Re: [PATCH 2/2] ath9k_hw: improve ANI processing and rx desensitizing parameters

[Rajkumar Manoharan <rmanohar@qca.qualcomm.com>]

On Thu, Feb 02, 2012 at 02:09:10PM -0800, Adrian Chadd wrote:
> Hi, questions below:
> 
> On 1 February 2012 08:05, Rajkumar Manoharan <rmanohar@qca.qualcomm.com> wrote:
> >
> > This patch improves ANI operations by switching among the immunity
> > levels based on PHY errors and beacon rssi which will adjust receiver
> > desensitizing parameters. The changes are
> 
> What environments are you seeing this bad behaviour in? I'd just like
> to be sure that this won't mess up the behaviour of ANI for other
> chips, other environments and other operating modes.
>
The actual test was done in AP mode in controlled environment with rx signal
level of -50 dBm and with interference of -45 dBm at STA side. In such scenario
beacon miss was observered and drastic hit in throughput at OFDM ani level 9 after
enabling the Weak signal detection. Not changing the any of weak signal detection
thresholds and retaining the ini values improved the throughput at higher ANI
levels.
> > * Irrespective of opmode, the Weak Signal Detection is with the current
> >  immunity level value.
> 
Again with AP mode enabling/disabling the Weak signal detection is not working.
It was verified from the register (use_self_corr_low). So that it will
configure the WSD with current ofdm immunity level, if there is mismatch b/w
configured WSD state and selected immunity level's WSD state.

> I didn't think that weak signal detection was a single bit in the PHY.
> It looks like you've just deleted all the threshold settings and are
> turning on/off using low self correlation.
>
As menstion above, the thresh values are continue to retain the init values.
Based on the test results, it is enough to update use_self_corr_low state.

> Have a read of the (very early, AR5212 era?) ANI patent:
> 
> http://www.freepatentsonline.com/7349503.pdf
> 
> If this has changed (ie, if in Osprey it's done differently) then it
> should likely be documented in the driver somewhere.
> 
> IF this all works then great, but I'd really like to try and better
> understand/document the "why" aspects of it. ANI is .. fiddly, and
> there aren't any tools available for ath9k (+freebsd) wireless
> developers to really understand the impact of ANI.
>
--
Rajkumar