I connect from HAP AC to the AX3 over 5ghz:
/interface wireless
set [ find default-name=wlan1 ] band=2ghz-g/n channel-width=20/40mhz-XX country=slovenia disabled=no distance=indoors frequency=auto installation=indoor mode=ap-bridge ssid=SSID wireless-protocol=
802.11 wps-mode=disabled
/interface wireless security-profiles
set [ find default=yes ] supplicant-identity=MikroTik
add authentication-types=wpa2-psk mode=dynamic-keys name=“station bridge” supplicant-identity=“”
add authentication-types=wpa2-psk mode=dynamic-keys name=wlan2-SSID-repeater supplicant-identity=MikroTik
/interface wireless
set [ find default-name=wlan2 ] band=5ghz-onlyac channel-width=20/40/80mhz-Ceee disabled=no frequency=5500 installation=indoor mode=station-pseudobridge security-profile=wlan2-SSID-repeater ssid=
SSID wireless-protocol=802.11
add disabled=no mac-address=C6:AD:34:32:D2:27 master-interface=wlan2 name=wlan3 security-profile=wlan2-SSID-repeater ssid=SSID
The problem is, I only get cca 100mbs of speed between routers and the distance between them is only a couple of metres. My notebook gets wifi speeds of more than 500mbs, connected to AX3, while the HAP AC only gets 100mbs.
You can already start making sure to use a frequency on AX3 which is clear of interference by other devices.
Auto means you don’t know what it will take and sometimes it is not the least occupied channel.
But how did you assess that speed when hap AC is connected ?
Computer to hap AC ? Are you sure it has a gigabit connection (allthough if it were fast ethernet, I would expect speeds in the 90-ish area). Doesn’t hurt to check.
And what’s the config like on that device ?
You also didn’t mention ROS version on your two devices. wifiwave2 received some love from MT devs lately and you probably should run at least 7.11.2 (if not 7.12 beta). On both devices.
Repeater mode on the same wireless radio (wlan2) , will at least reduce the throughput to 1/2. (receive, send) for every packet.
WLAN interfaces are known to be inefficient at high speeds. (Their AMPDU aggregation is limited to 3895 bytes per MPDU, and A-MPDU to 262000 bytes) This means many wifi access attempts (contention delays) to transmit a large file http://forum.mikrotik.com/t/wifiwave2-expected-performance/160075/3
So having 1/4 of the speed is no surprise. “wifiwave2” has big buffer sizes, so much less frequent overhead before transmitting. And the repeater function at least halves the “wlan” performance.
FYI only.
Forgot to just give some practical steps. (Explaining the reason for the lower speed is one thing, avoiding the slower rate is another)
The way to improve the rate after the 2nd MT (repeater) is avoiding the “repeater” send action on the same radio as the receiving radio.
OK 2.4GHz (40 MHz wide) is slower than 5 GHz (80 MHz wide), so better speed is not guaranteed when using 2.4GHz, unless there are more chains. (One per antenna)
Clients are often limited to 2S (streams), quite some MT AP as well.
Either use a MT with 4S to get higher # of interface chains (e.g. Audience) in https://mikrotik.com/products/matrix .
(Audience has even 2 separate 5GHz radio’s). To have this advantage with smaller MT AP’s, you need 2 AP’s, connected back-to-back with ethernet.
The whole point is to have receive and send on different (non overlapping) channels on separate radio’s.
There is no problem that the same SSID is on different channels on the original and on the repeater, the client device will adapt, when roaming.
Just practical, my installation … 5GHz ac 40MHz wide on 2S gives 400Mbps interface rate, and 260Mbps data rate with the WLAN driver between MT with NV2 protocol
5GHz ac 40MHz wide with 2S further down to a client device on different 5GHz 40MHz wide channel gives 400Mbps interface rate and still 180Mbps data rate.
There are 20 PtP links and 30 further AP’s in that area, so there is no room for wider channels. But with the WLAN driver higher interface rates wouldn’t help much.
(Test was 800Mbps interface rate, 320Mbps data rate). Wifiwave2 would do better,(like 500-600-Mbps data rate) but has no 4-address mode for PtP yet. (Tunnel needed)
@holvoetn, I used a gigabit adapter, connected to local speed test site (akostest.net). Might try with another adapter to see if there are any results. I will also scan the frequencies and set them manually. @mkx, both devices run 7.11.2, will try with beta, at least on HAP AC. @bpwl, thank you for a clear and thorough explanation! I’m stuck with HAP AC for the moment, at least for Ethernet connection. PC connected to AX3 works at higher speeds (500ish). Will also look into the articles you gave.
I don’t think akostest.net is the best testing tool. I just tried with my venerable andoid tablet: speedtest run against server hosted by my ISP yielded 175/100Mbps (down/up) while akostest.net yielded 62/83. Already latency was considerably worse when testing via akostest.net (6ms vs. 4ms).
A mid-class phone (Ssmsung A52 5G) was better: speedtest 307/106Mbps vs. akostest.net (283/104Mbps). BTW, speedtest app performance on this phone should not be a problem, I’ve measured 980Mbps DL over 5G network a few times.
Both tests were done using Audience (ac wifi running wifiwave2 driver) over the 4-chain radio.
Fun fact: when I connected the phone to an USB-C docking station, which has a Realtek gigabit ethernet card incorporated (and phone supports it), speedtest peaked at 150Mbps DL (give or take) while akostest.net went up to around 290Mbps. Go figure …
When doing same pair of tests from a wired (elderly) laptop running Win10 and Firefox, the results were similar for both test varieties (around 250Mbps downlink and 100Mbps uplink; my internet line is 1000/100), latency measured was even lower on akostest.net than speedtest. I should add that when running tests on laptop, it’s firefox performance that limits the result. Running same tests on chrome (same laptop) gives better results for speedtest (330/100) but not for akostest.net.
When running ookla speedtest app for windows, I get 695/95Mbps.
If I take away the GUI and run CLI speedtest on linux against same (ISP hosted) server, I easily get 830/102Mbps (and it’s a humble 2-core Intel Atom C3338 … laptop runs 2-core+HT i7-620M and according to some benchmarks it should be some 66% faster per core). And that’s when testing during peak hours (it’s 8pm here), when I ran CLI test during off-peak hours, I routinely got 1Gbps in DL (that’s wirespeed, my main router is hAP ac2).
BTW, just tried the CLI speedtest against IPv6 test server and I’m getting around 230Mbps with one CPU core of router fully loaded. That’s the difference between having fasttrack and not having it.