[Ninux-day] Diversity chip on Linksys, Redux

Mitar mitar at tnode.com
Sun Apr 25 05:55:41 CEST 2010


Hi!

This is getting funnier and funnier. But I do not have an explanation
for this.

So I have replaced Linksys with the new one. And attached to one of the
antenna ports (numbered 0 in wlc) external omni antenna (nothing big).
To the other port I connected original Linksys antenna. And everything
is on the roof. An old picture (without second antenna and without new
more strait antenna cable):

http://wlan-lj.net/wiki/Tocke/Solar

But what I noticed is that with original antenna connection is better
(more throughput if I do a data transfer over the link). And with our
system we plot graphs of average of all LQ in ILQ values reported by
OLSR for all links to peering nodes of this node through time. So I
switched from internal antenna to external antenna again to see how it
looks on the graph.

And I noticed that average LQ in ILQ switched around. Interesting. Then
I tried setting one antenna as receiving only and the other one as
sending only and the result was that both LQ in ILQ dropped. Then I
inverted and both got high. And ETX got high. So if I use for receiving
external antenna and for sending original Linksys antenna I get the best
result. How is this possible? (I have attached graph.)

The other thing I am wondering is about LQ in ILQ values. If LQ is good
when I have rxant and txant set to 1 and ILQ is good when rxant and
txant is set to 0, then if LQ means receiving quality and ILQ means
sending quality (how much peers receive) then best combination should be
rxant=1 and txant=0. But graph tells it should be just opposite.

We have went through our code and are almost sure that we do not switch
LQ in ILQ around somewhere. So is there a bug in OLSR (I doubt that in
this case) or is there something else going on?


Mitar
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rxant-txant.png
Type: image/png
Size: 19198 bytes
Desc: not available
URL: <http://ml.ninux.org/pipermail/ninux-day/attachments/20100425/90dbb957/attachment-0003.png>


More information about the Ninux-day mailing list