2.4 GHz vs 5 GHz for Wi-Fi CSI: coverage, subcarriers and interference

By Wu Jun · Updated 2026-10-29 · 5 min read

After bandwidth, the next choice is 2.4 GHz or 5 GHz. A rough summary is that 2.4 GHz propagates farther and 5 GHz usually offers wider channels and cleaner spectrum. For CSI, however, coverage, subcarrier count, measurement rate, interference and client compatibility must be tested together. Signal bars are not enough.

This page provides a reproducible A/B process. The 5 GHz side has a complete 10-minute capture; the 2.4 GHz side has one unresolved field anomaly, which we will not turn into a fake apples-to-apples benchmark.

0. Bottom line

• Test 2.4 GHz first for longer range or better penetration;
• test 5 GHz first for finer frequency resolution, more subcarriers and cleaner spectrum;
• in dense apartments, 2.4 GHz coverage can be cancelled by co-channel interference;
• choose from a same-device, same-client, same-position A/B, not from a product table.

1. What the band changes

"Usually" is not a device guarantee. Chip, firmware, regulatory domain, channel and antennas matter. Count data_num, measurement rate, RSSI/SNR and loss in your own capture.

2. Numbers from this project

The complete 5 GHz capture, checksum and scripts are public ^{[2] [3]}. We do not have an equivalent 2.4 GHz dataset, so the two columns are not a fair throughput or recognition-accuracy comparison.

3. The 2.4 GHz anomaly we observed

In one home environment, 2.4 GHz downlink entered a low-rate state around 5.6/8.7 Mbps even after CSI was disabled. An adjacent AP was present and the environment was congested. We confirmed the slow link, but not whether the cause was interference, client power save, rate control or the band itself. We also did not establish whether another 2.4 GHz channel would reproduce it or how CSI recognition accuracy changed.

That is what "unresolved" means. Turning one field observation into a firmware root cause would mislead later reproducers.

4. A fair 2.4 GHz vs 5 GHz A/B

4.1 Fix variables

• same AP, antenna and client;
• same room, height and orientation;
• fixed TX power, country and client power-save policy;
• controlled LAN traffic, not a public download;
• idle for 60 seconds before each run and record recovery time.

4.2 Scan before selecting a channel

# record neighbour APs, channels, RSSI and width
iw dev <iface> scan | grep -E "SSID|signal|freq|HT operation|VHT operation|HE operation"

# current interface and station state
iwinfo <iface> info
iw dev <iface> station dump

On 2.4 GHz compare 1/6/11 rather than accepting auto-selection. On 5 GHz record 20/40/80 MHz, primary/secondary channel, DFS state and any radar/channel-switch event.

4.3 Capture per run

Run at least three 10-minute sessions per band and alternate the order (2.4, 5, 5, 2.4) to reduce time drift.

5. What RSSI misses

A strong RSSI does not prove:

• complete subcarriers;
• a stable independent measurement rate;
• no long gaps;
• clean phase after rate or chain switching;
• low false alarms for the intended sensing task.

A 2.4 GHz link can have higher RSSI while suffering bursts from neighbour APs, Bluetooth or a microwave. A weaker 5 GHz link can still provide more stable frequency features because it has 256 subcarriers and cleaner spectrum. Keep signal strength and data quality separate.

6. Interference checklist

• 2.4 GHz: are 1/6/11 saturated? Are microwave, Bluetooth and wireless peripherals nearby?
• 5 GHz: DFS radar events, automatic channel changes or dynamic width?
• Are neighbour APs using the same 80 MHz primary/secondary channel?
• Is a large client consuming the airtime?
• Timestamp every scan, not just the failed run;
• store the raw scan output with the CSI capture.

Do not accept "it felt faster after changing channel" as evidence. Repeat six throughput runs and a 10-minute capture.

7. When to choose each

8. Common errors

9. Conclusion

• 2.4 GHz and 5 GHz trade propagation against interference and bandwidth;
• 5 GHz HE80 gives 256 subcarriers and finer frequency resolution here, but needs SNR and compatibility;
• 2.4 GHz coverage must be verified in the actual channel environment;
• a band A/B records direction, group rate, gaps, throughput and scans;
• unresolved field anomalies stay anomalies until repeated and explained.

Related: 80 versus 40 MHz · OpenWrt troubleshooting · dataset · reproduction guide