Show HN: I built a Wi-Fi 6 survey kit to see how common it was in my area
wirelessbits.net
wirelessbits.net
Assuming that we are talking about a 5GHz band 802.11ax ap inside an ordinary house, and you're on the exterior, you do not need to be very far away at all before the signal that you will see is something extremely weak like -84. Though if it's a dual band AP and is also running on a 2.4GHz channel at 20MHz wide on the same SSID you may see the 2.4 at a usable signal strength.
You will see slightly better results in areas of primarily wood framed single family houses, where the material to go through is wood, drywall, house siding, versus an area that has more concrete and steel construction.
As I'd like to be able to work sitting out on the deck, one of my spring projects is to trench some ethernet out to a shed (plywood and vinyl siding construction) and add another AP.
How much safer is using fiber, and giving up on battery backup? I'd also have to redo a bunch of power; the shed does have 120V but it's switched from inside the house and hardwired to a couple outdoor lights (on posts), so I would have to basically run a new power line either to the shed or to the lights; something I might eventually do but wasn't planning on this year.
The last I looked, the Intel 6e wifi cards did not work in AP mode, so building one was not an option. And sorting through the OpenWRT supported routers and trying to correspond that to 6e models from https://www.wi-fi.org/product-finder-results?sort_by=certifi... is rather challenging. Especially since vendors re-cycle model numbers..
Two weird caveats I had to overcome: 1. I had to compile OpenWRT from master myself so that I could configure the WiFi 6 interface using Luci (the web ui), because the latest stable release does not include that functionality. 2. I bought two of them for two floors and OpenWRT somehow managed to generate the same BSSID for both of those routers, which caused both of the networks to be inoperable. To fix this I had to edit a bash script which generates the BSSID in OpenWRT and change it to something else.
But the short answer for routers with OpenWRT 6E support is no. For m.2 6 GHz capable cards that have Linux drivers and support AP mode I know Qualcomm has a few models that fit that bill, in general just search any tri-band module from them and you should be set. Broadcom probably does as well but I'm not as familiar with those.
I passed dying on this hill long ago though. Even if you get such a card, get it working, and avoid bugs you're still going to be loading binary firmware to it. A much easier way to achieve the same level of openness is to buy a well supported router with AP mode and a properly designed antenna pattern then hanging that off a port on your open router; filtering any packets that source from the management MAC of the AP to completely isolate it.
Yeah but you're still trusting the black box to (a) implement security (WPA) correctly and (b) not get hacked and spew out packets onto the wired interface with forged MAC addresses claiming to be from your laptop's wifi card.
I've run into this problem before and currently I'm leaning towards just doing wide-open no security 802.11 with nothing-but-wireguard-packets firewalls on both ends of the wireless link. Basically s/wpa/wireguard/g.
That said if "attacking your adapter's/router's firmware flaws for an encryption/authentication vulnerability so they can hack in enough device control to fake being a valid client and further hack you from that launch point" is a real enough concern for you to be worried about then my first recommendation would be to stop using Wi-Fi in the first place, the entire design of the discovery and advertisement portion of the protocol is the anti-thesis to what you need.
However if you really think you can outdo everybody else by ground up building your own Wi-Fi system then a secure tunnel like wireguard or IPsec or so on will at least provide you a second layer that'd need to be attacked. That said such a layer is probably good for you in this case regardless and I'm not sure why you'd replace WPA* instead of use it as yet another layer of protection.
Have you spent any time reading the source code of wpa_supplicant?
I still have nightmares from that.
You have to look hard. It's basically only Atheros 9280 that can do all that with FOSS firmware. You also have to make sure the device has TWO RF front end chips (one for each band). Just looking for "AR9280" will get you a lot of single-band products.
