Horribly slow Wi-Fi on Mikrotik network

robmaltsystems · Wed Feb 09, 2022 10:30 pm

New client with a relatively simple setup:

Virgin Media broadband connection with 300Mbps
RouterBoard 2011UIAS
PoE 5 Port Gigabit Router RB960PGS
3 x cAP AC Wireless Access Point RBcAPGi-5acD2nD

The client has reported horribly slow Wi-Fi speeds. However, yesterday the speed was a reasonable 120Mbps although perplexed why it's never getting anywhere near the full speed of the broadband considering all the components are rated 1Gbps. The installation was a bit of a rush as they gave very little notice.

Getting time onsite and time for diagnosing is difficult as it's a bar open Mon-Sun 10am-midnight! I'm going to leave a spare laptop onsite so I can diagnosis it at my leisure but welcome any suggestions. I've deployed several Mikrotik routers but this is the first time I've dabbled with the cAP ACs.

erlinden · Thu Feb 10, 2022 8:19 am

Let's start with the configuration, can you please share (at least one) config of the cAP ac? /export hide-sensitive file=anynameyoulike
What would also be intereseting is the connection speed of the clients.

robmaltsystems · Thu Feb 10, 2022 11:59 am

Will do - going on-site later to install the laptop so I can generate the configuration. I'm also going to see if I can check the one thing I didn't install - the cabling...

gotsprings · Fri Feb 11, 2022 12:17 pm

Client to client forwarding and local forwarding will speed things up considerably.

Also... A 2011? Why would you use a decade old router?

Moving on... Restaurants and Bars are the only thing one of my contractors does.

Meaning I have over 70 stores to call for reference on what works and what doesn't.

The Mikrotik routers have no problems running those stores. But all the caps-man installs had to be removed.

Mikrotik wireless does not do well in noisy or crowded environments. Dropping some clients and not letting them reconnect. Or even having radios drop all traffic until restarted. Also the radios throughput is considerably lower than competitive products even under the best situations.

After months of fighting it and loosing a lot of clients... Mikrotik confirmed it was repeatable... And nothing could be done.

We had to give in and accept that Mikrotik wireless was stuck in 2014. And that while caps-man was great for management... The radios were not up to the task.

Went back to our old wireless vendor... All the trouble tickets stopped.

Had to replace cAP ACs and wAP ACs across the board... Residential and commercial. It was a very expensive lesson.

robmaltsystems · Fri Feb 11, 2022 3:00 pm

>Also... A 2011? Why would you use a decade old router?

Because that isn't advertised in front of you when ordering items. The price and specification was fine. But not first time this has happened - did with the same with Unifi AP which went EOL far too soon.

>The Mikrotik routers have no problems running those stores. But all the caps-man installs had to be removed.

The router is fine - speed testing at 350Mbps which is the internet speed.

>Mikrotik wireless does not do well in noisy or crowded environments.

I'm tempted to concur here - the 2.4GHz spectrum is horribly congested. The 5GHz isn't too bad esp. when using the higher channels. I suspect it could simply be congestion.

>Also the radios throughput is considerably lower than competitive products even under the best situations.
>After months of fighting it and loosing a lot of clients... Mikrotik confirmed it was repeatable... And nothing could be done.
>We had to give in and accept that Mikrotik wireless was stuck in 2014. And that while caps-man was great for management... The radios were not up to the task.

This is what I feared... that said, getting >100Mbps in a bar is more than acceptable but this leaves a rather bitter taste. I want to really like Mikrotik but this is disappointing.

>Had to replace cAP ACs and wAP ACs across the board... Residential and commercial. It was a very expensive lesson.

I hope that the network can be at least stable.

When you say there is a problem with Wi-Fi - is that the chipset used or the Linux OS driver? Would going to v7 help?

robmaltsystems · Fri Feb 11, 2022 3:02 pm

There are three AP covering the bar - whilst it's a big bar, I suspect we are a little OTT here and they could get away with two. So I'm tempted to take one down from the ceiling and bring it back so I can experiment and see what kind of speed I can get in the "lab".

robmaltsystems · Fri Feb 11, 2022 3:05 pm

Disabling CAPSMAN won't be the end of the world. There are only four AP in total. Once again, is CAPSMAN any better in v7?

The site I alluded to about with EOL Unif AP - I was tempted to replace them by Mikrotik as they now have a Mikrotik router (which works wonderfully). In light of this discussion, I might review that decision and stick with UniFi and the controller running on a laptop.

Love Router OS but as I type more here, I've always been disappointed in their Wi-Fi performance even on Wi-Fi 4 equipment. Cheap TP-Link access points can regularly get 50-70Mbps when close to the router. Mikrotik Wi-Fi always seemed a little lower

anav · Fri Feb 11, 2022 4:22 pm

What did you expect?
Have you seen anyone here in the last two years recommending MT wifi 5 hardware??
Capac is rated at 867 advertised speeds at 5ghz, the best you can do is approx 1/3 of advertised speeds.
thus you should be getting around 289 in a one way test. That is assuming a stable connection at full potential.
With my same vintage tplink eap245v3 (wifi5) I am getting around 400-500 downloads and uploads.

If you want performance for a business dont cheap out and get at least wifi6 ( a couple of HD620s would do it easy peasy up and running in 5 minutes, customers satisfied, and you can get on with life, dont waste another second of your time and get your reputation back )

https://www.walmart.com/ip/TP-Link-EAP6 ... /995619787 (usa prices)
https://www.amazon.ca/TP-Link-Wireless- ... B08TRPMC69 ( canada prices )
https://www.box.co.uk/EAP620-HD-TP-Link ... 74694.html ( Uk prices )

robmaltsystems · Fri Feb 11, 2022 4:33 pm

>What did you expect?

That it works as advertised?

>Have you seen anyone here in the last two years recommending MT wifi 5 hardware??

I don't spend my entire life in here...

>Capac is rated at 867 advertised speeds at 5ghz, the best you can do is approx 1/3 of advertised speeds.
>thus you should be getting around 289 in a one way test. That is assuming a stable connection at full potential.

Which would be fine as that's the kind of speeds I get here sat here right now under a VERY expensive Meraki access point with a 1Gbps backbone. But I've never seen >200Mbps with the Meraki access points.

>If you want performance for a business dont cheap out and get at least wifi6

This is public Wi-Fi in a bar so is perfectly fit for purpose but still doesn't stop me being curious.

robmaltsystems · Fri Feb 11, 2022 4:44 pm

I have just done a search on Wi-FI performance here and there are posts about performance issues with specific Router OS versions which suggests to me the problem is, partially, software related. Is the Wi-Fi chipset in the cAP ac the problem? I've read of Wi-Fi performance problems in Linux generally over the years so wouldn't be surprised if it's kernel related. Might have a go at upgrading one of them to v7 which I believe is built on a more modern kernel?

Fri Feb 11, 2022 4:57 pm

Most having issues on ROS7 with 5GHz see them not disappear but it least it gets better going back to ROS6.
Which version, seems to depend from device to device.
Some need to go back to even 6.48.<something>, others can use latest 6.49.2.

On the other hand, I am sure there are far more NOT reporting any problems because there aren't any.

robmaltsystems · Fri Feb 11, 2022 9:56 pm

CAPsMAN is clearly the bottleneck in this set-up although why it was horribly low that night is still a mystery. I've disabled CAPsMAN on the nearest AP to the laptop I have access to remotely and throughput is consistently higher. It's broken through 200Mbps and that's through a wall. Windows is reporting a connection speed of 866Mbps which is the maximum rating of the cAP AP. When using CAPsMAN, it's struggling to get above 120Mbps. The chipset in the cAP ac might be lagging compared to some others but 200Mbps is perfectly acceptable for this site. I think it's been recognised that the Wi-Fi drivers in v6 aren't optimal.

Whether this overload is outweighed by the ease of configuration depends upon the size of the installation. This site only has four AP so programming them manually isn't that much of an overhead esp. as it can be scripted.

As CAPsMAN is AFAIK a software protocol, it is a perfectly reasonable question to ask whether performance has improved in v7.

Another aspect I'm going to investigate is the wider 5GHz spectrum. When I got the 200Mbps speed, I had manually set the channel to 36 with Ceee. I'm going to let it run free and use auto with XXXX. This means it tends to get one of the higher channels. Are there performance issues up there? I'd expect not as they are less congested but I've tweaked Wi-Fi enough to know things are often not as them might seem.

troffasky · Sat Feb 12, 2022 12:33 am

You didn't respond to gotsprings's comment about local forwarding. That could account for lower performance with CAPsMAN.

anav · Sat Feb 12, 2022 1:17 am

Yeah with local forwarding should be able to get around 250 speeds....

robmaltsystems · Sat Feb 12, 2022 12:10 pm

>You didn't respond to gotsprings's comment about local forwarding. That could account for lower performance with CAPsMAN.

Thanks - I'd missed that, will have a look. Another bit of learning for me.

I happened to still have the remote session open at 1am before I went to bed - got well over 200Mbps and even hit 300Mbps once. Nobody else was using the network. This suggests that the problem is not specifically to do with the cAP ac themselves which is good to know. So keeps the finger pointing at either congestion over the air or performance of CAPsMAN when under load. As I part-time software developer, I can certainly accept that there could be performance problems in any bit of software. Mikrotik CPUs are good but not *that* good

Congestion over the air - I did do electrical engineering at uni but that was over 40 years ago so the knowledge is dim. I do understand how overlapping channels can impede throughput - and why access points on auto try to find the least congested channel.

But what I don't know is does the 2.4GHz spectrum interfere with the 5Ghz spectrum? That's where those 40 years are not helping me remember the basic theory.

Sat Feb 12, 2022 12:53 pm

Compare with normal radio:
AM doesn't interfere with FM either.

2.4 and 5 are too far apart to interfere each other in a noticeable way.

bpwl · Sat Feb 12, 2022 2:34 pm

and why access points on auto try to find the least congested channel.

But what I don't know is does the 2.4GHz spectrum interfere with the 5Ghz spectrum? That's where those 40 years are not helping me remember the basic theory.

"Manual channel planning yields better results than simple automation." https://metis.fi/en/2018/10/rrm-en/ (and some other documents from Metis)
Problem is also that "auto" sometimes choses the energy dip between 2 non overlapping channels, creating a new channel overlapping with both other channels.
Co-channel interference is preferred. Adjacent (overlapping) channel interference is destructive, corrupting packets transmitted.

That interference goes further than the usuable range of an AP! If the (slow spead) header can be decoded, devices will wait, even if the (faster speed) message is not receivable.

Klembord-2.jpg

r00t · Sat Feb 12, 2022 4:02 pm

Also using "auto" channel selection on Mikrotik is quite bad, because it evaluates channels based on number of APs, not signal strength. So it will choose channel with least amount of APs and happily select channel with one super strong AP. I haven't tested wave2 if it's any different in this regard.

mkx · Sat Feb 12, 2022 5:01 pm

But what I don't know is does the 2.4GHz spectrum interfere with the 5Ghz spectrum?

