How does routeros decide which frequency to use when AP is set to auto?
Does it scan and look for the frequency with the least noise? (If so; How often does it perform such scan?)
Does the connected clients affect the frequency selection in any way? If there are two AP's at same frequency (your own and one of your neighbors network), the neighbors AP may be out of range for AP (inside your house). But when you're outside in your yard the neighbors AP may interfere so much with your AP that your laptop struggles with connection. Would the auto channel selection switch to another frequency in such case? If so, how would it know that the laptop would prefer another frequency when the current frequency seems to be the best at the APs location?

It only scans once. It finds the frequency with the least number of other devices and sets it. That's all it does.

So it is not about signal levels and band utilisation and such things? Just about the number of other devices? OMG!

So it is not about signal levels and band utilisation and such things? Just about the number of other devices? OMG!

Yes, this is only based on number of networks in each frequency. It is only for home users that don't know which frequency to use.

Then it is clear now why I always finally set the frequency manually because automatic selection performed worse than expected.

There is normally only one problem with automatic frequency selection, and it is that it would not
adhere to your channel spacing method. I.e. when selecting manually you would use only
channels 1-6-11 or 1-6-13 and then the automatic selection puts it on channel 8,
for example. That will result in worse performance than just using channel 6 all the time.

The reason I asked is that a customer is in a location where channels are crowded. Even in the 5ghz band its hard to find channels. I've never used auto channel. At this place there may be channels available where the AP is located, but once moving 5m away the same channel is filled with other networks. The result is that what looks good at the AP is not usable at the client. Thats what makes the channel selection so hard.
To make it even worse neighbors are randomly changing channels so that even when I have found channels that performs well, my customers wlan start to perform bad after a while. In april they changed to cisco AP's, set them to auto channel, and their channel-issue has not reoccurred. I don't know how cisco decides channels, but it seems to be working.

Cisco have the clear channel technology.
Basically the Cisco AP is monitoring the channel if it detects a lot of interference it changes channels automatically.

Maybe that could be something for Mikrotik to implement as well?

is there a way of monitoring the channel and if it detects a lot of interference it changes channels automatically?

Or even some script that periodically scanned and could change channels if some aspect went over some threshold.

Anyone know how to restart the channel scanning process? Just as it does on boot?

One way could be to schedule a reboot. Preferably outside office hours.

Does the AP scan only at boot up, or would it scan any time the wireless is enabled? As in a script to disable/enable wireless at a selected time each day?


Sent from my iPhone using Tapatalk

when you disable/enable interface it will run channel select procedure.

Nordex, thanks for responding.

Would this be a recommended thing to do in a relatively high density environment? Run the script at 4am daily or some such thing, or weekly? I'm referring primarily to a single AP in a system, not multiples where you could end up with overlap.


Sent from my iPhone using Tapatalk

Running it at 4AM has the disadvantage that the situation at that time may not be representative.
Some access points in the neighborhood could be turned off at that time.
Anyway, once congestion is so high that you need to change channels the end is near. There are only 3 channels.
It is better to invest in a 5 GHz AP then.

Old thread, but it looks like a solution has arrived! I have not tested though, and it's not mentioned in the wiki.
ROS: 6.40.3
Capsman -> Channel -> reselect interval

The name sounds like something we've been looking for.
Seems like this cannot be set on a wlan interface not controlled by capsman.

So it is not about signal levels and band utilisation and such things? Just about the number of other devices? OMG!

Yes, this is only based on number of networks in each frequency. It is only for home users that don't know which frequency to use.

It's shocking to hear that the implementation is that weak. But then again can't say I'm surprised based on it's performance. Why oh why would you not make use of the frequency usage scanner as a minimum to find a channel that has the lowest usage and best noise floor combination?

One reason would be that (at least on some models) you have only one receiver so you cannot look around to see
what is going on on the other channels while still communicating with your clients. That would mean that there would
be an interruption just to check if another channel is less crowded, something some customers maybe would not like.
A more modern AP could be able to use a second or third chain to scan while still being connected, so that may not
always be a valid reason.