There are also devices with that chip plus an AR7010, which is a nifty little USB-to-PCie bridge so the whole thing fits on a USB stick, like this:
Also I think you meant 802.11n, while the card does support 802.11a that is 54 Mbps on 5 GHz only.
wlx00a096****** IEEE 802.11 ESSID:"XXXXXXXX"
Mode:Managed Frequency:5.18 GHz Access Point: XX:XX:XX:XX:XX:XX
Bit Rate=300 Mb/s Tx-Power=30 dBm
Retry short limit:7 RTS thr:off Fragment thr:off
Encryption key:off
Power Management:off
Link Quality=70/70 Signal level=-31 dBm
Rx invalid nwid:0 Rx invalid crypt:0 Rx invalid frag:0
Tx excessive retries:0 Invalid misc:0 Missed beacon:0
They don't have an expiration date, you know. We aren't talking about fruit here.
And yes, like everybody on planet earth other than the IEEE -- including the people at Qualcomm/Atheros who wrote the source code for my wifi card's firmware -- use "a" to mean "5ghz band". People need to abandon this silly crusade to insist that "5ghz band" does not have an 802.11 designator, as if it is a feature with no name or something.
The problem, which you're ignoring, is that there are multiple 802.11 standards for operating in the 5GHz band. Treating the oldest and slowest among them as if it's synonymous with all 5GHz WiFi is just plain wrong, because the successive standards for 5GHz WiFi have meaningful improvements in capability.
(Granted the person you were replying to also said 300Mbps on 802.11a, which is of course not possible.)
Not rotting isn't the same thing as being new. The best it can do is hold steady to the age it came out at with whatever enhancements can be provided in firmware. Ignoring ac(1/2)/ax and 6 GHz there are also a litany of intermediate standards (hence the letter skips) many of which are relevant to all PHY standards.
No matter how you slice it these things are ancient. That said it's great you can make it work for you though as there are no newer alternatives to fall back to.
Are there any USB-3 devices like that?
I suspect Wi-Fi 7 will be more common. Wi-Fi 6E was great but unless you're one of the rare consumers that is going to replace every client chip you can to gain use out of that 3rd band it's identical to the Wi-Fi 6 equipment many just recently upgraded to. Wi-Fi 7 will have the benefit of some phones/laptops already being 6 GHz capable at launch (even if only 6E devices at first) as well as being a general improvement and farther apart for most in their upgrade cycles.
I'd recommend just buying a nice wifi 6 system that puts clients and backhaul both on 5ghz.
I routinely pull 500mbps from a satellite on a wifi 6 mesh network(asus xt8), I don't see much need to upgrade.
01:00.0 Unclassified device [0002]: MEDIATEK Corp. MT7915E 802.11ax PCI Express Wireless Network Adapter
Bought from here: [1]
Works with no hassle as AP under stock Linux kernel.
[1] https://www.asiarf.com/shop/wifi-wlan/wifi_mini_pcie/wifi6-4...
Although this could conceivably be open source friendly with such controls only on the radio software, so far manufacturers have responded by simply locking down the devices.
As a prime example, ASUS merlin is actually open source and compatible with a number of WiFi 6 routers: https://github.com/RMerl/asuswrt-merlin.ng
However, directly in response to the FCC rules, ASUS has restricted the ability of end users to havk on their own devices, giving only one person the ability to publish signed builds.
I wanted to make a point to point connection for 500 Mbit/s and figured AC will be more than enough. Well, turned out I can squeeze at best 260Mbps between 2 fully optimized AC routers with 3 streams (3x3 antenna config) at 975 Mbit/s (40Mhz band) negotiated speed.
Which was really disappointing, I was thinking AC would be faster than N, which gets around the same real throughput of ~1/4 of the negotiated speed (i.e. 450 Mbps 3x3 = ~120 Mbps real bandwidth).
So, the only real way to get a higher speed is to use more streams (on both ends) or 80 MHz of band width, which is only available on higher end chipsets and is really finnicky in most conditions.
I tried all possible mkt config, including fully routed and fully switched, same result.
If you want an outdoor high speed five gigahertz point-to-point connection, I would recommend using a pair of dedicated radios designed for that purpose rather than regular Wi-Fi stuff. Such as ubnt af5xhd or similar.
Unfortunately I can't effectively test the AP-reported ~ 600 Mbps throughput on 867 negotiated, but it's pretty accurate at lower bandwidth.