It might, but it diesn't happen often.

There are higher harmonics, created by improper match between transmitter, feeder and antenna. Their frequency is multiple of base frequency, so 2.4GHz wifi transmitters can produce some noise in the band between 4.82GHz and 4.96GHz, which is luckily just below 5GHz band. 3rd harmonic falls to 7GHz+ (and is usually very much weaker).

Then there are RF mixer products (caused by non-linear faults due to some improper installations or faulty passive elements), the exact frequency emitted there is a bit more complicated to determine, but can fall into same band, and both lower and higher ...

gotsprings · Sat Feb 12, 2022 5:27 pm

43 clients connected across 2 radios.
A crap ton of nearby wireless networks chewing up the airtime.

robmaltsystems · Sat Feb 12, 2022 7:12 pm

>"Manual channel planning yields better results than simple automation.

Don't disagree but it's a bit of a moving target when you have so many adjacent routers set to auto. So you pick the channel that inSSIDer suggests if best but then one of the adjacent routers re-assigned it's channel. The 2.4GHz network on this site is horrible

I've got a temporary AP in there at the moment (the private). Should probably turn off 40Mhz wide channel... it's not playing nice.

This temporary AP is about 1m away from the laptop and it struggles >20Mbps. But that's an aside for this discussion.

bpwl · Sat Feb 12, 2022 8:32 pm

Nice case of overcrowded 2.4GHz band.
The 40 MHz channels are taking way too much, and Fountain-private is extremely strong. >-30dBm will saturate receivers as well, disabling them to properly receive a signal.
The 40MHz wide , at -48dBm on channel 6 is bad as well.
I don't know a method to set the "40 MHz intolerant" bit , to make 40MHz channels fall back to 20MHz.

Best I can see is channel 1, 20 MHz.wide. There are many of them and auto will not pick this up, but ok, no problem, co-channel interference is just about playing the "contention game" to get a transmit timeslot. They seem to play nicely around channel 1. To know how "bad" or "busy" it is , one should look with "Snooper" and "Freq usage".
viewtopic.php?t=182565#p907921
Even spectral scan over a longer period.... viewtopic.php?t=182565r#p908081

And what to do if it is all too busy? If somebody wiould stream a video, the MT would almost stop it's transmissions because it cannot get a free transmit window.. Reason is that video streaming is handled as priority with WMM. Priority means that in the "contention game" they use smaller values for the contention window wait, so they get a free transmit window more often.
Always wondered no one seems to use this to get a larger part of the available air-time just for all traffic. Oh wait ... it looks as they do (https://www.techjunkie.com/get-priority-wi-fi-network/)

MT has priority=0 ("Best Effort") on all traffic by default. So enabling WMM uses the long "Best effort" contention window settings, the same as without WMM. Only priority 0 has A-MPDU aggregation enabled. And that's important for the throughput: viewtopic.php?t=165698#p912622
To set the priority, one must do it on the AP, with FW mangle rules. Ah you can get "video" or even "voice" WMM priority for everything you want. But you lose the MPDU aggregation and speed, unless enabled. (!!!)

The CWmin/CWmax values are 802.11 standard, also in MT : viewtopic.php?p=901234#p864106
Setting all traffic to "Video" and enabling A-MPDU, is like driving with emergency vehicle lighting turned on. Not a 'nice' neighbor attitude, but may be needed not to be drowned by video-streamers. around. And Google wifi-routers do it also anyway.

bpwl · Tue Feb 15, 2022 2:44 am

viewtopic.php?p=913083#p913083

robmaltsystems · Tue Feb 15, 2022 2:02 pm

Thanks for the link. I'm suspecting that it's something to do with the 5GHz channel selection. I've just done the daily check remotely and the speed was 0.88Mbps down (!) and 20.69Mbps up. When I look at registrations on CAPsMAN, most devices have fallen back to 2.4GHz - I assume clients monitor the connection and disregard it if it's poor quality.

I've forced 5GHz onto 5180/Ceee and the speed has immediately increased.

Just also trying with "Skip DFS channels" set...

Tue Feb 15, 2022 4:20 pm

Because that isn't advertised in front of you when ordering items. The price and specification was fine.

My suggestion, check the wifi situation before buying anything. In this case, you could have foreseen the situation with overcrowded 2GHz and could have chosen a 5GHz device instead. Meaning for your location, specifications were not fine.

robmaltsystems · Tue Feb 15, 2022 7:14 pm

The devices are 5GHz and it's those that are struggling. The clients connected at 2.4GHz are completely fine - sure not fast but 20Mbps is perfectly acceptable in a bar. The weirdness that prompted this original post was horribly slow speed tests on the 5Ghz channels.

After turning off use of DFS channels, the laptop has been registered on the 5Ghz channel for over four hours now and is getting a perfectly acceptable 120Mbps throughput. If I had more time I'd try and diagnose which 5GHz channel is causing the problem.

Completely separate but connected, I have another client with UniFi Nanosation P2P links on the site. They were set to auto and very occasionally the packet loss across one of the bridges was horrendous (like 90%). The channel didn't look that bad but once again, picking another channel and packet loss disappeared. That site though is very close to Manchester airport and DFS channels are probably used.

All good fun

bpwl · Tue Feb 15, 2022 8:53 pm

DFS channels in use by radars don't give "bad performance" in that channel. They will force the AP to change frequencies and leave the channel! For a MT, operating without pre-scan of DFS channels, the full 1 minute "check for radars" is then mandatory. The AP will be silent for that full minute, the SSID is not detectable. If one of the "weather" channels (120-128) is in the selected DFS channels then that silent period is 10 minutes. If a radar is detected in that check time, the process and timer starts over again. There are more false radar detects than real radar detects in my experience.

Don't let this just happen. Give your smart channel list, after testing the environment for busy channels. A 80 MHz wide channel (XXXX,Ceee,eCee,eeCe,eeeC) goes very wide and uses 4 consecutive 20MHz channels. With XXXX there is no control of which of the 4 channels is the control channel defined by the frequency given.

The MT AP did not react as expected to a channel list. Hoped it would keep stepping down the list until a good entry was found. With radar detect it looks like the MT is restarting from the top more often than expected, and sometimes gets in a loop when first 2 lines cause "radar detect", without ever trying another line. [it were false radar detects, caused by a nearby wAP]

https://www.ekahau.com/blog/channel-pla ... ter-wi-fi/

robmaltsystems · Tue Feb 15, 2022 9:55 pm

>DFS channels in use by radars don't give "bad performance" in that channel.

I appreciate that but skipping DFS channels was a quick way to exclude some channels from the scan. This is all about me trying to diagnose/isolate the original reason for this post - a normally full functional Mikrotik network with a router, PoE switch and four cAP ac occasionally ends up getting very low throughout on the 5GHz spectrum.

I didn't personally test the wiring but I am assured it was tested. A problem I have with this site is that it's open 9am-midnight every day so access/chances to experiment are very limited.

mkx · Tue Feb 15, 2022 10:08 pm

DFS channels in use by radars don't give "bad performance" in that channel. They will force the AP to change frequencies and leave the channel!

That's what should be happening from the very beginning of the era when wifi was allowed to use DFS frequencis. But many vendors didn't implement DFS properly and some wifi APs weren't bothered by radars. It seems that some years ago (around 2018 give or take) aviation industry (with support from weather crowds) made enough fuss for FCC (and some other regulators) to press on vendors to do better job. It was the same time that MT implemented country regulations (e.g. EIRP limits) more stringently.
So it is possible to have some older wifi AP with old software to behave the way described by @robmaltsystems. It's illegal to operate it though, it's always owner's responsibility to make device operate within legal limits (same as driving a fast car ... one can't point at car producer if he's speeding).

And probably there are also better explanations of the behaviour.

mkx · Tue Feb 15, 2022 10:12 pm

A problem I have with this site is that it's open 9am-midnight every day ...

That still gives you time window to work full 8-hour shifts there (with some after-party goers as company)

robmaltsystems · Tue Feb 15, 2022 10:28 pm

>So it is possible to have some older wifi AP with old software to behave the way described by @robmaltsystems

I use a hAP ac lite here in the office for experimentation and training purposes. I've also got one at the client site running the tills & card readers so that when I'm diagnosing the main cAP ac network, they keep running - kind of critical in the bar!!

Interestingly, whenever I change the settings at the client site, it always goes into "detecting radar" which is exactly what it should do. Sometimes it finishes detecting in a short time - sometimes a lot longer. Assume this is that 1 min/10 min mentioned above.

However, on the identical hAP ac lite here at home, I've never seen this radar message... it goes from scanning for frequency straight to running ap. Why isn't this device doing the radar scan?

mkx · Tue Feb 15, 2022 10:43 pm

Whether AP runs radar detection procedure or not depends on which channel(s) it uses ... and with frequency set to auto channel selection is as good as random. Don't know how to check it in winbox, in CLI execute

/interface wireless monitor [ find where band=5ghz-a/n/ac ] once

to see which particular channel (and channel layout) device chose. Wikipedia has a good article with all channels (for all wifi bands) together with limitations in some countries/regions.

N.b. it might take more than 1/10 minutes for AP to settle on a frequency. E.g. it selects a weather radar frequency, it scans for radar signal for, say, 7 minutes, detects it. Selects a "normal" radar frequency, scans for usage of almost a full minute, detects it. Selects another weather radar frequency, scans for usage another 10 minutes ... 18 minutes gone and it finally starts ... after a few minutes it detects radar (can be false detection), drops all clients (they connect to other APs with worse signal levels) and starts the frequency selection procedure again.
So if you know there will be radars detected (because your venue is close to airport or weather radar or your experience says so), it's better to stear away from auto settings. At least set skip-dfs-channels=all.

bpwl · Tue Feb 15, 2022 10:53 pm

Frequency=auto .... you never know what comes out of this. It's the almost perfect random generator, but one location can tend to give non-DFS more often.
There are also non-DFS frequencies. If the selection is from that non-DFS range, 5GHz starts immediately. The selection is somehow based on the RF spectrum at that location.

While using installation "indoor" you choose from the follow 3 lines:

Klembord-2.jpg

I would advice to use installation "any" instead. There are no "outdoor only" frequencies, only "indoor only" exists, so the freq marked as "outdoor" are free to be used indoor!
By setting installation on "indoor" the outdoor frequencies are not selectable in the MT list. (Weird list or weird logic)

PS: check used freq in WinBox in TAB "Status" of the WLAN interface.

JanezFord · Thu Feb 17, 2022 10:00 am

Band : 5GHZ-A/N/AC

Switch to at least N/AC ... I seriously doubt there are any 5GHz A devices connecting to your network in 2022.

JF

robmaltsystems · Thu Feb 17, 2022 1:36 pm

>Band : 5GHZ-A/N/AC

Okay, done

I turned off WPA as inSSIDer suggested it but found the tills are stuck in the 2000s!

carl0s · Fri Feb 18, 2022 10:04 am

Hi Rob. If you are happy with ~20mbps then maybe try using 20MHz wide (no extension channel) on 5GHz and only putting in the non dfs channels as options. 5160, 5180, 5200, 5220. Set up one channel configuration with all those listed in the same conf (you can add multiple rows in the conf).

