Oscilloscope FFT for EMI pre-compliance: RBW, windowing, and probe tricks that actually help

Not everyone has a spectrum analyzer on the bench, but most of us have a scope with FFT. With some care it works surprisingly well for EMI pre-compliance. A few things I learned the hard way:

1. RBW is everything
Narrow RBW = lower noise floor but slower sweep. For conducted emissions (150 kHz-30 MHz), I use 9 kHz RBW to roughly match CISPR receivers. For radiated pre-scans, 120 kHz. Wide RBW hides narrowband spurs.

2. Window function matters more than people think
Use flattop when you care about amplitude accuracy (comparing against a limit line), Hanning for general spur hunting. Rectangular looks sharper but its sidelobes will lie to you.

3. Probe like you mean it
A near-field H-probe with 20-30 dB preamp gets you surprisingly close to chamber results for relative comparisons. Keep the probe ground lead short — better yet, use a ground spring. On cables, a common-mode current clamp tells you more than any near-field probe.

4. Know the limits
Scope FFT dynamic range is ~50-60 dB at best, nowhere near a real receiver. It is great for "did my fix help" A/B comparisons, not for certifying anything.

My workflow: near-field scan to find the hot spots, current clamp on cables to rank them, then fix, re-scan, compare. Quick and honest.

What's your go-to pre-compliance setup? Anyone correlating scope-FFT numbers against chamber data with a consistent offset?

Editorial reply — thanks for sharing a workflow that matches what we see on many benches.

One number that often surprises people moving from scope-FFT to a CISPR receiver: even when RBW looks "matched" (9 kHz / 120 kHz), the scope's noise bandwidth, detector (sample vs peak/QP), and window equivalent-noise-bandwidth still leave a systematic offset. In practice we have seen A/B "did the fix help" comparisons look solid on a scope FFT, while absolute levels versus a Class B limit were still 6–15 dB optimistic until a real SA (or at least a peak detector + known ENBW) was used. So your point on dynamic range (~50–60 dB) is the right caution for certification; for ranking cable common-mode with a clamp, the scope FFT is usually enough.

Two small add-ons to your probe section:
1. If you only have a 10× passive probe, treat anything above a few tens of MHz as qualitative — the tip ground lead is an antenna. A ground spring or coax pigtail to a nearby ground via changes the picture more than changing the window function.
2. For conducted pre-scans without a LISN, a current clamp on the DC input leads still catches many switch-mode harmonics; just don't call that a CISPR conducted result.

Open questions for other readers (and for you if you have data):
- Have you logged a consistent dB offset between your scope-FFT + clamp setup and a chamber/SA for the *same* cable fix?
- Which scope FFT quirk bites hardest in your lab: ENBW, window scallop loss, or aliasing when the timebase is too slow?

— rfforum 编辑部 / Editorial Team