There seams to be a bit more to it unfortunately.

One could theoretically use frequency-monitor or snooper to get an idea for how populated the other channels are to make an informed decision. But once you start going down this path you'll quickly see that these functions have been nerfed to make it practically impossible to make your own smart channel selection script. Frequency monitor does not support "do" or save to file so you can't get any of it's data out. Same type of issues with swoop. Either it is massively under-featured, or it's been intentionally removed to prevent users from writing their scripts that actually do what one expects in 2018. On the forum there is evidence that it once supported "do" so it seems like this was intentionally removed, which makes me lean towards the latter argument where they don't want us to do this.

One reason would be that (at least on some models) you have only one receiver so you cannot look around to see
what is going on on the other channels while still communicating with your clients. That would mean that there would
be an interruption just to check if another channel is less crowded, something some customers maybe would not like.
A more modern AP could be able to use a second or third chain to scan while still being connected, so that may not
always be a valid reason.

Expecting relative cheap Wifi devices like Mikrotik's indoor units, Ubiquity, TP-Link and name many other economic wifi routers for domestic use is the same as asking Formula 1 performances from a Hyundai I10 car.....

The problems with Wifi enviroments is that to find the best channel for the AP to work on depends on several variables:
- Signal strength at the receiving device. This can differ 'per channel' but AP needs to get this info from the client. But which client? One client can have a good signal where another gets poor and would be better off in another channel.
- Quality of the signal (S/N ratio) and thus connection rate. Same issue as signal strength.
- Clients on the move. So what was good once, can be poor just minutes later.
- Neighboring AP's and client devices. If these are idle (no data transport) they are hardly 'visible' for AP. So AP can decide to set certain channel and then suddenly neighbor goes on Netflix and his Wifi start working as a heavily interference source.

To have an AP coping with all these factors it needs to scan and test the medium continuously for all these variable and be able to change channel 'on the fly' without interrupting the data flow to clients that are communicating.

I am afraid that that is merely impossible with the present budget Wifi if with 802.11 Wifi at all...
Maybe it needs better Wifi protocol to start with and then better OS on more capable hardware.

Even brands like Ruckus/Cisco/Aruba and the like with there high end Wifi routers are still struggling to get the best results under all situations.

So for now imho the 'auto frequency' setting is only for dummy's that have no clue to start with and every serious installer/operator does do a lot of testing and monitoring on a regular base to see if he/she can find the best channel and keep tuned to the best channel.

On our fixed wireless network we almost quarterly check our AP's and back hauls if they are still performing the same as when we'd set them. Sometimes we do find indeed we'd better swap to another frequency.
When customers call with Wifi problems one of the fist things we'd ask to them is to do a scan and/or we change manually the channel to another one. 3/4 times this solves the problem.....
We never use the 'auto' mode....

Expecting relative cheap Wifi devices like Mikrotik's indoor units, Ubiquity, TP-Link and name many other economic wifi routers for domestic use is the same as asking Formula 1 performances from a Hyundai I10 car.....

The problems with Wifi enviroments is that to find the best channel for the AP to work on depends on several variables:
- Signal strength at the receiving device. This can differ 'per channel' but AP needs to get this info from the client. But which client? One client can have a good signal where another gets poor and would be better off in another channel.
- Quality of the signal (S/N ratio) and thus connection rate. Same issue as signal strength.
- Clients on the move. So what was good once, can be poor just minutes later.
- Neighboring AP's and client devices. If these are idle (no data transport) they are hardly 'visible' for AP. So AP can decide to set certain channel and then suddenly neighbor goes on Netflix and his Wifi start working as a heavily interference source.

To have an AP coping with all these factors it needs to scan and test the medium continuously for all these variable and be able to change channel 'on the fly' without interrupting the data flow to clients that are communicating.

