To be clear, this connector is tied to 12V regardless of the power source (DC barrel jack or USB PD).
Oh ok i didn’t look at schematic so didn’t know that was connected too
and the noise-levels are as reported by luci?
Yes. I left only the 5 GHz station up.
Nah, the extra 0.1 uF cap on the 3.3V line didn’t impact the Wi-Fi noise. Didn’t check the ripple with a scope, so it still might be worth it, as there are other devices on the 3.3V line beside the Wi-Fi NIC.
I’m probably going to.buy some caps to reduce ripple
My noise for 5ghz is -84dbm and -83dbm on 6ghz that’s with Taoglas antennas and both SFP slots populated
I may need to move the antenna cables so they point away from the WiFi card
I wonder what would happen if you put 2 47uf caps
I tried 100 uF with roughly the same result.
I tested if the extra load on the 3.3V regulator UP3 (MP8759) is what makes the SFP trick work. The M.2 slot for the SSD is powered by that regulator, too. So, I installed an SSD and created a load by continuously writing to it. That didn’t have any impact on the Wi-Fi noise level, but inserting and removing SFP modules still did. Moreover, until now I was using a 2.5GbE module and a DAC to fill the SFP cages, but then I swapped the DAC for a MikroTik S+RJ10 transceiver, and got a ~2 dBm improvement in noise. Thing is, that particular module is not getting power at all because the presence pin on it isn’t pulled to ground by default, which is required by the BPI-R4. After all, it’s the bulk of the SFP modules that makes a difference, not the electrical load on the 3.3V line.
Maybe there is a chance to add this or even more improved mass in/on the SFP cage or somewhere else on the board? Best way also maybe without spending ~65$ for a module.
Perhaps even a heat-sink on the SFP-cage… I will test, I have just received one.
I’ve bought some 50v1000uF capacitors too see if that makes any difference
I was able to point 3 of my pig tails away from the WiFi card and I went from -82/83dbm to -85/86dbm, Ill probably 3x 20 cm cables so I can route the remaining 3 away from the WiFi card
I tested with heatsink, but it did not provide better noise levels (perhaps because heatsink is still small and low mass and not connected to ground).
I measured with metal-bpi case - adding SFPs did indeed help for 5G and 6G networks. Also to note, I got the noise figures with iw phy-2g survey dump | grep -B 1 noise and looking at them, they vary massively depending on the frequencies, so there are some channels in 5G which report -90 dBm - I think to further test, we should always use the command-line and not luci. Depending on what OpenWrt chooses on automatic channel detection will massively affect the reported noise levels - and after each restart, the device chooses other channels, so there might not be consistency in reporting.
No SFP, no heatsink
- 2412 MHz → -80 dBm
- 5180 MHz → -66 dBm
- 6215 MHz → -76 dBm
With SFPs, no heatsink
- 2412 MHz → -79 dBm
- 5180 MHz → -77 dBm
- 6215 MHz → -86 dBm
With SFPs, with heatsink
- 2412 MHz → -81 dBm
- 5180 MHz → -76 dBm
- 6215 MHz → -88 dBm
Eg. here the noise floor reported by iw phy-2g survey dump | grep -B 1 noise for all bands:
for the 2G band, only one channel is reported, but 5G range from -76 dBm to -89 dBm and for the 6G band from -81 dBm to -91 dBm - not sure what luci reports - I guess it just always reports the first number found (so in this case -81 dBm)
frequency: 2412 MHz [in use]
noise: -81 dBm
frequency: 5180 MHz [in use]
noise: -76 dBm
frequency: 5955 MHz
noise: -88 dBm
frequency: 5975 MHz
noise: -89 dBm
frequency: 5995 MHz
noise: -89 dBm
frequency: 6015 MHz
noise: -90 dBm
frequency: 6035 MHz
noise: -91 dBm
frequency: 6055 MHz
noise: -91 dBm
frequency: 6075 MHz
noise: -90 dBm
frequency: 6095 MHz
noise: -90 dBm
frequency: 6115 MHz
noise: -90 dBm
frequency: 6135 MHz
noise: -88 dBm
frequency: 6155 MHz
noise: -88 dBm
frequency: 6175 MHz
noise: -88 dBm
frequency: 6195 MHz
noise: -88 dBm
frequency: 6215 MHz
noise: -88 dBm
frequency: 6235 MHz
noise: -89 dBm
frequency: 6255 MHz
noise: -89 dBm
frequency: 6275 MHz
noise: -90 dBm
frequency: 6295 MHz [in use]
noise: -81 dBm
frequency: 6315 MHz
noise: -91 dBm
frequency: 6335 MHz
noise: -91 dBm
frequency: 6355 MHz
noise: -91 dBm
frequency: 6375 MHz
noise: -91 dBm
frequency: 6395 MHz
noise: -91 dBm
frequency: 6415 MHz
noise: -90 dBm
I think LuCi must be reporting the channel in use, which indeed has the noise @ -81 dMb in this case. I, personally, never set the channel to auto once I have figured out the one that gives the best SNR.
I made an interesting discovery regarding the wifi noise. It seems that enabling ASPM for the BE14 reduces noise on all bands by ~5 dbm.
Without ASPM:
root@OpenWrt:~# iw dev 2G survey dump | grep -A1 "in use"
frequency: 2412 MHz [in use]
noise: -79 dBm
--
frequency: 5500 MHz [in use]
noise: -76 dBm
Directly after enabling ASPM:
root@OpenWrt:~# iw dev 2G survey dump | grep -A1 "in use"
frequency: 2412 MHz [in use]
noise: -85 dBm
--
frequency: 5500 MHz [in use]
noise: -81 dBm
You can change the ASPM policy and check the status via:
root@OpenWrt:~# echo powersave > /sys/module/pcie_aspm/parameters/policy
root@OpenWrt:~# lspci -vv | grep 'ASPM.*abled;'
LnkCtl: ASPM L0s L1 Enabled; RCB 64 bytes, LnkDisable- CommClk+
lspci: Unable to load libkmod resources: error -2
LnkCtl: ASPM L0s L1 Enabled; RCB 64 bytes, LnkDisable- CommClk+
LnkCtl: ASPM L0s L1 Enabled; RCB 64 bytes, LnkDisable- CommClk+
LnkCtl: ASPM L0s L1 Enabled; RCB 64 bytes, LnkDisable- CommClk+
Of course there’s the question if this is a good idea / the device is stable. So far I had no problem. Together with changing the CPU scaling_governor to powersave the power consumption of the device drops by ~3 W and it is much cooler.
I still need to test if there is a measureable difference in wifi range.
4 Likes
can confirm this powersave mode made a huge improvement, i am now at -88 and -87 from before -81 and -80
Yeah for me 6ghz on channel 65+ achieves -89dbm, even tho no one in my neighborhood uses 6ghz, so it’s all to me lol