Yes, I tested with a Fritz!Box 4060 with WiFi 6 (4x4 Multi-User MIMO) and a MacBook Pro 14". Confirmed on the Mac that 802.11ax is used. When standing right next to the router, I get between 400-700Mbs (it varies a lot). I also use a WiFi 6 repeater as an additional access point (hooked up to ethernet), so I am getting 200-400Mbs in most of the apartment.
So, it's ok-ish, but very variable. We have most of our apartment wired up with Ethernet, so Wifi is mostly for iPhones/IPad or when setting on the couch with a laptop.
Many of the changes aim at high density environments, such as trade fairs and conference centers. Once the majority of devices are upgraded, these situations will be handled much more effectively.
Lastly, OFDMA is capable of significantly lowering latency if all devices in a network support it.
If you are using Wifi6 at home with few devices connected, do not expect noticeable improvements.
The router is on the floor; I haven't tried putting it on a high shelf to see if that improves the outdoor coverage.
I haven't done any special configuration on the router or phone/laptop, just running everything as it came out of the box.
The three streams may not be helping much either, since there may not actually be much spatial separation point to point at several hundred meters. Best case, you're only going to get maybe a bit over half the link rate in real bandwidth. There's just a lot of overhead with wifi.
If you're using omnidirectional dipole antennas, you're losing a lot of signal strength at that distance by radiating mostly in the wrong direction.
Your best bet, I think, would be to use two sets of linearly polarized directional antennas offset by 90 degrees. That way, they're out of phase and you may be able to get 2 independent spatial streams. Alternatively, use two directional antennas but point one set indirectly at a big flat metal wall or something, in the hope that the signal reflects and intentionally gets some "multipath". Also, turn off "legacy" support so that you're using lower overhead greenfield preambles.
Doesn't seem to add up. The max rate for AC 3x3 40 MHz is 600 Mbps PHY rate not 975 Mbps. This is with GI at 400 ns and the maximum signal quality. 975 would only be reachable on AC via 80 MHz channel 3x3 with a GI of 400 ns and a less than optimal signal quality.
> Which was really disappointing, I was thinking AC would be faster than N, which gets around the same real throughput of ~1/4 of the negotiated speed (i.e. 450 Mbps 3x3 = ~120 Mbps real bandwidth).
That's a ~2.16x increase for a single generation jump!
> So, the only real way to get a higher speed is to use more streams (on both ends) or 80 MHz of band width, which is only available on higher end chipsets and is really finnicky in most conditions.
Depending on your environment bumping to an 800 ns guard interval may help, or at least lower the jitter from retries. Also for P2P having a good directional antenna can be more important than having more spatial streams/spectrum. Both of the latter increase maximum bandwidth but decrease signal quality while good RF design like a good antenna and placement will serve to increase signal quality.
Alternatively for P2P there are purpose built solutions that tweak a lot of the standard Wi-Fi assumptions to deliver better performance. A typical go-to is Ubiquiti's airmax line which is on the lower end of the cost spectrum and has options in the 2.4/5/60 bands. On the higher end there is the airfiber line which adds in the 24 GHz band as well as dedicated send/receive options. This makes the connection full duplex rather than half duplex which solves a litany of slowdownsz in one swoop. Unfortunately it also comes at a premium price point. If you don't like Ubiquiti then Cambium also has decent P2P gear, I don't recall if they have any cheap options though.
On airfiber I was able to achieve gigabit speed over a couple miles between buildings. We actually used it as the primary link for the IT office to the rest of the network as it performed overal better than the 300 mbps fiber line.
.
To your original question I've gotten real world >800 Mbps goodput on AC 2x2 80 MHz and real world >1 Gbps on AX 2x2 80 MHz. These are half duplex speeds of course. If the P2P use case is long distance I'd stay away from 6 GHz as there are some power limitations there.
Reading the replies I'm thinking better chipsets could've achieved higher speeds, but that doesn't quite make sense - if the negotiated rate is the same, why would the real speed differ between routers? I used Archer C7s, which I love because I'm poor, very cheap, very reliable with OpenWRT.
I ended up running fiber - it was surprisingly cheap and I didn't even need special tools for the connectors, just have to be veeery careful when cutting and inserting the cable.
6ghz though will surely make your statement true.