cAP AX 2Ghz problem

RouterOS Wireless Networking
1

Good day!

I have next config:
cAP AX
RoS 7.20.4 (tested on several versions from 7.19.6 to 7.21.beta8)
cAP AX with routeros and wifi-qcom packages
cAP AX was reset, get IP with DHCP

Trying to make 2.4GHz wifi (locally on cAP) with next (now simplest) config:

/interface bridge
add comment=defconf name=bridgeLocal
/interface wifi
set \[ find default-name=wifi2 \] configuration.country=Superchannel .hide-ssid=
no .hw-protection-mode=rts-cts .installation=indoor .manager=local .mode=ap
.multicast-enhance=disabled .qos-classifier=priority .ssid=HOME
.station-roaming=no .tx-power=14 datapath.bridge=bridgeLocal disabled=no
/interface wifi datapath
add bridge=bridgeLocal comment=defconf disabled=no name=capdp
/interface wifi
set \[ find default-name=wifi1 \] configuration.manager=local datapath=capdp
disabled=yes
/interface bridge port
add bridge=bridgeLocal comment=defconf interface=ether1
add bridge=bridgeLocal comment=defconf interface=ether2
/ip dhcp-client
add comment=defconf interface=bridgeLocal
/system identity
set name=CAP2
/system routerboard settings
set auto-upgrade=yes

Problem:
Clients are connecting to WiFi (2GHz G, 2GHz N, several with 2GHz AX) but actual speed (in registration table i can see 90/90, 142/142 Mbs ) is really LOW.
I.e i've checked with local speedtest portal (GitHub - librespeed/speedtest: Self-hosted Speed Test for HTML5 and more. Easy setup, examples, configurable, mobile friendly. Supports PHP, Node, Multiple servers, and more) on NAS connected to cAP AX directly (eth2): 0.5 - 4 Mbs (ping more then 100ms!!!)
When connecting to the same network with cable (100Mb) - about 80 - 90 Mbs (ping 4-10 ms)
When create 5GHz AX wifi - 400 - 500 Mbs (ping 4 - 10 ms)

Tried to set channel width, set specific channel (1,6,11) with no luck.

Same config* on RB951/RB962 (*cause there are no wifi-qcom drivers for that devices) - i've got normal speed (60 - 80 Mbs, 4-10 ms ping)

I can see several other WiFi AP's on air (10-12), but on different channels with low power (-60 .. -85 dBmn).

So it looks like problem exactly on cAP AX.
Are anyone faced with such problem too?
What I'm doing wrong, what i can do to check more?

Thanks in advance.

2

Spoiler: it's not cAP AX.

Obviously it's not the same config as RB951/RB962.
And despite your comment, this is not the simplest config either.
You are setting quite a bit of things which I have rarely seen being used.

Sometimes it can help to use smaller bandwidth ... forget superchannel and especially in an environment with 10-12 other APs transmitting. You are actually increasing chances of interference when using superchannel, especially on 2.4GHz range.

What if you use default config and ONLY set channel to a fixed frequency selecting the least occupied channel (scan your environment) ? You can also change SSID and security but don't touch anything else for now.

3

Actually You are right - i've posted not simplest config.

This is a consequence of trying to find best setting :slight_smile:
According to Superchannel - its in country choice (not channel actually ), to avoid any law limitations.

I've also tried simplest config with static channel (the least loaded):

set [ find default-name=wifi2 ] channel.frequency=2412 configuration.manager=local .mode=ap .ssid=HOME datapath.bridge=bridgeLocal disabled=no

And also tried with/without

channel.width=20mhz

Result is the same.

In fact, my config was much more complicated:
There E50UG as CAPsMAN and RB951 and RB962 as CAPs (since RoS6)
And now i bought 2 x cAP AX (to exchange RB951/962) to get heterogen WiFi6 network

And after faced problem with AX 2.4GHz - start from scratch for AX.

According RB951/962 - yes its totally different device (with different CPU arch and WiFi drivers).
But as i wrote above - with simplest config (just setting SSID and Security, no more else) in the same environment - they are work better.

I plan to test cAP AX on the clear place. But before it - asking if anyone faced similar problem.

4

Yesterday i've take AP to clear place (where there are no any WiFf) and did one more test:

  1. Reset AP (no default configuration)
  2. Configured IP 192.168.99.1 (and create dhcp) on wifi2 interface (2GHz)
  3. Enabled only wifi 2 interface (wifi1 - 5GHz disabled), and configured minimal parameters (SSID, passphrase)
  4. Connect two clients - Laptop Lenovo T410 and mobile phone Samsung A56

STAs was located less then 2 meters from AP and connected to AP with 2GHz N (laptop) and 2GHz AX (phone) - with Signal about -35 dBm and Tx/Rx rate more then 100Mb.

Start iperf server on laptop and try iperf client from phone.
Best speed - about 30Mbs, average 16 Mbs.

But more interesting, that ping between these to devices is too big - minimal 5 ms, average 30 ms, max - 3850 ms.
Same situation i've observed in real situation (when checked with librespeed).

With the similar configuration based on RB951G - 5 ms stable, and iperf about 80Mb.
May be this test is not relevant. What am i doing wrong?

5

Since both devices were using same radio interface, each IP packet traveled the air twice (once from sender to AP and then from AP to receiver). Which IMO means that actual AP throughput is double of what you've seen. Then if you were running bi-directional tests (both directions simultaneously), you have to sum both directions to assess AP throughput. And, as both stations tried to transmit to AP at the same time at least some portion of time, there were collisions (and consequently backoffs and retransmissions) making use of airtime inefficient.
You might want to repeat the test, but this time connect laptop to AP's ethernet port. And change config on AP (create bridge, add ether port and wif interface, set-up DHCP serve on bridge, etc.). See if the result is better. It should be pretty much better than what you saw: there won't be any congestion/collisions on wifi.
Additionally: laptop's N connection is a narrow bottleneck. N theoretically offers around 144Mbps of raw speed (2x2 MIMO already, not all N devices supported MIMO) and practical speed mostly reaches up to some 3/4 of theoretical (around 100Mbps). If you connect laptop to ether port, it will both rise the throughput for laptop to true 1Gbps (if laptop supports that) and will free up airtime for phone (which does use it much more efficient due to being AX - around 280Mbps when using 20MHz channel and 2x2 MIMO).

Devices running legacy wireless driver might have fared a bit better due to support of RTS/CTS (which takes care of collisions with the cost of a slight overhead ... still way better than dealing with actual collisions) ... and I'm not sure that wifi-qcom does support that (AFAIK it's not configurable).