This always baffled me. Mystery solved!
Also FYI you can rest assured that if you've engaged a semicompetent Wi-Fi professional to design the coverage in your office/campus/warehouse/etc then you don't need to worry about "the DFS problem" as it is well known and will have been accounted for in the design.
We all pretend that WiFi is this magic thing that allows our devices to transmit data, but in fact there are all these little details of reality "leaking through" that abstraction.
For example, it seems that instead of power cycling everything at once, WiFi should be turned on first to ensure that there is sufficient time for 5 GHz to be available. I sometime reboot my TV, Router, WiFi, and set-top-box simultaneously by simply turning the power board they're all connected to off and on. From now on I'll do them separately and in sequence...
In my setup there is a fiber modem (outside the house), a Wi-Fi router, and then all the connected devices.
Oddly, the one that I need to restart the most, seems to be my Mac. Realizing, of course, the problem could still be on the router side, but turning wifi on and off on the laptop usually solves my problem.
I rebooted one device at a time for about a year and made notes about each failure. I didn’t see the pattern until I had done this for a long time.
Obviously it would be illegal, and I promise I'm not gonna do it whatever your answer is :p
And regarding the reasons for DFS - it is not due to the radars posing interference for access points, it is the other way round. They are supposed to be able to pick up reflections from clouds and airplanes that literally carry less than a millionth of the power that a typical AP would radiate.
Here is an example: https://www.fcc.gov/document/fcc-fines-hinson-dangerous-misu...
If you do cause interference with weather radar, there's actually a chance that the authorities take action. WiFi interference isn't hard to scan for at all, even or you disable broadcasting your SSID on the channels you're interfering on.
depends, hence the checking.
> And is that sort of violation the kind of thing your local regulator would be scanning for and send someone out to shut you down?
highly doubtfull
People who operate radars (in particular radars for weather forecasts) are now really upset of the recurring interferences that come from wifi devices with (illegally) disabled DFS. Keep in mind that weather forecasts is not only about consumer people watching TV to know whether tomorrow will be sunny or rainy, but is also safety-related (e.g. for flights and navigation. You don't want your flight to be caught in a storm !).
Those issues (significant interferences that are now almost impossible to solve) are periodically raising debates within RF administrations about whether to require locked hardware (locked bootloader and signed firmwares) in order to ensure that DFS cannot be disabled. That would be a pity for any embedded linux geek (openwrt is great both with regards to freedom, privacy, long-term support with security updates and therefore avoiding obsolescence, etc.), but a lot of people now think that it is the only way...
I understand the shortcomings of DFS, yet you should really understand the 5 GHz wifi band is a band dedicated to radars as primary users where wifi has been tolerated under the condition of DFS, so wifi users should really not complain too much about the consequences and limitations of DFS (if there was no DFS, there would be no 5 GHz wifi band at all ! ;)
What does this mean? The DFS check doesn’t exist? Or they just reboot nearby devices several minutes apart to allow enough time for DFS to complete?
What's the best practise in other words
Then install OpenWrt (even on that TP-Link Archer C7 if you want), set "channels" to the list of channels that you identified (unfortunately this property is available only from the command line), and "channel" to auto. Then OpenWrt will select one of these channels after a one-minute delay, and jump away (but only to another known-good channel) if there is a radar.
Thanks!
Everything that really needs speed is connected by ethernet anyway.
They are low power, but unused, so they perform well.
https://www.ofcomnet.ch/api/rir/1008/12
https://kernel.googlesource.com/pub/scm/linux/kernel/git/sfo...
Dynamic Frequency Selection
To quote Elon: Acronyms Seriously Suck
Like this one for instance - have you actually learned anything now that you know it’s “dynamic frequency selection”? That’s not a good description of “an extra channel that sometimes isn’t available because radar uses it”.
DFS to a software crowd will always expand to depth-first search. I was trying to figure out if that's what was being discussed here (with plenty of skepticism, admittedly) until reading GP.
> This mechanism is known as Dynamic Frequency Selection (DFS) and is designed to mitigate interference to 5GHz radar by WLANs.
https://draytek.co.uk/information/blog/ofcom-relax-the-rules...
