[LUAU] wlan question
Jim Thompson
jim at netgate.com
Sun Jul 2 00:32:54 PDT 2006
There is no linux or Hawaii content in this message. Hit the Delete
key now if you don't care about a deep dive into WiFi's physics.
On Jul 1, 2006, at 5:20 PM, Tim Newsham wrote:
>> What will help is selecting the "strongest" AP, and then getting
>> your AP on the SAME CHANNEL.
>>
>> Try it, if you really want the math behind why it works, let me
>> know, but right now I'm remote without an AC adapter, so I
>> literally can't afford the time to write it down.
>
> I'm definitely interested in the "why." Btw, I gave it a shot. I'm
> still getting a bit of loss (about 35% right now, but I'm a bit
> further away than earlier). I think overall the results are on par
> with using a "free" channel that is two away from any nearby
> broadcaster.
OK, many people think that "WiFI" has "3 non-overlapping channels" in
the 2.4GHz ISM band.
In terms of transmit mask (the shape of the emissions envelope) this
is true. An 802.11b transmit mask is 22Mhz wide, and an 802.11g
(OFDM) signal mask is 20MHz wide. 802.11's "channels" are 5MHz
apart, so a 5 channel separation (1, 6, 11) results in frequency
centers spaced 25MHz apart.
But (and this is the tricky part) in terms of the selectivity of the
radio, there is but one channel in most real-world conditions!
The IEEE 802.11b (and below) minimum "adjacent channel
rejection" (channel 1 on channel 6, 6-on-1, 11-on-6, or 6-on-11) is
35dBm.
This means that, say, your AP running on channel 6 will "see" a
signal arriving on channel 11 35dBm "down" from what it would if it
were receiving on channel 11. Most chipsets used back in the day
(Aironet, Prism2, Orinoco/Lucent/Agere) were good for about 41dBm.
This is 6dB better than the spec. 6dB is "4X".
There is no "spec" for alternate channel (1 on 11, 11 on 1) rejection
in 802.11b, but the measured performance is about 20dB higher than
the ACR spec.
Now, having your radio suppress the signal from an adjacent channel
by 35dB is a good thing. Since the minimum sensitivity required to
recover an 802.11b signal (at a frame error rate of 10% for 1024 byte
packets) is -76dBm. 35dBm less than this is -111dBm, which is at
least 6dB below the noise floor for the frequency and signal
bandwidth used.
This made deploying 802.11b fairly straight-forward. The APs
wouldn't interfere with each other if located at the edges of their
(11Mbps) coverage area. A client could, occasionally interfere with
the operation of an AP, but typically the CCA function of 802.11
would keep this nearby client silenced while some STA (or AP) was
receiving a packet.
The sh*t just worked.
Enter 802.11g.
With 802.11g, you typically get a lot less. The IEEE *minimum* ACR
figure @ 6Mbps is 15dBm, and at 54Mbps is -1dBm.
If "15" sounds fine to you, note that this is a) the lowest
modulation rate and b) 100X worse than 802.11b. The alternate channel
rejection called for by the IEEE standard is about 16dBm better,
across the board.
For those of you not familiar with 'dB', remember that every 3dB is
double (if added) or 'half' (if subtracted). 10dB is "10X", 20dB is
"100X" and 30dB is "1000X".
"Why did this happen?", you ask... Well, radio architectures
changed. The first generation 802.11b cards (and those that proceed
them) were a "super-heterodyne receivers, and employed filters both
at the I/F (typically a SAW filter) and at the baseband (typically
just an anti-aliasing filter).
Because we all love cheap WLAN cards, the chip companies developed
"direct conversion" receivers, and these have no place to put a SAW
filter. Further, the SAW filters cost money, and in the head-long
race for a sub-$20 BOM (its not well below $10) for an 802.11g card,
a $1.10 SAW filter became evil incarnate.
Since SAW filters are electromechanical devices, there is little
chance that they can be made cheaper, ala the ever-decreasing price
of silicon gates, so they went bye-bye. Forever.
Now note that the minimum sensitivity (FER = 0.10, 1024 byte PSDUs)
for an 802.11g signal at 54Mbps is -65dBm, and at 6Mbps is -82dBm.
Redo the math and note that, given the -1dBm ACR spec @ 54Mbps, any
two signals arriving at a given STA, from the same range (distance)
*will* obliterate each other, even on adjacent channels.
Even at 6Mbps (min ACR: 16dBm, min RX sens: -82dBm) operation on an
adjacent channel results in effective degradation of the NF given
that the interferer is at the same range or closer.
