[ath9k-devel] Calibration data to print when Rx stucks

[Felix Bitterli <Felix.Bitterli@atheros.com>]

+nbd

Hi Senthil,

S/W should keep a history of measured NF values; S/W should scrap
unreasonable NF values, S/W should use the average of the history of
measured reasonable values and program that as the NF used by H/W. I
don't have a description of this, but I expect newma to be doing this.

Also Felix Fietkau ported this to ath9k a couple of months ago.

Best regards,
Felix(B)

On Tue, 2011-04-05 at 10:26 -0700, Senthilkumar Balasubramanian wrote:
> Thanks Felix. We shall try this tomorrow. also, who is the best person to ask regarding noise floor calibration. Our windows drivers seems to be using default noise floor (-95 dBm) for signal strength calculations and doesn't include the current channel noise floor and there were some comments in the code which says getChanNoise floor return values are not reliable and so use default noise floor.
> 
> Is there a document small README that explains noise floor calibrations.? With the recent patch that Felix has sent to wireless-testing, it would be an issue from support point of view as for the same distance, windows reports much higher signal strength and ath9k will report lower signal strength..
> ________________________________________
> From: Felix Bitterli
> Sent: Tuesday, April 05, 2011 9:46 PM
> To: Vasanth Thiagarajan
> Cc: Felix Bitterli; Senthilkumar Balasubramanian
> Subject: Re: Calibration data to print when Rx stucks
> 
> Hi Vasanth,
> 
> as first suggestion, use the watchdog registers, read attached document
> for interpreting these regs.
> 
> Please rename
>     "BB Panic Watchdog register"
> to
>     "BB Watchdog Timeout register",
> and make sure word "panic" does not exist anywhere.
> 
> Best regards,
> Felix
> --
> Vasanth Thiagarajan wrote:
> > Felix,
> >
> > We got a requirement from Sam (google) to print out some useful cal data
> > when there is beacon miss (possible Rx stuck). Other than NF values,
> > we are not sure about other cal_data/data which will be useful when there
> > is a possible rx stuck issue. Could you please point to data useful
> > in this condition to debug the issue?.
> >
> > Thanks,
> >
> > Vasanth
> >
>