I got a new WiFi USB adapter based on Mediatek 7921AU. Not only does 'survey dump' not work, but the card also hangs super easily.
I was running U22.04 with HWE kernel 6.8.
Gemini seems to imply the driver is more stable on U24.04, so I upgraded
to it. The upgrade hung midway — it seemed to be waiting for user
input — but luckily it could continue by running dpkg
--configure -a.
(I thought of upgrading to U26.04, since it is already out, but it uses Wayland, I don't know if I can VNC into the PC. I'm running it headless, so it is important.)
It didn't work. It is the same 6.8 kernel, just a few revisions newer. Gemini then again imply it'll work on HWE kernel 7.0, so I upgraded to it. It still didn't work.
Finally, Gemini said 'survey dump' does not work on mt7921. Technically, Gemini did not lie. Upon further questioning, Gemini said mt7921 supports 'survey dump', but the values are zero due to missing hardware support. :duh:
When I first asked which chipset supports 'survey dump', Gemini recommended Mediatek 7612. I got a USB adapter based on it, but the seller sent me a card based on rtl88x2 — same as my existing card. I returned it.
I was going to get another Mediatek card, but I thought why not look for an updated chip with AX support? Gemini said mt7921 supports 'survey dump'. I also thought mt7921 was a superset of mt7612.
The hang issue should be due to my motherboard. It cannot supply enough power — even for USB 3.0 ports. I tried the card on another computer and it seems fine.
Third time's the charm?
Part 2 is out and I bought it. But that's it, I don't think I'll buy more.
The story, when you read it, is pretty good. But if you give it some thought after that, it has too many coincidences and the relationship web is small and too connected.
It's still a good story. I wish the setting and the characters' thoughts are more fleshed out.
The 3 girls also spent very little time together. This is their individual stories before they met and how they met. I'll love to read more of their adventures where Youko is travelling incognito. Perhaps in a future book, but there's no such book.
|Std|Band|Max data rate (per stream)|Modulation|Innovation| |---|+--|+--|+--|+--| |G|2.4|54 Mbps @ 20 MHz|OFDM, 64-QAM, 3/4|OFDM at 2.4 GHz| | | |N|2.4, 5|72.2 Mbps @ 20 MHz
150 Mbps @ 40 MHz|OFDM, 64-QAM, 5/6|40 MHz channel
5 GHz band
MIMO| | | |AC|5|86.7 Mbps @ 20 MHz
200 Mbps @ 40 MHz
433 Mbps @ 80 MHz|OFDM, 256-QAM, 5/6|160 MHz channel (rare)
MU-MIMO
beamforming| | | |AX|2.4, 5|143.4 Mbps @ 20 MHz
286.8 Mbps @ 40 MHz
600.5 Mbps @ 80 MHz|OFDMA, 1024-QAM, 5/6|RU (Resource Unit)
UL/DL MU-MIMO
BSS Coloring
TWT (Target Wake Time)
Preamble puncturing (optional)| | | |AX-E|6|1.2 Gbps @ 160 MHz|–|6 GHz band| | | |BE|2.4, 5, 6|172.1 Mbps @ 20 MHz
344.1 Mbps @ 40 MHz
720.6 Mbps @ 80 MHz
1.44 Gbps @ 160 MHz
2.88 Gbps @ 320 MHz|OFDMA, 4096-QAM, 5/6|320 MHz channel
MLO (Multi-Link Op)
MRU (Multiple Resource Unit)
Preamble puncturing
AP coordination|
802.11ax is about overall efficiency. It can serve 9 clients (via 9 RUs) in a 20 MHz channel in one time unit vs 1 in earlier standards. It has lower latency and higher efficiency for small packets.
