Sub-$20 802.11n USB WiFi Adapter That's Linux Friendly

[caligula]

Here we come to what 9271 and 7010 are. AR7010 is basically USB-to-PCI-E bridge. Some microcontroller which got usb at one side and PCI-E lane at another, as well as some memory, etc.

AR9271 is basically something like AR7010+AR9285 radio in single IC package. It is good in terms of sensitivity and everything you expect from Atheros radios, i.e. decent support of virtually any mode one can imagine, up to working packet injection and so on, and more or less bug-free operation. Downside it is 1T1R, which means 150MBps at most in HT40 mode and half of that in HT20.

Best Atheros USB things picky one can find around is something like AR7010 paired with 2T2R Atheros 9xxx radio, giving everything one expects from Atheros drivers, opensource firmware, yet with 300Mbps speed (2T2R), good sensitivity and working with any more or less recent Linux out of the box.

One relatively known example would be something like TP-Link TL-WN821Nv3. Downside is that TP-Link always does same thing as other china/taiwan companies: now you can erroneously buy v4, which is totally different hardware under same name. Needless to say, it makes things a bit tricky.

I have that TP-Link one. It's great and didn't cost much. About $25. Anyways, I'd recommend realtek in general over shitty ralink and their proprietary crap.

[Henning]

I have that TP-Link one. It's great and didn't cost much. About $25. Anyways, I'd recommend realtek in general over shitty ralink and their proprietary crap.

I have tested the TP-Link TL-WDN3200 a little bit... dualband USB stick, 2*2 MIMO... and the Linux kernel driver seems to support all major Wifi modes.

[caligula]

I have tested the TP-Link TL-WDN3200 a little bit... dualband USB stick, 2*2 MIMO... and the Linux kernel driver seems to support all major Wifi modes.

Too bad it's not a sub $20 device unlike the one I listed is now.

[SystemCrasher]

(For reference, TP-Link label the hardware version at the end of the FCC-ID)

You see, nobody gives a fuck to FCC outside of USA. So you can see just "TL-WN821N" as description quite often, unfortunately. Without any hints about version at all. This makes such practice by *-link and somesuch quite annoying. Because if you're not inclined on tracking hardware versions, you can eventually buy completely different device under same label, which could be not what you want. Like if there is not enough letters for new model name when drastically changing hardware .

I have that TP-Link one. It's great and didn't cost much. About $25. Anyways, I'd recommend realtek in general over shitty ralink and their proprietary crap.

Realtek is known for manufacturing shitload of IC versions and revisions. Some of them are full of bugs. Some ICs are not supported by even recent mainline kernels by default (only if one dares to use staging drivers, etc). It maybe not a big deal if you only want some "station" (AP client) mode, but even then, drivers tend to be more bugged and it is not very well supported by each and every distro kernel out of the box, and extra messing with kernel and its modules is probably not what casual user would want to do.

In fact I can only imagine one reason why dongles manufacturers are so inclined on realtek. Traditionally, all Realtec ICs are cheap like a dirt. I guess that's the only reason why dongle MFRs are so inclined on Realtek. You see, if you save $1 per device when making 1M devices, you earn $1M extra. Which is good for MFR but not necessarily good for end user.

As for Ralink ... are they anyhow worse of Realtek? IIRC both tend to require some closed-source firmware for most of their USB devices, and both have relatively poor interaction with mainline kernel devs, aren't they? This leads to bugged/poorly integrated drivers. On other hand, ath9k is high-quality driver which rarely gives nasty surprises and well-supported in mainline kernels. And it has been like this for quite some years, so it usually works out of the box (Debian users may need to install firmwares though, even if I guess future Debians will support 9271/7010 out of the box due to opensource firmware nature).

The only known disadvantage of Qualcomm-Atheros is the way how they're treating regulatory stuff. In no way driver would extend device capabilities beyond what was stored in EEPROM, even if device capable of doing so. It would only further limit device capabilities, but never extend them. So if you bought "USA version" (where US country code stored in EEPROM), you can expect channels 12&13 disabled, even if you set EU, CN or other appropriate regdomain where chans12 and 13 are allowed. OTOH device with some EU or CN country code would gladly disable chan 12/13 once you set US regulatory, being able to resume chan 12/13 use after you stop using US regdomain. This makes "made-for-USA" devices purchase quite bad idea. Also, some chinese nuts do not read QCA manuals and sometimes believe writting 00 code indicates "worldwide". Yet it maps to USA, with channels 12/13 disabled. However most chinese adapters come with "CN" as eeprom country code, which is just fine (in terms of chan 12/13 and other regulatory crap).

[jabl]

ath9k FTW

Like someone already mentioned, if you want a IEEE 802.11n device with good Linux support, make sure it's one supported by the ath9k or ath9k_htc driver. At the moment, that's apparently the best driver, and the firmware for the device is fully open source. Also used e.g. by the cerowrt project (bufferbloat stuff etc.).

https://wireless.wiki.kernel.org/en/users/Drivers

[horizonbrave]

may I hijack the thread for asking if there's any friendly 801.11AC dongle someone tested?
it's a pretty sad situation from what I googled..
Thanks