This setup allows me to pretty much fully utilise an 80/20 fttc connection in coffee shops (70mbps from speedtest.net)

I did find that both of my APs were put on 5160 until I told capsman to reselect channel mind you.. probably easier to choose channels manually.

But I'm not sure of the need for 4 extension channels. Multiple spatial streams are used on the selected channel to increase performance as well.

Seems clients are happier to go for those lower channels. When I monitor the air I rarely see other APs in the higher channels except my MT stuff..

bpwl · Fri Feb 18, 2022 11:34 am

5160, 5180, 5200, 5220

Fully agree on using smaller channels. My list shows 5180,5200,5220,5240 as possible values in 5 GHz band.
If there is room for it 40 MHz would be 5180Ce and 5220Ce.

https://www.ekahau.com/blog/channel-pla ... ter-wi-fi/

5160 is in the full channel list and is allowed (https://en.wikipedia.org/wiki/List_of_WLAN_channels) but it does not show up in the drop down list of ROS
And cannot be used in ETSI regulatory domains. Maybe some other countries can use it.

Klembord-2.jpg

robmaltsystems · Fri Feb 18, 2022 4:37 pm

>I did find that both of my APs were put on 5160 until I told capsman to reselect channel mind you.. probably easier to choose channels manually.

The CAPsMAN interface is a little clunky isn't it around setting up multiple channels. The debate about "auto" is an interesting one. In the UBNT AirOS world, I've used auto for many years without any real issues(1) although there the central control via the controller is, in a way, less powerful as you have to edit each AP individually to set channel freq, width etc. But then again, that might be actually easier than CAPsMAN - don't you have to start assigning configurations by radio MAC?

(1) only time I tend to set channels is on PtP bridges.

carl0s · Fri Feb 18, 2022 4:58 pm

>I did find that both of my APs were put on 5160 until I told capsman to reselect channel mind you.. probably easier to choose channels manually.

The CAPsMAN interface is a little clunky isn't it around setting up multiple channels. The debate about "auto" is an interesting one. In the UBNT AirOS world, I've used auto for many years without any real issues(1) although there the central control via the controller is, in a way, less powerful as you have to edit each AP individually to set channel freq, width etc. But then again, that might be actually easier than CAPsMAN - don't you have to start assigning configurations by radio MAC?

(1) only time I tend to set channels is on PtP bridges.

well, you can override the channel conf, or any conf, in the CAP Interface area as well after they are provisioned. Doesn't have to be a separate provisioning profile/setting. That will override the provisioned config.

carl0s · Fri Feb 18, 2022 5:28 pm

5160 is in the full channel list and is allowed (https://en.wikipedia.org/wiki/List_of_WLAN_channels) but it does not show up in the drop down list of ROS
And cannot be used in ETSI regulatory domains. Maybe some other countries can use it.

That's interesting, and it's possible that it was 5180 that both of my APs were set to, not 5160. This was just yesterday that I was tinkering at a couple of the coffee shops. I hadn't realised 5160 might be being disallowed - I will have to experiment. cheers

carl0s · Mon Feb 21, 2022 7:41 pm

5160 is in the full channel list and is allowed (https://en.wikipedia.org/wiki/List_of_WLAN_channels) but it does not show up in the drop down list of ROS
And cannot be used in ETSI regulatory domains. Maybe some other countries can use it.
Klembord-2.jpg

It's strange.
IR 2030 lists Nominal Centre Frequencies starting at 5180, and says the reference is EN 301 893. EN 301 893 clearly seems to say 5160 to 5340, and 5480 to 5700.

ir2030.png

EN301893.png

robmaltsystems · Tue Feb 22, 2022 5:38 pm

>local forwarding will speed things up considerably.

I've set this up - the laptop I'm accessing remotely is in the comms rooms, not in open space to the cAP ac. Getting about 90Mbps which is of course perfectly adequate. Will drop in for a beer some time and do a test in open space. It's been a fascinating learning exercise in Wi-Fi and RouterOS so thanks everyone for taking their time to reply. Whilst not got to the bottom of the very slow Wi-Fi, I am now well placed to diagnose it if it happens again.

Summary of some of the changes made:

Changed channel width in 2.4GHz channels to 20MHz. Like most locations, this range is horribly congested. I'd left inSSIDer running all day and was horrified by the massive (like hundreds) of access points that it had found. Until I spotted that most of them were "Last seen" hours ago and realised that it was recording every single car and bus with Wi-Fi that drove past
Switched to G/N only and disabled B
Using auto on the 2.4GHz as it's so bad anyway that letting RouterOS count access points is as good as picking channels manually. However, I have fixed them to channels 1, 6 and 11
Specified the channels for the 5GHz spectrum by access point. Need to do a full site survey to find the perfect channels but the 5GHz spectrum isn't anywhere near as congested
Switched to AC only
Turned on local forwarding in CAPsMAN and understood what that's doing

hecatae · Wed Feb 23, 2022 10:52 am

Best I can see is channel 1, 20 MHz.wide. There are many of them and auto will not pick this up, but ok, no problem, co-channel interference is just about playing the "contention game" to get a transmit timeslot. They seem to play nicely around channel 1. To know how "bad" or "busy" it is , one should look with "Snooper" and "Freq usage".
viewtopic.php?t=182565#p907921
Even spectral scan over a longer period.... viewtopic.php?t=182565r#p908081

And what to do if it is all too busy? If somebody wiould stream a video, the MT would almost stop it's transmissions because it cannot get a free transmit window.. Reason is that video streaming is handled as priority with WMM. Priority means that in the "contention game" they use smaller values for the contention window wait, so they get a free transmit window more often.
Always wondered no one seems to use this to get a larger part of the available air-time just for all traffic. Oh wait ... it looks as they do (https://www.techjunkie.com/get-priority-wi-fi-network/)

MT has priority=0 ("Best Effort") on all traffic by default. So enabling WMM uses the long "Best effort" contention window settings, the same as without WMM. Only priority 0 has A-MPDU aggregation enabled. And that's important for the throughput: viewtopic.php?t=165698#p912622
To set the priority, one must do it on the AP, with FW mangle rules. Ah you can get "video" or even "voice" WMM priority for everything you want. But you lose the MPDU aggregation and speed, unless enabled. (!!!)

The CWmin/CWmax values are 802.11 standard, also in MT : viewtopic.php?p=901234#p864106
Setting all traffic to "Video" and enabling A-MPDU, is like driving with emergency vehicle lighting turned on. Not a 'nice' neighbor attitude, but may be needed not to be drowned by video-streamers. around. And Google wifi-routers do it also anyway.

They're in the United Kingdom, would Channel 13 be better, it looks cleaner than channels 1-11, 20mhz channel 13 should make the 2.4ghz fly.

bpwl · Wed Feb 23, 2022 1:32 pm

11 and 13 are only 2 x 5MHz channels apart, so they overlap.
In Europe and UK one could choose for 1-5-9-13 instead of 1-6-11 if b-protocol is not used, and if everyone is using that.
But most use 1-6-11 already, so the 13 is not really free.
Adjacent channel interference is worse than co-channel coexistence.

robmaltsystems · Wed Feb 23, 2022 4:54 pm

>Adjacent channel interference is worse than co-channel coexistence.

Interesting. So better for two overlapping access points to be on channel 1 rather than (say) 1 and 3?

bpwl · Wed Feb 23, 2022 9:00 pm

That is what most experts say: co-channel is better than adjacent channel: https://www.metageek.com/training/resou ... ls-1-6-11/

Logic behind this is that traffic has to wait in co-channel until the "ether" is clear to send. It's queuing, but no extra traffic. (Like a round about in traffic) The airtime is optimally used.

In adjacent channel interference one sender does not mind for the other, and both or more send at the same time.(Like in traffic with no rules) At the receiver end the signal will be disturbed and the higher MCS rates will be easily corrupted. The packet transfer fails, and has to be sent again, with probable new corruption.

Not in all cases is adjacent channel worse than co-channel, it depends on how much packets get corrupted. A weaker adjacent signal at the receiver corrupts less.

Cl3an · Wed Feb 23, 2022 9:23 pm

Try to change location from UK to United States to se if you get better signal strength. I live in Sweden and when I change my location my signal gets stronger

robmaltsystems · Thu Feb 24, 2022 6:38 pm

Well this makes no sense. I've been remoting into the laptop periodically, typically at least once a day, to do a speed test. Most of the time I get ~100Mbps but on *very* rare occasions I get >200Mbps on exactly the same set-up. Happened today:

In order to force the laptop to use the Wi-Fi, I've got a PowerShell script that disables the network card for a period of time (usually 2 minutes) and then re-enables it. Boy did I test that script a lot otherwise I could lose access

I thought that it hadn't disabled the wired connection (which is 300Mbps) but I sat there watching the status on the access point Wi-Fi interface and sure enough 200Mbps going through it.

But as you can see, 5 mins later and it's back to slower speeds.

Idle question, aside from other access points and DFS radar, does anything else interfere on 5GHz? I'm going to do some speed tests on between the access point and the switch - and the switch to the router.

bpwl · Thu Feb 24, 2022 8:04 pm

Idle question, aside from other access points and DFS radar, does anything else interfere on 5GHz? I'm going to do some speed tests on between the access point and the switch - and the switch to the router.

Wifi is a dynamic world (on the millisecond), as one is seldom allone using the "ether" to transmit things.

Most things have been said here already. So what follows is a summary and repeating.
Doing overall speed tests is good, as an overall quantification of what can be attained. But it is not good enough for differential diagnosis to find the root cause(s).
(All IT prolems become complex when there are multiple independent causes).

For wifi there are at least three levels: 1. interface rate 2. co channel interference 3. IP protocol

1. Interface rate
Very important to check ! It's dynamic, varies all the time in the real world. Send and receive are 2 separate values. It can be seen in the registration table, with some detail. Typical readout : "400Mbps-40MHz/2S/SGI". MT does not give the MCS level (can be looked up in https://mcsindex.com/) Here 400Mbps is MCS09 for a 40MHz wide channel (e.g. 20MHzCe) , with dual stream, short guard interval 0.4µs. All can vary : MCS rate, # of streams, guard interval , even the bandwidth can drop, dynamic in time. Wifi devices reduce the interface rate based on failed transmissions, and augment the rate again with succesfull transmissions. Each vendor has its own strategy. Retransmissions can be seen (HW Frames versus Frames), CCQ is a deduction from this. Transmissions fail because of adjacent channel interference, distortion, diffraction and noise with high MCS rates (=high QAM) where the signal cannot be decoded for that QAM. SNR is in the registration table.

Throughput is reduced by retransmission ratio. and It follows also the interface rate !

