will discuss two of these processes — video filtering and trace averaging.
Spectrum analyzers display signals as well as their own internal noise. Figure 1 shows an example of a signal plus internal noise.
Figure 1: A spectrum with no filtering, displaying a signal and noise
Smoothing, or averaging, the display helps to reduce the effect of noise on signal amplitude, such as that shown in Figure 2. A variable video filter performs this smoothing.
Figure 2: The display from Figure 1 with filtering
In a spectrum analyzer schematic, the low-pass video filter comes after the envelope detector and determines the bandwidth of the video signal. When a signal passes through the intermediate frequency (IF) chain, the envelope can rapidly vary. To reduce these variations, lower the cutoff frequency of the video filter to a point where it is smaller than resolution bandwidth (RBW). Figure 3 shows the effect of reducing the video bandwidth (VBW) to resolution bandwidth (RBW) ratio.
Figure 3: The effect of a VBW-to-RBW ratio of 3:1 (blue), 1:10 (yellow), and 1:100 (green)
When measuring noise, the effect (of what? Of reducing the VBW-to-RBW ratio?) is most noticeable when using a wide RBW. Peak-to-peak variations of noise minimize as you reduce the VBW. To ensure a smooth signal, stick to ratios of 0.01 or less. Don’t worry – the video filter will not affect parts of the trace that are already smooth, such as a sinusoid that is well above the noise.
The effect of video filtering changes based on your chosen detection mode. In positive peak detection mode, if VBW > RBW, changing the VBW does not make a significant difference in peak-to-peak noise levels. Additionally, if VBW < RBW, changing VBW causes small fluctuations in the noise level, altering the peak values of the smoothed noise envelope. When using average detection, the average noise level remains constant.
In trace averaging, smoothing takes place over two or more sweeps on a point-by-point basis. At each new display point, the spectrum analyzer averages the new value with the previously averaged data using the following formula:
Aavg = new average value
Aprior avg = average from prior sweep
An = measured value on current sweep
n = number of current sweep
To change the degree of smoothing, you can select the number of sweeps over which the averaging occurs. Figure 4 shows the effect of increasing the sweep number, n. While decreasing the VBW affects sweep time, trace averaging has no effect. However, it still takes about the same amount of time as video filtering due to the number of sweeps required.
Figure 4: Trace averaging for 1, 5, 20, and 100 sweeps, from top to bottom
Generally, it does not matter which type of filtering you use. — you will likely reach the same results. However, be sure to note the stark difference between the two: video filtering performs averaging in real time, while trace averaging does not. With video filtering, you see the full effect of the smoothing at each point on the display as the sweep progresses. The analyzer averages each point only once at a rate of about 1/VBW per sweep. Alternatively, with trace averaging, multiple sweeps occur to achieve the full degree of averaging. The averaging at each point takes place of the full time period needed to complete multiple sweeps.
On certain signals, you can get significantly different results depending on your chosen method of filtering. When analyzing a signal with a spectrum that changes with time, your average may be different on each sweep when using video filtering, depicted in Figure 5. If you use trace averaging over many sweeps, the value ends up much closer to the true average, shown in Figure 6.
Figure 5: Video filtering on a signal with a spectrum that changes with time
Figure 6: Trace averaging on the same signal as shown in Figure 5
Averaging helps reduce variations due to noise in your signal, enabling you to differentiate important spectral components from noise. While there are a number of methods out there, make sure to choose the right method for your specific signal when differences in averaging occur. For more background information on spectrum analysis basics, start at Part 1.
Next time, I will go over time gating — an essential topic for spectrum analysis both in classic spectrum analyzers and modern, advanced signal analyzers. For a cumulative overview and in-depth reading on spectrum analysis, check out Application Note 150.