"Fred Marshall" <fmarshallx@remove_the_x.acm.org> hath wroth:
>Thanks for the reply. Good thoughts / insights. I'd not thought of the
>file transfer test. It all makes sense. Somehow I had thought that the
>signal strength reported by the softwarewas a "system" measure (i.e. 2-way)
>but I guess that was only a hopeful notion per your comment. I'll try it.
The Signal Strength and S/N ratio are reported by both the access
point and the client radios. Unfortunately, the DWL-3200AP firmware
scribblers elected to only supply the signal strength.
|
http://support.dlink.com/emulators/d...lientInfo.html
The S/N ratio is more important because it is used to set the speed of
transmission. The lower the S/N ratio, the slower the data rate. Of
course, Dlink also didn't include the speed. Sigh.
Anyway, some of the management packets include both the signal
strength and S/N ratio, which can be sniffed and extracted. Methinks
Kismet (Linux) is the appropriate tool. See:
|
http://www.remote-exploit.org/index.php/Auditor
>No, I didn't try it side-by-side. Thus my question.
It's difficult to guesstimate how far you can RELIABLY communicate.
The data sheet has a table of ranges:
802.11g
(Full Power with 5dBi Gain Diversity Dualband Dipole Antenna)
Indoors:
98ft (30m) @ 54Mbps
112ft (34m) @ 48Mbps
128ft (39m) @ 36Mbps
154ft (47m) @ 24Mbps
184ft (56m) @ 18Mbps
217ft (66m) @ 12Mbps
259ft (79m) @ 9Mbps
325ft (99m) @ 6Mbps
Outdoors:
367ft (112m) @ 54Mbps
820ft (250m) @ 18Mbps
1640ft (500m) @ 6Mbps
Methinks they're being a bit optimistic, but it does show what
obstacles (indoors) can do, and approximately the effects of speed
changes. At the range extremes, you're probably at 1 or 2 Mbits/sec
if you have that enabled.
>So, "a couple of blocks" is around 500 feet. I wasn't out there with a tape
>measure - just walking around the streets in an area where there is medium
>building density of 1-story wooden buildings (as in perhaps 70% occupied
>ground without counting the streets).
In other words, you had obstructions between you and the access point.
You probably also had reflections and definitely had some trees or
building with the Fresnel Zone. Sounds like a challenge.
>I should add:
>A couple of years ago there was a survey of coverage published with
>coverage test results from a number of mainstream commodity wireless routers
>(B or G) and D-Link stood out. Are there any similar, more recent, reports?
>Where might I find them?
The best reports I've seen are at:
http://www.practicallynetworked.com
The various reviews are VERY illuminating. Tim Higgins goes out of
his way to produce accurate reports using realistic conditions and
reproducible procedures. However, the various web pages are designed
to deliver advertising, so each report is chopped into 10-20 pages.
It's worth the effort slogging through them. Nothing on the
DWL-3200AP.
>I've not tried Netstumbler so I'm off to do that! In surveying a very high
>wireless site density in a small city area I found that the site survey
>software would not drop sites off the list as I moved away from them. That
>made it really hard to figure out where the sites were likely located and
>what sites were likely sources of interference.
Methinks Kismet would be better, but Netstumbler will suffice. Note
that there are several Signal Strength, S/N ratio, and noise level
numbers shown (scroll horizontally). These do NOT come from the
access point. They come from the NDIS5 driver on whatever laptop
you're using for testing. Therefore, an overly strong access point
xmitter will show up as a very strong signal with NetStumbler, while
the ability to communicate reliably is limited by the laptop xmitter,
which is not shown by Netstumbler.
>Regarding interference: I did find that setting the DWL-3200AP in an
>automatic mode for deciding the best channel at any time seemed to be *not a
>good idea* in a high-density location.
Yep, you've noticed. The big problem with the channel hopping scheme
is "what happens when there are 20 connected clients, the access point
changes channel, and some of them don't notice the change"? Well, the
one's that missed the channel change have to re-associate. Some
wireless clients do this very well (i.e. Intel Proset). Others suck.
>That's of course, the opposite of
>what I'd expected. It was creating lots of client dropouts. Performance is
>much better, and remains so today, to just manually select what appears to
>be a mostly unused channel and leaving it there. This is in that same
>high-wireless-density location.
Yep. You might also try fixing the wireless speed to 12Mbits/sec.
This locks out the 802.11b only clients, but is a good compromise
speed. I've found that disconnects and weird stalls are reduced (but
not eliminated) by fixing the speed.
>Do you find that most interference is from other 802.11x sites, other
>related wireless devices or totally unrelated device EMI? Or is it a mix?
Almost all the interference I see come from other access points or
municipal networks. These are fairly easy to see and find with
Kismet. The worst are glass wall office buildings with literally
hundreds of access points and clients spraying RF all over the view.
During lunch time, the same building also has plenty of leaky
microwave ovens running. I've also run into a few 2.4GHz wireless
security cameras. Non-802.11 interference is where I get involved. I
drag out the dish and spectrum analyzer, and do the transmitter
hunting thing.
I have seen some EMI, but usually it's very local. Switching power
supply wall warts sometimes generate lots of RFI. One particular
monitor had a huge field that affected any nearby electronics. See
the FAQ for a list of probable interference sources at:
|
http://wireless.wikia.com/wiki/Wi-Fi#Interference
--
Jeff Liebermann
jeffl@comix.santa-cruz.ca.us
150 Felker St #D
http://www.LearnByDestroying.com
Santa Cruz CA 95060
http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
Wireless Routers and Access Points 
Linear Mode
