Interesting results, although I’m wondering if copper is better or aluminium as a RF shield
copper should be very good: RF Shielding Material Guide - it is just more expensive than others, so it is probably only used in higher level gear.
2 Likes
Ok, I also tried shielding the backside and made sure everything is grounded (I placed the aluminium foil next to the screws so when I screwed-in the wifi-board the whole foil got grounded. I measured 0 Ω ground-to-foil).
But at the end I have not seen any improvement, I’m still at approx -84 dBm. But as I just taped down the foil, I still have some air-gap (or tape gap 
) - you seemed to punch holes for the antenna cables?
Perhaps -84 dBm is the max we could achieve with this board? - or I’m just lucky to get -84 dBm also with no shielding at all.
But I will nevertheless try a more proper copper shielding - even if it does not help, it might not do more harm.
1 Like
I just fed the pigtails between the top and bottom sheets and compressed the foil around them. What I’d try next is make the sheets larger so that they extend beyond the board area and can be sandwiched together around it. Theoretically, It could be that you’ve hit the limit of what ghetto shielding can do to improve noise if the environment isn’t particularly noisy to begin with (which it seems to be) and now only a properly fitted ‘real’ RF shield will make a difference.
1 Like
At first, I brushed off the idea of bad pigtails. The same 1.13 mm coax cable is used everywhere. Still, it might be worth it to remove the pigtails temporarily (both with and without shielding). Although, I can’t remember if OpenWrt displays any radio stats with no clients connected…
1 Like
Mine does for the router side not the client side
Just so everyone knows it’s the Taoglas antennas and the RG178 cables, also my ISP router (which is emitting a WiFi signal) is sitting next to my router and 1 TV and 1 monitor and a raspberry pi so yeah I got some serious interference
1 Like
I’m using the same antenna’s + pigtails, range is decent but noise is pretty bad in my case. I think we really need a proper rf-shielding solution from @sinovoip
2 Likes
Yup agreed, one that we shouldn’t pay for due to the fault being on BPI end
1 Like
I did some further testing, now with better control of confounding variables. At baseline the router was out of the housing, no RF shielding, pigtails grounded on the SMA end. Noise level at -80 dB.
Front side RF shielding on the Wi-Fi board made no difference on its own. Back side shielding, however, dropped the noise level to -85 dB. Notably, it didn’t matter whether the foil was grounded or not. Finally, I left a smaller piece of insulated 0.2 mm copper sheet between the Wi-Fi module and the main board and buttoned it up in the housing. Now the noise is staying at -80 to -81 dB, which is the best I ever got from a fully assembled BPI-R4. Although, still far from perfect.
BTW, the type of pigtails (1.13 mm or RG178) didn’t impact noise, at least on the 5 GHz band.
I’m not an electronics engineer, so I might be talking out of my a** now. But to me it looks like the main source of noise is internal to the router since the worst case scenario I found was the stock metal housing. If no extra shielding is added, it could be reflecting the interfering signal right onto the Wi-Fi board. Taking the thing out is already an improvement. Apparently, the BPI-R4 is too small for its own good with that stacked PCB layout.
As for the proper RF shielding, there’s the problem of clearance between the Wi-Fi module and the main board. A solid shield can won’t fit there at all because of the height of SMD components on both sides of the gap. Small cutouts will be necessary to accommodate those. I’ll make a model/drawing of what it should look like.
SinoVoip should at the very least start putting RF shielding on both sides of the BE14 in future batches. Better yet, also make a DIY shielding kit available for a couple of dollars. I doubt they’ll do either, though. I only hope the upcoming BE19 doesn’t end up like this.
1 Like
I guess the bpi-r4 is somehow special, no other bpi product has this layered setup, where the wifi module is on a separate board, so this might be their first such product.
As mentioned, not a single bpi board has any shielding, while you still see left-over shielding traces (would-be shields ) on those boards - so I guess they got away with no shielding so far, and on bpi-r4 it seems to fail.
I have not received a response from OpenWrt One users, but I did a search for “noise” on the forum and found some people posting wifi-stats (with noise level) - and those looked not bad (and again: no shielding on those devices either!):
- BPI-R3-Mini: Noise: -91 dBm
- BPI-R64: Noise: -92 dBm
- BPI-R3: Noise: -99 dBm to -96 dBm (not 100% sure, if they talk about client side noise)
Only a sample size of 2… but nevertheless, they seem to work pretty good with no shielding.
