h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

My home connection to the internet is a 30Mb connection (cable) with a wireless N router and wireless N connector on my computer. It's not that far away (prob only 10 meters or so) but there's a couple of walls and wiring and plumbing in the way. Plus I'm in a residential area with a few other wireless networks around (none that I can see of that strong a signal) and at least one CB style aerial a few houses up.

When I look at the download speed profile on something like Steam or just watch the connection speed I've noticed that it rarely just sits at one level. Indeed, it is far more like the top half of a sine wave.
It'll gradually increase to a max then drop down again, sometimes to nothing before picking up again. Very occasionally, if nothing else is running, and it has been going for a while, it may smooth out. Then again it might just drop out completely.

So my question is two fold:
Is this an artifact of the sampling process that produces the stats? Which is, after all, a sampling and averaging over time at key time points.
Or is it actually that the connection speed is rising and falling in the manner indicated?

And either way, why should this be so?
It does seem that connection speed drops with high transfer as the pc connector (usb) gets rather warm. Disconnecting, allowing it to cool, and reconnecting brings it back up again. So could it be some sort of temperature cycle such that high speed, warms up, reduces speed, so it cools again, so it speeds up etc. Or is that unlikely since these things are designed to run at such speeds.

Experiments:

Keep some kind of notes which make it easy to see performance changes.

Wireless involved at all:
Does the same thing happen with a straight cable connection to the router?
(ie could it be the website is 'lumpy' or that bandwidth available fluctuates when there's a lot of demand from other cable users)

Other local wireless networks (or other non-constant electronics) interfering:
Does the same thing happen in the small hours of the morning?

Temperature:
Does the same behaviour happen with a desk fan directed at the USB adapter?
Does the casing get warm in such a situation?

(Not necessarily conclusive - if insides aren't in good contact with case, cooling case may make relatively little difference to internal circuit temperature.)
Personally I'd have thought temperature would tend to lead to a work:fail case rather than 'work less well when warmer' one.

Do you have detailed information on the wireless connection (ie the max speed of the connection at any instant)? - is the max speed fluctuating as if there may be time-varying interference with the connection?

Cheers Potholer.

Unfortunately, wired connection test is not practical - router and cable connection is upstairs whilst computer is downstairs plugged into the telly.

Timing wise. Early morning (or late night) the connection strength picks up significantly (from 1/2 bars up to 4/5 bars) and the highest attainable downl;oad rate increases accordingly. But it still does the wave pattern. So instead of cycling between a max of 1.3mbps down to 200kbps (or zero) it'll cycle between 2.3mbps down to say 1mbps.

The casing doesn't really get warm - but the USB connector does.

Wireless connection strength fluctuates and but the reading on the connection speed seems to be a bit..erm.. random. It's a Cisco n connector (actually linksys) and from experimentation it seems to be piggybacking the windows zero config wireless stuff. Past connectors I had would use one or the other so you could turn off the windows zero but with this one, although it has its own drivers, if I turn off (disable) the windows one it re-enables it.

The reported connection speed is usually starts off in the high 200s to 300. It can then reduce to somewhere in the hundreds. When it starts getting problems it goes down to 6.5 or even 5.5 and I then have to reconnect it. Yet I can still get very good download rates even with a low connection strength and low connection speed. I have a suspicion that it isn't reporting the overall connection speed but just what is left over after windows zero config apps have taken their share. I do notice it drops when things are auto updating in the background.

I haven't checked max speed in real time though so that is the next logical step.