I am afraid that that is merely impossible with the present budget Wifi if with 802.11 Wifi at all...
Maybe it needs better Wifi protocol to start with and then better OS on more capable hardware.

Even brands like Ruckus/Cisco/Aruba and the like with there high end Wifi routers are still struggling to get the best results under all situations.

So for now imho the 'auto frequency' setting is only for dummy's that have no clue to start with and every serious installer/operator does do a lot of testing and monitoring on a regular base to see if he/she can find the best channel and keep tuned to the best channel.

On our fixed wireless network we almost quarterly check our AP's and back hauls if they are still performing the same as when we'd set them. Sometimes we do find indeed we'd better swap to another frequency.
When customers call with Wifi problems one of the fist things we'd ask to them is to do a scan and/or we change manually the channel to another one. 3/4 times this solves the problem.....
We never use the 'auto' mode....

I disagree enormously.

1. Mikrotik's hardware already supports everything that it needs to offer a considerably better product. Have a look at frequency monitor and wireless snooper - in relation to frequencies, noise floors, signal strength, channel usage etc.
2. There is no need to monitor the signal strength at the client side. Using the best signal around the access point is good enough, and light years better than what is currently implemented.
3. "Quality of the signal (S/N ratio) and thus connection rate. Same issue as signal strength." Already available on Mikrotik.
4. So what if clients move, you choose a channel based on the AP not on the clients that might be moving around. Otherwise you would have to constantly change channel. A lot of installation are fixed installations as well.
5. Neighboring AP's and client devices - yes there are blind node issues. But if the Mikrotik looks at the CCQ as well as the factors mentioned above it can switch channel and find a better channel that is not congested. Again light years better than what is has now.
6. Wow, so you scan once every 3 months in 2018? WiFi channels and their performance change hourly, and you need to move quickly to adjust.
7. I prefer to resolve issues before my customers have to call.

Summary:
Mikrotik has all the features it needs to offer a considerably better auto-channel selection. In fact in just a few hours I've scripted up a considerably better algorithm which we'll switch over to shortly. I feel for the other Mikrotik customers who can't write their own scripts.

Wyz4k, feel free to share it with the community would for sure help a lot of us (me included) unfortunately I do not have the experience in that kind of scripting...

I disagree enormously.

1. Mikrotik's hardware already supports everything that it needs to offer a considerably better product. Have a look at frequency monitor and wireless snooper - in relation to frequencies, noise floors, signal strength, channel usage etc.
2. There is no need to monitor the signal strength at the client side. Using the best signal around the access point is good enough, and light years better than what is currently implemented.
3. "Quality of the signal (S/N ratio) and thus connection rate. Same issue as signal strength." Already available on Mikrotik.
4. So what if clients move, you choose a channel based on the AP not on the clients that might be moving around. Otherwise you would have to constantly change channel. A lot of installation are fixed installations as well.
5. Neighboring AP's and client devices - yes there are blind node issues. But if the Mikrotik looks at the CCQ as well as the factors mentioned above it can switch channel and find a better channel that is not congested. Again light years better than what is has now.
6. Wow, so you scan once every 3 months in 2018? WiFi channels and their performance change hourly, and you need to move quickly to adjust.
7. I prefer to resolve issues before my customers have to call.

Summary:
Mikrotik has all the features it needs to offer a considerably better auto-channel selection. In fact in just a few hours I've scripted up a considerably better algorithm which we'll switch over to shortly. I feel for the other Mikrotik customers who can't write their own scripts.

1.
Might well be that MT already has these raw tools. But to make an intelligent AP that based upon the info found in these tools and taken in a continuous process, needs apart from a second radio (or use of a chain that is underused) an intelligent OS plus probably a bit better hardware just to make the comparisons and calculations happening. And due the variable nature of all that data as such we also need to build a history (=memory) so the OS can make decisions based upon time of the day etc. All this 'intelligence' is going to need hardware too. This simple fact alone will drive prices up.

