Various issues with Hap AC2, mostly wifi performance

RouterOS Wireless Networking
1

I got an Hap AC2 today and I have some trouble with setup…

I am testing from Windows 10 on a MacBook air, which has a wifi AC card.

  • version 6.45.9 (long term)
  • wifi connection status says it’s connected with good speed, e.g 866 Mbps
  • speedtest.net shows only ~16 Mbps download and 200 Mbps upload, should be around 500/500.
  • same laptop card is capable of higher speed, I restarted in MacOS and test was 370 mbps / 440 Mbps
  • router browser interface loads slow, images under login form load slow
  • after entering the user/pass, web page loads forever
  • when I try to login with WinBox (v3.24) using router IP, I click connect and get “Downloading descriptors…” forever, not logging in
  • other devices, e.g an iPhone 11 connected to this same router, speed test shown almost 500Mbps up/down :open_mouth:

What can this indicate ? Is the router OK, should I focus on the laptop and reinstall OS maybe ? :frowning:
Download is 10 times slower than upload, which is strange.

Wifi settings:

  • on wifi2 interface (5Ghz), I set frequency to 5500 Mhz, because letting it to auto took ages to boot up the 5G :confused:
  • Channel width: 20/40/80MHz XXXX
  • Auth Types: WPA2 PSK
  • Unicast Ciphers: aes ccm, tkip
  • Group Chiphers: aes ccm, tkip
2

It is not that clear what you’re testing: the wired or the wireless connection?
And where is the test destination: in WAN, in LAN, or in WLAN?
Same with the source: is it the wired connection or the wireless connection?
Check your routes on your PC. The upload seems to go over the wired connection.

You should test individually first the wired connection with wireless disabled on your PC.
Then test the wireless connection with network cable plugged off, or wired network disabled on PC.

And: try other web browser, for example FireFox or Chrome.

3

Other devices have good wifi, and your registration status is OK. Do you really see “866Mbps/2S/80Mhz/sgi” in registration RXrate and TXrate ? So CCQ is 100% and P-throughput is high?
“auto took ages” … if you leave it on auto then you can end up at whatever frequency. If the selection contains channel 120,124 or128 (5600-5640MHz) then the DFS wait time is 10 minutes.
And that feels like forever. This is a mandatory wait time in the 802.11 standard for DFS channels of the weather radar frequencies. Other channels in the 5500-5700MHz range wait 1 minute.

So the PC is suspected, but remove the tkip from the Ciphers. (tkip will slow down the throughput considerably)

4

I was able to fix this, now I can start experimenting more with the settings, to optimize speed.

Windows 10 automatically updated my MacBook Wifi driver (Broadcom 802.11ac) to a driver dated 2019/10/10, this one caused the problems.
I got the “BootCamp” drivers from Apple, installed this driver, (dated 2015) and speed is much better now, around 350Mbps down / up.
I can also login to WinBox now without getting stuck at “Downloading descriptors…”.


I was testing the wifi AC speed, but doing a speedtest.net test, WAN is an 1000/500 Mbps but wifi AC test only shown 16Mbps, on cable it works 1000/500 Mbps.

I think under “registration” it shown a small Mbps if I remember correctly, but in Windows 10 under connection status it shown around 800 Mbps.
Yes, the PC was the cause :slight_smile:

5

On hap-ac2 (ARM based) I recommend you upgrade to 6.46.6

There have been quit some improvements on ARM devices in the recent SW releases, as well as some Wifi improvements.
(but don’t expect miracles on the wifi side, its more stability etc. than throughput)

As previously said keep the “auto frequency” setting on 5Ghz.
Most commercial AP switch by default to the non-DFS bands, so these frequencies are pretty congested nowadays.

6

I will try the latest one to see how it goes.
On this problematic Mac AirBook, it gets around 200Mbps on auto and around 300Mbps on some manual frequencies.
When I test with the interface, there is just one neighbor on a 5Ghz, rest are clear.
Do frequencies affect the speed (besides being slow down in case of noise) ?

7

Affect on speed depends on the country/region you are in. Setting on “auto” is not a good idea. You don’t know what you get.So unlees you look at the frequency choosen the results of any test will vary.
Don’t be afraid of DFS. If there are radar signals you will move out of the way, mostly only once. This radar signal quite often is a false alarm but DFS cannot be disabled in a certified device. Real radar signals are very rare indoors. Avoid channels 120,124 and 128 as the “listen” time for radar is 10 full minutes.

The affect on speed has to do with the different TXpower that is allowed according to the country/region. US has everything on 30 dBm, however Europe (etsi) has 23 dBm, 20dBm and 27 dBm as limits. Outdoor frequencies can freely be used indoor. Text “indoor” should read as “do not use outdoor”. Be aware that “installation=indoor” in Mikrotik will limit to “indoor” only !
Stay away from your neighbors 5 GHz! Using XXXX does not specify if the extension channels are left of right from the set control channel. Using Ceee, eCee, eeCe or eeeC gives more control.

interface wireless info country-info etsi
ranges: 2402-2482/b,g,gn20,gn40(20dBm)
5170-5250/a,an20,an40,ac20,ac40,ac80,ac160,ac80+80(23dBm)/passive,indoor
5170-5330/a,an20,an40,ac20,ac40,ac80,ac160,ac80+80(20dBm)/dfs,passive,indoor
5250-5330/a,an20,an40,ac20,ac40,ac80,ac160,ac80+80(20dBm)/dfs,passive,indoor
5490-5710/a,an20,an40,ac20,ac40,ac80,ac160,ac80+80(27dBm)/dfs,passive

Having a higher TXpower, will result in a higher RSSI and higher SNR. SNR and signal quality defines the highest usable MCS encoding, what defines the interface rate.

8

I appreciate the detailed info, I will look over it again while I experiment with settings.

I have a different question…
From windows command prompt, I started to ping the Hap Ac2 router with ping 192.168.1.1 -t, normally it should be always 1ms, but when I start a SpeedTest in the browser (or speedtest app), this ping starts to drop, up to 15ms during download and 140ms during upload.
Doesn’t this indicate a problem ? I was expecting it to remain stable.
If someone could also test on their end, it would be great.

9

No, that’s very normal behavior. It’s dynamic, not static.

10

I know it is not static like an IP, and that it is the latency, just that I am surprised it slows down the response that much.

11

But your previous statement was:

normally it should be always 1ms

But I think you could use some QoS (rate-limiting, traffic-prioritizing etc.) to control that behavior.

12

1ms because network is good, I wanted it to remain like that :smiley:
I will look into QoS, I wasn’t sure that is the one that does this /balancing.

13

Check this on QoS in wireless traffic: https://mum.mikrotik.com/presentations/KE18/presentation_5035_1517536100.pdf