L
Admin
[Technical article / FAQ, rfforum Editorial Team]
A near-field chamber can produce a beautiful pattern plot that is not yet a far-field gain number. The transform, probe correction, and sampling all own the claim; skipping any of them is how a "we scanned it" file becomes a datasheet gain. This is not an argument against near-field methods — it is a traveler discipline for the last 1–2 dB that starts fights.
1. What actually has to be true before the plot is gain
- The scan must be at a known, recorded probe height or radius, not "about a wavelength away." A 5–10 mm height error on a mmWave AUT already moves amplitude and the apparent phase centre.
- Sampling density must satisfy the spatial Nyquist for the aperture and the highest frequency in the claim. A 2–3 mm step that was fine at 6 GHz is already coarse at 28 GHz on a multi-wavelength aperture; undersampling aliases into sidelobes and a shifted peak.
- Probe correction (pattern, polarization, and the probe's own multiple reflections) has to be inside the transform, not applied "by eye" on a screenshot later.
- The far-field transform, including the assumed quiet-zone / truncation / window, must be named. A raw near-field amplitude map is not gain, and a "smooth" plot is not a sampling-density check.
2. Freeze list before you accept a far-field gain or sidelobe claim from a near-field scan
- Probe PN, polarization, and calibrated height or scan radius
- Sample step in both axes, frequency list, and whether density was checked at the highest frequency (not only at a mid-band default)
- Transform type, truncation or window, and whether probe correction was on
- Overlay or check: a known-gain standard antenna, or at least one far-field cut on the same AUT — not only "the plot looks smooth"
3. Habit
Never freeze "antenna gain = X dBi" from a near-field file whose traveler has no probe height, sample step, or transform note. A 1–2 dB argument on a mmWave AUT is often a sampling or probe-correction argument, not a radiator argument. If two scans of the same AUT disagree by more than about 0.5–1 dB in peak gain, first compare probe height and sample density, then the transform settings, before you respin the antenna. If sidelobes move when you shrink the scan area, you are looking at truncation, not at the AUT.
4. Practical gate
- Write height, step, and transform on the same sheet as the gain number
- A sentence that only says "we used a near-field scanner" is not a method
- Keep a known-gain check in the same campaign when the number will be published or used in a link budget
Discussion: on your last NF versus FF disagreement, did the fix land in probe height, sample density, or probe correction — and do you forbid freezing far-field gain from a scan that has no transform note?
— rfforum 编辑部 / Editorial Team
A near-field chamber can produce a beautiful pattern plot that is not yet a far-field gain number. The transform, probe correction, and sampling all own the claim; skipping any of them is how a "we scanned it" file becomes a datasheet gain. This is not an argument against near-field methods — it is a traveler discipline for the last 1–2 dB that starts fights.
1. What actually has to be true before the plot is gain
- The scan must be at a known, recorded probe height or radius, not "about a wavelength away." A 5–10 mm height error on a mmWave AUT already moves amplitude and the apparent phase centre.
- Sampling density must satisfy the spatial Nyquist for the aperture and the highest frequency in the claim. A 2–3 mm step that was fine at 6 GHz is already coarse at 28 GHz on a multi-wavelength aperture; undersampling aliases into sidelobes and a shifted peak.
- Probe correction (pattern, polarization, and the probe's own multiple reflections) has to be inside the transform, not applied "by eye" on a screenshot later.
- The far-field transform, including the assumed quiet-zone / truncation / window, must be named. A raw near-field amplitude map is not gain, and a "smooth" plot is not a sampling-density check.
2. Freeze list before you accept a far-field gain or sidelobe claim from a near-field scan
- Probe PN, polarization, and calibrated height or scan radius
- Sample step in both axes, frequency list, and whether density was checked at the highest frequency (not only at a mid-band default)
- Transform type, truncation or window, and whether probe correction was on
- Overlay or check: a known-gain standard antenna, or at least one far-field cut on the same AUT — not only "the plot looks smooth"
3. Habit
Never freeze "antenna gain = X dBi" from a near-field file whose traveler has no probe height, sample step, or transform note. A 1–2 dB argument on a mmWave AUT is often a sampling or probe-correction argument, not a radiator argument. If two scans of the same AUT disagree by more than about 0.5–1 dB in peak gain, first compare probe height and sample density, then the transform settings, before you respin the antenna. If sidelobes move when you shrink the scan area, you are looking at truncation, not at the AUT.
4. Practical gate
- Write height, step, and transform on the same sheet as the gain number
- A sentence that only says "we used a near-field scanner" is not a method
- Keep a known-gain check in the same campaign when the number will be published or used in a link budget
Discussion: on your last NF versus FF disagreement, did the fix land in probe height, sample density, or probe correction — and do you forbid freezing far-field gain from a scan that has no transform note?
— rfforum 编辑部 / Editorial Team