Going from interface rate to data rate can be calculated: viewtopic.php?t=165698&hilit=spreadsheet#p912622

2. co channel Interference

It's seldom one is alone on the channel. Co-channel interference or coexistence is gently waiting for each other, based on the timing in the packet headers. Packet headers are received further away than real data can be decoded. Mikrotik is not giving "channel busy" information. Other brands do, and send it even in the AP beacon as QBSS. The contention window (CW) parameter defines the chances to get a free transmission slot. CW parameters are different with WMM for video and voice. (they use smaller values and get much better chances that way).

Some time is lost in management packet overhead. (0.5% for every SSID on every AP if set at 6 Mbps, 3% if basic rate set at 1 Mbps)

Only one can talk in a channel. Data throughput is reduced to free airtime.

3. IP protocol.

The influence of TCP congestion avoidance is not to be underestimated. There are many protocols in the clients (https://en.wikipedia.org/wiki/TCP_congestion_control ). The impact can be huge, getting lower than 10% usage. Varying wifi transmission conditions can trigger unwanted congestion avoidance actions.

DFS does not slow down the data rate as above. If "radar detected" is triggered, then the AP must change frequencies at once. For a Mikrotik this is for at least 1 full minute no communication possible if another DFS channel is selected.

dbraddon · Mon Feb 28, 2022 10:31 pm

This is fantastic information, and you are really well educated on this topic. So I thank you so much for taking the time to answer my posts.

I take it from your response that there are no setting changes that I can make to the wireless interface that are going to increase bandwidth above what I am seeing? Is that the take away?

robmaltsystems · Thu Apr 14, 2022 12:07 pm

This one rumbles on and I hadn't quite appreciated how under-powered the RB2011UiAS is CPU and RAM wise. The horribly slow Wi-Fi speeds that triggered this long interesting thread continued to occur. We've now added a scheduled task to reboot the router every night and, touch-wood, this has really helped. Maybe something was just building up over a busy day. We're considering replacing the router with something more powerful but the chip shortage means our supplier is on 1-2 month delivery schedules

anav · Thu Apr 14, 2022 3:37 pm

If implemwnting wifi for a business (bars etc), I wouldnt touch wifi5 equipment at this point, wifi6 devices are quite common.

robmaltsystems · Thu Apr 14, 2022 3:47 pm

Bars doesn't really mean or merit business connectivity. Are you in sales for Mikrotik?

To be honest, Wi-Fi 4 is still more that satisfactory for a bar situation. That said, the split bandwidth use of Wi-Fi 6 to make more effective use of the bandwidth is a useful feature and should allow more devices per access point. Never made sense of a fridge to use up the entire channel width.

But AFAIK Mikrotik doesn't offer Wi-Fi 6 yet? They really do need to get it on their roadmap. With TP-Link offering Wi-Fi 6 access points for just over £100, the price differential isn't that much - esp. if you can get away with fewer access points. This bar has three - could have quite easily got away with two.

anav · Thu Apr 14, 2022 8:31 pm

Bars doesn't really mean or merit business connectivity. Are you in sales for Mikrotik? To be honest, Wi-Fi 4 is still more that satisfactory for a bar situation. That said, the split bandwidth use of Wi-Fi 6 to make more effective use of the bandwidth is a useful feature and should allow more devices per access point. Never made sense of a fridge to use up the entire channel width.

But AFAIK Mikrotik doesn't offer Wi-Fi 6 yet? They really do need to get it on their roadmap. With TP-Link offering Wi-Fi 6 access points for just over £100, the price differential isn't that much - esp. if you can get away with fewer access points. This bar has three - could have quite easily got away with two.

I dont care about the nomenclature used for APs, business class what have you, what matters is customer satisfaction which means coming back, word of mouth, ie spending more time and thus money at the bar. If decent wifi is part of that equation, then not sure what your problem is...... No nothing to do with sales, but I do consider value point of products and thus concur with the latter half of your comments starting at "That said,..)", prior to that ... garbage.

robmaltsystems · Tue Jun 14, 2022 10:05 pm

I continue to monitor this site as it's working fine now we're rebooting the router every week and the access points every night. We're regularly getting 250-300Mbps throughout the site which is more that enough for a bar. Before these reboots, speed would be fine for days but then randomly drop from the aforementioned speed to <20Mbps.

This doesn't smack to me of a hardware or even a driver limitation but a software fault. A memory or other resource leak.

When I've got a time slot free when the bar isn't open (rare!), I'm going to upgrade to RouterOS 7.

gotsprings · Wed Jun 15, 2022 1:17 am

I have Tiks that have not been rebooted since 2018...

Wed Jun 15, 2022 2:17 am

i think Wi-Fi diagnostic without site survey and spectrum analysis is a guessing game

Ca6ko · Wed Jun 15, 2022 9:08 am

RB2011UiAS is CPU and RAM wise.

You need to set up the capsman on the cap-ac, This is to unload 2011.
You didn't specify the number of clients per access point anywhere. I have a cafe with similar systems. When there are more than 25 clients per access point, wifi network performance drops dramatically.
I totally agree with chechito.
It is very common that network engineers are not able to configure the radio network parameters, rely on automatic settings and get frustrating results.

Znevna · Wed Jun 15, 2022 9:51 am

[...]
When I've got a time slot free when the bar isn't open (rare!), I'm going to upgrade to RouterOS 7.

Don't do that mistake. It will trash whatever performance RB2011 can still bring to the table thanks to the route-cache still present in the older kernel.
What I would do: ditch CAPsMAN altogether and setup each AP as standalone, 3 APs aren't that hard to configure.

robmaltsystems · Wed Jun 15, 2022 12:14 pm

>You didn't specify the number of clients per access point anywhere.

Assuming that everyone is spread out throughout the bar, then the maximum I've seen (at busy times) is 20 DHCP entries per access point. However, when I look at the list of registered clients (i.e. currently attached to Wi-Fi), then it's a handful of active clients.

robmaltsystems · Wed Jun 15, 2022 12:18 pm

>i think Wi-Fi diagnostic without site survey and spectrum analysis is a guessing game

I don't think that would achieve much as we know the spectrum is noisy and busy esp. in the 2.4GHz spectrum.

Everything is fine for days on end with perfectly acceptable speeds. However, speed will occasionally *massively* drop from 250-300Mbps consistently to <20Mbps. I saw this last night. As the network can limp along at this speed, I tested it every 15 minutes for an hour. Speed never greater than 20Mbps.

But, and this is the key point, the instant I rebooted the CAP ac, speed jumped immediately back to 250-300Mbps. Nothing else changed in the environment.

That must be some problem with the CAP ac itself - either hardware or software. Everything is suggesting that isn't the environment.

robmaltsystems · Wed Jun 15, 2022 12:19 pm

Don't do that mistake. It will trash whatever performance RB2011 can still bring to the table thanks to the route-cache still present in the older kernel.
What I would do: ditch CAPsMAN altogether and setup each AP as standalone, 3 APs aren't that hard to configure.

Noted... it did cross my mind. We're going to replace it when we can BUT lead times in the UK on the replacement is currently August :- another reason I'm moving away from Mikrotik

Znevna · Wed Jun 15, 2022 12:32 pm

You don't have to replace anything.
Just reconfigure 3 APs with some basic config.
CAPsMAN is a nice tool to use when you manage many more devices, not just 3. Otherwise it doesn't bring any added-value.

Ca6ko · Wed Jun 15, 2022 1:00 pm

I rebooted the CAP ac, speed jumped

I have seen this effect several times on these points. The reason was always too high installed power, and accordingly strong interference. Or the point is not set correctly - on the wall, not on the ceiling and close to the corner.
In both cases the access point overloads itself with the reflected signal.

robmaltsystems · Wed Jun 15, 2022 2:55 pm

You don't have to replace anything.
Just reconfigure 3 APs with some basic config.
CAPsMAN is a nice tool to use when you manage many more devices, not just 3. Otherwise it doesn't bring any added-value.

I'm already using CAPsMAN although early on this thread, it was recommended to use local forwarding on the access points themselves and not via the router. Which I've done.

CAPsMAN is okay but for a small set-up is probably not worth the effort - it's not a trivial process. It's complicated when you want to specify per AP which Wi-Fi 6 channel to use. For large installations, I use UBNT UniFi access points because their central management is far easier to use esp. the automatic rollout of new firmware.

robmaltsystems · Wed Jun 15, 2022 2:59 pm

I rebooted the CAP ac, speed jumped
I have seen this effect several times on these points. The reason was always too high installed power, and accordingly strong interference. Or the point is not set correctly - on the wall, not on the ceiling and close to the corner.
In both cases the access point overloads itself with the reflected signal.

Yes, I would imagine that this can be a factor but doubt it in this case. The access points are ceiling mounted in the middle of open spaces. If something is intermittently causing reflected signals, they it would go away itself. It doesn't... Also, would that explain a drop in speed from 250Mbps to <20Mbps? That's a *lot* of interference.

Again, if it was interference, why would a reboot of the AP cause it to instantly fix itself?

Znevna · Wed Jun 15, 2022 3:30 pm

I was trying to say that it's not worth running capsman just for 3 APs, and you should consider not using it anymore and configure the APs as stand alone, normal APs.
Sorry if it got lost in translation.

carl0s · Wed Jun 15, 2022 10:07 pm

etc. But then again, that might be actually easier than CAPsMAN - don't you have to start assigning configurations by radio MAC?

No.
What you need to do, is make sure your provisioning conf is not set to Dynamic.
Change it from create dynamic enabled, to just create enabled.
Then you can tinker with the capsman interfaces after they have been created. You can change channel, extension channel, output power, etc (actually everything) from capsman interface list.
I still think it is much better than ubiquiti, and TP link omada, just because it is properly managed. Just a shame the WiFi drivers are outdated.

robmaltsystems · Wed Jun 15, 2022 10:11 pm

>Just a shame the WiFi drivers are outdated

Hence the interest in RouterOS 7 because I assume the drivers/kernel have been updated?

Wed Jun 15, 2022 10:12 pm

Wifi is updated to wifiwave2 but only for limited devices.
No wifi 6 on the horizon yet.

carl0s · Wed Jun 15, 2022 10:22 pm

>Just a shame the WiFi drivers are outdated

Hence the interest in RouterOS 7 because I assume the drivers/kernel have been updated?

Yes, wifiwave2-driver support for more devices (existing hardware), and support for capsman, is what I am waiting for.

Right now you can move to rOS v7, but you can't move to wifiwave2 drivers with capsman, and also the wifiwave2 driver package only supports few devices like audience, hap-ac3, and rb4011 or something. Makes the hap ac3 interesting anyway, but no capsman support on wifiwave2 'new driver'.
Certainly no proper Poe ceiling APs announced yet anyway, and mikrotik are asking about people's ideas for a management protocol (see other thread) which might mean they are not looking to extend capsman. And they are very early days on the new protocol/system judging by the sort of feedback being requested.

robmaltsystems · Sat Oct 22, 2022 11:56 pm

The site keeps rumbling on. Further up, the performance of the router we're using was questioned (quite rightly). Just had further confirmation that it's underpowered. We went to put a queue on the public Wi-Fi. Dead easy to do except that it caused throughput on the private unthrottled network to also drop from 350Mbps wired and ~300Mbps wireless to around 100Mbps. Turn Fasttrack back on and back to full speed. Caused of course by the additional overhead when Fasttrack is disabled to allow queues to work.

