Re: 802.11b/g amps

What a fun video, too bad its wrong in many places.

The "three non-overlapping channels" is a bit of a distraction.  Due to 
the low(er) ACR of 802.11g radios, you don't even get 3 non-interfering 
channels.  The 4-channel spacing never did work, even with good 11b 
gear.   You can probably fix the problem on the AP end, using channel 
filters, but the clients will interfere with each other, and your range 
and/or throughput will suffer.

The 802.11a slide says that there is an EIRP limit of 52dBm if PTP in 
U-NII3, which is almost true.  Its 53dBm (23dBi antenna gain and 30dBm 
of tx power) if operating under the U-NIII rules (47 CFR 15.407), and 
unlimited except for the laws of physics under the ISM rules (47 CFR 
15.247). The FCC places no limit on antenna gain in 5.725 - 5.850GHz 
for PTP links, so you could, in theory, run a 29dBi antenna attached to 
a 30dBm radio, for 59dBm EIRP.

52.0 dbm is 158.48W
53.0 dbm is 199.52W
59.0 dbm is 794.33W

Use a 29dBi antenna at the other end, and use 6-18Mbps modulation, 
where a good 802.11a chipset gets -89dBm rx sensitivity, and you can 
overcome 177dBm of path loss, modulo losses.  Even with 10dB of losses 
(connectors, antenna efficiency) and 10dB of fade margin, you're still 
able to overcome 157dBm of path loss.  It takes approximately 150 miles 
LOS to generate this much free space path loss at 5.8GHz.

Now, you probably don't want to build towers high enough to have free 
space between two radios located 150 miles from each other, so the 
result will typically be that you up the modulation rate.   You're 
going to need a phat PA at both ends to overcome the PAPR issues, but I 
find it quite likely that a well-engineered 802.11a system could get 
10-20 mile LOS links at 54Mbps.  Add a bit of software work to enable 
frame bursting, (or discard the 802.11 MAC altogether) and it could be 
possible to get real-world throughputs that are a substantial fraction 
of the modulation rate.  Maybe 35Mbps or more.

You'll also need to find a way to overcome the significant out of band 
emissions that this solution will generate, but its all possible, I 
tell ya.  Perhaps not commercially available, but *possible*.

Lets continue with the video.  You're not going to co-locate "eight 
802.11a APs" in a room in the lowest 2 U-NII bands and get 8X capacity. 
  You just won't.  The speaker sort-of explains 'why' without saying so. 
  To quote, "the remote receiver may be hit with a billionth of a watt" 
and still want to recover the signal.  If there is a sender on an 
adjacent channel, the ACR of most 802.11g radios at 54Mbps is much 
better than the IEEE spec, -1dBm  (By comparison, the Prism2/2.5 
chipset has 41dBm of ACR.)  This is one of the reasons why 802.11g 
networks are a completely different beast than 802.11b networks.  (One 
of the other reasons is that the CCA boundry is quite a bit 'farther' 
for 802.11g than it is for 802.11b.

Back to the video, the "license holders" (in 5GHz) are (ahem) "mobile 
platforms" (read: airplanes and ships) with RADAR.  Read "U.S. 
Military".  The last thing they want is for a high EIRP AP to blind 
their target RADAR after a lock.   Lord knows where the missile might 
go then.

OFDM does not "fall apart" at higher power levels, but it does require 
a LOT more of the PA (power amplifier), due to the higher Peak to 
Average Power level of OFDM modulation.  If you look at an OFDM signal 
on a spectrum analyzer that is run through a PA with insufficient TOI 
(or which has otherwise reached compression) it will appear to "bloom", 
so maybe this is what the speaker mean by "fall apart".)

Nor is the problem "chipset related", its a requirement of (more) 
linearity from the PA.   (M-codes have been proposed to reduce PAPR in 
OFDM, but I don't know of an 802.11 chipset that uses them, since they 
would make for a proprietary solution.)  Vivato engineered a 11g system 
that will output up to 29dBm (see the FCC docs) at the lower 
modulations, and IIRC, 24dBm at 54Mbps, and I believe that its now 
well-known that this 11g product from Vivato uses Atheros miniPCI 
cards.

The other issue is that most external PAs used with 802.11b equipment 
'sense' the transmitted signal from the radio and 'switch' the PA on.  
(If the PA is not on, the switch in this PA is in the 'receive' 
position.)   The issue here is one of timing.  The 802.11g/802.11a 
preamble is *MUCH* shorter than the DSSS preamble.  By the time the 
circuit detects RF energy from the radio, sets the switch, and the PA 
is 'on' and ready to transmit, too much of the preamble hasn't been 
amplified, (or, due to the rx/tx switch having not been set to 'tx' 
yet, hasn't even made it to the antenna!) and the (remote) receiver 
doesn't have enough information to properly demod the incoming frame.

Atheros has demoed 90Mbps of TCP throughput (using netperf) between two 
"turbo mode" APs.  This is data dependent (they use LZ compression for 
part of the gain) but the other benefit comes from the channel bonding 
("2 channels") as well as (nearly standard) 802.11e enhancements, 
including frame bursting.   The units don't need to be "3 feet" from 
each other, though.   BTW, the "2 channels" are 2 of the 802.11b/g 
channels (typically 4 and 5) not two of the non-overlapping channels 
(say, 1 and 6 or 6 and 11).  (Same thing in 802.11a, btw.)

There are other errors in the video, but I already have enough of a 
reputation as an assh*le, thanks.

Jim

On May 23, 2004, at 9:36 AM, Frank Keeney wrote:

>
> You'll find that 802.11g is not as ideal as "b" for amplification. We 
> had
> Cisco out to our SOCALWUG meeting last month
> (http://www.socalwug.org/april_22nd_2004.htm) and he discussed the 
> reasons
> why it's a problem. Streaming video is here:
> http://stream.lpbn.org:8080/ramgen/socalwug042204.rm
>
> Frank Keeney
> Tel: 888-259-5110 x0
>
> Wireless Antennas, Amplifiers, Cables and Equipment:
> http://www.wlanparts.com
>
>
>> -----Original Message-----
>>
>>
>> Anyone know of a site that has a list of the best and the least 
>> expensive?
>
> _______________________________________________
> BAWUG's general wireless chat mailing list
> [unsubscribe] http://lists.bawug.org/mailman/listinfo/wireless

_______________________________________________
BAWUG's general wireless chat mailing list
[unsubscribe] http://lists.bawug.org/mailman/listinfo/wireless