L
Admin
Near-field H- and E-field probes show up in almost every desense / spur / enclosure-leak hunt, but most benches treat them as magic sniffer wands. A few habits separate “found the hotspot” from “invented a hotspot with the probe cable.”
**1. Relative first, absolute later**
Most commercial near-field probes are sold with a conversion factor vs frequency. That number is useful for *ordering* sources and for repeating a measurement on the *same* geometry. It is not a substitute for a calibrated field probe + documented setup when you need absolute V/m or A/m for a compliance argument. Habit: label every plot as **relative dB (probe + geometry)** unless you have actually done a transfer against a known field.
**2. Freeze geometry or the ranking lies**
Probe tip height, orientation (H-loop normal vs E-tip), and which face of the package you are over all move the reading more than most people expect. Mark the scan path. Hold tip height with a foam spacer or a simple fixture. Re-seat once and demand the same hotspot ranking before you redesign a shield can.
**3. The probe cable is often the real antenna**
A flexible coax behind an H-loop radiates and picks up. Symptoms that the cable is in the measurement:
- Grabbing the cable moves the spectrum.
- Rotating the cable 90° moves nulls.
- Hotspot “follows” the cable route more than the IC pin.
Fixes that actually help: clip-on ferrites on the first λ/4 behind the probe, keep the cable dressed the same every sweep, and prefer a short semi-rigid or phase-stable jumper only for the last hop if you must.
**4. What a near-field scan can and cannot prove**
- Can: rank radiators, show which seam / via / connector shell is loudest, guide where to add stitching or a gasket.
- Cannot alone: prove far-field compliance, prove antenna efficiency, or replace a conducted desense test with USB traffic on / off.
Pair the probe map with a **system metric** (throughput, PER, RSSI, NF, or a conducted spur) so you know the hotspot you killed actually moved the product number.
**5. A short “cal” that is honest for relative work**
Once per campaign I do a transfer check: fixed probe height over a known microstrip trace or a small loop driven at a known power, record the analyser reading at the band of interest, then re-check after lunch or after swapping the probe cable. If that reference moved >1–2 dB, I do not trust today’s ranking until geometry / cable / analyser attenuator are restored.
**Discussion**
On your last desense or enclosure-leak hunt, did the probe ranking survive a cable dress change — and do you treat the vendor conversion factor as absolute, relative-only, or something you ignore and normalise to a local reference each time?
— rfforum 编辑部 / Editorial Team
**1. Relative first, absolute later**
Most commercial near-field probes are sold with a conversion factor vs frequency. That number is useful for *ordering* sources and for repeating a measurement on the *same* geometry. It is not a substitute for a calibrated field probe + documented setup when you need absolute V/m or A/m for a compliance argument. Habit: label every plot as **relative dB (probe + geometry)** unless you have actually done a transfer against a known field.
**2. Freeze geometry or the ranking lies**
Probe tip height, orientation (H-loop normal vs E-tip), and which face of the package you are over all move the reading more than most people expect. Mark the scan path. Hold tip height with a foam spacer or a simple fixture. Re-seat once and demand the same hotspot ranking before you redesign a shield can.
**3. The probe cable is often the real antenna**
A flexible coax behind an H-loop radiates and picks up. Symptoms that the cable is in the measurement:
- Grabbing the cable moves the spectrum.
- Rotating the cable 90° moves nulls.
- Hotspot “follows” the cable route more than the IC pin.
Fixes that actually help: clip-on ferrites on the first λ/4 behind the probe, keep the cable dressed the same every sweep, and prefer a short semi-rigid or phase-stable jumper only for the last hop if you must.
**4. What a near-field scan can and cannot prove**
- Can: rank radiators, show which seam / via / connector shell is loudest, guide where to add stitching or a gasket.
- Cannot alone: prove far-field compliance, prove antenna efficiency, or replace a conducted desense test with USB traffic on / off.
Pair the probe map with a **system metric** (throughput, PER, RSSI, NF, or a conducted spur) so you know the hotspot you killed actually moved the product number.
**5. A short “cal” that is honest for relative work**
Once per campaign I do a transfer check: fixed probe height over a known microstrip trace or a small loop driven at a known power, record the analyser reading at the band of interest, then re-check after lunch or after swapping the probe cable. If that reference moved >1–2 dB, I do not trust today’s ranking until geometry / cable / analyser attenuator are restored.
**Discussion**
On your last desense or enclosure-leak hunt, did the probe ranking survive a cable dress change — and do you treat the vendor conversion factor as absolute, relative-only, or something you ignore and normalise to a local reference each time?
— rfforum 编辑部 / Editorial Team
