Channel Bonding Caveats – Over and Above Spectrum Hogging


Popular literature on 802.11ac describes 40 MHz and 80 MHz operation (channel bonding) as doubling and quadrupling of the data rate, respectively. Every time I saw that mentioned, the following question came to my mind.

When radio transmits over 40 MHz (or 80 MHz) channel, is the total transmit power proportionally increased over 20 MHz to maintain the SNR (signal to noise ratio)? And, how is the data rate multiple with channel bonding distributed over the cell?

This question nagged me like a little stone in the shoe that is impossible to ignore. My subsequent findings from the lab tests show that the popular literature is only partially true. Read on to find out why.

Hit on SNR with Channel Bonding

Look at the spec of any enterprise 802.11ac AP and you will notice the same transmit power for the radio in 20 MHz, 40 MHz, and 80 MHz. That makes sense, because you would always use the maximum power that the radio can transmit irrespective of the channel bandwidth. In OFDM, the total transmit power of the radio is almost equally distributed across the subcarriers. So, a subcarrier in 80 MHz would have half the transmit power (and hence half the receive SNR) than that of one in 40 MHz, and the one in 40 MHz would have half the receive SNR than that of one in 20 MHz. What is the net effect of these two countering forces – more subcarriers due to channel bonding, but reduced SNR per subcarrier at the same time?

Different Net Effects – Close/Far from AP

Performance tests with 3×3 enterprise AP and 2×2 Macbook Air provide insights into the answer. Figures below show the observed distribution of MCS (Modulation and Coding Scheme) indices for different channel bandwidths in the two regions. At about 50 feet NLOS (Non Line Of Sight) from the AP, transmission is dominated by 256 QAM to 64 QAM. At about 100 feet NLOS from the AP, transmission is dominated by 16 QAM to QPSK. Closer to the AP, the MCS is maintained even with channel bonding. However far from the AP, the MCS drops with channel bonding.

Why so? My explanation is that when clients are closer to the AP, there is headroom in SNR (incident power higher than receive sensitivity) that gets utilized when the channels are bonded. Step away from the AP and you lose the headroom, so MCS drops with channel bonding. The drop is only somewhat countered by the higher transmit power rating for the radio at lower MCS and some additional frequency diversity due to wide channels.

MCS Distribution Close to AP

MCS Distribution Close to AP

MCS Distribution Far from AP

MCS Distribution Far from AP

Different Data Rate Multiples – Close/Far from AP

Figure below shows the data rate averaged over frames in the two regions of the cell. Closer to the AP, there is 2x data rate multiple for every doubling of the channel bandwidth. Far from the AP, the multiple is smaller – in this case about 1.5x for every doubling of the channel bandwidth.

Data Rate Multiples at Different Distances from AP

Data Rate Multiples at Different Distances from AP

Wide channels result in spectrum hogging and have associated complexities, as I discussed in my earlier post “802.11ac (Wave-1): Network Engineering Insights.” In addition, the above analysis shows that one cannot expect the data rate to simply double and quadruple with 40 MHz and 80 MHz channels as the popular literature states. Keep the above caveats in mind while designing networks with wide channels.

Have you experienced these issues in your 802.11ac deployments?

Hemant Chaskar

Hemant Chaskar is Vice President for Technology and Innovation at AirTight. He oversees R&D, product strategy, and intellectual property.Hemant has more than 15 years of experience in the networking, wireless, and security industry and holds several patents in these areas.

Twitter LinkedIn YouTube 

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>