WLAN Lab · Wi-Fi 7 sensor · controlled channel sweep

Antenna Shootout

Eight RP-SMA antennas, characterised on 2.4, 5 and 6 GHz by sweeping one controlled Ruckus R670 across channels — recording received signal, throughput and the ambient band at every channel. One rig, one client, every antenna measured the same way.

Captured 2026-09-03 → 09-04 Metric p95 RSSI (dBm) Client NanoPi · Intel BE200 Source Ruckus R670 · 9 channels · MANUAL Throughput iperf3 → wired NUC

It depends where the scanner lives

The answer

Raw power favours the big whips — but they only matter on a desk. For a pocket scanner the question is the best small antenna, and there the compact circular-pol Osprey trades blows with the bigger Dixingtek. Two picks, then the raw per-band winners.

Per-band leaderboard

Controlled RSSI · band average

Mean p95 of the R670's beacons at the NanoPi across the three swept channels per band (2.4: 1/6/11 · 5: 36/149/157 · 6: 37/101/165). The air was clean (0 tx-retries), so received power is the honest antenna metric.

What sampling one channel per band hid

The channel sweep · compare across the band

Signal and throughput at every swept channel, 2.4→6 GHz. This is the whole point of sweeping the certified AP rather than sampling it once: the shape — and the shared dip at 6 GHz ch101 — only appears when you walk the band. Tap any set of antennas to compare them head-to-head on both charts.

Received signal p95 dBm · higher = stronger

iperf3 throughput Mbit/s → wired NUC

▚Bands shaded amber / blue / violet. The dip all eight antennas share at 6455 MHz is mostly the AP itself running 3 dB lower there (see below) — flip to Source-normalized to remove it. With nothing selected, all eight show faintly.

The 6455 dip is mostly the AP, not the antenna

What the sweep caught

Every one of the eight antennas dips at 6 GHz channel 101 (6455 MHz) — none peaks there. That's the fingerprint of a source effect, and the beacons prove it: the R670's own TPC-report power reads 19 dBm at 6455 versus 22 dBm at 6135 and 6775 (identical on every capture). So 3 dB of the dip is the AP running lower there — flip the chart to Source-normalized and that much flattens out. Antennas that dip only ~3 dB (Eightwood 6E, Foxeer long) are fully explained by it; the ones that dip more (Oscium, Osprey, MSI) carry an extra path/pattern null on top. Either way it is not antenna quality — and the old single-6 GHz-point method, sampling 6135 only, would never have seen it. The same beacons show 6 GHz sourced 2 dB hotter than 2.4/5 GHz overall, which the normalized view also corrects for.

iperf3 to the wired NUC

Controlled throughput · band average

NanoPi associated to the R670 on each swept channel, 80 MHz EHT, 15 s each way; values are the per-band mean of ↓ / ↑ in Mbit/s.

Antenna	2.4 ↓	2.4 ↑	5 ↓	5 ↑	6 ↓	6 ↑

The eight, by form factor

The contenders

Grouped pocket → desk, with polarization tagged. Photos and buy links where the antenna came from a listing — the three bare whips (Oscium, Ducky, MSI) shipped without one; MSI came bundled with a Wi-Fi card.

Polarization is a non-event for throughput. RHCP, LHCP and linear all land within ~10% on 5 and 6 GHz — Osprey (LHCP) even tops the linear Dixingtek on 5 GHz. The circular-pol antennas' only deficit is 2.4 GHz, and that's because they're 5.8 GHz-tuned FPV parts, not a handedness penalty.

How every number was made

Method & provenance

The rig

UNITNanoPi + Intel BE200 (wlp1s0f0); only this antenna changed between sets.
SOURCERuckus R670, stepped MANUAL across 9 channels via the cloud API — ChannelFly off at the venue so the pin actually holds; each move verified over the air before measuring.
PER CH60 s beacon capture (RSSI + ambient 3rd-party APs) then an iperf3 run to the wired NUC.
POWERAP pinned to MAX; actual per-channel TX power read back from each beacon's TPC report (20 dBm on 2.4/5, 22 on 6, 19 at 6455) and used for the source-normalized view.
SERVERNUC, wired, iperf3; reverse-path filter cleared each run.

Why this replaces the first cut

6 GHzA generic radio can't legally beacon 6 GHz (No-IR/passive), so only the certified R670 sources it — the reference AP must be the one that's swept.
SWEEPEvery antenna is walked across the same channels, so a per-channel null is visible and can't be mistaken for antenna quality.
FRESHBuilt only from this 09-03→09-04 sweep campaign. No data from the earlier single-point run is carried over.

Original research and documentation by WLAN Geeks. Sources and credits · Report a documentation issue