It's a shame can Fasttrack can only be turned off for a subnet? Ohh hang on, can it?

robmaltsystems · Sun Oct 23, 2022 12:21 am

Once again, impressed with flexibility. Yes, can put source address of the private subnet in to Fasttrack so it's used for that traffic but the throttled public traffic queue works. Still need to put a more powerful router in though

robmaltsystems · Sun Oct 23, 2022 3:10 pm

and mikrotik are asking about people's ideas for a management protocol (see other thread) which might mean they are not looking to extend capsman. And they are very early days on the new protocol/system judging by the sort of feedback being requested.

Sorry missed this back then... I'm not surprised. CAPsMAN seems, to be honest, a bit more trouble than it's worth esp. for an installation of maybe a handful of access points. I found the configuration pretty messy when I tried to put specific access points on specific channels. Maybe in somewhere like a bit hotel where overlapping channels isn't that much of a worry, then it might make sense.

gotsprings · Sun Oct 23, 2022 3:38 pm

robmaltsystems

I couldn't disagree more. If you even have 2 Mikrotik WAPs... Caps-man is awesome. It's speed and flexibility would make it one of the best wireless managers I have ever seen.

The crap radio performance, snaps me back to reality real quick thou.

robmaltsystems · Sun Oct 23, 2022 4:14 pm

I can certainly see it's worth for large installations but not for a three AP large bar

mkx · Sun Oct 23, 2022 5:30 pm

I guess that for working WiFi roaming (read: 803.11 r/k/v) in Mikrotik network, one will have to use CAPsMAN even if network infrastructure consists of only 2 APs. So better get used to CAPsMAN

gotsprings · Sun Oct 23, 2022 8:16 pm

I can certainly see it's worth for large installations but not for a three AP large bar

What?

A 3 AP BAR.
You can use caps-man for:
Per client passwords for the managers.
PCC enforcement of bandwidth.
Scripting to shut down the guest wifi on a schedule
Scripting to shut down the get SSID when your primary gateway goes down... And you limp along on bandwidth metered, limited speed back up connection.
Proper roaming between APs.
Priority for wifi calls.
Etc
Etc
Etc

However, like I said... The radios inabilities to deal with Multiple Users or deal with nearby interference taught me a PAINFUL LESSON.

gotsprings · Sun Oct 23, 2022 8:20 pm

I guess that for working WiFi roaming (read: 803.11 r/k/v) in Mikrotik network, one will have to use CAPsMAN even if network infrastructure consists of only 2 APs. So better get used to CAPsMAN

I got used to ordering WAPs from other manufactures.

robmaltsystems · Sun Oct 23, 2022 9:01 pm

>deal with nearby interference taught me a PAINFUL LESSON.

Please tell me more, this is worrying!

robmaltsystems · Sun Oct 23, 2022 9:01 pm

>I got used to ordering WAPs from other manufactures.

This is indeed VERY worrying. Without decent Wi-Fi access points, might as well stick with everything from another manufacturer.

gotsprings · Mon Oct 24, 2022 4:48 am

>I got used to ordering WAPs from other manufactures.

This is indeed VERY worrying. Without decent Wi-Fi access points, might as well stick with everything from another manufacturer.

Absolutely not.

I try to use the best parts I can.
When it comes down to routing... Mikrotik has the least amount of "yeah I am sorry... I can't do that." And when the Ubnt distributor for the Illinois, Missouri, Wisconsin, Minnesota region needed to interconnect their branches. They may have UniF--k switches in the rack and WAPs on the ceilings... But the routing was way out side what Ubnt Edge or UniF--k could do. So they had to buy some Mikrotik routers and pay some d--k in Maryland to program them...

EdgeSwitches have been my preferred switch for years.

When it comes down to wifi... That's the critical last link to clients devices.

And when your wifi is still using 2014 protocols in 2022... It's a little ridiculous.
I found out after it happened again and again... There are radio cards that Mikrotik has problems with. If you happen to be stuck with one... No amount of tweaking will save you.
I learned that the 2.4 radios in the WAPs will "just give up" if you put them in a noisy enough environment.
The lack of even ACv2 really limits the abilities of multiple devices to be handled well. I mean that improvement to AC was ratified in 2016... And we have what 3 radios that can use Mikrotik's "in process Wave2 driver". And forget about caps-man... it only works with the old driver.

I had been using Ruckus for years when I got sucked into "look at all the cool s--t I can do with caps-man". In the lab and in really low density installs... The speed limits were the biggest issue. As long as people had slow enough WAN and they were in the middle of nowhere... Mikrotik wireless might be OK.

Then I made the mistake of letting it into a few installs in the city. Nearby WAPs overwhelmed the Mikrotik WAPs and they refused to accept clients until rebooted. They would work for a while then do it again.

The next kick in the teeth came from the automation companies preferred choice of remote controls. The wifi chip in them would not play well with the Tik WAPs. Often showing as connected but not actually passing traffic. Until restarted again.

The commercial installs were just one trouble ticket after the next. The interference and inability to deal with client density caused full melt downs.

When I first shifted over to Mikrotik radios... They could compete with UniF--k. Even handle more clients than the counts that locked up UniF--ks. But as more and more devices became wireless... Neither one would have done it.

To this day I am constantly annoyed by other trades showing up with their UniF--k WAPs, thinking they are the smartest Mother F--kers in the World. I warn them that the nosie where we are working around is gonna render their WAPs useless... They never believe me. A week or two later they are begging me to put their SSIDs on my radios. Because while their APs start to get flakey and orders don't go through... I am carrying a couple hundred guests on our wifi with the managers using video calling on their SSID...

Update/edit:
Holy f--king s--t.
There is the actual thread where Mikrotik officially told you... "Wrong tool for the job."

Kaldek · Mon Oct 24, 2022 7:33 am

I guess that for working WiFi roaming (read: 803.11 r/k/v) in Mikrotik network, one will have to use CAPsMAN even if network infrastructure consists of only 2 APs.

802.11 r/k/v are protocols that should be deeply understood before being concerned about the lack of them.

802.11r only really benefits handoff when using EAP methods. No real benefit when using Pre-Shared Keys.
802.11k provides information about all the access points for the purpose of keeping the load on an AP low. Clients can (and do) ignore 802.11k reports.
802.11v is similar to 802.11k. Clients can (and do) also ignore 802.11v data.

My point is, Mikrotik does have some history here, but the lack of 802.11r/k/v is a red herring for most environments. For a corporate environment that uses EAP and needs fast handoff roaming, the lack of 802.11r can be a problem that does impact user experience. If you have a lot of roaming whilst performing real-time voice/video.

Rox169 · Mon Oct 24, 2022 8:13 am

Hi,

MT already support 802.11 r/k/w only v is missing. But the question is if it works as should be. The 802.11 r is cripled to only one AP which is client connected.

robmaltsystems · Mon Oct 24, 2022 11:40 am

>I learned that the 2.4 radios in the WAPs will "just give up" if you put them in a noisy enough environment.

That's very sad to hear and is the most likely explanation so far as to why this bar situation sometimes just becomes unusable requring a reboot. It's incredibly noisy. Not only do they have the 20+ access points in surrounding businesses, but they also have a stream of cars & buses fitted with 2.4Ghz Wi-Fi going right past the window. Leave inSSIDer running all day and it records hundreds of access points.

I'll will keep this in mind when doing the next install and will use UBNT access points which don't seem as unstable.

robmaltsystems · Mon Oct 24, 2022 11:44 am

>The is the actual tread where Mikrotik officially told you... "Wrong tool for the job."

Having read to the bottom of your post now, this is probably the most important post I've read in the long time

Question - is the CAP ac problem hardware or software?

gotsprings · Mon Oct 24, 2022 12:21 pm

There is a post from Mikrotik in this thread where they told you and everyone else... Wrong tool.

Also...
I sent Mikrotik a support file this weekend. They got back to me this morning with the typical "can you update the device" or change some random bulls--t.

If I keep everything else in my network the same but swap access points to another manufacture... I have no problem keeping 2.4 clients connected.

Let me qualify that. I have had issues with Ruckus and Cambium over the years. When I have an issue with Cambium... They look at the support file and FIX THE DRIVER. The next software release corrected the situation. No changes were needed on my system.

With Ruckus... They don't really pay attention anymore. I have to test each firmware from the ground up. There are releases that my "canary" can't connect to. If I put that firmware out in the field... Complaint tickets start coming in. So when I find a working firmware... It goes on all new deployments despite being 17 months old. (They finally got a good firmware out with the release of the H550.)

But as to the question of hardware or software...
The cAP AC doesn't have sufficient memory to run the WAVE 2 driver. There for the answer must be "both". (Unless you put openwrt on them. Or some other hack I was reading about months ago...)

bpwl · Mon Oct 24, 2022 12:56 pm

"Leave inSSIDer running all day and it records hundreds of access points."

Be aware of the SSID/beacon overhead in a channel. Each SSID transmitting its beacon in 802.11b mode or at 1 Mbps, will consume 3% of the available air-time. (There are overhead calculators on the internet: https://revolutionwifi.blogspot.com/p/s ... lator.html ). 6 AP's with 6 SSID each in the same channel, and ALL air-time is gone. Disconnects will happen.

A 802.11b client/device will force higher level (g,n,ac,ax) 802.11 AP's to support 802.11b mode, even if 802.11b is not enabled. (Check with InSSIDer)

jbl42 · Mon Oct 24, 2022 10:15 pm

It is and was a sad story:
Never put MT WiFi APs into noisy environments, especially not if you are the one being called if things do not work:

The 2.4Ghz Radio will kind of lock up every few hours or days, requiring a reboot to get clients connecting again

For 5Ghz, if using DFS channels (what is important in noisy environments), it will detect bogus radars on random channels. Causing it to constantly jump channels and sometimes also to lock up similar to the 2.4Ghz radio.

Being a MT router proponent, people often do not believe me that MT WiFi sucks for noisy/high density environments.
All of them learned it the hard way.

And yes, replacing the WiFi AP in such locations with other brands at the same location, keeping everything else unchanged, suddenly makes WiFi work.
MT support does know this. If pushed hard enough, they even admit it.

ivicask · Mon Oct 24, 2022 11:21 pm

Exactly same observation my side to the last word!

gotsprings · Wed Oct 26, 2022 1:47 pm

The last email from support... they recommended changing the lease time on the router and Extending the Key Exchange time out to something longer than the 5 min default.

Both are now done.

My 2 2.4 test cameras that I use as a canary in the coal mine. Both connected to the Audience 2.4 radio. All settings are pretty standard fair.

Both showed as connected with a signal in -65 range.

