Spurious WiFi signal and speed readings

[The late XV105]

Question
What causes spurious speed and signal strength readings when hanging on to a WiFi connection by the fingernails, please? Just curious.

Background
Whilst away on holiday I was given access to a ASUS 520g wireless access point from a neighbour about fifty clear line of sight metres down the street. Their house is largely wooden in construction, which I assume helps the radio signal propagate, though my holiday house was of 50cm solid stone construction, rendered on the outside, and with just one wooden framed, secondary glazed (15cm air gap) window in the room I used that could help the radio signal.

Using the standard XP Pro "WiFi Status" indicator, during the day the signal was a weak but rock steady 1 bar and the indicated speed between 2.0 and 5.5 Mbps. Perfectly fine for basic browsing, MSN messenger, and MS Exchange Server with only minimal patience needed. As night fell, and TVs and things got switched on and especially when the neighbour switched the chains of Christmas light adorning his house on (an almost perfect grid that had an instant negative effect on the signal), the speed fell to between 1.0 and 2.0 Mbps with signal remaining at 1 bar. Connection remained rock steady. Much later in the night (everyone else in the street in bed and appliances switched off) it rose to a consistent 11 Mbps, still at 1 bar and still rock steady with no disconnections.

Then the snow came and laid a thick blanket on the roof of the neighbour's wooden house, the eaves of which reach down to adult head height and are therefore lower than his upstairs WiFi access point. At this point connection became dire, usually being completely unavailable despite a 1 bar signal still shown and repeated attempts to "Repair" (usually failing at the connection stage, sometimes getting further but failing on IP allocation). When connection suceeded though, it did strange things. Chief amongst these was registering either an untrue 54Mbps speed (but still the usual 1 bar signal) a few seconds before disconnection, or an untrue excellent signal strength likewise a few seconds before disconnection.

Unrelated to my question but for completeness of background, yes I did try different channels (especially 1,6, and 11) and whacking the radio power up to maximum but it didn't help and performance appeared the same.

[mixture]

aah XV105,

We meet again....

Happy new year !

Quick answer your question : I believe we've previously discussed WiFi vs Copper/Fibre. :ok:

Long answer: I don't know. Probably a combination of dodgy drivers with, ahem, "uncomplicated algorithms" for calculating strengths and displaying details .... combined with consumer grade transmitting equipment and the associated firmware which may have bugs.

[mad_jock]

I think it will be to do with the snow reflecting the signal back.

It will be getting a strong signal, it will be your signal going 2 feet out the wall and back again.

[The late XV105]

Hi mixture! Thanks for that, and a happy new year to you, too! :ok:

The gigabit network out to the garage roof where the NAS lives is working great, thanks. Chuffed to bits, actually, and even my wife wonders how we managed without it (both an always-on NAS and a gigabit network!) :) I've even got a pukka sheathed lightning conductor cable bonding the two buildings as well so hopefully no fried equipment at the other end in the event of a strike to either of them. Just a shame that running a cable down the street wasn't an option on holiday!

Thanks, mad jock, but I would be surprised. No snow on my wall, only the distant wall (low hanging roof to be exact) of the neighbour's distant house.

[mad_jock]

I wouldn't be suprised with anything to do with EM data transfer. I proberly was talking bollocks about the snow. But the whole thing with troubleshooting Wifi is it has so many variables that can effect it.

If you fixed a wired network problem before and got something working normally if the same problem came up again you were 99% certain all you had to do was your original fix and away you went.

Not so with Wifi in my experence.

[Capetonian]

Wifi is very odd and unpredictable. I have a Belkin router/modem. My desktop is connected to it with an Ethernet cable and we have three other computers in the house. The closest is in my son's study, it's less than 5 metres from the router, and has an internal wall to go through, and is not line of sight. There are no variables, but the signal on his computer varies from 0 to full strength, and there is no clear pattern of it being better during hours of light or darkness.

I often use my laptop on the balcony, there are 4 walls for it to go through and again it is not line of sight. The signal is fairly consistently 2 or 3 out of 5 on the indicator, but sometimes, for no obvious reason, it is impossible to get a signal there.

Sometimes I move into the lounge if it gets cold outside and the signal there is worse than on the balcony, yet it is closer to the router and there is one less wall for the signal to go through.

