CAP ax Gen 6 mediocre performance

RouterOS Wireless Networking
1

Home network with lots of iot and a spread out old home.

I have been using two engenius EAP1300’s, one hAP ax2 and one hAP ac2. These latter are used for special purposes not general wifi. I also had a OpenWRT running on a Linksys 1900, that I needed to replace.

I bought a CAP ax gen6 expecting its performance to be good.

I got it configured and was underwhelmed, signal strength was mediocre and speed was as well, so I did a head to head test, putting it and an engenius in exactly the same place and testing with a laptop and cell pone about 40’ away.

At 5ghz the CAP had better receiption (-74 vs -85dBm) and similar transmit (-75 vs -73dBm at my cell phone), this was despite the CAP running 24mw (defaulting) and the EAP running 20mw. But the speed difference was weird – the CAP was 51 down and 93 up, the EAP was 125 down and 47 up (wired speed is 946/948 so not firewall or ips limited).

The EAP’s are really old and low end. I was expecting significantly better performance from the CAP.

I put the Linksys there and tried again, got -72 / -83, but 244 down and 94 up.

Is there anything I may be simply missing? These really old, mediocre AP’s should not be doing better than the CAP?

The CAP has firmware 7.14.2, and pretty much everything disabled except the AP functions. Below are what I think are the relevant portions of the config. Ethernet 2 is not plugged in yet. The upstream switch is the same one I can test at 1gbs speeds and same port configuration.

Am I missing anything? Does the config look correct?

Linwood

/interface bridge
add admin-mac=xx auto-mac=no comment=defconf name=bridge vlan-filtering=yes
/interface vlan
add interface=bridge name=vlan1 vlan-id=1
/interface wifi security
add authentication-types=wpa2-psk,wpa3-psk connect-priority=0 disabled=no name=GuestPSK wps=disable
add authentication-types=wpa2-psk,wpa3-psk connect-priority=0 disabled=no name=InternalPSK wps=disable
/interface wifi
set [ find default-name=wifi2 ] channel.band=2ghz-ax .skip-dfs-channels=10min-cac .width=20/40mhz configuration.country="United States" .mode=ap .ssid=Reboot2 disabled=no name=Reboot-2 security=InternalPSK security.authentication-types=wpa2-psk,wpa3-psk .connect-priority=0 .ft=yes .ft-over-ds=yes
set [ find default-name=wifi1 ] channel.band=5ghz-ax .skip-dfs-channels=10min-cac .width=20/40/80mhz configuration.country="United States" .mode=ap .ssid=Reboot5 disabled=no name=Reboot-5 security=InternalPSK security.authentication-types=wpa2-psk,wpa3-psk .connect-priority=0 .ft=yes .ft-over-ds=yes
add configuration.mode=ap .ssid=RebootGuest2 disabled=no mac-address=xx master-interface=Reboot-2 name=Guest-2 security=GuestPSK security.connect-priority=0
add configuration.mode=ap .ssid=RebootGuest5 disabled=no mac-address=xx master-interface=Reboot-5 name=Guest-5 security=GuestPSK security.connect-priority=0
/interface bridge port
add bridge=bridge comment=defconf interface=ether2
add bridge=bridge comment=defconf frame-types=admit-only-untagged-and-priority-tagged interface=Reboot-5
add bridge=bridge comment=defconf frame-types=admit-only-untagged-and-priority-tagged interface=Reboot-2
add bridge=bridge comment="hybrid port" interface=ether1
add bridge=bridge frame-types=admit-only-untagged-and-priority-tagged interface=Guest-2 pvid=134
add bridge=bridge frame-types=admit-only-untagged-and-priority-tagged interface=Guest-5 pvid=134
/interface bridge vlan
add bridge=bridge tagged=bridge untagged=Reboot-2,Reboot-5,ether1,ether2 vlan-ids=1
add bridge=bridge tagged=ether1,ether2 untagged=Guest-2,Guest-5 vlan-ids=134
/ip address
add address=192.168.130.214/24 comment=defconf interface=vlan1 network=192.168.130.0
2

Maybe connected, maybe not, but the usual recommendation is to NEVER use vlan 1 because it can create issues in many configurations.

3

I thought VLAN 1 had more to do with security? I’ve got a lot of devices configured with VLAN 1 as the main internal network, I hate to change it.

