SpecAn and Channel Bonding

I was experimenting with Aruba Networks ability to use their AP units to spectrum analyze the RF and I thought of writing this post with regards to channel bonding capabilities of 802.11n and 802.11ac.  I know that many of my clients do not understand how channel bonding works – at least not completely.  So, I thought I would pass along some insights.

First off you must remember that all management traffic (beacons and control frames) must be sent in 20MHz wide channels.  This is to ensure backwards compatibility with devices that do not support channel bonding.  They will be sent on the Primary channel so that all devices can hear them.  When a device that can support the channel bonding wants to send traffic using the bonding channels, it will send out the Contention Free Start frame.  This tells all devices within “earshot” to set their back-off timers so that it can maintain access to the medium for the time that it is transmitting.  When it is complete, it will send the Contention Free End frame.  All of this is done on the Primary 20MHz channel.

Once the CF-Start frame is sent, the device that wants to use the channel bonding then starts sending the data on both the Primary and Secondary channels (or more than just two if using higher than 40MHz).  Now, in most busy networks and especially in networks with multiple AP units using different channel sets you likely will not be able to discern this from the spectrum capture easily so I decided to use my relatively quiet home network to demonstrate.  It is set to use 40MHz with Channel 153 as the Primary and 149 as the Secondary channel.  With that in mind, examine the image below:

5GHz_SpecAn

Note the heavily colored Ch 153 and the sporadic colored bands in Ch 149.  For those that do not know, the colors indicate the amount of airtime being utilized – i.e. traffic intensity, if you will.  It does this by representing the strength of the signal in a visual color such that when the radio is transmitting, especially as close as this was, the dBm value is high (red) versus the dBm value being low (Light blue – Blue).  Since this is happening extremely fast the color variation is an easy way for our eyes to interpret how often the radio is transmitting and how strong, therefore the red areas indicate many 1’s and 0’s being sent versus other areas where its not so much – in the yellows and greens – to very little in the light blues.

So why is the Ch 153 column so clearly colorful while the Ch 149 column is not?  Simple – the areas you see color in the Ch 149 column is where a device in my network was utilizing the 40MHz channel bonding capabilities to transmit its data!  We can also capture it using the Real Time FFT graph, although it is a little harder.  See the images below:

RealTimeFFT_20MHz_Size

RealTimeFFT_40MHz_Size

The top graphic shows a moment in time when the system was transmitting using only 20MHz while the bottom using 40MHz.

Okay, so why did I think this was important enough to blog about? Well think about this on a bigger scale.  I have been to client sites where it was thought that channel bonding was a good feature to use so they set their system to use 40MHz bonding.  Unfortunately, their automatic radio management set their AP units based on the number of channels available – and sadly they did not enable the UNII-2c or more commonly known as the UNII-2 Extended channels.  Their deployment also had more AP units than available channel sets so there was plenty of channel re-use.

All this helped to initiate poor Wi-Fi performance even for those client devices that could use the Channel bonding – and some tried  – mainly due to the lack of channel diversity.  Picture this, AP Alpha is using ch 44- (Ch 44 is Primary and Ch 40 is Secondary).  Nearby AP Bravo is using ch 40+ (Ch 40 is Primary and Ch 44 is Secondary) Remember those CF-Start and CF-End frames sent on the Primary channels?  When Device #1 sends the CF-Start frame for transmission to Alpha, Device #2, associated to AP Bravo, doesn’t hear it and can start transmitting when Device #1 starts sending data frames, potentially corrupting them.

I did not have visibility into their WLAN system configurations at first; however, since I was using a spectrum analyzer I noticed a couple of items right away.  First, they had too much channel overlap for channel bonding. Second, I could see some intensity differences within the bands that indicated to me they likely had more 20MHz capable clients than 40MHz.  A quick check within the reporting capabilities of their WLAN system proved I was correct.  After taking my recommendation to go back to using 20MHz channel widths their users’ overall experience went up even though they were not using the maximum PHY rate available by channel bonding.

Since we rarely encounter full Greenfield deployments, especially in the Enterprise space, it is increasingly important to educate our users and clients in the prudent use of the features and functionality of these more advanced systems.  Additionally, I see this occurring more as we update infrastructure and increase density in our designs without additional spectrum space as Wi-Fi usage increases in importance.

One final thought, I find it increasingly prevalent that users believe more PHY speed equals better performance… ah a discussion for another day!

7 thoughts on “SpecAn and Channel Bonding

  1. Very interesting blog post, thanks for sharing.

    Although maybe I misunderstood, but aren’t the CF End/Start frames for networks using PCF only? Also, wouldn’t RTS/CTS prevent two bonding BSS:es (different primary channels) from colliding?
    / B

    Like

    1. CF frames exist in DCF too. RTS/CTS are sent on the Primary channel only as well so if you have two AP units that have adjacent Primary channels clients won’t hear the transmission on the other Primary as they are only listening to the one they are currently using for their traffic.

      Like

      1. I’m confused, is there some document/book that explains this in detail? I did some quick research and I got the impression that CF start/end is PCF only frames (and I can’t find it when reading up on DCF). For exampe: https://mrncciew.com/2014/10/02/cwap-802-11-control-frame-types/

        Regarding CTS/RTS: a HT/VHT client would duplicate and send these frames in non-HT format on all 20MHz channels, correct? So that other AP unit on the adjacent channel would back off.

        Like

      2. Quick clarification – you are correct in that CF Poll frames are part of the PCF portion of the standard. PCF is an optional portion added on top of DCF. Once the AP (the Point Coordinator in PCF) gains access to the medium using normal DCF processes, it sends out the Management frames which includes the CF frames on just the Primary 20MHz channel. However, only data frames are sent in HT and VHT format. Therefore, only devices that are listening on that 20MHz channel would hear the CF frame. There isn’t any repeating outside the original Primary channel as I think you are suggesting. So, if two AP units are on adjacent channels as their Primary ones, a client would not hear the CF start frame on the other channel that it is not listening on and could potentially start transmitting simultaneously as the HT or VHT data frames from the other channel start. This potentially could cause collisions and impact performance, especially in a high density situation. This is one reason why going higher than 40MHz in high density deployments is something that is carefully and thoughtfully implemented, if at all. It is also why Overlapping BSS’s (OBSS) is something that should be carefully managed and may not be handled very well when using automated systems.

        I have personally seen this occur on at least a few occasions. And they don’t even have to be physically right next to one another – they could even be a floor or two away and still have an impact depending upon other factors.

        Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s