2.4 GHz vs 5 GHz for Wi-Fi CSI: coverage, subcarriers, and interference
2026-10-11 · 1 min read
After bandwidth, the second decision in a CSI setup is which band. The trade-off is blunt: 2.4 GHz reaches further, 5 GHz gives you more frequency resolution and cleaner spectrum.
1. The numbers
| 2.4 GHz | 5 GHz | |
|---|---|---|
| Typical width | 20 MHz (HE20) | 80 MHz (HE80) |
| Subcarriers (this hardware) | 64 | 256 |
| Range bin (c/2B) | ~7.5 m | ~1.875 m |
| Penetration | better through walls | worse [1] |
| Noise floor | crowded (microwave, BT, neighbours) | usually cleaner |
| Channel overlap | 1/6/11 only | many non-overlapping |
Doppler resolution is unaffected by band — that comes from your measurement rate.
Ad slot (AdSense/Ezoic once live)
2. What we measured
- 5 GHz, 80 MHz, dedicated AP: 256 subcarriers, 95 records/s for 10 minutes, 57,293 records, no gap > 5 s — grade E, raw capture and scripts published [2] [3];
- 2.4 GHz: we also observed a downlink collapse on 2.4 GHz, but the location was congested (a neighbouring AP on an adjacent channel), so we treat it as unresolved rather than a firmware finding. On 2.4 GHz you are fighting for airtime with everything else in the house.
3. Practical rules
| Situation | Pick |
|---|---|
| You need resolution (respiration, fine motion) | 5 GHz, as wide as your clients allow |
| Sensors far away / through walls | 2.4 GHz, accept coarse resolution |
| Legacy clients that must stay usable | keep them on 2.4 GHz, reserve 5 GHz for the CSI link |
| Dense apartment building | 5 GHz (and check DFS channels) |
4. Two implementation notes
- Do not put the CSI AP on the same radio as the clients you care about if you can avoid it — a dedicated interface (or a dedicated radio) keeps the measurement rate stable.
- Measure your noise floor before you blame the algorithm: scan both bands from the AP itself
(
iw dev <iface> scan), log RSSI of neighbours, and keep that with your dataset — it explains most "mystery" degradation.
Related: 80 MHz vs 40 MHz · OpenWrt slow after idle · dataset
References and data sources
Every external claim carries a source and an evidence grade (grading scheme). Grade E marks first-party, reproducible measurements from this project.
- Grade B · peer-reviewed
Measurement and Modeling of Penetration Loss in the Range from 2 GHz to 74 GHz
Y. Du, C. Cao, X. Zou, J. He · 2016 IEEE Globecom Workshops (GC Wkshps) · 2016
doi:10.1109/glocomw.2016.7848958 · citations 18 · verified: CrossRef API, 2026-10-11 · Measured and modelled penetration loss for common building materials above 2 GHz — used here to bound how much a wall attenuates. - Grade E · first-party measurement
This project's 10-minute raw capture (57,293 records / 600.0 s, SHA-256 00373f83…95e2)
Router CSI Lab · 2026
csi.okwujun.pw/zh/datasets/ · verified: Reassembled chunks verified byte-identical (2026-10-11) - Grade E · first-party measurement
Router CSI Toolkit — capture/analysis/watchdog scripts (MIT)
Router CSI Lab · 2026
csi.okwujun.pw/zh/tools/ · verified: Repository bundle has a SHA-256; scripts run on real hardware