I also realize that it is more normal in Mikrotik to have trunk ports not include a native PVID (hybrid port), but I’ve got a lot that do that and spend most of my day job around cisco where it is common. I would not think that would matter (would it?).

I ran the speedtest option between the CAP and an hAP AX that is wired as well (though with two swtichs in between). All up/down speeds were at 924-952mbs (on 1gbs switches and wiring), so I do not think the issue is on the lan side, but the wifi side.

4

I did another test this time sitting right beside each AP. The Engenius was 230/314 down/up at 5g. The CAP Ax was 313/486.

So the CAP is faster with a very hot signal.

The other thing I found in experimenting is that for a more distant signal (the original 40’ or so) the orientation does matter for signal, I lose about 6dBm received signal (at the client) if the AP is mounted vertically (wall) rather than horizontally (same height sitting on a stand).

This conflicts a bit with a lot of postings I have read (I think from Mikrotek) that the radiation pattern inside is basically spherical and wall/ceiling do not matter. 6dBm is not much, but it is in the significant category, however my speed tests did not change significantly with orientation so maybe “not significant” is apropos.

Anyway… I may keep this. My goal is to have enough AP’s that everyone gets a decent signal, but I am disappointed and hope someone sees a problem in my configuration.

5

Random thoughts for making tests simpler to interpret:

  • disable/unset WPA3, FT and connect priority
  • setting band to AC, unset width
  • setting the same frequency on the two ap, but switching off one of them when testing the other
  • reporting the results obtained with both devices (smartphone and laptop)

Rationale for unsetting: to allow the device using the defaults (probably more battle tested)
Rationale for disabling WPA3 and setting AC: see if there are incompatibilities with newer standard that you cannot see with the old devices.

6

Notwithstanding what may have been written here or there, it makes no sense that the pattern of a Cap Ax is “spherical”, the signal at the back of the device should be zero or near zero, there are even sort of reflectors behind the antennas. So, the pattern is more “semi-spherical”, but the device is (or should be) intended to cover an as large as possible area from the ceiling of a room, so from a height of some 3 m, if It was semi-spherical, either the radius would be some 3 m (and thus cover a very small area) or much more, let’s say 9 m (and thus cover with wifi one or two storeys below :open_mouth: ).
It would be more logic to design the shape as as a spherical cap:
https://en.wikipedia.org/wiki/Spherical_cap
with h much smaller than a, and this would be coherent with the results of your tests/measurements.

7

I would reset configuration and perform a speedtest with default configuration.

8

Source please. Can’t make accusations of this nature without backing it up with sourcing.

9

They are all over the place in discussions of wall vs ceiling, e.g.

http://forum.mikrotik.com/t/must-cap-xl-ac-be-installed-on-ceiling/156219/2
http://forum.mikrotik.com/t/missing-radiation-pattern-from-specifications/154717/2

And same long before:

http://forum.mikrotik.com/t/antenna-diagram-for-cap-ac-access-point/117089/9

My 6dBm is not inconsistent with the note that said 3dBm.

10

I went ahead and bought a second CAP ax gen6. Not crazy about the distant performance in 2.4ghz, but the setup seems solid. They are in and configured, so I think all good. Thanks for the discussion.

11

Hmmm…

I was running 20mhz channels because it was all IOT (on 2.4ghz) and figured I did not need the bandwidth, and it might cause less interference (to my neighbors or even my own).

But I was having trouble with one device pretty far away from the furthest AP. Out of curiosity I changed to 20/40Mhz. The channel stayed the same (it shows 2412/ax/Ce).

That devices RSSI went from -89 to -74 (occasionally up to -79). That seems like a huge difference.

Why would running a wider channel cause a signal strength increase? Is it some artifact of how it’s measured? Or does the actual power go up that much (it did, and does show 23dBm as the Tx power under status).

12




The only thing that comes to mind is that the application may not be measuring power directly but providing some kind of quality metric from which the RSSI is extrapolated. For example, it could be converting the agreed-upon PHY rate into RSSI.

Typically, using a wider channel with the same power and distance results in higher throughput, albeit with diminishing returns.

For instance, consider the difference in throughput between 20 MHz and 40 MHz with an RSSI of -74. You get about a 55% increase in throughput (going from 29 Mbps to 45 Mbps), even though you are using 100% more bandwidth.

Now, let’s compare 40 MHz to 80 MHz at -71 RSSI: you go from 60 Mbps to 97.5 Mbps, an increase of 62% in throughput but using 100% more bandwidth.