On the other hand, since nobody uses it, it's generally totally clear, and if your AP and client devices support it (and you can live with the lower bandwidth) it can be a nice "secret" channel to use in a noisy 5GHz environment.
Not everyone has a spectrum analyzer of course to verify this, but if you have a reasonable assurance that the channel is not being used and your devices are AC/AX that can make use of at least 2+ spatial streams, that more than makes up for the truncated channel width in most cases.
Where I live, the channels 36-48 - which is really just one channel once you go 80 Mhz - is chock-full of networks and completely unusable. The DFS channels are also complete unusable, because they jump around like crazy due to some noise source (I checked, and there's no radars around). Trying to use them means you see disconnects ten+ times a day.
My router can use channels 149+, which are not DFS and most consumer routers don't use them so that part of the spectrum is clear, but those channels have smaller allowed broadcast power so I get slower speeds. Or I could use 2.4 which works reliably, but is even slower than the low power 5 Ghz channels.
Interesting consequence is that while my ISP started offering gigabit connection, I see no reason to upgrade, since my Wi-Fi is now the limiting aspect, unless I start using wires or replace my devices with ones that can do Wi-Fi 6e or something. Even the 300 Mbps I pay for I only get in the living room.
Also, don't use hidden SSIDs on a DFS channel: https://badfi.com/blog/2016/2/15/yet-another-reason-avoid-us...
Every time the main route was unavailable, due to wind or something I would lose my Wifi because of DFS. That was really fun to debug until I finally figured it out.
Unifi especially may take hours to get the 5GHz network back for some reason.
https://docs.engenius.ai/cloud-white-papers/zero-wait-dfs#wh...
Switching channels due to DFS is not "frequency hopping" in that sense – that would be many orders of magnitude faster.
Any actual radar would probably be far away from my home 5GHz Wi-Fi anyway, which has relatively low tx power. Is there a realistic scenario where a 5GHz-scanning radar sends a pulse from far away, and my home Wi-Fi's signal gets picked up by it?
Same idea with radar aimed at an urban area. Thousands of WiFi devices in a high rise will add up. Tens of watts, hundreds of watts of noise. A large and dense downtown core will be transmitting kilowatts of radio noise from millions of devices. More noise means less SNR. And less SNR in radar means a blurry/noisy radar view, and lower effective range.
https://journals.ametsoc.org/view/journals/bams/97/7/full-ba...
The FCC has a list of WISP companies they've gone after for misconfiguring their wireless equipment:
https://www.fcc.gov/general/u-nii-and-tdwr-interference-enfo...
It's interesting. If you're a typical consumer and use default settings, you end up with a mostly working setup that's glitchy but not quite broken enough for anyone to notice because Netflix still works. If you're a tech enthusiast and tweak seemingly harmless settings that should unambiguously improve things, you end up with a completely broken network.
But Skype/Teams/Zoom gets visibly glitchy, which may be annoying during remote schooling or work. In the past two years, a sizeable fraction of "typical consumers" started to use real-time applications a lot.
Not to mention, while Netflix "still works", TikTok may not. Or YouTube, if you happen to switch between videos at the moment. Or Facebook or Twitter or anything that relies on the "infinite scroll" dark pattern.
Point being, regular, non-tech users definitely notice. We may think that they don't, but that's because they don't know how to frame what's going on, and have been conditioned to accept that digital technologies are just shitty and glitchy. They think it's their fault, or the problem with "their Internet", or that their computer "has viruses" - and won't tell you until you're close enough with them they feel they can vent to you, or hope you can fix it for them.
Cherry picking one example very recently, we discovered a certain router model shipping in calibration mode such that only 1-2 STAs could transmit at any given time. Essentially rendering an AC router into the performance class of a mixed-mode a/b/g router from 2004. Working with the upstream vendors to fix this took several weeks.
Sometimes though its not even phy or mac layer bugs we find, its more banal nonsense like a poorly configured firewall that eats up all the NPU clock cycles when it sees a 30 pps rate of unsolicited UDP ingress. Or my personal favorite was the erase block management on flash that would remount the data partition as read-only whenever a single ECC cycle would fail on a given sector and permabrick the unit until reboot.
