Luci is not correct, known so far. But here SSh session is not correct either.
sta0: TX power 0dBm
sta1: TX power 127dBm
wlan0: TX power 20dBm
wlan1: TX power 23dBm
Measured signal strength values have no relation with what is shown in SSH.
... and fluctuates ...
Setting in GUI is reflected in SSH dBm, but is not related to the measured signal.
Was not OK in 4.3.21, but strong enough to be used. Not OK in 4.7.2beta and extremely low.
Not saying it is only visual. It is real and frustrating.
Just saying that numbers in LUCI and SSH session show no problem!
But the problem is there for sure, and measurement clearly shows this.
(This is measured by using inSIDDer on windows, same results with Wifi Analyzer on android, or data in Mikrotik registration table. )
Graphics uploaded to show the fluctuations in power.
Problem is very real, it is there in 4.3.21 as well, but there it has no dramatic consequences because the signal level overall is much higher. People using ethernet uplink (or any other non-repeater uplink,) will not be disturbed as much, because the client devices are mostly within a meter distance from the GL.inet router). The uplink AP might be further away, getting -82dBm and lower (-90dBm) and will disconnect frequently.
Wifi adapts to low signal condition, by reducing MCS encoding number (what lowers the interface rate,) and introduces more redundancy bits in the encoding, to make it work, but much slower.
With inSIDDer use the Export Summary, analyse the output (collected beacons from all SSID instances) with Wireshark ... a ton of usefull wifi information !
That mt798111 or mt798112 as seen in 4.7.2beta file is that something for the MT3000 and also for the SFT1200? I expected other hardware in SFT1200, something like SF19A28 ????
Is this Firmware Image for the Beryl MT3000 and not for the SFT1200 ????
So far all digital settings seem OK, but we have no view on the radio (RF,analoge) settings by the driver. Driver is probably complex, ... doing calibration, heat control, etc.
Just looking for possible clue ... was the txpower_lvl default value changed with 4.7.2beta? It is not set in the etc/config/wireless file AFAIK, so it takes the default then?
Or is this manual not the proper place to look ?
> config wifi-device 'radio0'
> option type 'mac80211'
> option country 'CN'
> option txpower_lvl '2'
> option channel '1'
> option band '2.4G'
> option hwmode '11g'
> ```
https://siflower.github.io/2020/08/12/wifi_architecture_and_configuration_manual/#22-wifi%E4%B8%8A%E5%B1%82%E9%85%8D%E7%BD%AE%E8%AF%B4%E6%98%8E
Something must be wrong in the driver. Setting the TXpower in the GUI to LOW gives the highest emitted power signal in 2GHz band. (parameter with reversed bits ?)
The other settings give strongly fluctuating signal (overheating, power protection, radiation limiting, ... ?)
Only Siflower (and GL.inet) have enough information.
OK. I'll wait for the newer version.
Going back to the stable version 4.3.24 now, until this beta becomes usable.
Learned a lot on GL.inet software and on OpenWRT with all this searching and testing.
Unfortunately this 4.3.24 the current latest stable is still something with awfull performance.
Tested very long to get stable connection, (with BSSIDlock, only one known network, no Multiwan checks, had it stable, but now looking at the speed, it's not good at all.
Just using the built in tool "LUCI-Status-Realtime Graphs" as double check, with setup PC at 3 meters from the SFT (preventing any receiver saturation) , stable signal, this is just bad performance. (Green graph- interface rate)
Too bad. SFT with ALL rev 4 firmwares I have tested and tweaked (2 months, 3h per day) make the SFT1200 no better than a "paper weight"/"presse papier".
Further regression done to firmware 3.216 and this actually works. Missing the new features, but at least this works so far. Heh ... sort off working.
If this also starts to show major defects, the SFT1200 will go back in the box, as my Mikrotik mAP Lite and Cudy TR1200 serve me personally well enough.
Hope to be able to recommend the SFT1200 or other GL.inet to our tenants (21 families), but for now that would be a wrong recommendation.