Comparing 80 MHz to 160 MHz at -73 RSSI, you go from 65 Mbps to 65 Mbps, showing no increase despite using an additional 80 MHz. However, if you had an RSSI of -68, you’d go from 130 Mbps to 195 Mbps, a 50% increase in throughput, at the cost of 100% more bandwidth.

Lastly, compare 20 MHz to 160 MHz at -70 RSSI: you go from 43.3 Mbps to 130 Mbps, an increase of 300% but at the cost of an 800% increase in bandwidth.

As you can see, the wider you go, the more throughput you get, but with diminishing returns. Generally, for each doubling in bandwidth, you get a 50% increase in throughput.

13

Not to forget that if transmitters work at EIRP limitation, then each doubling of bandwidth will reduce RSSI by 3dB. So going from 20MHz to 160MHz will end up with reduction of RSSI by 9dB … so one should actually compare those lines from the table … eg. 144.4Mbps (20MHz @RSSI of -64dBm) vs. 180Mbps (40MHz@-67dBm) vs. 195Mbps (80MHz@-71dBm) vs. 130Mbps (160MHz, RSSI of -73dBm). All the figures are for 2x2MIMO and SGI.

Only when RSSI is high (even excessively for narrow channels - e.g. 20MHz@-40dBm) one gets somehow linear speedups with using wider channels.

14

I thought that RSSI was the total power received, not the power per unit frequency. I searched the internet and couldn’t find any information that it’s tied to the bandwidth.

15

I get most of that, it was the RSSI going up with change of frequency width that got me, but your speculation that it is a derived number from bandwidth is a possibility. This is an IOT device where I have no real visibility into what it does internally.

The issue is as you get into the 80 and above range, they tend to disconnect periodically to try to find a better signal even if what they have is actually working. So changing to 40mhz got me a good result (stopped disconnects) even if it did not yield an actual, real better signal.

Or something inside the CAP is weird and it really did output more signal. Any possibility they are pumping out more energy to hold (or try to) signal spread over wider range? That seems wrong, so I doubt it.

But anyway… weird result. Possible explanation.

16

From the wikipedia article detailing RCPI measurements (different acronym than RSSI but according to description this is what’s actually being measured these days):

For the most part, 802.11 RSSI has been replaced with received channel power indicator (RCPI). RCPI is an 802.11[5] measure of the received radio frequency power in a selected channel over the preamble and the entire received frame …

Now: the way wide channels are done (C and a few e) and due to backward compatibility (to allow g devices or a devices), those preambles are transmitted principally on C channel. And since total Tx power has to be spread over the whole channel bandwidth (between 20MHz and 160MHz or whatever), it’s clear that preambles are transmitted at lower power when total bandwidth is higher.

17

If true, it would be an “isotropic radiator”
https://en.wikipedia.org/wiki/Isotropic_radiator
So I think we can conclude that it isn’t.

18

Would be nice if Mikrotik then published correct diagrams, and did not give incorrect answers. If your conclusion is right of course, that it isn’t basically spherical.

I also recently had another interesting result - I moved it down off the ceiling to about 8" below, and got a lot better power distribution. I’m thinking this may be more about the ceiling materials and especially the bad just below the ceilings than the radiation pattern, but am now a bit torn about mounting location (this one in particular I’m trying to get a lot of 2.4ghz reach for IOT stuff in the garage and outside).

19

Unsure how big your home is. I have 1 cAP ax set up in a 2 story 2500 sq ft home, not including the basement, and it serves me outstanding. AP is placed on the top of my wall mounted rack in the basement and I get coverage throughout the house and garage where my wifi sprinkler controller is. From my garage, I just did a speed test and got almost 250 mbps up/down on my cell phone on a 500/500 connection. That’s on the 5 GHz radio with a 80 MHz wide channel. Really cannot ask for more. Coverage, from what I can tell, is equal to the coverage I had with a Ubiquiti AC PRO

20

Indoors (bar the case of a single huge room that is “easy”) there can be a lot of differences due to reflections, obstacles and what not, it is very hard to predict how this or that device will perform or behave or find the better spot to place it, with devices with external antennas you can play a bit with orientation and/or move the antenna with an extension cable (if WAF is not involved) but with a Cap Ax or any “internal antenna only” devices such as the Ax2 your options are very limited.