As soon as I connect to one of the cameras... it kind of hangs then in wireless... the signal shows -88 and the Tx and Rx show as 1M.

That will be the end of communication with that unit.

ivicask · Wed Oct 26, 2022 1:55 pm

Doesnt help i set that long time ago...

gotsprings · Thu Oct 27, 2022 12:32 am

There was a version back on 7.1 with the WiFi Wave 2 driver that the cameras connected to the Mikrotik Audience BETTER than my other WAPs.

One version of Ruckus firmware was so messed up... The audience was there to ONLY CARRY MY OLD CAMERAS. It could handle the house full of devices at that time too... I forgot to note the exact version.

The Audience would reboot suddenly from time to time. Support confirmed it was the Audience asked for too much power from the switch. Ended up plugging in it's power supply and it was "stable", well for Mikrotik wireless.

jbl42 · Sun Oct 30, 2022 3:46 pm

The last email from support... they recommended changing the lease time on the router and Extending the Key Exchange time out to something longer than the 5 min default.

This is sheer desperation. There are long standing issues with key exchange on MT WiFi, but this is not related to noise/HD problems.

Being an electrical engineer with some RF design experience, I dare to say that the observed issues are clearly showing the symptoms of poor RF circuit design: Lack of SNR, channel selectivity, and RF amplifier linearity. This is not fixable in SW.
The main reason for price differences to professional HD/high noise capable WAPs is the RF radio circuit design. Good RF circuits require lots of experience and add to the BOM.
You get what you pay for.

robmaltsystems · Sun Oct 30, 2022 6:46 pm

>You get what you pay for

But the price differential between MT and UBNT isn't that much - cap AC is £92 inc. VAT in UK and UAP AC Long Range comes in at £109. £20 for something that works isn't worth the saving.

jbl42 · Sun Oct 30, 2022 9:42 pm

The UBNT are much more stable (and you have to like the UniFi controller thing), but if you need an AP running reliable in a busy bar or office, with 10+ neighbor SSIDs fighting each other, they still have issues.
In such locations, it is worth to spend some £200 - £250 for a lower end professional business WAP. It might sound expensive at first, but it will be worth it. Such devices play in another RF performance league.

gotsprings · Fri Nov 04, 2022 1:12 pm

How I miss being able to pepper a location with H510s.

gotsprings · Sat Jan 07, 2023 11:42 pm

Quote of the day...
"It doesn't matter what it costs, if it DOESN'T F--KING WORK!"

The boss was arguing with the Point of Sales guys... As a test we put their SSID on our radios and unplugged theirs... Strangely, all the handhelds the waiters use actually worked.

Weird.

robmaltsystems · Sun Jan 08, 2023 12:32 pm

So the overall feedback to Mikrotik here appears to be "Great software. Not flawless but exceptionally powerful and flexible. But please fix the radios & highlight capacity issues in low-end kit/get some more powerful mid/low-range kit out there esp. access points"?

anav · Sun Jan 08, 2023 9:22 pm

You sir are the slowest learner I have ever seen. I do appreciate the experience and feedback from gotsprings it keeps me coming back to read.

DarkNate · Sun Jan 08, 2023 10:09 pm

hAP ax2 works fine for me. 850Mbps download peak performance and around 920Mbps upload peak performance with minimal bufferbloat using FQ_Codel.

For $99, good luck finding a gigabit AP other than hAP ax2.

gotsprings · Sun Jan 08, 2023 11:19 pm

hAP ax2 works fine for me. 850Mbps download peak performance and around 920Mbps upload peak performance with minimal bufferbloat using FQ_Codel.

For $99, good luck finding a gigabit AP other than hAP ax2.

More like good luck finding a hAP AX2. At least, none of the distributors I registered with have got any yet.

robmaltsystems · Mon Jan 09, 2023 1:27 am

>You sir are the slowest learner I have ever seen.

And you are one of the rudest people I've had the misfortune to meet on here. Tell you what, if you've not got anything nice to say, don't say anything at all. What do any of your comments add to the discussion?

robmaltsystems · Mon Jan 09, 2023 1:39 am

More like good luck finding a hAP AX2. At least, none of the distributors I registered with have got any yet.

Unlike some Miktotik kit, we can get the hAP AX2 in the UK:

https://linitx.com/product/mikrotik-hap ... d-tc/17189

The hAP AC2 is also a good price:

https://linitx.com/product/mikrotik-hap ... -psu/15370

bpwl · Mon Jan 09, 2023 1:58 am

So the overall feedback to Mikrotik here appears to be "Great software. Not flawless but exceptionally powerful and flexible. But please fix the radios & highlight capacity issues in low-end kit/get some more powerful mid/low-range kit out there esp. access points"?

+1

There are some things MT could do, maybe, if it fits the tight memory.

- getting TPC certification, so as to be allowed to use the TPC EIRP levels with the classic MT wifi drivers. (wifiwave2 driver uses the +3dB TPC level for EIRP power limit, like everything else I know)
- fix the A-MSDU and A-MPDU aggregation , and bring it to the higher values of the 802.11 standard, like wifiwave2 and what all others do. Not to waste that much air-time, because of the smaller transmissions (losing quite some throughput at the high interface rates)
- make WMM work, based on DSCP tags by default. (and keep aggregating also for the non-BE priority by default)

anav · Mon Jan 09, 2023 4:29 am

>You sir are the slowest learner I have ever seen.

And you are one of the rudest people I've had the misfortune to meet on here. Tell you what, if you've not got anything nice to say, don't say anything at all. What do any of your comments add to the discussion?

Really, your the ashole who screws a bar over by providing crappy wifi service despite the advice here, and further forget about the owners, you continued to provide people at the bar with crappy service, effing shame on you mate.

wsunardy · Tue Jan 10, 2023 1:26 pm

hAP ax2 works fine for me. 850Mbps download peak performance and around 920Mbps upload peak performance with minimal bufferbloat using FQ_Codel.

For $99, good luck finding a gigabit AP other than hAP ax2.

May I know what need to configure in ax2 to get this speed?
My ax2 only can get 600Mbps max

DarkNate · Fri Jan 13, 2023 2:20 am

May I know what need to configure in ax2 to get this speed?
My ax2 only can get 600Mbps max

Use single bridge configuration where your LAN ports and both wireless interfaces are in the same bridge, segregate them with VLAN filtering if you need to. Then enable bridge fastpath/forward and configure the wireless parameters like this:

/interface wifiwave2
set [ find default-name=wifi2 ] channel.band=2ghz-ax .skip-dfs-channels=all .width=20/40mhz configuration.chains=0,1 .mode=ap .ssid=Dark_Nate .tx-chains=0,1 .tx-power=40 disabled=no \
    l2mtu=2290 mtu=2290 name=2GHz_1 security=Main_VLAN_Security security.authentication-types=wpa2-psk,wpa3-psk .wps=disable
set [ find default-name=wifi1 ] channel.band=5ghz-ax .skip-dfs-channels=all .width=20/40/80mhz configuration.chains=0,1 .mode=ap .ssid=Dark_Nate .tx-chains=0,1 .tx-power=40 \
    disabled=no l2mtu=2290 mtu=2290 name=5GHz_1 security=Main_VLAN_Security security.authentication-types=wpa2-psk,wpa3-psk .wps=disable

I benchmarked using Apple products supporting Wi-Fi 6 natively. All showed around 700Mbps average in the worst case like through three thick walls with a metal container in-between.

Don't forget to configure FastTrack for input/forward/output chain or just use mangle table and do it for prerouting, output, forward etc.

For FQ_Codel do it on interface level:

/queue type
add kind=fq-codel name=FQ_Codel
/queue interface
set ether1 queue=FQ_Codel
set ether2 queue=FQ_Codel
set ether3 queue=FQ_Codel
set ether4 queue=FQ_Codel
set ether5 queue=FQ_Codel
set "2GHz_1" queue=FQ_Codel
set "5GHz_1" queue=FQ_Codel

mrigi · Fri Jan 13, 2023 2:17 pm

For $99, good luck finding a gigabit AP other than hAP ax2.

Actually market is full of other brand AX AP's for half of that.

ANTONBORODA · Fri Jan 13, 2023 3:26 pm

May I know what need to configure in ax2 to get this speed?
My ax2 only can get 600Mbps max
Use single bridge configuration where your LAN ports and both wireless interfaces are in the same bridge, segregate them with VLAN filtering if you need to. Then enable bridge fastpath/forward and configure the wireless parameters like this:
Code: Select all
/interface wifiwave2
set [ find default-name=wifi2 ] channel.band=2ghz-ax .skip-dfs-channels=all .width=20/40mhz configuration.chains=0,1 .mode=ap .ssid=Dark_Nate .tx-chains=0,1 .tx-power=40 disabled=no \
    l2mtu=2290 mtu=2290 name=2GHz_1 security=Main_VLAN_Security security.authentication-types=wpa2-psk,wpa3-psk .wps=disable
set [ find default-name=wifi1 ] channel.band=5ghz-ax .skip-dfs-channels=all .width=20/40/80mhz configuration.chains=0,1 .mode=ap .ssid=Dark_Nate .tx-chains=0,1 .tx-power=40 \
    disabled=no l2mtu=2290 mtu=2290 name=5GHz_1 security=Main_VLAN_Security security.authentication-types=wpa2-psk,wpa3-psk .wps=disable
I benchmarked using Apple products supporting Wi-Fi 6 natively. All showed around 700Mbps average in the worst case like through three thick walls with a metal container in-between.

Don't forget to configure FastTrack for input/forward/output chain or just use mangle table and do it for prerouting, output, forward etc.

For FQ_Codel do it on interface level:
Code: Select all
/queue type
add kind=fq-codel name=FQ_Codel
/queue interface
set ether1 queue=FQ_Codel
set ether2 queue=FQ_Codel
set ether3 queue=FQ_Codel
set ether4 queue=FQ_Codel
set ether5 queue=FQ_Codel
set "2GHz_1" queue=FQ_Codel
set "5GHz_1" queue=FQ_Codel

How do you get such speeds with ax2 through 3 brick walls? In my case, with ax3, a single brick wall kills the signal strength to like -70 dbi and speed drops to like 80 mbps at best. Constant disconnects included. The settings are pretty much the same as yours.

gotsprings · Fri Jan 13, 2023 4:43 pm

For $99, good luck finding a gigabit AP other than hAP ax2.
Actually market is full of other brand AX AP's for half of that.

The F--k it is?

The cheapest competitor in the US would be UniF--k AP lite. Which is $100.

rextended · Fri Jan 13, 2023 4:50 pm

Actually market is full of other brand AX AP's for half of that.

Actually forum are full of users than just write for write.

DarkNate · Fri Jan 13, 2023 5:26 pm

How do you get such speeds with ax2 through 3 brick walls? In my case, with ax3, a single brick wall kills the signal strength to like -70 dbi and speed drops to like 80 mbps at best. Constant disconnects included. The settings are pretty much the same as yours.

