Using Spectrograms to Visualize Spectral Changes

Figure 1. The spectrogram shows a history of
change in a spectrum and highlights variations
in frequency or amplitude.

In our last post, we discussed spectral analysis of RF in the lab as part of developing situational awareness. Another tool for spectral analysis is the spectrogram.

The spectrogram is a display composed of the most recently acquired 256 spectra all stacked in a persistence display. It is a feature of the SPECTRUM-1 and SPECTRUM-PRO-2R options that highlights variations in acquired spectra, making dynamic changes immediately visible. 

The spectrogram in Figure 1 shows the timing dynamics of a power rail load variation in a three-dimensional (3D) plot with color persistence. The same data could also be rendered in a flat, two-dimensional (2D) spectrogram display, or using monochrome instead of color persistence.  

Spectrograms are useful for any type of spectral analysis, although they shine wherever it is especially important to view dynamic changes, such as in switched-mode power and power rail analysis. For example, one of the common tests to perform on a power rail is load regulation. The test is quite simple: you monitor the rail voltage while changing the load current. Normally, the output voltage is measured as a function of the load current, and load regulation is computed. But the frequency spectrum of the signals emitted by the power supply change as a function of load, as well. 

Figure 2. Spectrums of a load regulation test
on a 5-volt USB power module with
load currents of 100 mA, 200 mA and 330 mA.

Figure 2 shows the spectral responses to load variation on a simple, five-volt USB power module at 100 mA, 200 mA and 330 mA load currents. The switching frequency of the power supply is shown as the spectral line at just above 20 kHz in the 100mA acquisition in the top grid in the figure. The spectral lines to the right of the switching frequency are harmonics at just above 40, 60, 80 and 100 kHz. As the load current increases, the power supply switching frequency increases and the conducted spectral amplitude increases.

From an electromagnetic compatibility standpoint, it is necessary to study these variations. However, the spectrums look…much the same.  Only by inspecting them carefully can we see the change between loads. Dynamic response to load changes can be difficult to read from traditional power or magnitude spectrum plots, because these plots represent a single acquisition—a single moment in time, not variation over time, which is the forte of the spectrogram.

The 3D spectrogram packs a good deal of the information in multiple spectrums into a compact display. The horizontal axis represents frequency, and the vertical axis represents the spectral amplitude—same as on any spectrum—but the third axis represents the timing of the acquired spectrums, with the newest acquisition in front and previous acquisitions behind it. The 2D spectrogram would eliminate the vertical deflection, displaying frequency and relative time.  

The encoding of the amplitude information depends on the selection of monochrome or color display. For the monochrome display, higher amplitudes are represented by more intense (brighter) shades of the same color.  The color display maps amplitudes as variations in color, with the higher amplitudes shown in the hotter colors (red the highest) and lower amplitudes shown in successively cooler hues (violet the lowest).

Figure 3. The colored ridges in the spectrogram
show the timing of frequency changes.

The spectrogram in Figure 3 shows the higher-level amplitudes from 10 Hz to about 10 kHz in red. As the spectrum amplitude fall offs, color varies from red through orange, yellow and green. The power supply switching frequency appears as the red ridge that varies from 20 to 30 kHz as the load is varied. The timing of the frequency changes is indicated by the position of the ridge in the spectrogram. The current spectrum in the front of the display shows a middle frequency value of about 26 kHz. Before that, the frequency was at its highest value about 30 kHz. Going further back in time, the frequency dropped to its lowest value of about 23 kHz. Harmonics also appear in each acquisition, their color varying between orange to yellow as they change over time. 

Because it shows variations in the spectrum over time, the spectrogram is a useful tool for visualizing spectral changes due to dynamic events, whatever the source of the acquisition.

Watch Dr. Eric Bogatin incorporate spectrograms into the Real-time Spectral Analysis of Power Rails in his on-demand webinar.