R
Member
A pulsed signal can look perfectly reasonable on one setup and several dB off on another, even when nothing about the device under test has changed. In my experience, the first suspect is usually the wrong RBW and sweep combination, not the device itself. The analyzer is not lying; it is simply showing what the chosen settings allow it to see.
1. RBW versus pulse width
If the resolution bandwidth is wider than roughly the inverse of the pulse width, the analyzer smears the pulse in the frequency domain and the displayed peak power reads low. The pulse energy is spread across the filter, so the peak no longer represents the pulse itself. A practical check is to narrow the RBW step by step until the displayed peak stops moving. When it stabilizes, you have a setting that is at least resolving the pulse rather than averaging it away. Whatever value you settle on, record the RBW alongside every number you report, because a peak power without its RBW is not really comparable between benches or setups.
2. Sweep time
A sweep that is too fast across a pulsed source can miss pulses between display points. The trace then depends on which pulses happened to coincide with the sweep, and repeated sweeps disagree for no obvious reason. Use a sweep slow enough to catch several pulses per display point, so each point is supported by real pulse events rather than luck. Alternatively, use zero span with a capture long enough to see the pulse train directly. That time-domain view often explains a confusing spectrum trace in seconds: you can see the pulse width, the repetition interval, and whether the source is truly steady.
3. Detector choice
The detector decides what the displayed point means, so it must match the quantity you actually want to report. Sample detectors can lie badly on pulsed signals, because a single sample may catch a pulse, miss it, or catch only part of it. Peak detectors are useful for finding the maximum, but they overstate average power if you then treat the result as average power. Average and RMS detectors only make sense once you state the averaging window relative to the pulse repetition interval. If the window covers many pulses, the result is closer to true average power; if it covers only part of a pulse cycle, the answer depends on where the window falls. State the detector, the averaging window, and the intended quantity together.
4. Video bandwidth and triggering
A narrow video bandwidth smooths away pulse edges and can make a sharp, bursty signal look like a steady one, hiding the very behaviour you need to measure. Free-running sweeps on a bursty source have a similar problem: they produce pretty but unrepeatable traces, because each sweep starts at a different point in the pulse cycle. Where the analyzer supports it, trigger on the pulse or use gated sweep so that each sweep measures the same part of the signal. Repeatability is not a cosmetic concern; it is what makes a measurement trustworthy.
So, what RBW and detector combination do you trust for pulsed measurements on your bench, and what caught you out first?
1. RBW versus pulse width
If the resolution bandwidth is wider than roughly the inverse of the pulse width, the analyzer smears the pulse in the frequency domain and the displayed peak power reads low. The pulse energy is spread across the filter, so the peak no longer represents the pulse itself. A practical check is to narrow the RBW step by step until the displayed peak stops moving. When it stabilizes, you have a setting that is at least resolving the pulse rather than averaging it away. Whatever value you settle on, record the RBW alongside every number you report, because a peak power without its RBW is not really comparable between benches or setups.
2. Sweep time
A sweep that is too fast across a pulsed source can miss pulses between display points. The trace then depends on which pulses happened to coincide with the sweep, and repeated sweeps disagree for no obvious reason. Use a sweep slow enough to catch several pulses per display point, so each point is supported by real pulse events rather than luck. Alternatively, use zero span with a capture long enough to see the pulse train directly. That time-domain view often explains a confusing spectrum trace in seconds: you can see the pulse width, the repetition interval, and whether the source is truly steady.
3. Detector choice
The detector decides what the displayed point means, so it must match the quantity you actually want to report. Sample detectors can lie badly on pulsed signals, because a single sample may catch a pulse, miss it, or catch only part of it. Peak detectors are useful for finding the maximum, but they overstate average power if you then treat the result as average power. Average and RMS detectors only make sense once you state the averaging window relative to the pulse repetition interval. If the window covers many pulses, the result is closer to true average power; if it covers only part of a pulse cycle, the answer depends on where the window falls. State the detector, the averaging window, and the intended quantity together.
4. Video bandwidth and triggering
A narrow video bandwidth smooths away pulse edges and can make a sharp, bursty signal look like a steady one, hiding the very behaviour you need to measure. Free-running sweeps on a bursty source have a similar problem: they produce pretty but unrepeatable traces, because each sweep starts at a different point in the pulse cycle. Where the analyzer supports it, trigger on the pulse or use gated sweep so that each sweep measures the same part of the signal. Repeatability is not a cosmetic concern; it is what makes a measurement trustworthy.
So, what RBW and detector combination do you trust for pulsed measurements on your bench, and what caught you out first?
