AR9280 reported signal strengths are flat (comparison with bcm4322 included)

AR9280 reported signal strengths are flat (comparison with bcm4322 included)

[George Nychis]

I am experiencing unexpected signal strength values reported by an
AR9280.  Basically, they always tend to be flat no matter how far or
close I am from the device that is transmitting.  Roughly between the
-65 and -55dBm range on average.  The exact AR9280 I am using is:

     Sony UWA-BR100 802.11abgn Wireless Adapter [Atheros AR7010+AR9280]

I have compared this behavior to a bcm4322 and the behavior I see with
the bcm4322 is as expected.  Far away I get low signal strength
values, and close to the device I get high signal strength values
(-20dBm).

Here is a comparison of the two (taken at the same time) as I walk
closer to the transmitter.  I record the signal strength from
wireshark using radiotap header field radiotap.dbm_antsignal.

http://www.flickr.com/photos/gnychis/7203437510/in/photostream

Surprisingly, the average signal strength seems to decrease with the
ar9280 as I get closer, yet remains relatively flat.  The bcm4322
shows a clear increasing trend as I get closer.

Does anyone have any insight as to what might be going on?  I'm using
the most recent firmware and compat-wireless as of May 14, 2012.

Thanks!
George

Re: AR9280 reported signal strengths are flat (comparison with bcm4322 included)

[Holger Schurig]

This is probably due to ANI (adapative noise immunity). I think ath9k
does too much here, in other words: it doesn't just kick in if there
are too strong signals there. Instead it seems to always adjust the
input attenuator, needed or not.

Re: AR9280 reported signal strengths are flat (comparison with bcm4322 included)

[George Nychis]

Hi Holger,

Thanks a bunch for your response!  I had read about ANI before on the
list, it would be interesting to know if this is the cause for this.
It definitely seems plausible. I will dig around the ath9k code and
see if I can narrow down this to the cause.


On Wed, May 16, 2012 at 2:31 PM, George Nychis <gnychis@gmail.com> wrote:
>
> Hi Holger,
>
> Thanks a bunch for your response!  I had read about ANI before on the
> list, it would be interesting to know if this is the cause for this.  It
> definitely seems plausible. I will dig around the ath9k code and see if I
> can narrow down this to the cause.
>
> - George
>
>
> On Wed, May 16, 2012 at 2:08 AM, Holger Schurig
> <holgerschurig@googlemail.com> wrote:
>>
>> This is probably due to ANI (adapative noise immunity). I think ath9k
>> does too much here, in other words: it doesn't just kick in if there
>> are too strong signals there. Instead it seems to always adjust the
>> input attenuator, needed or not.
>
>

Re: AR9280 reported signal strengths are flat (comparison with bcm4322 included)

[George Nychis]

BTW- is ANI _not_ applied on broadcast traffic?  I notice that
monitoring the signal strength of beacons does not show this behavior.
 Maybe not applied for localization reasons?

- George


On Wed, May 16, 2012 at 2:32 PM, George Nychis <gnychis@gmail.com> wrote:
> Hi Holger,
>
> Thanks a bunch for your response!  I had read about ANI before on the
> list, it would be interesting to know if this is the cause for this.
> It definitely seems plausible. I will dig around the ath9k code and
> see if I can narrow down this to the cause.
>
>
> On Wed, May 16, 2012 at 2:31 PM, George Nychis <gnychis@gmail.com> wrote:
>>
>> Hi Holger,
>>
>> Thanks a bunch for your response!  I had read about ANI before on the
>> list, it would be interesting to know if this is the cause for this.  It
>> definitely seems plausible. I will dig around the ath9k code and see if I
>> can narrow down this to the cause.
>>
>> - George
>>
>>
>> On Wed, May 16, 2012 at 2:08 AM, Holger Schurig
>> <holgerschurig@googlemail.com> wrote:
>>>
>>> This is probably due to ANI (adapative noise immunity). I think ath9k
>>> does too much here, in other words: it doesn't just kick in if there
>>> are too strong signals there. Instead it seems to always adjust the
>>> input attenuator, needed or not.
>>
>>

Re: AR9280 reported signal strengths are flat (comparison with bcm4322 included)

[George Nychis]

Well, I believe that I forced ANI to be disabled and I still see this
same behavior.

To disable ANI on my AR9280, I added this to __ath9k_hw_init() in hw.c
  if (AR_SREV_9280(ah))
    ah->config.enable_ani = false;

I believe this works because in ath9k_hw_post_init() I have checked
that ah->config.enable_ani is set to false.  Therefore,
ath9k_hw_ani_setup() and ath9k_hw_ani_init() are never called.

