So it is a bug in Luci? Wanna try to find where it is located?
Basicly you can modify uboot environment using fw_setenv from openwrt userspace,but if anything goes wrong you are stuck without usb2serial…because of this nearly all guides require the adapter to handle possible errors or at least see whats wrong.
Ok that would explain why I could not really reproduce the reduction of noise level, even when I placed everything on a table and tried to route antenna cables in different directions… for me the noise level stayed approximately the same. I even don‘t saw any change when I „live“ changed positions of the antenna cable.
So I could give it another try!
1 Like
Meanwhile, I’ve tried using mini PCIe extenders to create distance between the BE14 and the main board. Sadly, even with the shortest cables provided by the kit, the device won’t boot at all (despite the BE14 getting power). I don’t have a UART cable to diagnose the failure, but I doubt anything could be done to make the extenders work. Most likely, the tracing tolerances are just too tight.
1 Like
Did you try wrapping extenders in tin foil?)
Those are somewhat shielded by interleaving ground wires, like IDE cables, and extra shielding wouldn’t help with signal loss regardless.
Did you measure voltage?
The power LED on the BE14 came on. I might check the voltage to make sure next time I take my R4 apart.
Nope, it’s still a problem
I’ve checked all the lines on the extender, and every one is connected from one end to the other. The flex cable, though, doesn’t have any shielding contrary to what I originally thought. I really liked the idea of using extenders, but even if it worked with extra shielding around the flex cables, that would only mean the PCIe bus is barely stable at this length, adding yet another unpredictable variable into the mix.
@sinovoip has seen this thread now as my original title was edited by sinovoip, hope we get some news now
1 Like
Honestly, I don’t expect an official solution as the BE14 is merely a stopgap before the BE19 comes out. What I’d like to have for DIY is a DXF of the BE14 board to skip on prototyping RF shield cans or, better yet, a complete can model if Sinovoip made one.
But not everyone can afford to throw another 90$ for another card, we should have some sort official solution
1 Like
Hi, thank you all for the great info. I’m running shielded WiFi 6 radios but have been unhappy with the noise floor. I wish I could re-route my antenna cables away from the PCIe bus like the other poster did. Unfortunately, I found trying to route eight RG178 cables rather challenging and do not want to try moving them again.
My starting noise floor for the 5.8Ghz radio was a disappointing -80dbm on channel 149 (40Mhz width). I cycled through the DFS range and the noise floor did improve some, but I didn’t want to run it in DFS. My other AP, a pc engines apu2 with the same radios was running on a channel 36-40 at -91dbm noise floor. Following a recommendation from another post on this forum, I upgraded the power supply to the MeanWell GST90A19-P1M and my noise floor became -84dbm. I then added some ferrite beads to the Ethernet cable and some additional beads to the new power supply cable and the noise floor now reports as a stable -85dbm.
I have the recommened Taoglas antennas on order and hope to see some improvement with those as well.
My new AP build details:
- BPI-R4 v1.1 board and oem case.
- AsiaRF AW7915-NP1 4x4 radio set to 2.4Ghz.
- 2nd AsiaRF AW7915-NP1 4x4 radio set to 5.8Ghz.
- Gelid 15W Ultimate Thermal Pads 0.5mm for the wifi radios to contact the case.
- 8x TUOLNK RG178 coax cables.
- Antennas taken from my old ASUS RT-AC87U.
- PSU MeanWell GST90A19-P1M 19V. (changed out from the original 12V 3A cheap PSU that is blundled when you just buy the BPI-R4 board and not the wifi 7 card.)
- OpenWrt 24.10.0
1 Like
Even with the Taoglas antennas it’ll suffer from RX noise, it solves TX but not RX
IMO in real life the SNR may not be a perfect factor to indicate connection status, one reason would be many clients (especially apple devices) will automatically reduce wireless power when network activity is low and this would cause the SNR looks very bad (sub 10, I’ve seen 2db on iphone15promax connected to bpi r4 6G band at close distance) but when you do speedtest, it rise up to 25+. It may be a good parameter to look at in RF chambers with analyzers but in reality it makes much less sense. Dont forget there are much more variables when doing real life tests. Maybe there are other reasons.
If, the wireless performance issue is really due to the inner-module interference, the best practice would be , as someone already tried , to shield the RF chips. but this is more than to cover a layer of metal around the BE14 module. it also requires the BE 14 PCB design to contain a full layer of copper as ground, and many vias to make sure good contact between the copper layer and the shield on the top, to form a faraday cage. without modifying PCB layout those efforts would not help so much-even a small leak can destroy these efforts.
some other suggestions, usb3 has strong interference to 5g-6g wireless. if you are using the on-board USB, try without it. i have a cheap asus motherboard and whenever I plug in a device to the USB port close to the WiFi NIC, I see a significant speed drop in a speedtest. before figuring out it was the USB issue I used to think the router was bad.
1 Like
I think sinovoip already put a ground pour on the PCB of the WiFi card cause it doesn’t cost anything and reduces emi, and the fact it has the outline for a shield showed that they were going to add a shield but scrapped it last minute I think
I don’t use USB yet signal noise is quite high, and it seems to affect RX mostly
I see your points, but it is not SNR what we are measuring, it is the absolute noise floor which is to high. In all other commercial routers I have seen, this is usually in the -88 dBm range - here for BPI-R4 with the BE14 Wifi-Card we see -78 dBm.
Of course because of that, the SNR is also lower - because we already give up on 10 dBm, just because of that high noise floor. E.g. I have -65 dBm RSSI in one room, so best SNR I can possibly get is 13 dBm, even if my laptop is sending with full/allowed power. With my old ASUS router and a noise floor of -88 dBm, I have 23 dBm SNR on the same spot here.