I hardly use these tools anyway since it's not even accurate. Frequency monitor sometimes doesn't see Wifi radio I know they are there or it 'sees' frequency use while a scan doesn't reveal anything.
Apart from all that, the freq monitor only 'sees' the center ("pilot") frequency. It doesn't tell me if a channel is 5, 10, 20, 40 or 80Mhz wide. So in fact it has little use.
So all these tools could well be bond in a 'ultimate' AP but it will drive the price up to beeing 'high end' or beyond and probably the bulk of buyers will not be able or willing to pay for that...

2. How come signal at the client is not important? It is THE most important factor. If the client has poor signal than it has poor conn. rate and thus poor throughput. If there are other signals around the S/N will be bad too with a low signal. So yes, for clients signal is one of the most important items. Now it differs if we work with a fixed network with fixed antennas. Here we won't see a lot of difference in the strength of the signals apart from maybe atmospheric or weather phenomena s. But MT AP must also work as 'Hotspot' alike function where devices from other manufacturers and/or mobile moving CPE's are the clients. An intelligent AP cannot start changing channel to make things good for one device if that means another one will deteriorate.... It would need a highly sophisticated AP that can decide what is the best radio parameter setting to service most clients best in an ever changing environment. (Amount of clients versus data usage versus interference versus S/N ratio etc. etc.) It needs an AP that builds a sort of 3D spectral dimension for all its connected clients 'on the fly' while at the same time supplying and receiving data....pffff

3. And therefore the same issues.

4. Not an issue at fixed networks, a BIG issue in networks with moving, coming and going clients. Mikrotik, like all others, need to serve both worlds.

5. CCQ is a percentage measured on the combination of time plus data(loss). High data transport over weak to moderate signals often give better CCQ's then almost idle connections. To distill a good working decision process based on that needs again a lot of intelligence as mentioned under 1.

6. We have a fixed network with 40+ AP's and some 30 back hauls in an area with at least twice as much of the same from other providers and most of these channels are 'visible' for each other. But luckily most radio's don't change a lot of frequency channel so usually we monitor the network and act on basis of client calls about poor performance. My 'once every 3 months' is more an educated guess since some AP's and links work fine and for years are at the same frequency but other need a much more regular adjustment. Every update cycle of the ROS we also randomly check if a chosen frequency and or bandwidth and or protocol (NV2 sucks usually, Nstreme works best sometimes, 802.11 works in 60% of our cases best) is still the best for that AP/Backhaul.
But this all is time consuming and client disturbing (all the 'tools' MT has break the connections unless its 802.11) and just too much for a small business as us. So we have to act on 'ad hock' monitoring. No other way of doing it unless you can throw a truck load of more resources into it...
I think we represent the bulk of the clients in the WiFi-WISP industry in that respect.

7. Me too. Like I have more dreams... Reality is a bit more different.

And too be honest, I think while we work in a very complicated environment with the many radio's in the very limited Wifi (5Ghz) band, we are doing the same or better then the competition. With our simple MT hardware tools and our simple brain.
75% of the issues clients have is 2,4Ghz home wifi related and here the issue is even more unsolvable. Even main ISP's struggle with that issue since this is a market where no client want's to spend on Wifi routers thus the provider provides and goes economic.
We have a part of our network running mimosa and they have their G2 Wifi router for indoor use. As a rare example they have continuous channel scan and change to get the best results but it doesn't work very well. With their limited antenna and radio gain they not seldom lose the battle against the cheap 1,5-3dBi antena high power 30dB radio devices supplied by the standard provider. These G2 wifi units already cost triple or more then the standard Wifi for the extra technology but it has to bow its head against raw unlimited power in many instances.

So my point is, to get the ultimate wifi AP that can beat all nowadays issue it needs a highly intelligent device with a lot more hardware power to get there. I know a company like Ruckus and Cisco (and some others probably) trying to work in that direction but even they haven't found the holy grail in Wifi yet....

Ok, let's step back and look at the facts.

