Ways to increase the reliability of wireless networking wanted
 

I'm still suffering with unreliable wireless networking. So often the link just drops.

A few baseline hardware info and figures:

The NIC is a D-Link DWL-G510 PCI card.
The driver is the rt61pci driver.
The router is the "O2 Wireless Box II", which is a rebranded Thomson SpeedTouch TG585v7

Link quality, as reported by iwconfig, is typically around 28 to 34/70. I've never seen it above 40 (though since I check when there are problems, there's some sample bias). Signal strength is typically around -74 to -82 dBm, and similarly, I've never seen it above -70. I've no real idea, but I get the feeling these figures are lowish.

There are two solid brick walls between the router and the PC. I know this isn't helpful, but there's really no way to change it. I don't have any laptop with wireless networking any more (the USB adapter I used to have bit the dust), so I can't try and map the signal around the house to see just how much the walls attenuate it.

I've tried various physical things - moving the router and reorienting its antenna, reorienting and changing the antenna on the PC. (Making sure PC and router match on antenna orientation seems to boost the signal by 6-10 dB.

What in the way of software changes might help my connection? Things like iwconfig settings, router settings, and so on.

I know the problem could be that the router sucks. But the thing here is I really don't want to spend £40 or more on a new router and find my wireless is STILL unreliable. (I suppose I could buy from Argos and return it if it doesn't help)

If no one will suggest you better idea, I would like ask you when you notice that connection was dropped, please type in console: "dmesg | tail -n 50". It will gives you output of systems logs, and you and we can (I hope) see why there was disconnection.

I found that adjusting the rate on wlan0 to 6mb/s gave me more reliable wireless. Also using NDIS-wrapper was the BEST solution for my atheros wireless. I used NDIS-wrapper with Mandriva which was simple and it worked just like it had with vista, which was stronger and more reliable than any of the OS drivers I've used. On Ubuntu NDIS-wrapper proved to be a PITA so I adjusted the rate.

Brick walls are still killers. Relocating the router is still probably the best option if there is any way to do it, such as moving it to an attic or a window.

EDIT:
BTW there are WIDE ranges in the range of different routers. I am assuming that you are somewhere in the UK because you seem to be using pounds instead of dollars, and I've never heard of your current router so I don't know that I could recommend anything that would be available to you, but the buy and return option might not be a bad idea either. Also, USB dongles tend to suck-the-big-one compared to built in wireless. I don't have a lot of experience with desktop wireless cards, but I frequently see new customers using high gain antennas that I believe use USB. Theses must work fairly well, because I see them used to steal the neighbors signals at decent ranges (100-200 feet between houses) I'm usually only there because the neighbor moved or put a password on their router. I know D-link makes one of these antennas.

Thanks, I'll try that. (Of course, one issue is actually measuring the result of any change. I think I should be able to grep the iwconfig output in some sort of script, but it will require many hours if not days to determine if any change has helped)

Well the router location is constrained to be near enough to the telephone socket. I can only really move it around within the same room.

The range available at retail is pretty much the same as in the US to my knowledge. D Link, Netgear, Belkin, etc.

Well when it comes to my laptop, it HAS no built-in wireless.

http://www.bonanzle.com/booths/hitec..._magnetic_base Is an example of what I was talking about. I guess they aren't usb, but just an additional antenna for when you have a have a replaceable antenna to begin with.

Ah, I see. I currently have two antennas - the common stick type, and a free standing one with about a metre of cable to the plug. The latter seems to work a bit better since I can move it around.

Another question: when logging wireless signal strength, how often should I do it?Like, every 5 seconds, every minute, what's best? Obviously too long an interval risks losing short-period fluctuations, as well as possibly giving sampling bias. But too short an interval just leads to too much data, and I'm not sure if running iwconfig every second is a good thing.

I'm more of a "beat it till it fits" kind of person. At work when I'm looking for the cause of a problem I look about twice a second (manual refreshes) and look for spikes here, drops there. It's usually obvious within a few minutes.

From your outputs it looks like your signal is marginal at best to begin with which would make it harder to deduce I'd think. Maybe every 30 seconds for an hour? Every 15?

What does say 100 pings to your gateway look like?

Could you attatch your bigger antenna to the router maybe?

you should consider the fact that there could be adjacent transmitters using the same frequency/ chann that you are. you can observe this using:

iwlist scan

and see if theres an offending signal.

also lower frequencies by the nature of physics tend to be more intelligable at the same output power but at a lower frequency. therefore lower freqs go thru walls better. ie use 2.4 GHz band instead of 5.0 GHz and try its lowest channel.

also consider that non-wifi sources can interfere, ie cordless phones, microwave ovens and maybe leaky CATV adapters.

finally - if you suspect that there no other signal affecting your quality, then do a low budget mod to double the signal level from the router... just get a flat metal surface and put the router antenna a few inches in front of it, the metal surface's face should point directly toward the computer antenna.

hope this helps.

Maybe the phone or the microwave, but CATV should be operating at or below 1000MHZ and should cause no interference.

When I see the neighbour's networks they're usually on channel 6, I'm on 11.

Hardware only supports b/g anyway

None of them in our house. And I can't do much about what the neighbours use.

