Bonding to 40 MHz Next to a Loud Neighbour
A 5 GHz WLAN serves an open-plan office of 60 workstations. The AP runs a 20 MHz channel at 17 dBm into an 8 dBi omnidirectional antenna, and the farthest desk measures -70 dBm RSSI. A spectrum sweep at that desk shows the noise floor sitting at -95 dBm across the band, plus a neighbouring tenant's AP transmitting continuously at -78 dBm on the 20 MHz channel immediately above yours. Users complain that file copies crawl. Your manager has read that channel bonding doubles throughput and asks you to enable 40 MHz so that the two 20 MHz channels become one bigger pipe.
1. Calculate the SNR at the farthest desk today and after bonding to 40 MHz. What do you recommend, and why?
Read the worked answer
Today (20 MHz): SNR = -70 - (-95) = 25 dB. Comfortable for 256-QAM / high MCS rates.
After bonding to 40 MHz: the neighbour's -78 dBm signal now sits inside your channel, and the thermal noise also doubles (+3 dB) because you are measuring over twice the bandwidth: -95 + 3 = -92 dBm.
Combine them: 10·log10(10^-9.2 + 10^-7.8) = 10·log10(6.3×10^-10 + 1.59×10^-8) ≈ -77.8 dBm
New SNR = -70 - (-77.8) = 7.8 dB.
Recommendation: do not bond. SNR collapses by roughly 17 dB because a bonded channel is only as clean as its dirtiest 20 MHz half. At 7.8 dB SNR the link falls back to the lowest-order modulations (BPSK/QPSK), so you would trade a wide channel for a slow one. Fix the interference first - move to a 20 MHz channel well away from the tenant's, ideally in a different UNII band - then re-measure before considering wider channels.
2. A colleague proposes raising the AP from 17 dBm to 20 dBm to win back the lost SNR. Evaluate that proposal.
Read the worked answer
Three separate problems:
- It cannot work numerically. 3 dB of extra AP power against a 17 dB SNR deficit recovers less than a fifth of what was lost. The dominant term is the interferer's -78 dBm, and raising your own power does nothing to it.
- It is one-directional. Transmit power at the AP improves only the downstream RSSI at the client. The client still transmits at its own power (typically 14–15 dBm), so the AP's view of the client is unchanged. You would create an asymmetric link where the AP can reach the client but the client cannot reliably reach back - the classic cause of associations that succeed but traffic that stalls.
- It worsens the interference. Every dB you add is a dB of co-channel energy you impose on the neighbour (and on your own second-channel reuse), which invites retaliation and raises the noise floor further.
The right levers are channel choice and interferer removal, not power.
3. Before enabling any wider channel, what would you measure, and what would make you reject a channel that looks quiet on an amplitude plot?
Read the worked answer
Sweep the entire 5 GHz band at both the AP location and the client location - a clean AP site does not mean a clean client site, and the client's neighbourhood is what determines its SNR.
Look at channel utilisation (airtime duty cycle), not just amplitude. A -78 dBm interferer at 60% duty cycle destroys throughput; a -70 dBm interferer at 2% duty cycle is nearly harmless. Amplitude alone cannot distinguish these.
Also confirm before committing:
- The AP radio supports the wider channel in that specific sub-band (some bands and some regulatory domains restrict width).
- The clients actually support 40/80 MHz - a bonded channel is wasted on 20 MHz-only clients.
- Whether any radar-avoidance (DFS) obligations apply to the band you pick, since a DFS channel will vacate on detection.
A channel that looks quiet but shows high utilisation, or that is only quiet at the AP and not at the client, should be rejected.