89

Wi-Fi congestion, especially in the 2.4GHz range, is a serious problem in some areas. It is widespread enough that there are many guides to choosing a less congested channel. E.g. https://www.howtogeek.com/197268/how-to-find-the-best-wi-fi-channel-for-your-router-on-any-operating-system/

Given that most routers default to automatically choosing their channel and the hardware seems capable of detecting conflicting networks, why don't they do a better job of channel selection?

  • 13
    Solution. Use 5GHz not 2.4. – Tetsujin Apr 5 '18 at 18:26
  • 28
    @Tetsujin That's a bit blunt, not really a solution. – Ultrasonic54321 Apr 5 '18 at 18:27
  • 4
    The problem is the environment can keep changing and it depends on the access point how often it checks the best channel. If it's just at boot and the device isn't restarted often, then it could be months between setting the best channel. So it depends on the device. – HelpingHand Apr 5 '18 at 18:32
  • 6
    @Ultrasonic54321 - it pretty much solves everything. If you live in a high-density urban environment, just check how many access points are on 2.4 & how many on 5. Here it's about a 50:1 ratio. I'm the only person in 'view' on 5, the other 50 I can see from here never changed it up from 2.4. – Tetsujin Apr 5 '18 at 18:45
  • 7
    Everyone has a different situation. Your remarks are not wrong, but they aren't right either. – Ultrasonic54321 Apr 5 '18 at 18:53
133

The failure of Wi-Fi APs to pick 2.4GHz channels well comes down to a small handful of issues:

  • Most only pick a channel at boot time, but a channel that was good when the AP was last rebooted may have become a poor choice days, weeks, or months later.
  • Most do not want to delay booting by spending long enough to truly evaluate every channel, so they use poor heuristics like "just pick the channel where we see the fewest APs", which doesn't necessarily correlate to which channel will provide the best throughput and reliability. Even worse, these oversimplified heuristics can cause problems like choosing a channel that partially overlaps with channels other APs are on, which will cause APs to interfere with each other without being able to cooperate with each other like they would if they were on the exact same channel.
  • Most don't even have the spectrum analyzer hardware necessary to truly evaluate the RF interference on each channel; they have Wi-Fi radios and focus on interference from other Wi-Fi devices, and are fairly ignorant of interference caused by non-Wi-Fi devices such as Bluetooth, microwave ovens, cordless phones, wireless subwoofers, baby monitors, wireless cameras, and more.
  • Creating an AP that has the hardware and the algorithms to choose channels well not just at boot, but to keep re-evaluating the channel choices later, and change channels when there would be benefit to do so, is both expensive and fraught with potential interop problems. Not all clients are great at honoring channel switch announcements from the AP, so an AP that changes channels on the fly risks having clients fall off the network every time it does so.
  • 33
    And firmware developers are too chicken to lock the choices to 1/6/11. – chrylis Apr 5 '18 at 23:25
  • 8
    Add in that the channel congestion where the router is situated may differ from the channel congestion where your end device is located. – Gary Apr 6 '18 at 2:27
  • 13
    @detly 1/6/11 are the only channels you should be using because they are the channels that don't overlap at all. If, for example you choose 3 because there is a lot of people on 1 and 6, you are now congesting everything between 1 and 6. I think. – Matt M. Apr 6 '18 at 3:22
  • 11
    @CaptainMan Any time the batteries died in my Gameboy while in the car I'd end up reading that label over and over. – Adonalsium Apr 6 '18 at 13:14
  • 4
    You need to shorten days to 'the next time you have a power cut'. Think about what happens when 80 AP's all boot up after a power cut, all see that channel 3 is free completely and all then decide to use it, until the next power cut ;) – djsmiley2k Apr 8 '18 at 10:57
49

The overarching problem here is that the 2.4GHz band is completely saturated in any moderately populated area. In addition, there are only 14 channels, depending on country, available to use. Out of those 14 only 3 channels don’t overlap and interfere with each other. And that is only true if the device uses only 20MHz of bandwidth and not the 40MHz bandwidth available on some access points.

All properly configured Wi-Fi routers should only use channel 1, 6, or 11 at 20MHz bandwidth. An access point stomps on the signals of any nearby access points for at least 2 channels higher and 2 channels lower from itself. Worse if it’s on 40MHz bandwidth.