Make sense? Not to me.

[Pontius Navigator]

Heavy snow is just solid water and will act as a barrier to EM.

I remember many moons ago watching TV, the old B&W stuff, when it rained. It rained supercooled dropplets. The picture got noisier and noisier then was snuffed out.

Outside everything was sheathed in ice. You can get the same loss in an aircraft with heavy icing.

[Saab Dastard]

When connection suceeded though, it did strange things. Chief amongst these was registering either an untrue 54Mbps speed (but still the usual 1 bar signal) a few seconds before disconnection, or an untrue excellent signal strength likewise a few seconds before disconnection.

I think that this might be down to a combination of sampling frequency and Fallback data transfer rates, aka Adaptive Rate Selection.

The fallback rates are:


802.11b: 11, 5.5, 2, 1 Mbps, auto-fallback;
802.11g(Normal mode): 54, 48, 36, 24, 18, 12, 9, 6 Mbps, auto-fallback.


Note that a .b / .g wifi access point actually has a separate modulation mechanism for .b and .g - a "b" client" on a "g" network will significantly reduce the throughput on the .g network.

It looks as if your wifi card was set to 802.11b/g, and since it connected reliably at 2 / 5.5 Mbps it had "locked on" to an 802.11b signal (this could either be because the wifi router was .b only, or both .b and .g, and had reduced to .b because of poor signal strength).

If the signal was lost, your card might then have tried to establish a .g connection, which would "start" at 54 Mbps and drop to nothing when the signal became too poor.

The (very approximate) graphical indication of wifi connection speed & strength samples and displays at fairly long intervals compared to the actual re-try attempts going on behind the scenes, so perhaps just showed the beginning (strong signal, 54Mbps) and / or end of the process (no signal, no connection).

SD

[mad_jock]

Show off :p

[Mike-Bracknell]

Wifi is very odd and unpredictable.

No it's not. It's just that in order to interpret the results you need to spend more money than is acceptable to the home (and half the business) users in order to sort it.

e.g. employ a few people with PhDs, a bunch of airwave analysis software and a whole bunch of work and it's understandable. Unfortunately, it's much cheaper to just waggle the aerial about a bit amid mutterings of the mantra: "goddammit who invented this sh1t"

:ok:

[The late XV105]

Thanks, SD. :ok:
I like learning and this evening I learned some more!

I already knew about b and g coexisting dramatically reducing throughput (my home network is set to g only for this very reason) but I hadn't stopped to think that actually to have them coexist means two signals.

Whilst on holiday I found on the web the User manual for the access point I was using and it advised that b range is something like 2 or 3 times that of g. It therefore fits perfectly that I was initially connecting to a b signal given the distance and obstacles involved, hence the fall back rates you quote fitting with what I saw too.


Cheers!
XV

[Saab Dastard]

It's only a hypothesis - I'm not a PhD in wifi! :O

SD

[jimtherev]

... employ a few people with PhDs, a bunch of airwave analysis software and a whole bunch of work and it's understandable. Unfortunately, it's much cheaper to just waggle the aerial about a bit amid mutterings of the mantra: "goddammit who invented this sh1t"

:ok:
Are you implying that a bunch of people with PhDs etc never mutter "who invented this sh1t"?
Oh Mike, they do, they do! :)

[ShyTorque]

My Wifey is also very odd and unpredictable; I've been having this trouble for years.

[Keef]

Like wot MikeB says, RF propagation follows simple, clearly-defined rules. I passed the theory exams in those when I was a nipper.

The trouble is, when you get up into frequencies in Gigahertz (like WiFi), the interactions that affect that propagation become incredibly complex. This is the point where you wave the aerial around a bit and enquire of the parentage of the inventor.

I can squirt RF all round the globe (and have done, frequently) - but there are a load of conditions that need to apply. Hardly any of those apply to WiFi. With WiFi, I work on the simple principle that it will percolate through about 50m of free air, or one stone wall, or two stud walls, or two glass windows (as long as the glass isn't metallised) - or any combination thereof pro-rata.

If you are lucky and the humidity is right and the electrons didn't get stoned last night, you may get some random better results than that. But that's not the way to bet.