Background

It always kind of bothered me that tech companies list the bandwidth of wireless devices way the hell higher than what they're really capable of.. It's like storage devices where they list a gigabyte as 1,000,000,000 bytes instead of 1,073,741,824 bytes which is a real gigabyte. Sorry, got off track there. Anyway, I decided to test out different levels of wifi network and see what the actual data rate is. This isn't definitively accurate, like the RFC 2544 test used to qualiy carrier ethernet would be, but it gives you a good idea of how it will handle real internet traffic in a best case scenario.

Test Setup

For the test setup, I used two Mikrotik Routerboard RB911G-5HPac 802.11a/n/ac radios with their output power lowered to 1dBm (1mW), connected directly to each other with a short length of Andrew Heliax low-loss, low delay coaxial cable and a 40dB microwave attenuator. I also tried connecting them through a section of WR159 flexible waveguide, but it didn't lower the RSL or change any of the test results.

Both radios reported a -40dBm received signal level (RSL), which is equivalent to an "Excellent" signal level on a Windows PC, but not so high that it would overwhelm the receiver and degrade performance, which starts to happen at signal levels over -20dBm. The radios reported a noise floor of -107dBm for a signal to noise ratio (SNR) of 67dB.

According to the published documentation, 802.11n and 802.11ac only report their maximum bandwidth of 300 Mb/S and 866.6 Mb/S respectively with two channels. Ideally, these channels would be connected to separate antennas oriented orthogonally (crossed) to one another on both the receive and transmit sides. The only situation where this is likely to happen in the real world is in a fixed point-to-point link with dual polarized antennas. In the context of an access point this is much less likely to happen, since the antennas on the AP are usually oriented vertically and the orientation of the antenna(s) on the client device can change depending on how it's being held. So, I started off by measuring single channel bandwidth.

Test Results, Single Channel

Mode/Channel Width	Reported Bandwidth	Actual RX Bandwidth (RX Only)	Actual RX Bandwidth (Simultaneous TX + RX)
802.11a/20 MHz	54 Mb/S	23 Mb/S	12 Mb/S
802.11n/20 MHz	72.2 Mb/S	52 Mb/S	30 Mb/S
802.11n/40 MHz	150 Mb/S	99 Mb/S	59 Mb/S
802.11ac/20 MHz	86.6 Mb/S	62 Mb/S	36 Mb/S
802.11ac/40 MHz	200 Mb/S	125 Mb/S	74 Mb/S
802.11ac/80 MHz	433.3 Mb/S	200 Mb/S	84 Mb/S

As you can see, the actual bandwidth is much lower than what's claimed. Where did the excess bandwidth go? The payload for this test was randomly generated TCP packets. If I had used UDP packets, the results would have been much higher; for example, I got 337 Mb/S of receive only bandwidth on 802.11ac/80 MHz using randomly generated UDP packets. The reason I used TCP packets is because most internet traffic runs over TCP, including web, file transfer, and email. UDP packets are mostly used where a real-time response is more important than message integrity, such as live video, VoIP, and the like.