An overview of how the various modes apply to frequency band, the number of channels available, and the expected data rates and throughput.
This slide helps to explain the first two 802.11n innovations with a picture. Channel bonding is analogous to taking two pipes of the same diameter, and making one pipe that has twice the diameter of the original ones. The bigger the pipe, the more water can flow. Similarly, by taking two channels and “bonding” them into one channel that’s twice as wide results in much greater speed. The price to be paid for higher speed is less unique, distinct channels. The 2.4GHz example here shows how 3, 20 MHz channels reduce down to 1, 40MHz channel and 1 remaining 20MHz channel. So… higher speed per channel, but less channels overall. The mixed-mode effect is seen next. If only “new” .11n clients are allowed to operate in the bonded 40MHz channel, then the maximum throughput of 802.11n is possible. However, as soon as “old” .11g devices are allowed to mix with the n devices, the throughput can drop to as little as a quarter of before. The same phenomenon happens in 20MHz channels. The bottom line: understand what clients you will need to support and in what band. Channel bonding and legacy support, in isolation, are a good thing. In practice, in a system context, things may not be so clear.