Assuming your configuration is A to Z like mine in the most important parts and follows best practices as per official MikroTik docs. And also note, I'm not a wireless network engineer nor am I an expert in hardware and silicon production aka I'm not a physicist or a chemist unlike others in this forum who pretend to be “experts” that know it all, wireless networking is not my specialisation other than installing APs in my house to enjoy 1Gig performance in the bathroom, so take my opinion with a grain of salt. But the following is what I personally keep in mind when I deal with anything in tech including Tik Wi-Fi APs/Routers.

1. With every MikroTik hardware that I buy fresh off the shelf or a used/refurbished product, I always perform a clean, netinstall of the latest ROS v7 at the time of doing it. And I never restore config using the “backup” feature. I do it via script/terminal by hand. For some reason, it seems even fresh out-of-the-box hardware from Tik tends to be buggy if you don't perform a clean netinstall. Why? I don't know, ask MikroTik. But in general, this rule of thumb for “Clean install” applies even to x64 machines and laptops and even the almighty Apple M CPU series laptops, I've had the same problem with factory installs for decades, I always clean installation on a fresh purchase/ownership regardless to avoid problems in the long run and this paid off for me.

2. Hardware quality/electrical components quality and the underlying chemistry varies from batch to batch on the simple fact that constant absolution does not exist. Everything is a variable. Perhaps you got lower quality batch, than I did, a lot of possibilities there, I ain't about to examine every purchase in an electron microscope.

3. The way your house is constructed, material/thickness/consistency of your walls obviously vary from mine. I'm not a civil/wireless networking engineer and hence I will not pretend to know how this affects wireless signals, but it clearly does play a role.

4. I put my AP(s) in an elevated position, as close as possible to the ceiling on each floor. Always remember that wireless signals falls to the ground, it doesn't fly to the sky. Therefore, higher elevation helps improve SNI ratio and is better than placing it on the floor in contrast. I once tested this, but putting a single AP on the roof of the building and compare it with on the ground of the building, the former allows better SNI/signal strength compared to the latter as I moved from floor to floor.

5. Point 2 applies to all electrical devices such as your phone, NIC, laptop etc. I have two Apple M CPU series laptops, both show difference in performance even though both support Wi-Fi 6 natively.

biomesh · Fri Jan 13, 2023 5:42 pm

Actually market is full of other brand AX AP's for half of that.
The cheapest competitor in the US would be UniF--k AP lite. Which is $100.

I would not say it is full of $50 AX APs but you can find a Zyxel wifi 6 AP on amazon for $75 USD. Its not what I would buy, but there are options.

rextended · Fri Jan 13, 2023 7:32 pm

AHAHAHAHAH!!!!!!!!!!!!!
I just looked at it, they give it to you for a discount if... you make a periodic purchase!!!

Have olny 3 LAN, not published any relevant spec (also on site)
Full price is ~60€ if you have amazon prime

mkx · Fri Jan 13, 2023 7:41 pm

4. I put my AP(s) in an elevated position, as close as possible to the ceiling on each floor. Always remember that wireless signals falls to the ground, it doesn't fly to the sky.

While your observation is pretty much spot on, the explanation why it's so is complete garbage. Gravity on Earth surface is several orders of magnitude too small to have any notable effect on how electromagnetic waves travel through space (on short distances of WiFi coverage).
The more likely reason for better signal propagation from ceiling-mounted APs is that most often there are no obstructions in the upper part of building floors (other than walls obviously) while lower part is usually full of obstacles (closets, racks, chairs, plants, people, donkeys, you name it) ... and all of those obstructions degrade both signal strength and cause reflections (which usually translates to self inflicted interference and consequently drop in SINR).

DarkNate · Fri Jan 13, 2023 8:45 pm

While your observation is pretty much spot on, the explanation why it's so is complete garbage.

Which part of the following is unclear?

And also note, I'm not a wireless network engineer nor am I an expert in hardware and silicon production aka I'm not a physicist or a chemist unlike others in this forum who pretend to be “experts” that know it all, wireless networking is not my specialisation other than installing APs in my house to enjoy 1Gig performance in the bathroom, so take my opinion with a grain of salt.

Regardless of possible eye defects from the expert in question. Although I only learnt physics up to high school, I am certain gravity does pull down radio signal towards the ground, the impact on the “bent” may be small, but it is not zero, and it certainly is noticeable on $99 equipment versus a million dollar military/NASA grade radar or wireless antenna, source:

https://qr.ae/pvBSiM

https://www.qrg.northwestern.edu/projec ... waves.html

Also, it appears the physics community has yet to agree on a solid scientific theory on how gravity affects electromagnetic waves such as radio:
https://astronomy.stackexchange.com/a/21495

Unless you're the next Einstein who have solved and explained away quantum weirdness and perhaps proved the existence of gravitons, I rest my case on the layman-understandable fact that gravity affects radio signals regardless of whether there are objects in the path or not.

Fri Jan 13, 2023 9:01 pm

Unless you calculate with 25 digits after the decimal sign, the effect is negligible.
Otherwise we would never be able to communicate with space devices in or out of our galaxy...
Think about it.

Obstructions on lower levels of a floor are MUCH more responsible for signal loss then gravity.

Pure logic. No Einstein needed.

mkx · Fri Jan 13, 2023 11:26 pm

Although I only learnt physics up to high school,

My university degree is in physics and I worked 17 years as radio network engineer for major MNO in my country (half of that as senior radio engineer). So I gather I do know something about effect of gravity on EM waves. Unlike you (or so it seems).

As @holvoetn wrote: light passing galaxies (with black holes in their centres) bends by a few degrees ... now compare those dimensions (mass of a galaxy ... or only of a black hole) bends light by (say) 20° over distance of many light years ... comparing to mass of Earth (neglectable compared to black hole) and distance of a few metres. Do you really think any of your computing devices can hold enough decimal places to show how much that radio signal bends towards Earth centre? (probably its about a diameter of oxygen atom core ... just a guess, don't quote me on that).

rextended · Sat Jan 14, 2023 12:00 am

the deformation is really insignificant due to gravity,
more significant is the fact that the earth, solar system and the galaxy rotates within the just generated magnetic field,
the wave moves away at the speed of light, but at the same time the departure is further forward (or further back) because the earth and all the galaxy is rotated...
obviously this count only if the devices are a few light day of distance, on same apartment is like nothing happen...

bpwl · Sat Jan 14, 2023 12:47 am

Well having interest in the physical working of wifi is important to understand what causes the reduced performance. Nothing wrong with that, on the contrary. But there is always the possibility to misunderstand some phenomena (that's why engineering studies take long, and why there are always discussions between students, even between experts.)

Earth Gravity will not bend the time-space universe in a noticable way, this is something for black hole masses and galaxies.

But this down tilt of the wifi signal is something very realistic and important, certainly for outdoor and for long distance connections.
This down direction is created on purpose, by design, in some antenna, and is sometimes significant. It's not due to gravity, but done with phase shifts in antenna segments.

I have 3 texts that might interest you :

https://www.kpperformance.com/Antenna-D ... rview.html
https://www.cnet.com/home/internet/home ... i-network/
https://www.huffpost.com/entry/wifi-rou ... _n_6943024

I have not seen documented electronic towntilt in the Mikrotik APs, so installing outdoor under an angle as for the mANTBox is quite common
Well, not correct, the cAP certainly has it! And the cAP XL certainly has it, as there is even a reflecting mirror inside the XL variant.
Diagrams of the antenna patterns can be found ... for some devices
viewtopic.php?t=182762#p909850

ANTONBORODA · Sat Jan 14, 2023 1:16 pm

How do you get such speeds with ax2 through 3 brick walls? In my case, with ax3, a single brick wall kills the signal strength to like -70 dbi and speed drops to like 80 mbps at best. Constant disconnects included. The settings are pretty much the same as yours.
Assuming your configuration is A to Z like mine in the most important parts and follows best practices as per official MikroTik docs. And also note, I'm not a wireless network engineer nor am I an expert in hardware and silicon production aka I'm not a physicist or a chemist unlike others in this forum who pretend to be “experts” that know it all, wireless networking is not my specialisation other than installing APs in my house to enjoy 1Gig performance in the bathroom, so take my opinion with a grain of salt. But the following is what I personally keep in mind when I deal with anything in tech including Tik Wi-Fi APs/Routers.

1. With every MikroTik hardware that I buy fresh off the shelf or a used/refurbished product, I always perform a clean, netinstall of the latest ROS v7 at the time of doing it. And I never restore config using the “backup” feature. I do it via script/terminal by hand. For some reason, it seems even fresh out-of-the-box hardware from Tik tends to be buggy if you don't perform a clean netinstall. Why? I don't know, ask MikroTik. But in general, this rule of thumb for “Clean install” applies even to x64 machines and laptops and even the almighty Apple M CPU series laptops, I've had the same problem with factory installs for decades, I always clean installation on a fresh purchase/ownership regardless to avoid problems in the long run and this paid off for me.

2. Hardware quality/electrical components quality and the underlying chemistry varies from batch to batch on the simple fact that constant absolution does not exist. Everything is a variable. Perhaps you got lower quality batch, than I did, a lot of possibilities there, I ain't about to examine every purchase in an electron microscope.

3. The way your house is constructed, material/thickness/consistency of your walls obviously vary from mine. I'm not a civil/wireless networking engineer and hence I will not pretend to know how this affects wireless signals, but it clearly does play a role.

4. I put my AP(s) in an elevated position, as close as possible to the ceiling on each floor. Always remember that wireless signals falls to the ground, it doesn't fly to the sky. Therefore, higher elevation helps improve SNI ratio and is better than placing it on the floor in contrast. I once tested this, but putting a single AP on the roof of the building and compare it with on the ground of the building, the former allows better SNI/signal strength compared to the latter as I moved from floor to floor.

5. Point 2 applies to all electrical devices such as your phone, NIC, laptop etc. I have two Apple M CPU series laptops, both show difference in performance even though both support Wi-Fi 6 natively.

Well, looks like my houses' bricks are iron containing then. I have my AP mounted on the ceiling and it's still performing like garbage

wsunardy · Sat Jan 14, 2023 1:49 pm

May I know what need to configure in ax2 to get this speed?
My ax2 only can get 600Mbps max
Use single bridge configuration where your LAN ports and both wireless interfaces are in the same bridge, segregate them with VLAN filtering if you need to. Then enable bridge fastpath/forward and configure the wireless parameters like this:
Code: Select all
/interface wifiwave2
set [ find default-name=wifi2 ] channel.band=2ghz-ax .skip-dfs-channels=all .width=20/40mhz configuration.chains=0,1 .mode=ap .ssid=Dark_Nate .tx-chains=0,1 .tx-power=40 disabled=no \
    l2mtu=2290 mtu=2290 name=2GHz_1 security=Main_VLAN_Security security.authentication-types=wpa2-psk,wpa3-psk .wps=disable
set [ find default-name=wifi1 ] channel.band=5ghz-ax .skip-dfs-channels=all .width=20/40/80mhz configuration.chains=0,1 .mode=ap .ssid=Dark_Nate .tx-chains=0,1 .tx-power=40 \
    disabled=no l2mtu=2290 mtu=2290 name=5GHz_1 security=Main_VLAN_Security security.authentication-types=wpa2-psk,wpa3-psk .wps=disable
I benchmarked using Apple products supporting Wi-Fi 6 natively. All showed around 700Mbps average in the worst case like through three thick walls with a metal container in-between.

