Re: Dropouts, dropouts and more dropouts!

jaypeecee <jaypeecee.2z9nt7@no-mx.wirelessforums.org> hath wroth:

>This is my first posting on this forum and I'm sorry it's rather
>lengthy. From the title above, it's clear why I'm asking for
>suggestions.

Detail is good, but I'm not thrilled with you spending more words
describing the title, than simply duplicating it. I have my
newsreaders setup to NOT display the title, so it's an extra step for
me to re-read the title.

>For several months, I had a reliable wireless connection.

Good. That means that it was working. The trick is to see what has
changed.

>Just recently, it's all changed. I get a lot of dropouts, particularly
>when I click to go to a new web page.

OK. Time to do a simple test. You didn't mention anything about your
wireless client, it's hardware, or it's operating system, so I'll
assume a desktop running XP Home.
Start -> Run -> cmd <enter>
ping -t 192.168.2.1 (the IP address of your wireless router)
You should see continuous pings, with a very consistent 1-2 msec
latency. However, if you get something like this mess:
Reply from 192.168.1.1: bytes=32 time=76ms TTL=128
Reply from 192.168.1.1: bytes=32 time=1ms TTL=128
Reply from 192.168.1.1: bytes=32 time=8ms TTL=128
Reply from 192.168.1.1: bytes=32 time=1ms TTL=128
then you are experiencing packet loss and retransmissions. Try this
with various wireless clients to eliminate the possibility that the
packet loss is being caused by some program (i.e. worm or virus)
causing the client to become busy.

>I carefully monitor other networks
>in my neighbourhood using WirelessMon.

WirelessMon is a good program but it doesn't show "hidden" networks
that do not broadcast their SSID. They also do not show MIMO routers
and a variety of non-802.11 sources. See list at:
<http://wireless.wikia.com/wiki/Wi-Fi#Interference>
Windoze based sniffers are a good start but you'll see the missing
SSID's in addition to client radios using Linux based Kismet. Download
the Backtrack 2 Linux LiveCD from:
<http://www.remote-exploit.org/backtrack.html>
Make sure you have a compatible wireless card. Boot it on your
laptop. Run Kismet.

>Right at this moment, there are
>no less than eight networks visible. I'm using channel 6 and there is a
>network on channel 7 but its signal strength is very low.

Please note that the direct sequence signal is about 22MHz or 5
channels wide. Therefore, a signal on Ch 6 will occupy Ch 4 - 8. To
be sure that you're clear to use Channel 6, Ch 4 - 8 must also be
clear. That's also why the recommended non-overlapping channels are
1, 6, and 11 as any channel in between will receive interfernce from
both of the other channels.

>One of the
>things I like about WirelessMon is that it gives signal strength in dBm,
>which allows meaningful comparisons to be made between networks.

Too bad it's rather inaccurate. Same with their tx power output
readings. It's not the fault of Passmark, but is more of a design
problem. The dBm reading is based on the RSSI value of 0-255 or
roughly 10*log(RSSI). That gives fairly good resolution at the low
end, and large granularity at the high end. In addition, the current
all digital wi-fi chips are not very accurate or linear at determining
the RSSI. It's certainly good enough for comparing gains and signal
strengths for a given radio, but I would not try to use the RSSI for
absolute measurements or for comparing wireless clients.

>So, the
>received signal strength from my router (BT Voyager 2091) hovers around
>-47dBm (on ch 6) and the channel 7 network is less than -81dBm. The only
>other channels that are in use at present are ch 1 and ch 11.

-47 is very strong. -81 is barely detectable. However, the signal
strength doesn't matter with interference. Even with a strong signal,
interference can cause problems by filling in the gaps between packets
with junk and trashing the ACK's.

>From the information above, it would be reasonable to assume that I
>should not get interference from other networks.

Wrong. You're not getting interference from networks that you can see
with Wirelessmon. It won't show hidden nodes. Turn off SSID
broadcast on your router and WirelessMon will not see it.

>So, there are two
>possibilities that come to mind. Firstly that there may be interference
>from other 2.4GHz sources or that the received signal of -47dBm at my
>desktop network card (Belkin F5D6001) is too low to maintain a reliable
>connection.

There are plenty of other possibilities. However, I don't want to
list them in detail until you do a few simple tests and updates.
1. Update the firmware on your unspecified model wireless router to
the latest available. Disclosing the maker and model would also be
nice.
2. Disconnect the router. Take it and your wireless desktop to some
place where wi-fi cannot penetrate. A basement or closed room without
windows will suffice. Run the ping test and use IPerf to test for
thruput.

If that works as expected, then your hardware is clean and not the
source of the problems. If that fails, then there's something broken
in either the router or the client computah. You can determine which
by substitution with known working units.

IPerf is a thruput test that will require an additional wired (not
wireless) computer to act as a server.
<http://dast.nlanr.net/Projects/Iperf/>
I've posted instruction on how to do the thruput test at least 4
times, so you should be able to find it with Google Groups Search.

>Unfortunately, I don't have a Wi-Spy 2.4Ghz sniffer so I
>can't eliminate the first possibility.

I have the original Wi-Spy. It's lack of sensitivity (inherent
because of the lack of about 10dB of processing gain for spread
spectrum) will only show very strong source of interference. If
possible, borrow a real spectrum analyzer and big dish antenna from
one of the local ham radio operators.

>And I don't know what signal
>strength your average PCI card needs for a good connection. Does anyone
>have an idea?

Yes. The base sensitivity varies with connection speed. Worst case
is 54Mbits/sec which usually runs about -70dBm for a rather lousy BER.
At least a 10dB fade margin is usually required so the minimum usable
signal is about -60dBm. See graph at:
<http://802.11junk.com/jeffl/rx-sens/receiver%20sensitivity.htm>
Incidentally, everyone lies about their rx sensitivity numbers so
don't treat these numbers as accurate or even close. Many of them are
just the chipset sensitivity, without any concern for what goes
between the antenna connector and the chip.

Overload level varies radically with chipset. I have some numbers,
but I'm not sure they're any better than the manufacturers. My guess
is anything over about -20dBm will begin to cause distortion and
corresponding errors. Some chipsets overload at -30dBm. Some of the
really good outdoor routers can handle up to +10dBm. Besides being a
bad guess, the overload point is not an abrupt reference, but rather
the start of slow deterioration. It may take 20-30dB more signal
before anything worse than a few retransmissions occurr.

>I could, of course, buy a new router (of the MIMO variety)

MIMO will buy you more speed. It will NOT buy you more range or more
interference immunity. There are also two types of MIMO. The beam
steering or beam forming variety (RuckusWireless), will have some
benifit is that it can insert a null in the direction of the
interference source.

>but I can't be sure that it will solve the dropout problem. And maybe
>there's something else that I'm overlooking.
>
>Suggestions please.
>
>JPC

--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558