What is implemented at the moment:
An algorithm that does a frequency sweep and literally counts how many MACs it sees in each frequency. It then takes one of the channels with the fewest MACs. Probably not even a random one, probably the top one.

What is better:
Literally anything that actually looks at the signal strength of those various MACs to determine how close they are. Ones with a signal strength of -10dB are on top of you and that channel should be avoided at all costs. One with a signal strength of -80dB is negligible.

You do point out some valid issues above, but again that just proves that a better algorithm is required. If you set your interface to be 802.11 it will only look for 802.11 signals on its scan. If you set it to nv2 it will only show nv2 etc. If you set it to any it will show all of the ones that it understands, but other ones it won't pick up in its scan. A noisefloor (which is also already available) will pick up any signal, whether it understands it or not.

If you have 1 client you can try to find a golden middle-way But if you have 2 clients, or 10 clients that becomes exponentially difficult and is not practical. The AP would literally have to switch to one channel, wait for all clients to reconnect, do an analysis, move to the next channel, repeat. Cycle through all of the channels and then come up with the best channel. This produce will take several minutes and will result in significant downtime and interrupt. Does that sound practical to you? Can you give me any example of an access point that does that?

You talk about improving the wifi specification. That is way beyond the scope of Mikrotik and is an impractical suggestion. Mikrotik follow the standards once they have been released, they have no say into the standards themselves more than what you or I have. I agree that it would be great if future standards provide more functionality in this regard, but ultimately that is beyond the current scope.

Would it make your life easier if the "auto" channel selection algorithm is improved to the point where you don't have to monitor the links every 3 months, or randomly when there are major issues? Then why are you fighting against my request for them to use the tools already at their disposal to improve the algorithm from something that is truly shocking.

Wyz4k, feel free to share it with the community would for sure help a lot of us (me included) unfortunately I do not have the experience in that kind of scripting...

I would have if I had developed this during my private time, but as this was developed during company time I can not. Sorry.

It only scans once. It finds the frequency with the least number of other devices and sets it. That's all it does.

I know this thread is old but still relevent... Is there a way to know which frequency the "auto frequency" detector selected?

After selection ?
Most likely there are easier/better ways to get the required info but a raw way:

legacy wifi:
e.g.

/interface/wireless> monitor wlan1 once
status: running-ap
channel: 2432/20/gn(14dBm)
wireless-protocol: 802.11
noise-floor: -100dBm
overall-tx-ccq: 98%
registered-clients: 0
authenticated-clients: 0
wmm-enabled: yes
current-tx-powers: 1Mbps:0(0/0),2Mbps:0(0/0),5.5Mbps:0(0/0),11Mbps:0(0/0),6Mbps:0(0/0),9Mbps:0(0/0),
12Mbps:0(0/0),18Mbps:0(0/0),24Mbps:0(0/0),36Mbps:0(0/0),48Mbps:0(0/0),54Mbps:0(0/0),
HT20-0:0(0/0),HT20-1:0(0/0),HT20-2:0(0/0),HT20-3:0(0/0),HT20-4:0(0/0),HT20-5:0(0/0),
HT20-6:0(0/0),HT20-7:0(0/0)
notify-external-fdb: no

wifiwave2:
e.g.
> /interface/wifiwave2/monitor wifi3 once
state: running
channel: 2462/ax/eC
registered-peers: 3
authorized-peers: 3
tx-power: 15
available-channels: 2462/ax/eC

If using Winbox, go to status tab for the selected interface.

If using Winbox, go to status tab for the selected interface.

Yep, there it is in Winbox Wireless Status tab. Thank you!

Whenever you are using "auto channel selection", you must always make sure you have a "channel list" of valid non-overlapping channels for that band.
E.g. on 2.4 GHz you should have a channel list of 2412,2437,2462 (channel 1,6,11).
When not doing that, the selected channel will be one of the others e.g. 3 or 8, where it may see less valid traffic but actually it works less efficient because it is always colliding with traffic on the standard non-overlapping channels.