[Prism54-users] Field report - performance at a distance

[Jeff Bobzin]

--10 Second Drops Revisited--

I know some people probably thought I was crazy when I did a post here =
that reported seeing bursts of dropped packets at exactly 10 second =
intervals. Well the performance diagrams in this article from =
SmallNetBuilder shows this same behavior.

 (article is http://www.smallnetbuilder.com/Sections-article43-page1.php =
)
 (diagram is =
http://www.smallnetbuilder.com/myimages/NTK/11gntk_intersil_newfw_vs_wrt5=
4g.jpg )

Apparently this is characteristic of Intersil  firmware but does not =
show up in Broadcom-based devices. I wonder what the firmware is doing =
every 10 seconds - maybe housekeeping?

--Client Performance at Distance--

I was reviewing this great article for tips on my broader concern - =
performance at a distance.  I get great bandwidth at line-of-sight =
distances but it falls off dramatically through a single wall.  I have =
been able to compare prism54-based units with windows-based units =
connecting to  a standard SMC "G" wireless router at the other end of =
the building.  The windows driver reported 40% signal, the prism54 =
reported  minus 60db.

It is difficult to get consistent test results as the signal fades =
unpredictably,  but I don't see any consistent difference between the =
drivers.  It does seem as if the prism-based unit is more prone to drop =
connection and has problems reconnecting.

Control of connection is in the hands of the firmware which is the same =
between units. But there may be private configuration IOCTLs that the =
windows driver authors had access to configure draft-G performance =
variables. Its possible private variables may influence decisions to =
drop connection on errors. This is pure speculation though.

--AP Performance-

Comparing a Prism54-based AP vs SMC gives the same signal strength to =
clients located at distance. Clients have a lot of trouble connecting =
though. It is difficult to directly compare performance because the SMC =
Router has the diversity dual antenna. Doing 'ping -A' reveals similar =
transmission times with periodic jumps in time (performance drops) for =
both.  The periodic performance loss seems a little worse connected to a =
prism54 AP. Updating the SMC router to the latest firmware (nitro!) =
seemed to make it more prone to connection drop and more similar to =
prism-ap.

--Hawking Antennas--=20

I am also testing the Hawking Omni and Directional High Gain antennas.  =
Iwconfig does report a numerically stronger signal but the connection =
drops continue. Of course better antennas collect more noise too.   I =
can give a more complete report at a later date if desired.

--Conclusions--

The only suggestion I can give the developers is to investigate adding =
more private IOCTLs - if we can get the current owners of Intersil to =
reveal them. Certainly there must be several additional firmware =
controls for master mode.  The g-only or g+b control for example.

Also - I am open to a suggestion to move this discussion to a forum,  =
perhaps called "Performance & Tips",  if the mailing list manager =
desires.

--Jeff