Channel Bonding and Frame Aggregation

What it is

Channel bonding makes a channel wider, and frame aggregation makes each transmission carry more data.

Both ideas raise throughput, and both have a price. A wider channel uses more spectrum and picks up more noise. Larger aggregated frames save overhead, but a lost frame costs more to repeat.

They are the two main ways in which 802.11n gets more from the same air.

How it works

Bonding joins a control channel and an extension channel, and aggregation packs several frames into one transmission.

Channel bonding

A normal channel is 20 MHz wide, but 802.11n can join two neighbouring channels into one 40 MHz channel.

The first is the control channel, which carries management and the legacy parts. The second is the extension channel.

The frequency setting names the control channel. The extension channel lies either above or below it. You tell RouterOS which with the channel width value:

/interface wireless set wlan1 channel-width=20/40mhz-Ce
/interface wireless set wlan1 channel-width=20/40mhz-eC
/interface wireless set wlan1 channel-width=20/40mhz-XX

Think of the letters in order of frequency. In Ce the control channel comes first, then the extension. In eC the extension comes first.

The 20 in the value shows that a 20 MHz fallback exists. If the extension channel is busy, the link can use 20 MHz.

Bonding and the spectrum

Two things follow from bonding. First, a 40 MHz channel uses the room of two 20 MHz channels, so there are fewer places to put it. Second, a wider channel collects more noise, so the signal to noise margin is smaller.

Both ends must agree

A station uses the width that the access point offers, within what it supports. On a point to point link between RouterOS devices, set the same channel width and extension on both sides. If the extension channel is noisy, check it with the frequency tools, or move the control channel.

Frame aggregation

Every Wi-Fi frame has fixed overhead, such as headers, waiting times and acknowledgements. When frames are small, the overhead is a large part of the airtime. Aggregation joins several frames so that the overhead is paid once.

802.11n has two methods:

With A-MSDU, one lost bit means that the whole large frame is lost and must be repeated. With A-MPDU, each frame is protected on its own, so a loss costs less. A-MPDU is usually the more robust choice.

Aggregation settings

RouterOS gives some control over aggregation:

/interface wireless set wlan1 ht-ampdu-priorities=0,1,2,3
/interface wireless print detail where name=wlan1

The first command allows A-MPDU for the listed priorities. The second shows the aggregation settings that are in force.

• ht-amsdu-limit is the largest A-MSDU that the interface may build.
• ht-amsdu-threshold is the largest frame that may be placed into an A-MSDU. Larger frames are not aggregated.
• ht-ampdu-priorities lists the packet priorities for which A-MPDU may be used.

The priorities are the same priorities that WMM uses, so the aggregation choice connects with traffic marking.

When aggregation helps

Aggregation helps most with a steady stream of data, such as a large download. It helps less with small, sparse packets, such as voice, because there is nothing to join and joining would add delay.

On a link with many errors, large aggregated frames can hurt. Each loss costs more airtime. If a noisy link performs badly with aggregation, test with smaller limits.