When access points can see each other, on the same channel, they will cooperate and share the air space. If two access points are using nearby, but different channels, then they stomp on each other and each collision results in lost data.

Unfortunately, most modern Wi-Fi routers, for simplicity, default to auto-channel selection. However, they do not adhere to the 1, 6, or 11 rule. Instead they use a proprietary algorithm that is probably based on the usage of each channel. This causes severe and unavoidable interference of nearby networks, practically rendering the 2.4GHz band useless in some areas. In addition, the auto-channel selections usually only happen during a reboot or rarely at all. So the channel selection can quickly become stale as nearby access points also jump channels and compete to find the “cleanest” channel. To make things worse, the channel selection is based on what the AP hears, and not what the client hears, which may be closer to a different set of APs.

So, the problem is not the selection mechanism, but the fact that the 2.4GHz band is just completely saturated. Not only by Wi-Fi access points, but by cordless phones, microwaves, Bluetooth, baby-monitors, wireless cameras, and any number of other technologies.

The answer is to use the 5GHz band. There are dozens of 5GHz channels available. None of which overlap with others if the standard 20MHz bandwidth setting is used. This means that all devices using the 5GHz band can cooperate with each other without interfering. Unfortunately, Wireless-N and especially Wireless-AC allow for wider channels which overlap in an attempt to provide greater throughput. So, even in the 5GHz band, you should be conscious of co-channel interference and choose your settings wisely, rather than utilizing auto-channel selection.

In a densely populated area, the use of wide channels will provide little, if any benefit and could actually make things worse.

  • 5
    Suggesting people use tiny old 20MHz channels in 5GHz is poor advice unless you warn them that it'll cut their 802.11ac performance to less than a quarter of what it could be. – Spiff Apr 5 '18 at 19:11
  • 1
    Also, what the AP sees/hears is different to what the AP clients here, so it is making decisions on slightly wrong information. – davidgo Apr 5 '18 at 19:26
  • 1
    @davidgo Agreed. Channel selection should be planned. Auto selection only causes issues. However, the 5Ghz band eliminates a lot of it. – Appleoddity Apr 5 '18 at 19:38
  • 6
    @Spiff I appreciate that it might reduce performance. But minimizing co-channel interference is one of the biggest goals in wifi planning. In any urban, or moderately populated area, this is going to mean using 20Mhz bandwidth channels. Attempting to use anything more will either make things worse, or not do much of anything at all. 802.11AC can use per-frame bandwidth decisions to avoid co-channel interference but is rendered useless by wireless-N devices. Ultimately, you will see little difference trying to use larger bandwidth channels in dense areas. – Appleoddity Apr 5 '18 at 19:46
  • 2
    The sources are down. It seems very conservative to use 25 MHz spacing, when the actual channel is using 16.25 with the rest being guard band already. The reality is different, where 1,5,9,13 is actually usable. The other sources in that paragraph are either old (Cisco using 22 MHz) or unavailable or don't test 1,5 (or similar) channels. – user3549596 Apr 8 '18 at 10:55
10

As Spiff mentioned, channel selection is usually only done during boot time, as periodic re-evaluation of utilization of alternative channels requires additional or better hardware. There is also no accepted standard on how APs should cooperate when selecting their channel. What would happen if all the APs in an area suddenly see that channel 6 is being less utilized than channels 1 and 11? Right. A few seconds later channel 6 has become unusable, and every AP is jumping back to channels 1 and 11...leaving channel 6 open as the prime target for the next AP invasion.

In the 5GHz band, Dynamic Frequency Selection (DFS) may be required for some channels (channels 52-64 and 100-140 in Germany and the USA). This is, however, not meant to improve the cooperation of APs, but to prevent APs to affect weather radars. An AP using DFS has to constantly monitor the channel for weather radar and, if it detects something which could be a weather radar, has to leave that channel immediately (typically switching to a channels from 36 to 48, as these are not used for weather radar and do not require DFS...in other words, the AP does not select the best alternative channel, but rather just a channel which is guaranteed to be safe from weather radar).

It may be possible that some manufacturers of APs have algorithms which can optimize the channel assignment when an area is covered by a number of their (and only their) APs. A "Rogue Access Point" (which is not taking part in this optimization process) can significantly disturb the network. Some companies periodically perform hunts for Rogue APs on their premises.