802.11ax, for the first time, acknowledges that there can be multiple uncooperative APs. BSS coloring is a good idea, but it doesn't work as well in the real world for two reasons:
Preamble puncturing is rarely implemented in consumer routers. However, it is still very limited even in 802.11be. It allows one 20 MHz hole for 80 MHz channel, one 20 MHz or 40 MHz hole for 160 MHz channel, one 40 MHz and one 80 MHz hole for 320 MHz channel.
Most of these combinations are useless in home environments because people normally use 80 or 160 MHz at 5 GHz.
802.11be adds AP coordination. Now unrelated APs can cooperate when using the same channel.
4096-QAM is achievable under ideal, short-range (1.5m) and line-of-sight. It needs 40 dB SNR. If floor noise is -90 dBm, then the min signal strength needed is -50 dBm (for 20 MHz channel). Normally, 256-QAM (30 dB) and 64-QAM (24 dB) are more realistic.
2.4 GHz is crowded. Even non-DFS 5 GHz is pretty crowded. The question is, how much?
The RT-AX82U has two metrics: noise and channel utilization. Right now, it is showing -89 dBm for noise and 18% utilization for 2.4 GHz channel 9. Both are very good!
Unfortunately, it is only for the current channel.
As far as I can find, the lowest-end router/AP that has dedicated spectral analyzer is the U7 Pro XGS (replaces the Pro Max model). It can map all bands and interference from all sources — including non-WiFi ones, this is crucial — "simultaneously". It uses only one radio to do so, I presume it takes some time to sweep through all the channels.
I'm surprised this is not more common — I thought nerds would be interested in such info. Maybe they don't live in such high-density housing as Singapore. We can expect 10 neighbours in HDB/condo housing.
The U7 Pro XGS plus a PoE++ injector is pretty costly, though.
After many days of googling for alternatives, I came across this command:
sudo iw dev wlan0 survey dump
It shows noise level and busy time. Busy time includes non-WiFi interference — this is key. You can calculate channel utilization from it (busy time / active time). There is also 'receive time' to tell how much of 'busy time' is WiFi traffic. Neat!
I have a spare WiFi USB adapter — I used it to connect my notebook to my home network when its network card is flaky, but I notice it is dropping connection much less with UDR 7.
I should be able to collect some stats from my Linux box! :lol:
First problem. The 5.x kernel was too old. It did not have the driver out of box. Luckily, the kernel could be upgraded to 6.x by itself.
With anticipation, I ran:
sudo iw dev wlan0 survey dump
Nothing appeared. It turned out the WiFi card was based on Realtek 8822BU. Realtek chipset does not support wireless stats. What a letdown.
This works:
sudo iw dev wlan0 scan
It gives standard info from beacons that APs are broadcasting. I'm able to tease a bit more info compared to SSID scanner apps — probably because they want it simple for layman to use.
Some APs indicate channel utilization (from their PoV) and no. of clients.
It is a little more info, but it is not very useful. Will need a WiFi adapter that supports 'survey dump'.
Is it always right to choose channel 1, 6 or 11?
|Chn|Utilization|Interference (dBm)| |---|---|---| |1|41%|-66| |2|33%|-54| |3|37%|-78| |4|18%|-86| |5|19%|-83| |6|45%|-75| |7|23%|-54| |8|23%|-78| |9|9%|-96| |10|27%|-74| |11|46%|-22| |12|15%|-86| |13|11%|-88|
I have another router at channel 11, that's why interference is so high.
I'm beginning to wonder if it is better to use channels 3 - 4 and 9 - 10, especially if channels 1, 6, 11 are relatively weak, e.g. -75 dBm and below. Use neighbouring channels to overpower the weak signals instead of being forced to cooperate with them.
Notice that channels 4 and 9 are exceptionally clean.
By default, routers send beacon frames at 1 Mbps for 802.11b compatibility — this has not been needed for years! This uses ~2.34% airtime per SSID (assuming default 10 Hz broadcast rate). If changed to 6 Mbps (802.11g), it becomes 0.39%.
You can expect around 10 neighbours in HDB housing, with 1 to 3 SSIDs each.
