In my continuing saga with wireless networking I’ve learned quite a few things.  I’d previously mentioned that I was having trouble keeping my wireless network up and running.  I originally suspected interference from a wireless phone, perhaps in the house next door (which had new residents).  It turns out that it wasn’t a wireless phone, but a combination of faulty equipment and other interference.

After waiting about three weeks for my Wifi Detector(¹) to arrive, I was finally able to use it to get a reading during an outage.  What I found surprised me.  I was getting no signal whatsoever, because the router/AP was going offline.  The router I was using was a Linksys WRT54GS v2 that I purchased in late July to replace my old WRT54G v1 that had died while I was on vacation.  The old one had been rock-solid up to its untimely death after two years of service.

Guessing that something was amiss with the router, I went ahead and obtained a WRT54G v5 (²).  This one had the exact same symptoms as the other one.  I started doing some Google searches and ran across some interesting forum entries about the WRT54G[S] line.  Some people had been experiencing overheating that caused intermittent wireless outages and reboots.  They cured it in some cases using fans and heatsinks (and one guy even tilted his router up so that heat would flow better).  Another thread found that there appears to be a bug in the ethernet driver in the router that causes it to lock up after a certain amount of data had flowed (and/or when it was under heavy load).  Many users were able to very accurately predict failure by copying large files.  The router would lock up in the same place each time (although the amount of data flow varied).  Finally, one user found that there was a certain byte sequence in the data stream that would cause the router to lock up and reboot.  He was able to reproduce the problem with only a 1600-byte file.

Needless to say, I was less than thrilled with this, so after a morning spent with intermittent disconnects I went over to Staples and acquired a Belkin Pre-N router.  Installation was not pain free, as the thing has some weird quirks with regards to setting it up in AP-only mode (I found myself having to use a laptop with an ethernet cable and switch cables between the LAN and WAN ports after changing to AP mode).  It turns out that it needed an upgrade to the firmware to work correctly as an AP³.  Once I had that ironed out it seemed to work quite well, though.  The MIMO technology seems to provide better signal coverage as my office bridge was reporting 100% signal, where it had only gotten 76% with the Linksys. 

All seemed well for about a week.  Then I started having intermittent slow-downs and an occasional disconnect.  Some browsing on the bridge showed that during these periods it was seeing other networks in the area:

What’s interesting is that these networks don’t show up most of the time.  At the time this snapshot was taken, the bridge was working fine, even with the other networks visible (note the 100% signal strength).  What I noticed during the outages is that the signal strength on my network will go down to between 70% and 50%, and the 2Wire network will show up.  If the 2Wire isn’t visible, I don’t have any problems. 

A little digging on Google unearthed a possible explanation.  Some 2Wire AP’s are capable of transmitting at 400mW.  It is my understanding that “normal” AP’s transmit somewhere between 100-200mW.  However, it would seem that since my network is on channel 1 and the 2Wire is on channel 6 that there shouldn’t be a problem.  It turns out that this isn’t completely true.  The typical bandwidth for wireless-g is 44MHz (or so I’ve been told), which means that it uses frequencies from 22MHz below the channel center to 22MHz above the channel center.  This means that for 802.11b/g that channels 1,6, and 11 almost touch each other (e.g. channel 1 is at 2412MHz and channel 6 is at 2437MHz; adding 22MHz to channel 1 gets you to 2434MHz).  Here’s a diagram (borrowed from the IEEE 802.11b-1999 document) of the channel layout in North America.

Typically, the adjacent signals should not interfere with each other, because the IEEE specification shows a -50dBr signal strength at +/- 22MHz from the channel center, as shown in this diagram:

However, since I only seem to have trouble now when the 2Wire is operational, and since it’s capable of transmitting at much higher power than usual, I can’t help but wonder if something is out-of-spec on the 2Wire that’s causing interference on channel 1.  Although -50dB is a significant difference in power, it should be noted that it is a relative measurement.  If you take a 400mW signal and attenuate it by -50dB, it’s still greater than a signal at 200mW attenuated -50dB.  I’m not a radio expert, though.  I know just enough to be dangerous.  But my empirical observation is that the 2Wire is the likely culprit, and I’ve seen some information via Google that suggests that others are having similar problems.

Now it remains for me to see if I can pinpoint the house with the 2Wire unit and encourage the user to move the device to channel 11.  At least that would give some breathing room for my network, and perhaps show once and for all whether the 2Wire is the problem. 

Previous entries:
Channelling Annoyance
Wired to the Channel

¹ I suppose I shouldn’t complain too much, but I think Amazon’s free shipping has spoiled me for free shipping offers from other companies.  Amazon usually delivers early on their free shipping offer (or at worst they will be right on time with their estimate).  In this instance, with Buy.com, I thought perhaps the item was lost, since I got a shipping notice on 10/12 and didn’t get the package until 10/28.  In fact, I was just about to initiate a “lost package” request with them when I read their conditions.  You have to wait 21 days on a “budget shipping” item to report it lost.  If I was more cynical I’d suspect they do this on purpose to encourage you to purchase their other shipping options…

²  Linksys has a habit of keeping the same model number and just adding a revision code on the bottom of the router to indicate what hardware revision you get.  With the WRT54G[S] series, sometimes those revisions can be fairly drastic.  Version 5 of the WRT54G series has a much smaller motherboard, less flash memory, and a new operating system (VxWorks instead of Linux).  Further, because of the changes, it will not work with alternate firmware, such as DD-WRT.

³  Why would you buy a router and use it as an AP?  Fios TV requires me to use the D-Link router they supplied, and it’s a wired unit.  Further, it turns out that it’s often less expensive to buy a wireless router than a AP.  I’m not sure why it’s that way, but that’s the way it is in the marketplace (perhaps because in the home/home-office segment more people buy wireless routers than AP’s?).