Hmmm...interesting results from this. Metal sheet 3cm (1/4 wavelength) from router antenna - no overall improvment but seemingly more fluctuation. Metal sheet 3cm from computer antenna - perhaps a boost. The thing is doubling the signal strength is only 3 dB, and the 'resolution of the signal strength is 2 dB.

I did, however, find the perfect place for my computer's antenna, giving around a 10 dB improvement. The middle of my floor right where my computer chair goes :-(

Lol

cf:
(ignore the router designation: it doesn't exist)

that's on the 'a' (5G) band and seems to be working fine...at least fine for simple stuff, like browsing the 'net, although if I was trying to stream video, I might have another opinion.

My signal level is similar to yours, but the noise is really rather low (that -127 dBm sounds like the theoretical background level and may be a bug in the driver, I suppose). But the 'a' band is relatively quiet, and I don't see drop-outs even with a similar level to you.

I know most of my neighbours are on the 'b' standard/band, which is why I'd rather stay on 'a' and 'a' tends to be quieter, anyway.

The brick walls are the brick walls. You are right that they aren't helpful, but if they are there, they aren't likely to go away any time soon (you'd hope, given the problems of brick walls suddenly going away).

10 dB is a bit more than I had expected...

It might have been interesting to try leaving everything running in the absolute best condition (is that two separate 10 dBs, or is there just one 10 dB available?). In any case, with your best combination, you should have a healthy signal, so does everything work well with the healthiest signal available, or are there still problems?


...PCWorld offer a similar deal, FWIW. There are some on-line suppliers that do, too. If boosting the signal by 10dB+ makes the problem go away, then there is a good probability that an 'enhanced' router work work well. OTOH, if the 10dB+ doesn't do it, the better router is probably still going to struggle.

One supplier calls its enhanced b/g products 'RangeMax', and that sounds like the kind of thing that you'd want, although I have to emphasise that I am in no position to compare the virtues of one supplier against the others, who have similar, but different, proprietary technologies available.

If you have a replaceable antenna, an antenna with gain has real advantages (google, for example, cantenna). However, be aware that the physics of propagation mean that the more gain that you get from the antenna, the more directional it becomes. So, it starts to get difficult to ensure that the antenna alignment doesn't start to become a problem.

The point that 'lower frequencies go though walls better' is, up to a point, reasonable (although, the actual cause is not as stated), but:

• The percentage difference between the highest and lowest frequencies on the b/g band are so low that the effects within that band can be swamped by wave interference (or, alternatively, electrical interference) ones.
• the 2.4 G band has far more users, so has a higher level of interference/background noise
• A large problem for 2.4 G is that it is the first water resonance (that's why it was easy to get as an ISM band...its useless for long-range communication) so, if one of the walls could be damp, that could tilt the balance

Microwave ovens, bluetooth 'phones, yes. Cordless 'phones have their own band, and, unless they have moved on to bluetooth for some reason, shouldn't be an issue (they hadn't last time I looked, but that was 15+ years ago). But, it is an ISM band and therefore free for unlicensed use up to a certain rather low power level, so anyone/anything could use it within the limitations and you have no recourse to anyone to fix the 'problem' for you, unless the interferer is breaking the regs.

I regularly use 2.4 Ghz to go 10 miles at work. So to say it's useless for long range communication is a little off.

two ideas immediately strike me. other signals could be interfering and low signal strength.

to map out interfering signals you may want to consider getting a spectrum analyzer - try a WiSpy usb device on the cheap - just a couple hundred dollars vs tens of thousands.

if signal strength and quality is still unacceptable after relocating devices to a quieter location, increase the directivity of accesspoint and computers wifi device - use a pair of old 18inch sat dishes with the wifi antennas each a the focal point (where the sat LNB would go). the dish isnt optimized for the freqs in use but would give massive directivity gains - atleast 40db when pointed directly at each other.

but who would ever put that inside their house, besides me, lol.

AFTERTHOUGHT: use a different freq band, or repeater or higher power transmitters.

A couple of hundred dollars is NOT cheap in my book! One could probably build a receiver or downconverter for much less, but the challenge would be ensuring linearity across the 80MHz of bandwidth.

There is one software solution that would help from a usability perspective. At present when the network goes down it STAYS down. It would be much better to have something automatically try and re-establish the link. I could knock up a simple bash script based on the ifconfig or iwconfig output (perhaps put the functionality in the same script I use for stats logging), but does anyone have suggestions for existing solutions? (There seem to be many 'wireless connection managers' and such around, so it's hard to know what's best.)

It's worth bearing in mind that high-gain antennas risk hitting legal issues with exceeding ERP. I actually have a ham license, so I could run high power wi-fi - but then I'd be prohibited from using any encryption (which would include https), so it's not suitable. Directionality even without an increase in ERP still brings benefits though. Mind you, aiming a narrow beamwidth antenna like a dish indoors really wouldn't be practical. And frankly, such elaborate setups should not be necessary.

Also, coming back to something mentioned before
Condensation might explain the seemingly daily variation in signal quality experienced by myself and another (in another thread). It's colder at night, which promotes condensation. OTOH if there was damp in the walls I'd have thought I'd have noticed it.