Finally, power levels in 802.11g cards tend to be lower, especially
as the modulation rate goes up, due to issues with Peak to Average
Power Ratio. (The higher rate modulations can result in more
"carriers" peaking at the same time, and this means that either the
manufacturer must spend more money on a more linear PA, or spend more
money (and increase power consumption) for a "bigger" PA. In the
world of ultra-competitive BOMs, this isn't going to happen in the
primary markets (those who build miniPCI cards for OEMs to use in
laptops), which is ... one of the reasons that Netgate has a business.
Anyway, power levels coming out of the card (and AP) are lower. As
the modulation rate goes up, as much as 8dBm lower than a typicaly
802.11b AP or card.
So:
the power leaving your laptop (and the AP) is lower
the required signal strength (min rx sensitivity) required to
demodulate the signal is higher (especially at the higher modulation
rates)
decreased ACR means a higher likelyhood that a signal on another
channel lowering the SINR enough to cause you to not be able to
demodulate the signal.
Remember that even in LOS conditions, the "path loss" at 2.4GHz is
41dB in the first meter, approximately 60dB at 10m, and 80dB at 100m.
The signal leaving your laptop (which has no effective antenna gain)
is approximately 15dBm (may be lower at higher modulation rates).
At 10m it is 60dB lower, so -45dBm. At 20m this falls to about
-51dBm and at 40m its around -57dBm. (Note the "inverse square"
law. Doubt the distance and you get 1/4 (-6dB) the signal.)
The AP must be able to decode at 54Mbps signal down to -65dBm, so you
would be good to go here except for the operation of the other APs
(and their clients) and a couple other technical details that I won't
mention here.
I've said roughly similar things elsewhere online for years now.
Just google my name and "ACR" or "sensitivity" and they should pop
up. Also, if the APs (and/or clients) are operating "closer" than 5
"channels" apart, then the situation is a LOT worse, because one
shoulder of the transmit mask overlaps the original signal.
> It's possible that this isn't an interference issue, but that seems
> to be the most logical explanation so far (unless someone knows of
> interop issues between dlink and various laptop builtin drivers,
> such as intel 3945abg).
Sometimes when I forget and knock the Airport (extreme) on top of the
cable modem, it eventually gets hot enough that its performance goes
all to heck. A simple power cycle (and removing it from the top of
the cable modem) serves to restore things.
Might check that your AP isn't getting hot.
Jim
>> Jim
>>
>> On Jul 1, 2006, at 10:50 AM, Maddog wrote:
>>
>>> You might try a third party firmware. There are several out
>>> there. I know that fixed an issue i had with my Linksys router
>>> and it added so many cool new features. Google for third party
>>> firmware for your model. There are quite a few hacks sprouting up
>>> out there for D-link too.
>>> MD
>>> ----- Original Message ----- From: "Tim Newsham" <newsham at lava.net>
>>> To: <luau at lists.hosef.org>
>>> Sent: Saturday, July 01, 2006 10:39 AM
>>> Subject: [LUAU] wlan question
>>>> I've been having some wlan issues here off and on for a while
>>>> and it seems like it has been getting worse. Where I'm at I can
>>>> see about about 7 APs, including my own. Sometimes a few more,
>>>> sometimes a few less. Typically the other APs are on channels
>>>> 2,6 (several), 10 and 11 (several). I have tried various
>>>> channels trying to avoid the crowds, usually 8, 9 or 4 (since
>>>> the 6's and 11's are closest to me, at least in terms of signal
>>>> strength). I am sitting about 8 feet from my AP right now and
>>>> still getting 15-25% packet loss! I'm receiving about -57db from
>>>> my AP and -79db from the nearest competing AP, so its not clear
>>>> to me that the interference is necessarily coming from other
>>>> APs. One other thing to note, sometimes I start getting a lot
>>>> of drops (usually when I'm further away from my AP) and
>>>> resetting the AP (d-link, 802.11g) seems to "fix" the problem
>>>> for a while.
>>>> Any ideas on what might be causing the poor performance? Is
>>>> this normal performance for this amount of congestion? Is this
>>>> due to a cheap AP? Is there anything I can do to track down the
>>>> interference or make adjustments that will increase my reliability?
>>>> Tim Newsham
>>>> http://www.lava.net/~newsham/
>>>> _______________________________________________
>>>> LUAU at lists.hosef.org mailing list
>>>> http://lists.hosef.org/cgi-bin/mailman/listinfo/luau
>>> _______________________________________________
>>> LUAU at lists.hosef.org mailing list
>>> http://lists.hosef.org/cgi-bin/mailman/listinfo/luau
>>
>> _______________________________________________
>> LUAU at lists.hosef.org mailing list
>> http://lists.hosef.org/cgi-bin/mailman/listinfo/luau
>>
>
> Tim Newsham
> http://www.lava.net/~newsham/
> _______________________________________________
> LUAU at lists.hosef.org mailing list
> http://lists.hosef.org/cgi-bin/mailman/listinfo/luau
More information about the LUAU
mailing list