1 Like
I have 2xBPI-R3 and a BPI-R4. All in default, unmodded cases. Noise levels below, from
iw <if> survey dump
R3:
f: 5500 MHz, noise: -92 dBm
f: 5180 MHz, noise: -92 dBm
f: 2412 MHz, noise: -80 dBm
f: 2412 MHz, noise: -90 dBm
R4:
f: 2447 MHz, noise: -72 dBm
f: 5180 MHz, noise: -76 dBm
f: 5955 MHz, noise: -82 dBm
My R4 is not seeing much use because of the unreliable WiFi. This might explain why. Hope it helps.
2 Likes
Here’s a SketchUp model of shield cans with mounting holes for Harwin S0961-46R clips. The front side can is 2.1 mm tall, which allows for 0.1 mm inside clearance between the metal and the SMD chokes. The back side can is 1.5 mm, this is the absolute minimum for correct mounting and clearance. Insulating film on both sides is a must for safety. Not a fan of cutouts to account for groups of components on the main board, but I’d not risk making them any smaller.
The wireless ICs on the front side are 0.9 and 1.1 mm tall. Meaning a gap of 0.9 and 0.7 mm for thermal pads. 0.6 mm left between the can and the stock metal housing.
The best way to manufacture those (short of crowdfunding for factory costs) would be laser-cutting sheet metal, then pressing/bending it into shape with a CNC-machined die. That’s what I’m going to do in the next couple of months after some prototyping to verify dimensions. Scissors, pliers and other home tools might work to an extent, but I tried them and it was miserable.
shield_can.skp (350.0 KB)
1 Like
I just checked by R4 and only have devices connected to 2.4GHz currently.
I have a noise of -91dbm and signals at -31, -44 and -51 on IoT devices around my home. Is that good or bad?
The antennas are currently just thrown together in a distribution box for multimedia devices. (I hope this makes sense, I don’t know what it’s called in English. Basically a big fuse box, but without fuses and instead space to mount multimedia devices)
I plan to drill six holes in the door and mount the antennas on the outside in a ring. Alternating 2.4, 5 and 6Ghz antennas. This should improve the signal a lot I assume, as now the antennas are behind a metal sheet door.
The R4 itself has no case. Plenty of heatsinks. The official set on the CPU with fan. And a heatsink on the SFP cases and a big heatsink covering all three wifi chips.
I have a noise of -91dbm and signals at -31, -44 and -51 on IoT devices around my home. Is that good or bad?
Better than my R4 and what people in this thread are getting, surely. Would you mind checking the noise with 5 GHz enabled too?
Looks like you have the best noise level reported so far!
BTW what is your version of the wifi board? Mine is V1.0 as stated on the bottom-left corner. And how you power the device? (For testing I power over USB-C, perhaps I should give it a try using a proper power adapter - EDIT: No change in reception noise on USB-C or 19V power adapter)
All radios are enabled, just didn’t have a device connected at the moment.
My laptop on 5Ghz Wifi shows in Luci -71/-90. And my Pixel 6 phone connected via 6Ghz Wifi shows -78/-90 in Luci.
On the phone I have an app called WifiAnalyzer running. It shows signal strengths at -61 for 2.4Ghz SSID, -65 for 5Ghz SSID and -70 to -75 for 6GHz SSID.
But as I said before, my antennas are just “lying around” in that box. I have to fit them properly and check again.
I’m using these antennas: Bingfu Dualband WLAN Antenne 2,4 GHz 5 GHz 5,8 GHz 3 dBi MIMO RP-SMA-Stecker + 2 x 30 cm U.FL auf RP-SMA-Kabel für WLAN-Router Wireless Mini PCI Express PCIE-Netzwerkkartenadapter: Amazon.de: Computer & Zubehör (German page)
Thanks for those detailed sketches! - I will also try to build one and plan to fabricate a “mold” with my 3d printer and “deep-draw” the 0.2 mm copper foil into it, by pressing the formed molds against the foil - not sure if it will work, but should be fun at least - but I probably plan to solder it directly without clips…
1 Like
BTW, I had a chat with ChatGPT 
And it gave me some new ideas, not sure if someone has experience with it, as I never saw them in the wild yet: There exists non-conductive RF absorber mats, e.g.
- 3M AB7000 Series → useless, only up to 4 GHz
- 3M AB6000 Series
- TDK IFM Series
Perhaps we could also experiment with those and try to shield as much as possible from the main-board. It is not as effective as an copper-shield, but would be much easier to handle…
These look promising too. Perhaps even as an addition to the can shielding. But they still have a conductive metal layer or embedded metal flakes. One has to be careful not to puncture the mat with sharp leads sticking out of the main PCB.
Well they’re quite expensive but the AB7000 series one is useless as it only operates upto 4ghz, where as the AB6000 goes upto 18ghz