RSSI fluctuations in SWX-SRC cards in 802.11a

[advait]

Hi,
I am advait.
I am using SWX-SRC card for my link characterization experiments. I am operation on channel 40, 6mbps data rate and using madwifi 0.9.4 drivers. I observed that the RSSI of the received packets fluctuates a lot even that is for a sequence of packets received first I get 4-5 packets to the correct expected value then all of a sudden the RSSI drops down by 15-20 dBm. Some what like this
-64
-64
-80
-64
-64
-64
-80
-63
-65
-64
-81
-62
-64
-64
-79
-62
-64
-65
Please keep in mind that these are the result of controlled experiments i.e with the use of RF cables and attenuator.
We also did the same for few indoor experiment and the result was same. We made sure that there is no interference(in 5 Ghz) around.

and the most important thing is that we observed this behavior across different cards of SWX-SRC as well as across few minipci SR5 cards. In all the experiments we transmitted around 4 pckts of size 1000 bytes in 1 second for 1 minute. We have also repeated this experiment for long durations.

So does that means that there is some hardware quirk......please suggest

[UBNT-keba]

For me it looks like it uses antenna diversity. Check madwifi driver sources related on antenna switching.

[advait]

I already have taken care or diversity at the time of experiments by setting
diversity to 0 and
txantenna to 1
rxantenna to 1

[UBNT-keba]

Have you took a look to madwifi driver code? As I remember there was some issues with diversity (default antenna switching even the diversity is off). Not sure how it looks like now.

[advait]

Ya it does so until diversity is set
if u do smthing like this

echo 0 > /proc/sys/dev/wifi0/diversity
echo 1 > /proc/sys/dev/wifi0/txantenna
echo 1 > /proc/sys/dev/wifi0/rxantenna

during I have also verified it from athstats this shows how many packets are being transmitted from every antenna


this disables the diversity

Infact if in case(lets assume) the diversity is on then in that case every alternate packet should be affected not every 1 packet per second.

[UBNT-Zy]

Hello advait,

We are sure that there is no hardware quirk.
As you said you had monitored "how many packets are being transmitted from every antenna" however the RSSI is calculated on received packets not the transmitted ones. The root cause is within the the driver code itself as it does not disable the diversity even after "echo" routines are initiated.

[advait]

We have infact tested the presence of these spurious packets using a power meter(A Agilent power meter) as a receiver. So that eliminates the impact of RSSI calculation at driver level.

If in case the problem is due to the diversity then that should cause every second(or alternate) packet to to have that effect. But contrary to this we are getting such packets one every 1 second.