Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email firstname.lastname@example.org
From: Brian Capouch (briancpalaver.net)
Date: Sat Jun 23 2001 - 14:57:59 CDT
I decided that I'm going to write up the complete, very long description
of this project and put it on the web, in case someone out there winds
up wanting to try to follow my progress sleuthing out this problem.
Even at that, this message is going to be longish. I hope it's worth
I am trying to set up a three-machine, P-T-P longish-haul network using
PCI342 cards. This is using Ben's driver under Linux 2.4.5(-pre6) on
low-end Pentium desktops. Site One and Site Two are about 1.2 miles
apart, and Site Three is 2 miles from Site Two.
We will limit our discussion to the present problem, which only involves
two of the machines, one attached to a 24db directional antenna and the
other to a 8db omni.
Before permanent installation, I put the omni on a 20' pole and leaned
it up against a building at Site One, then went to Site Two (just over 1
mile away, with bare line of sight) and set up the directional aimed at
the omni. Over a two-week period, in every kind of weather, I could
reliably get 5.5Mbs on this link and sustain it, doing things like
intense X windows activity to make sure it would hold up under sustained
load. At one time I used it intensely like this for over an hour,
without a bit of problem.
The test network mimicked the final installation in every respect except
that there is a 100+' grain elevator at the Site Two, so after testing
it was the omni we put up on the tower there and we moved the
directional over to Site One. We have a very short (12') LMR200 cable
on the omni, and a 25' foot LMR200 on the directional. This is actually
*less* cable length at the omni end than we had on the testbed.
Now the two sites cannot talk at all. I do not even show the reception
of a single beacon from the omni (using ACU) at the directional site,
and visa versa. I even moved the directional about 300 yards closer to
the omni, where this absolute total line of sight. Still nothing.
Now comes the convoluted part, which I relate in full because perhaps
some part of it may trigger a light bulb. Everything reported here has
been done repeatedly, with essentially the same results each time.
We boot the omni box and I put a second test machine underneath the
tower out about 100 yards, with only a "rubber duck" type antenna.
Using iwconfig to set up the interface for 1Mbs, there is immediate
sync, and I can then move to any supported speed and after a few seconds
I see beacons and things are fine.
However, in the 95% case, after about 10 minutes (+/- 2 min) of
continuous 1-sec pings at 11Mbs it freezes up. If I slow to 5.5M still
nothing, ditto for 2M, but once I get back to 1Mbs it comes right up,
and then I can walk upwards through the speeds and get back to 11M where
it will again run for about 10 minutes before repeating this behavior.
One time out of perhaps 20 tests it didn't freeze up (at least for an
hour when I ended the test) but every other time I've tried it this has
been the behavior. Note that once it freezes I have to go back to 1M
before it will sync at any other speed. No abnormal log messages, and
ACU doesn't show any received beacons after the freeze until I have
throttled back to 1M.
I move the test machine to Site One and put on the directional.
Nothing. No beacons at any speed, and moving the antenna through as
many permutations of "here" as I can think of yields nothing.
So now I go back to Site Two and find that I can no longer talk even
with the rubber duck antenna--at all, just like at Site One. For the
first three iterations of this test I figured that somehow the machine
up on the tower had gone down, so I rebooted it, and found that in each
case the behavior mentioned above was reproducible, starting with good
signal to a piddly antenna, which then went away after I had gone to
Site One, which has never yielded a bit of anything.
On the fourth go-around it struck me to see what would happen if I tried
the directional antenna right under the tower, and to my surprise, the
card isn't dead; it's just a lot pickier about signal, because after
careful aiming of the directional (note that: I have to set up the ping
and wander around with the directional aimed up at the tower) it will
sync up just fine and I can talk at any speed and there were (in two
tests) no lockups.
So today we figured the omni was either bad or underpowered, so we
replaced it on the tower with a 15db omni. To my great surprise (and
chagrin) the behavior is *exactly* the same: initially I get a good
signal with a rubber duck, but with frequent lockups. Then I take the
test system to Site One where nothing happens, and when I return to Site
Two now I have to use the directional again to get any signal out of the
(second) omni. And using the directional like this it is *extremely*
finicky about the placement of the directional; any slight movement once
it's set up and the link goes dead. Moving it back restores the link,
BUT ONLY IF I AGAIN START AT 1M and work back up.
All these problems are in such distinct contrast to the easy and
flawless connectivity I got when my setup was so much cruder. That is
the most disheartening aspect of this case.
Sorry for the length of this post. I don't have any sophisticated test
equipment and this is the first 2.4Ghz for the antenna person, so he
doesn't either. If anyone could give me some "try this" stuff I would
sure be obliged. I am going to continue to experiment and if anyone
shows any interest in being kept up-to-date on the situation I'll be
glad to send along what I find.
Thanks for your patience and any ideas that may come forth.
Aironet mailing list - Aironetcsl.cse.ucsc.edu