Unfortunately, I still see this same flat line trend without strong
signal strengths near the device.


On Wed, May 16, 2012 at 2:35 PM, George Nychis <gnychis@gmail.com> wrote:
>
> BTW- is ANI _not_ applied on broadcast traffic?  I notice that
> monitoring the signal strength of beacons does not show this behavior.
>  Maybe not applied for localization reasons?
>
> - George
>
>
> On Wed, May 16, 2012 at 2:32 PM, George Nychis <gnychis@gmail.com> wrote:
> > Hi Holger,
> >
> > Thanks a bunch for your response!  I had read about ANI before on the
> > list, it would be interesting to know if this is the cause for this.
> > It definitely seems plausible. I will dig around the ath9k code and
> > see if I can narrow down this to the cause.
> >
> >
> > On Wed, May 16, 2012 at 2:31 PM, George Nychis <gnychis@gmail.com> wrote:
> >>
> >> Hi Holger,
> >>
> >> Thanks a bunch for your response!  I had read about ANI before on the
> >> list, it would be interesting to know if this is the cause for this.  It
> >> definitely seems plausible. I will dig around the ath9k code and see if I
> >> can narrow down this to the cause.
> >>
> >> - George
> >>
> >>
> >> On Wed, May 16, 2012 at 2:08 AM, Holger Schurig
> >> <holgerschurig@googlemail.com> wrote:
> >>>
> >>> This is probably due to ANI (adapative noise immunity). I think ath9k
> >>> does too much here, in other words: it doesn't just kick in if there
> >>> are too strong signals there. Instead it seems to always adjust the
> >>> input attenuator, needed or not.
> >>
> >>

Re: AR9280 reported signal strengths are flat (comparison with bcm4322 included)

[George Nychis]

Might anyone else have some insight here? I've tried digging through
the code to see if AGC is not properly taken in to consideration
during the RSSI calculation, but I think this is done in the firmware
which is out of view of my eyes.

On Wed, May 16, 2012 at 11:36 PM, George Nychis <gnychis@gmail.com> wrote:
>
> Well, I believe that I forced ANI to be disabled and I still see this
> same behavior.
>
> To disable ANI on my AR9280, I added this to __ath9k_hw_init() in hw.c
>  if (AR_SREV_9280(ah))
>    ah->config.enable_ani = false;
>
> I believe this works because in ath9k_hw_post_init() I have checked
> that ah->config.enable_ani is set to false.  Therefore,
> ath9k_hw_ani_setup() and ath9k_hw_ani_init() are never called.
>
> Unfortunately, I still see this same flat line trend without strong
> signal strengths near the device.
>
>
> On Wed, May 16, 2012 at 2:35 PM, George Nychis <gnychis@gmail.com> wrote:
> >
> > BTW- is ANI _not_ applied on broadcast traffic?  I notice that
> > monitoring the signal strength of beacons does not show this behavior.
> >  Maybe not applied for localization reasons?
> >
> > - George
> >
> >
> > On Wed, May 16, 2012 at 2:32 PM, George Nychis <gnychis@gmail.com>
> > wrote:
> > > Hi Holger,
> > >
> > > Thanks a bunch for your response!  I had read about ANI before on the
> > > list, it would be interesting to know if this is the cause for this.
> > > It definitely seems plausible. I will dig around the ath9k code and
> > > see if I can narrow down this to the cause.
> > >
> > >
> > > On Wed, May 16, 2012 at 2:31 PM, George Nychis <gnychis@gmail.com>
> > > wrote:
> > >>
> > >> Hi Holger,
> > >>
> > >> Thanks a bunch for your response!  I had read about ANI before on the
> > >> list, it would be interesting to know if this is the cause for this.
> > >>  It
> > >> definitely seems plausible. I will dig around the ath9k code and see
> > >> if I
> > >> can narrow down this to the cause.
> > >>
> > >> - George
> > >>
> > >>
> > >> On Wed, May 16, 2012 at 2:08 AM, Holger Schurig
> > >> <holgerschurig@googlemail.com> wrote:
> > >>>
> > >>> This is probably due to ANI (adapative noise immunity). I think
> > >>> ath9k
> > >>> does too much here, in other words: it doesn't just kick in if there
> > >>> are too strong signals there. Instead it seems to always adjust the
> > >>> input attenuator, needed or not.
> > >>
> > >>