Don't forget to configure FastTrack for input/forward/output chain or just use mangle table and do it for prerouting, output, forward etc.

For FQ_Codel do it on interface level:
Code: Select all
/queue type
add kind=fq-codel name=FQ_Codel
/queue interface
set ether1 queue=FQ_Codel
set ether2 queue=FQ_Codel
set ether3 queue=FQ_Codel
set ether4 queue=FQ_Codel
set ether5 queue=FQ_Codel
set "2GHz_1" queue=FQ_Codel
set "5GHz_1" queue=FQ_Codel

Thanks, let me check my device configuration and compare

mrigi · Sun Jan 15, 2023 3:14 pm

The cheapest competitor in the US would be UniF--k AP lite. Which is $100.

BS.
There are zillions competitors. You may not like these because they don't have fancy software options. But hardware-wise there are plenty of AX options below $99.
TP-LInk has at least Archer AX1500, AX23, AX53.
Xiaomi does AX1800, AX3200, Mi AIoT, AX3600, etc.
Asus does RT-AX1800U, RT-AX53U, RT-AX56U, etc.
Huaiwei AX2, AX3, etc
Zyxel EX3301-T0, NBG7510, etc
Tenda TX2 Pro, RX2 Pro, etc
And ton of others.
All these AX WiFi AP's for less that $99. Many of them for twice less.
And some of them can run OpenWrt just fine. For better software.
And there is tp-link omada as a competitor to unifi.

gotsprings · Sun Jan 15, 2023 3:58 pm

For $99, good luck finding a gigabit AP other than hAP ax2.
Actually market is full of other brand AX AP's for half of that.

1/2= $49.50

For your statement to be correct... Thing have to be LESS THAN $49.50.

I have not seen most of the brands you listed on US shelves. Certainly not in Best Buys. Probably not in Microcenter. Etc.

P7p6zDrs2gzSCa · Mon Jan 23, 2023 9:14 pm

@DarkNate, thank you very much. Your insight helped me get up and running. Bravo!

Joe

Use single bridge configuration where your LAN ports and both wireless interfaces are in the same bridge, segregate them with VLAN filtering if you need to. Then enable bridge fastpath/forward and configure the wireless parameters like this:
Code: Select all
/interface wifiwave2
set [ find default-name=wifi2 ] channel.band=2ghz-ax .skip-dfs-channels=all .width=20/40mhz configuration.chains=0,1 .mode=ap .ssid=Dark_Nate .tx-chains=0,1 .tx-power=40 disabled=no \
    l2mtu=2290 mtu=2290 name=2GHz_1 security=Main_VLAN_Security security.authentication-types=wpa2-psk,wpa3-psk .wps=disable
set [ find default-name=wifi1 ] channel.band=5ghz-ax .skip-dfs-channels=all .width=20/40/80mhz configuration.chains=0,1 .mode=ap .ssid=Dark_Nate .tx-chains=0,1 .tx-power=40 \
    disabled=no l2mtu=2290 mtu=2290 name=5GHz_1 security=Main_VLAN_Security security.authentication-types=wpa2-psk,wpa3-psk .wps=disable
I benchmarked using Apple products supporting Wi-Fi 6 natively. All showed around 700Mbps average in the worst case like through three thick walls with a metal container in-between.

Don't forget to configure FastTrack for input/forward/output chain or just use mangle table and do it for prerouting, output, forward etc.

For FQ_Codel do it on interface level:
Code: Select all
/queue type
add kind=fq-codel name=FQ_Codel
/queue interface
set ether1 queue=FQ_Codel
set ether2 queue=FQ_Codel
set ether3 queue=FQ_Codel
set ether4 queue=FQ_Codel
set ether5 queue=FQ_Codel
set "2GHz_1" queue=FQ_Codel
set "5GHz_1" queue=FQ_Codel

aacury · Sat Jan 28, 2023 9:36 pm

Hello,

As already reported in this forum, I'm also having some issues when it comes to download speeds at 5GHz with my device (RBcAPGi-5acD2nD). Strangely, the upload speeds are higher than the download's. In fact, I'm getting about a third of the max download speed I should get, and I've got no idea of what is causing this.

There are no firewall rules, and no substantial interference with other local networks around.

I'm sharing my current wlan2 config. However, if you need further information to help me with this, please let me know.

# jan/27/2023 07:03:30 by RouterOS 6.49.7
# software id = DS4X-PPP2
#
# model = RBcAPGi-5acD2nD

/interface wireless security-profiles
set [ find default=yes ] supplicant-identity=MikroTik
add authentication-types=wpa-psk,wpa2-psk eap-methods="" management-protection=\
    allowed mode=dynamic-keys name=Alexandre supplicant-identity=""
/interface wireless
set [ find default-name=wlan1 ] band=2ghz-b/g/n channel-width=20/40mhz-eC \
    disabled=no frequency=auto mode=ap-bridge security-profile=Alexandre ssid=\
    Alexulia
set [ find default-name=wlan2 ] band=5ghz-onlyac channel-width=20/40/80mhz-XXXX \
    country=brazil disabled=no distance=indoors mode=ap-bridge preamble-mode=\
    short security-profile=Alexandre ssid=Alexulia+ tx-power=25 tx-power-mode=\
    all-rates-fixed wmm-support=enabled
/interface wireless cap
set bridge=bridgeLocal discovery-interfaces=bridgeLocal interfaces=wlan1,wlan2

And the general config:

# jan/27/2023 07:28:52 by RouterOS 6.49.7
# software id = DS4X-PPP2
#
# model = RBcAPGi-5acD2nD
/interface bridge
add admin-mac=2C:C8:1B:C1:07:7C auto-mac=no comment=defconf name=bridgeLocal
/interface wireless security-profiles
set [ find default=yes ] supplicant-identity=MikroTik
add authentication-types=wpa-psk,wpa2-psk eap-methods="" management-protection=\
    allowed mode=dynamic-keys name=Alexandre supplicant-identity="" \
    wpa-pre-shared-key=***** wpa2-pre-shared-key=*****
/interface wireless
set [ find default-name=wlan1 ] band=2ghz-b/g/n channel-width=20/40mhz-eC \
    disabled=no frequency=auto mode=ap-bridge security-profile=Alexandre ssid=\
    Alexulia
set [ find default-name=wlan2 ] band=5ghz-onlyac channel-width=20/40/80mhz-XXXX \
    country=brazil disabled=no distance=indoors mode=ap-bridge preamble-mode=\
    short security-profile=Alexandre ssid=Alexulia+ tx-power=25 tx-power-mode=\
    all-rates-fixed wmm-support=enabled
/interface bridge port
add bridge=bridgeLocal comment=defconf interface=ether1
add bridge=bridgeLocal interface=ether2
add bridge=bridgeLocal interface=wlan1
add bridge=bridgeLocal interface=wlan2
/interface wireless cap
set bridge=bridgeLocal discovery-interfaces=bridgeLocal interfaces=wlan1,wlan2
/ip dhcp-client
add comment=defconf disabled=no interface=bridgeLocal
/system clock
set time-zone-name=America/Sao_Paulo

Any ideas? Thanks in advance.

mkx · Sat Jan 28, 2023 9:56 pm

Is the signal strength, as measured by client, decently good? What is the figure?

aacury · Sat Jan 28, 2023 10:29 pm

It is. See the picture below

Sat Jan 28, 2023 10:32 pm

Move to a less crowded channel.

aacury · Sat Jan 28, 2023 10:58 pm

Already tried it and didn't work, unfortunately.

In fact, I've tried different configurations in the wlan2 interface window, but nothing changed.

craigxau · Sun Jan 29, 2023 3:04 am

I run multiple cAP ACs in a domestic environment with 6 heavy users. My areas is very WiFi dense lots of traffic and noise.
The only way we had good performance and success was coverage.

Lots of Power does not work. Wide channels only make the problem harder to solve.
What works for us.
* multiple WAPS
* managed by CAPSMAN
* Low power
* Max 20 channel width
* Well seperated channels
* Dont ignore DFS ( it makes your wifi dissapear from time to time) Stay out of those bands if you can.

I have close to 200 devices on my network (wireless) plus lots of wired. My kids are harsh critics and they love our WiFi and dont complain.

Less Power more coverage less channel width more channel spacing.

Hope this helps

aacury · Sun Jan 29, 2023 5:05 am

Thanks for your inputs.

However, my environment is much simpler than yours. My connected devices are way fewer (around 5, in total) and all of them are very close to the MT.

I honestly fear that there might be some hardware issue since I can't find a way to solve it configuration wise.

bpwl · Sun Jan 29, 2023 1:13 pm

Move to a less crowded channel.

Yes. But first check what is a busy channel. This is not necessary the channel with many SSID active. A channel with just one other SSID can be very busy. An idle SSID just takes 0.5% of the available air-time in that channel. (@ 6Mbps basic rate) for its beacons.

Extra tools to be used are "Freq Usage", "Snooper". Even "Wireless sniffer" for deep inspections. "Scanner" only shows the AP, not the clients. Even data volume is not the important criteria. It is "air-time". One slow connection can take most of the airtime in a channel.
viewtopic.php?t=185432#p929139

Don't like that "vila3" SSID , it might create disruptive adjacent channel interference (in contrast to co-channel interference/coexistence when channels match.)
(802.11ac is smarter in this than g and n, so it might be better than I expect)

robmaltsystems · Thu Feb 01, 2024 11:39 pm

I'm going to resurrect this thread due to something that's come to light whilst implementing a hAP ac2 at home discussed in detail in another thread. to recap - this particularly congested Wi-Fi channel wise client reported slow internet. We've been all around the houses with this but put it to bed after a nightly reboot of the access points seemed to help.

In the other thread, it's been said that if you specify a single 5GHZ channel that's in the DFS range (which is most of them) and a radar event occurs after initial DFS/CAC scan, then the 5GHz channel is disabled. In this scenario, I kind of expected RouterOS to switch to another non-radar channel after scanning the spectrum. But this belief might be wrong.

It would possibly explain the symptoms were seeing if the 5GHz interface is shutting down - they'll be falling back to the horribly congested 2.4Ghz interface. I've extended the log size to 10,000 entries to see if I can spot radar events.

It would also explain why rebooting the access points every night seemed to help - I've never seen a radar event during the initial CAC after reboot. So in the morning, 5GHz interface will be up. Later in the day, a radar event might knock it out.

Caveat - my hAP ac2 is running ROS v7. The client site is running ROS v6 and I don't know whether the same scenario applies in that version, i.e. if you specify channel = 5500, it's that channel or nothing.