[Cu-wireless] status report; weekend meeting?

David Young dyoung at ojctech.com
Wed Oct 30 01:29:05 CST 2002

The growth of our network is temporarily stuck. I am bringing the
problem points to the group for your solutions. I suggest some possible
solutions, too.

We have 3 stations, the Indymedia Center's, Peter M.'s, and Sascha's.
Call them IMC, PM, and SM.  They are at 218 W. Main, 403 W. Green,
and 412 W. Illinois, respectively.

PM and SM are within about 100m of each other. IMC and PM are within
about 400m of each other. At IMC, there is broadband Internet, so we
are anxious to establish a reliable connection there.

At PM, we receive IMC very weakly, at -87dBm. The radio at PM only has
a receive sensitivity of -80dBm, meaning we must be getting zillions of
bit errors.

As you can imagine, we have a poor link between PM and IMC.  We may have
to improve the radio at PM, or we may have to shorten the distance by
installing a new station. I have a feeling that the link intermittency
could have its cause in software deficiencies, too.  I will explain.

Software deficiencies
-------- ------------

One software deficiency is that Prism firmware seems to choose a new
BSSID every 10 seconds when it does not hear any IBSS station using
the same SSID. So it could be that when IMC does not hear from PM for
a while, it chooses a new BSSID. This partitions the network. When it
hears PM's beacon again, it syncs with PM's BSSID. At least, it syncs
if it implements WECA's BSSID merge standard. I think I am seeing WECA
merges on the network, beause I do not know how else to explain IMC and
SM choosing the same BSSID. Anyway, it helps to explain interruptions
if IMC very often "gives up" on its old BSSID and chooses a new one.

Another software deficiency is that Prism firmware seems to switch
aggressively from low to high speeds. When it detects packet loss,
it *gradually* reduces speed to 1Mbps. I think that 1Mbps is the only
semi-reliable rate between IMC and PM, but if the firmware behaves as I
think it does, IMC will only operate at 1Mbps occassionally. That will
also help to explain the communications difficulty.

Another possibility is that the IMC station, which has a super-high
receive sensitivity, does not have a proportionally insensitive
clear-channel assessment, so it is a reticent transmitter. Peter Folk
tells me he thinks he sees this problem in his network.  If the problem,
at bottom, is a reticent transmitter, then that also helps to explain
why our link is intermittent.

Solutions: software
---------- --------

I prefer to solve software problems to hardware problems because it
saves cash, and because I am a "software guy." I have solutions in mind
for two of the software problems.

There is a command to set a fixed transmit rate for a Prism radio.
Our radios will not honor the command "out of the box," but a firmware
upgrade fixes that. I will upgrade the firmware at SM, PM, and IMC,
and then I will fix all the transmitters at 1Mbps.

It is possible that if the problem is a "reticent transmitter," then
there is a command I can write to change the clear-channel assessment
mode. If not, then maybe the programmer's manual documents a value in
the firmware that I can tweak.

Solutions: hardware
---------- --------

In the hardware department, I have also thought of a few things to try.

We can install a new station between IMC and PM, thus reducing the length
of PM's next hop to the IMC and the Internet. I like this choice because
it grows our network at the same time as it connects it to the Internet.
We have prospective station hosts at Race & Green and near Elm & Cedar,
but Elm & Cedar is a slim prospect, and Race & Green's sight of IMC is
probably blocked by several buildings.

We can install a more sensitive radio at PM. It will be a costly radio
(at least $100), which will require a costly pigtail (at least $25).
We will also need to pull a new cable and to work some on the roof.

We can also choose a new Internet uplink and hop there through new
stations. That will prolong the Internetlessness.

Summary & Action Items
------- - ------ -----

We are close. We need another station in a sweet location, an improved
radio, or some software tweaks, and then we are on the Internet. We
may as well try all the software solutions, but we should also seek new
station hosts. If all else fails, we will have to improve the connection
with our choice of radio.

Can anyone come to a meeting on Sunday afternoon at 1pm, to try to
debug the network we have now, and to make some decisions affecting the
network's growth?


David Young             OJC Technologies
dyoung at ojctech.com      Engineering from the Right Brain
                        Urbana, IL * (217) 278-3933

More information about the CU-Wireless mailing list