Though one will never learn without at least some modicum of living dangerously.
As a counterexample, 2.4 GHz is completely unusable in our penthouse office. I'd rather deal with limited cell size than continuous interference issues.
Not in the US. While incidental interference with unlicensed use of spectrum is legal in the US, intentionally blocking, jamming, or interfering with authorized (including unlicensed) radio communication is prohibited.
However, there are other laws related to jamming (wilful or malicious interference) that seem likely to cover the situation. https://www.fcc.gov/general/jammer-enforcement Perhaps you get away if caught if you could show you didn’t have criminal intent?
An especially important example might be if you caused someone’s VoIP phone to not work, since that is related to “safety”.
Because you only need a short signal, it would be very hard to detect, unlike noisy full-time jammers.
One place to check on the legalities would be conference centres that have WiFi systems that interfere with guest hotspots - I seem to remember that Cisco||Meraki had that functionality? Presumably the manufacturers would have thoroughly vetted the legalities - especially since selling jamming devices has heavy penalties.
As you can see in the article, the radar pulse sequence to trigger DFS requires pretty short pulse widths, but you want to use pulse compression (with longer pulses) in modern radars, so you won't trigger DFS. Additionally, you may scan over particular location only once per 5 minutes, so again you have lower chance to trigger it. And another challenge are ever-changing atmospheric conditions - maybe sometimes you won't see the wifi device and so you won't trigger the DFS (https://en.wikipedia.org/wiki/Anomalous_propagation) and then the conditions change and you suddenly start getting interference until the DFS on the remote end re-evaluates.
We have a radar on the Czech-German border and it's interesting that we have way worse interference coming from the Czech side. But this may also be because of some local conditions.
In the comments, it is suggested to transmit a pseudorandom code and then correlate it with the received signal to filter out uncorrelated interference. I'm of course doing this, but it helps for point targets (towers, airplanes) - not so much for distributed targets (clouds), especially when they are not stable in time (the droplets all vibrating in turbulences) - so the replicas you receive are distorted by various means, instead of a single reflection you get from e.g. an airplane.
What helps A LOT is a wifi packet detector, which completely blanks the data when the remote station is transmitting, so our radar is basically operating in the gaps between the packets. For some products (such as reflectivity), we have enough oversampling so we can interpolate the gaps; for other products this degrades the data. But at least you get gaps, not giant "lightsabers".
I've moved to current firmware as a test and it is no longer a problem for me.
Not a bad day too, nice and sunny, drinking a coffee from the burger van, watching the boats racing up and down the loch (one end of the link was to the "timing tower", a cabin with all the timekeeping equipment at the far end and built a couple of hundred metres off shore), watching the planes stacking at the Lanark holding-point for Glasgow airport, wow look, there's three of them just circling now, isn't that cool? Anyway no signs of bother from the netwo<PLUNK>
Scrolling up the logs, an error message about "cannot allocate DFS channel", and lo and behold all the channels are flagged as "RADAR interference".
I dialled the DFS hold time down to five minutes from an hour, and the problem went away. The spot wasn't usually that busy with planes and they typically weren't stacked for more than about ten minutes ever.
My intuition so far was always to put the AP as high as practical...
[1]: https://en.wikipedia.org/wiki/IEEE_802.11d-2001
[2]: https://w1.fi/cgit/hostap/plain/hostapd/hostapd.conf
Fun fact: Access points can also advertise a timestamp and timezone.
[3]: https://openwrt.org/docs/guide-user/network/wifi/basic#bss_t...
Also, regarding WiFi equipment. Finding out the quality of consumer WiFi gear is near impossible before buying, because of so many other factors affecting wifi performance (especially that I use mostly 2.4ghz). If anyone has a foolproof method for this please let me know.
It also shows the maximum power (in mW or dB) for the channels, which is also key to get a good reception. The dB notation is for example (20), which corresponds to 100 mW. Every 3 dB is a factor 2 in power, so (23) is 200 mW.