10

Just adding a visual representation about the 2.4GHz congestion vs the 5GHz band to the already excellent answers.

I live in an European capital with a strong Internet and Wifi market penetration.

Furthermore, most local ISPs also add an extra roaming SSID/network by default in their router/modems/CPE and so often it is at least 1 SSIDx2 per home/neighboor. Do keep in mind, that besides the APs broadcasting signals, clients also broadcast.

So as an example, only listening with a normal notebook without any amplification in a fixed point in my bedroom, without walking around home, I can see at least 136 SSIDs (around 70-90 APs). It would not be a long stretch that led me to suspect I might have around me aprox. 200 equipments (APs+clients) broadcasting signals in the 2.4GHz band.

Compare the graphics in the left side, 2.4Ghz, with the right side, in the 5GHz band.

wifi

  • 1
    That graphic is cool - how is it made? – krubo Apr 9 '18 at 10:06
  • @krubo It is WiFi Explorer in MacOS – Rui F Ribeiro Apr 9 '18 at 11:32
  • @krubo I would try InSSIDer (should be free to use) – Jan Ivan Apr 10 '18 at 8:43
  • InSSIDer is paid nowadays too. I bought WifiExplorer for the graphics. – Rui F Ribeiro Apr 10 '18 at 18:56
4

In a high-congestion area where there are dozens of APs on 2.4GHz channels 1, 6 and 11, I sometimes get a more reliable connection by forcing 802.11b (the slowest mode), especially on less-used channels like 4 and 8. The bandwidth overlap diagram from wikipedia (below) suggests tantalizing clues why this might work, since the round bandwidth profile of 802.11b (DSSS) makes it look like it would care the most about the middle of its own channel even if overlapping channels were present. Of course this approach is too zany for the router to do on its own. Your mileage may vary.

enter image description here

-7

The real reason is that 2.4gHZ is a junk band and it never should have been used for anything. And the reason is it is a junk band is that it is the same frequency as the water molecule, it resonates. That is why radio astronomers use the band extensively searching for exoplanets and nebula. Corporate didn’t want the band because they knew it was useless. So thanks to the corporate toadyness of the FCC, 2.4ghZ became public domain by default. Sort of like the town dump being inagurated as a city park, but without the improvements.

The problem of the water molecule can’t be understated. Anything wet will interfere, including humans, dogs, house plants, microwave ovens, aquariums, snow and plastic water pipes. Competing transmitters cause “interference bubbles” that wander around over the course of minutes. There is no solution to this wandering, its part of the resonance of 2.4 with the water molecule. The more competing transmitters in your neighborhood the worse the wandering gets.

Sorry to say it but the only solution is really to move up the spectrum to 5.6.

  • 6
    Actually, 2.4GHz was largely chosen because it was already an ISM band and was low enough in frequency to make the RF chipsets cheap. The water thing is not actually true, water having many, many resonant modes, most of them fairly low Q. The absorption bands do not become a major issue until you hit the oxygen one at about 60GHz which can cost you 15dB per km of increased path losses. 2.4GHz was a compromise based of magnetron size, harmonic considerations and some experiments on penetration into foodstuffs back in the 1940s, an oven works perfectly well anywhere between about 1 and 20GHz. – Dan Mills Apr 6 '18 at 17:18
  • 1
    There is usefulness to having bands that get absorbed. The whole point of having low power APs in such a band is that they're not going to interfere with others too far away. Unfortunately, there's a ton of stuff in that band. – Brad Apr 7 '18 at 23:47
  • Let me quickly check the resonant frequency of uncondensed water vapor...right...: "The peak is 10GHz to 50GHz depending on temperature." I think that the resonant frequency at room temperature is around 22GHz. This resonance doesn't affect WiFi much, but...it affects 5GHz signals more than 2.4GHz signals. Even worse for 5.6GHz and higher, obviously. Note that 5GHz WiFi has (at the same output power) a smaller range than 2.4GHz WiFi. However, the limited range of 2.4 and 5GHz WiFi is beneficial...as a longer range would mean that you'd pick up more interference from neighbors. – Klaws Apr 11 '18 at 8:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.