A spectrum analyzer is designed to measure the frequency components of electronic signals. It measures the magnitude of an input signal against frequency, so each signal component appears at its own frequency. For a fuller description, see what is a spectrum analyzer. Common types of spectrum analyzers include:

• Swept-tuned spectrum analyzer: Sweeps across a selected frequency range and measures the signal level at each point. It is a traditional tool for RF design, service and maintenance.
• Real-time spectrum analyzer: A real-time spectrum analyzer calculates FFTs continuously and builds the spectrum as the signal arrives. Keysight's probability of intercept note calls real-time analysis crucial for detecting, observing and identifying transient signals, and lists the N9021B, N9032B and N9041B among Keysight's signal analyzers equipped for it.
• Vector signal analyzer: A vector signal analyzer captures magnitude and phase, so it can demodulate a signal and measure modulation quality. It is used for demanding work such as designing and testing wireless communication systems.

Digital oscilloscopes primarily measure signals in the time domain, but models with FFT capabilities can also display their frequency components. According to Keysight's Infiniium help, the sampling rate must be at least twice the highest frequency component; strictly, it must be greater than twice to avoid aliasing (see step 1). FFT frequency-bin spacing also depends on the sampling rate and the number of captured points. See the features to look for before buying an oscilloscope.

A signal generator can produce a tone at a known frequency and send it through a system. A spectrum analyzer can then measure the output to identify changes in signal level or unwanted frequency components. These differences may indicate how the system behaves or reveal a fault, provided you've accounted for harmonics and spurious signals produced by the generator itself.

MATLAB and Python libraries such as SciPy and NumPy include Fourier transform functions for analyzing a recorded signal.

Unlike a spectrum analyzer, which measures a signal's frequency components, a network analyzer measures how devices such as filters, amplifiers, cables, and antennas respond across frequency. A vector network analyzer (VNA) supplies its own test signal and measures both magnitude and phase, as explained in the network analyzer versus spectrum analyzer guide.

1. Capture the signal. Use an oscilloscope, signal analyzer or similar instrument to capture the signal. When digitizing a signal directly, the Nyquist sampling theorem requires a sample rate greater than twice the highest frequency component to avoid aliasing, which can make high-frequency signals appear at lower frequencies. Many RF spectrum analyzers first mix the signal down, so for them the sampling requirement applies to the bandwidth they digitize, not to the input frequency. Before you connect a signal, check the analyzer's maximum input level in its data sheet.
2. Choose the analysis technique. The Fourier transform is a common choice because it breaks a signal into its constituent frequencies. Use the short-time Fourier transform or a wavelet transform when the frequency content changes over time.
3. Apply a window function. Apply a window function to reduce spectral leakage, which occurs when a captured signal does not complete an integer number of cycles within the measurement window. This causes energy from a frequency component to spread into nearby frequencies. Common windows include Hann, Hamming and Blackman. The choice depends on the trade-off between frequency resolution and leakage that you can accept.
4. Perform the transform. The result shows the frequencies, their amplitudes and, with some transforms, their phase.
5. Interpret the results. Look for significant peaks, relate them to source components, and check harmonic relationships. Unexpected frequency components may indicate interference, distortion or signal instability.

When you use an FFT, frequency resolution depends on the sample rate, record length, and window function. According to Keysight's Infiniium help, dividing the sample rate by the number of points gives the frequency-bin spacing. The window function also affects how well the FFT can distinguish closely spaced signals. Infiniium can add zeros to a record to reach a power-of-two FFT length. This changes the displayed bin spacing but does not improve the ability to resolve closely spaced signals. For an FFT without zero-padding, the bin spacing is:

Frequency resolution = sample rate ÷ number of points

Keysight's Infiniium help identifies 2 contributions to FFT frequency accuracy: an uncertainty of ± half a frequency bin and an additional error from the oscillator that generates the sample clock.

At a sample rate of 100 MS/s, 1,000 captured points give a bin spacing of 100 kHz, while 1,000,000 points give 100 Hz. At a fixed sample rate, a longer capture provides finer frequency resolution, provided the signal remains stable throughout the measurement.

In telecommunications, spectrum analysis supports designing, testing and troubleshooting communication systems. It lets engineers examine signal quality, measure bandwidth and find sources of interference.

Spectrum analyzers help engineers set up and maintain television and radio transmitters by measuring signal bandwidth and checking for unwanted emissions outside assigned frequency bands. Spectrum analyzers also help engineers identify and troubleshoot electromagnetic interference (EMI). For more information about conducted and radiated testing, see what is EMI testing. Keysight offers an EMI measurement application for X-Series analyzers. For receiver measurements to standards such as CISPR 16-1-1, see the PXE EMI test receiver.