Saturday, 10 February 2018

Noise Floor Penalty of Wider Channels in Wi-Fi Networks

I’ve been told a number of times that although wider channels in a Wi-Fi network generally provide a higher connection speed (and hopefully a higher throughout), it comes at the cost of increasing the perceived noise floor of the client device. I thought it would be interesting to test this out for myself.

With the advent of 802.11n, it became possible to bond together the 20MHz wide channels of earlier standards in to 40MHz channels (though in reality, this was only practically feasible on the 5GHz band).

Several years later, 802.11ac enabled us to bond together even larger chunks of contiguous channels and achieve 80MHz and 160MHz wide channels on the 5GHz band. Though 80MHz channels are not feasible in many environments and 160MHz is limited to very niche scenarios, they nonetheless are options.

Theoretically, each time we double our channel width, we are going to double our connection speed and our throughput (there are some protocol efficiencies achieved which mean we may slightly more than double our throughput, but lets keep it simple).

However, each time we double our channel width, so the theory goes, we also double the amount of noise our Wi-Fi station is exposed to. I guess this is (kind of) intuitive…

So here’s a quick check using my trusty Macbook Pro:

First of all, I connected to an SSID that is configured for a 20MHz channel width and took a look at the noise floor. In this instance, the noise is reported as –98dBm:

Screen Shot 2018-02-04 at 15.11.35

Fig 1 – 20MHz channel width

Next, I reconfigured the same SSID for a 40MHz channel width. Now, the noise floor was reported as –95dBm. Our noise floor had gone up by 3dB. (Sidenote: an increase of 3dB indicates a doubling of power).

Screen Shot 2018-02-04 at 14.57.30

Fig 2 – 40MHz channel width

Then, I reconfigured the SSID to double the channel width of the same SSID to achieve an 80MHz channel width. The noise floor increased by another 3dB (i.e. doubled again), changing from –95dBm to –92dBm:

Screen Shot 2018-02-04 at 15.17.31

Fig 3 – 80MHz channel width

The observed effect fits the theory. There, that’s 10 minutes of your life you’ll never get back….but never mind, at least you didn’t choose accountancy as a profession.

OK, so why are we really interested in this anyway? Well, the other aspect to this is that if you look back at the results again,  you can see that the received signal level has remained the same in each instance, but our noise floor has crept up. Our signal to noise ratio (SNR) has gone down by 6dB as we moved from 20MHz to 80MHz:

  • Rx_sig_level – Noise_level = SNR
  • 20MHz: –54dBm - -98dBm = 44dB
  • 40MHz: –54dBm - -95dBm = 41dB
  • 80MHz: –54dBm - -92dBm = 38dB

Our SNR values are  still very good in this example, but in an environment with a higher noise floor, a 6dB difference could make a significant difference. Many higher MCS values for 802.11n and 802.11ac rely on having high SNR values for successful operation. This means that a lower SNR could, in some environments have an impact on the higher speeds that are achievable across a network.

Well, that’s pretty much it. Hope this has been of interest.

(By the way, if you wonder how I got the additional Wi-Fi connection info on the Mac, when you click on the Wi-Fi symbol on the Mac top-bar, press the ‘alt’ key at the same time to see the enhanced connection info).