802.11N confusion

[tmetro-blu-5a1Jt6qxUNc at public.gmane.org (Tom Metro)]

David Kramer wrote:
> So I'm shopping for a new Wireless router to replace my WRT54G...
> My needs are simple enough; Wireless N, 4 or more ports, would be nice
> if it was DD-WRT compatible.  Recommendations welcome.

I pity the average consumer going to buy a wireless N router.

A while ago I replaced my trusty Linksys WRT54G wireless router (being
used as an access point) at home with a TRENDnet TEW-652BRP 802.11n
router. (The reviews were a bit mixed, but it was cheap and reports were
that it was supported by at least some of the open firmware projects,
and had decent CPU and RAM, so if nothing else it could be repurposed.)

The reality was that after replacing my WRT54G, neither my laptop or 
Nokia N810 connected as reliably, the laptop didn't show any better 
signal strength, and worse of all, it still showed that it was 
connecting at 801.11G speeds. (With kernel updates and antenna position 
tweaking, my Ubuntu laptop now connects pretty reliably, but the N810 
still struggles to establish a connection, unlike with the WRT54G.)

That's when I went back to the research drawing board and learned more 
about 802.11N, how it utilizes two different frequency bands, and that 
products that use only one of the two bands still get labeled 802.11N, 
so it's possible your router and laptop aren't even on the same 
frequency, despite both being N products.

It seems the Draft N marketplace is a rather confusing mess. Have you
ever looked at the routers for sale at say Staples and wondered why
there are 3 similarly described Linksys N routers, each at a different
price point ranging from $50 to $130?

The devices can vary by the number of streams, number of independently
operating antennas, number of bands (2.4 GHz and/or 5.8 GHz), and width
of the bands, providing theoretical speeds from 6.5 Mbps to 600 Mbps.
See the data rate chart:
http://en.wikipedia.org/wiki/802.11n#Data_rates

There's an attempt to distill the most relevant bits into a label of the
format a x b:c, where a=transmit antennas, b=receive antennas, and
c=streams. See:
http://en.wikipedia.org/wiki/802.11n#Number_of_antennas

Good luck making that consumer friendly. So far manufacturers haven't
event bothered to report this info.

A consequence of this is that you can buy an 801.11n router to go with
an 802.11n notebook and experience: 1. no connection if the two run on
different frequencies (most single frequency devices use 2.4 GHz, so
unlikely), or 2. reduced speed because one or both devices don't have
enough antennas, streams, or don't operate at 5 GHz where you're more
likely to be able to operate at full bandwidth without interference.

Then I read about "single stream" pseudo-N (they're not officially N 
compliant) routers being brought onto the market as a cheaper alternative:
http://www.smallnetbuilder.com/content/view/30804/100/

See also:

SmallNetBuilder's Wireless FAQ: The Essentials
http://www.smallnetbuilder.com/wireless/wireless-basics/31083-smallnetbuilders-wireless-faq-the-essentials

5 Ways To Fix Slow 802.11n Speed
http://www.smallnetbuilder.com/wireless/wireless-basics/30664-5-ways-to-fix-slow-80211n-speed


All that lowered my expectations of seeing N speeds, but I attempted to 
look into ways to diagnose the wireless link to my laptop, such as 
determining what frequencies it and the router supported, and checking 
to see which channels were more crowded in my area, but I didn't get far 
before running out of time.

I was disappointed to see that the common WiFi scanning tools for Linux 
wouldn't report the channel that a discovered network was on. (I know 
the war driving tools, like Kismet, report this, but it shouldn't be 
necessary to use something that specialized just to look for the least 
crowded channels.)

Similarly there was no apparent way to determine what 802.11 standard 
was being employed by your wireless connection in Ubuntu, other than 
inferring it from the reported connection speed.

The Intel WiFi driver:
iwlagn: Intel(R) Wireless WiFi Link AGN driver for Linux, 1.3.27ks
iwlagn: Copyright(c) 2003-2008 Intel Corporation
iwlagn: Detected Intel Wireless WiFi Link 5100AGN REV=0x54

doesn't seem to report that info when a connection is established:

wlan0: authenticate with AP 00:...:1d
wlan0: authenticated
wlan0: associate with AP 00:...:1d
wlan0: RX AssocResp from 00:...:1d (capab=0x431 status=0 aid=2)
wlan0: associated
ADDRCONF(NETDEV_CHANGE): wlan0: link becomes ready
wlan0: no IPv6 routers present

Changing the router's channel width from 20 MHz to "auto 20/40 MHz" had
no effect on the connection speed. Not surprising if it is connecting in
G mode, as that's fixed at 20 MHz.

I did try a variety of different channels, and different bandwidth 
settings on the router, but never achieved anything better than the G 
speeds.


> ...even the $100+ units have a striking number of very low reviews...

I noticed that too. I think a big factor is that these are RF devices, 
and RF is highly dependent on the environment - interference and 
obstructions. As the number of frequencies and bandwidth goes up, as is 
the case with N, this will only get worse. (At least until the next 
technological paradigm shift, like spread spectrum, or MIMO provided in 
the past.)

Then you have to factor in inter-vendor interoperability. Most of the 
time the client machine is using a wireless interface from a different 
vendor, and although most interoperability issues have been worked out 
for 802.11G, they still seem to exist for the newish N.


> I know a wireless router is not a trivial device (it's pretty much a
> single board computer), but how can top manufacturers produce expensive
> units that regularly lock up or disconnect?

They're building a complex, self-contained, low-power computer with a 
multi-band radio transceiver, plus CPU, RAM, and OS that probably rivals 
what you had on your desktop 8 or 10 years ago, and manufacturing it for 
probably less than $30. You can bet they've cut some corners.

On the up side, I can report that the TRENDnet TEW-652BRP has been 
stable for the most part, having only been rebooted once since I 
deployed it. (I have it sending syslog messages to another box, and 
after about 125 days of uptime it spewed out some cryptic kernel errors 
that sounded like it was running low on memory. TRENDnet support 
recommended reflashing the firmware (to the same version) and resetting 
the configuration. I just rebooted, and it hasn't occurred again in the 
last 60 days since.)

  -Tom

-- 
Tom Metro
Venture Logic, Newton, MA, USA
"Enterprise solutions through open source."
Professional Profile: http://tmetro.venturelogic.com/