Slashdot Mirror
How Best To Deal With WiFi Interference?
marciot writes "I live in a condominium where I get interference from my neighbors' WiFi. I understand that 1, 6 and 11 are the only non-overlapping WiFi channels, but how does this translate into real-life best practices? When you must overlap, is there a 'good' way to do it? With nine access points, for example, is it better to have three APs each on 1, 6 and 11, so that each completely overlaps with only two others? Or is it best to distribute those APs across nine channels such that they only partially overlap others (but potentially overlap more APs in total)? Do use patterns affect interference? For example, is it best to overlap a channel with multiple APs that rarely transfers data, or to share a channel with one person who downloads torrents 24/7? Does maximum data rate affect interference or robustness to interference? I found out by accident that setting my access point to '802.11b only' mode appeared to give me a vastly more reliable connection that leaving it in 'mixed 802.11b/g.' Is this a fluke? Or does transmitting at 10 Mbps when everyone else is using 54 Mbps (for their 3 Mbps DSL pipes!) give you a true advantage?"
You can't trust what you can see in the wireless network list to be an accurate representation of the noise level.
As another poster pointed out, the re are plenty of other devices that mucks up the signal in a condo. For instance, in my previous flat some caring neighbor bought a wireless surround system and since he was on the dole he pretty much watched TV all the time he was awake. And his sleep pattern was plaid.
The big problem with these automatic systems is that some of them will dynamically allocate a channel for itself when it is turned on and any channel you have previously chosen might be garbage now.
You can find autodetecting systems for wireless, but you might have to dig around a bit to find them.
Me, I use to hop channels and instead of trusting the channel strength and such I run a ping to a known host outside for each channel and then select the one with the least interference. But if your neighbor gets a noisy microwave or an anarchistic stereo, that could become a rather tedious hobby.
The answers are generalities, since each situation is unique. As others have already said, the real solution to your problem is spelled "5 GHz." However, if we add the condition that you must remain at 2.4, here we go:
In general, the former is best. Most site planning is done this way, with the (I hope obvious) additional condition that the cochannel APs are physically separated as much as possible.
Yes, use patterns affect interference. In general, the former is best, since the channel has more idle time available for "your" data.
Maximum data rate has a major effect on interference robustness. As you've found, in general lower rates can tolerate higher levels of interference than can higher rates. More explicitly, there's a range of interference levels (low) at which both will work. Above this is a range of interference levels (medium) at which the low rate will work and the high rate won't. Above this is a range of interference levels (high) at which both will not work. What you've found is that you're in the medium category, in which your system will work at 10 Mbps in the presence of interference from your neighbor's 54 Mbps system, but your system will not work at 54 Mbps in the presence of the same interference.
A second phenomenon may also be present, one specific to the 802.11g standard. To make it backwards compatible (i.e., so that an 11g AP would work in a network having one or more 11b devices) the 802.11g folk mandated a behavior in which an AP checks first to see what's around it. If it hears an 11b device, it downshifts into 11b. This, of course, slows the entire 54 Mbps network down to 10 Mbps. You may be experiencing a side effect of this -- all the checking and upshifting and downshifting takes time, so if 11b devices come and go frequently (as they might in your scenario) the net throughput can be less than if one stayed at 11b speeds in the first place.
It's not the first time I've heard this, but it comes from people who've observed degraded performance after increasing their AP's power output (usually with a 3rd-party firmware).
What's going on here is that:
a) Clients are still transmitting at normal power, so the AP can't hear the clients.
b) Many APs are built with circuitry that doesn't like to be pushed very far beyond factory defaults with transmit power: the signal really does get "noisy" at high power settings.
Too bad the default Kamikaze 7.09 OpenWRT firmwares kills any and all (six!) WRT54GL routers that I put it on (previously ran White Russian brilliantly). Apparently flashing these things with said firmware out of the box defaults the output power to 150mW (default is 28mW), and fries the transmitter circuitry. There's no option to fix this, you're supposed to install a package onto the router called "wl" and hack a call to this utility in the init script for yourself that sets the output power at bootup.