[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Adrian Chadd]

Hi,

FreeBSD's AR9160 support enables all three RX chains in legacy mode.
The ath9k codebase doesn't - it calibrates all three chains, but the
ADC calibration results in legacy mode give all 0x0's for chain 1 and 2.
(And seemingly invalid results for chain 0 in 11g mode, but that's another
issue which I am hoping someone will comment on.)

Is there any benefit or detriment to leaving RX chains 1 and 2 enabled
when in 11a/11bg mode? Does the unit do any kind of multipath
discrimination?


Adrian

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Luis R. Rodriguez]

On Sat, Sep 25, 2010 at 08:14:43PM -0700, Adrian Chadd wrote:
> Hi,
> 
> FreeBSD's AR9160 support enables all three RX chains in legacy mode.
> The ath9k codebase doesn't - it calibrates all three chains, but the
> ADC calibration results in legacy mode give all 0x0's for chain 1 and 2.
> (And seemingly invalid results for chain 0 in 11g mode, but that's another
> issue which I am hoping someone will comment on.)
> 
> Is there any benefit or detriment to leaving RX chains 1 and 2 enabled
> when in 11a/11bg mode? Does the unit do any kind of multipath
> discrimination?

Good question, not sure, but I can see the benefit of enabling other
chains for selection diversity for legacy, perhaps we just never
bothered implementing that on ath9k on Linux.

  Luis

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Felix]

On Mon, 2010-09-27 at 11:19 -0700, Luis R. Rodriguez wrote:
> On Sat, Sep 25, 2010 at 08:14:43PM -0700, Adrian Chadd wrote:
> > Hi,
> > 
> > FreeBSD's AR9160 support enables all three RX chains in legacy mode.
> > The ath9k codebase doesn't - it calibrates all three chains, but the
> > ADC calibration results in legacy mode give all 0x0's for chain 1 and 2.
> > (And seemingly invalid results for chain 0 in 11g mode, but that's another
> > issue which I am hoping someone will comment on.)
> > 
> > Is there any benefit or detriment to leaving RX chains 1 and 2 enabled
> > when in 11a/11bg mode? Does the unit do any kind of multipath
> > discrimination?
Leaving all RX chains enabled will give a huge gain for all OFDM rates,
they get MRC combined
(http://en.wikipedia.org/wiki/Maximal-ratio_combining) in the frequency
domain(after the FFT).

  Felix

> Good question, not sure, but I can see the benefit of enabling other
> chains for selection diversity for legacy, perhaps we just never
> bothered implementing that on ath9k on Linux.
> 
>   Luis

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Luis R. Rodriguez]

On Mon, Sep 27, 2010 at 12:25:36PM -0700, Felix wrote:
> On Mon, 2010-09-27 at 11:19 -0700, Luis R. Rodriguez wrote:
> > On Sat, Sep 25, 2010 at 08:14:43PM -0700, Adrian Chadd wrote:
> > > Hi,
> > > 
> > > FreeBSD's AR9160 support enables all three RX chains in legacy mode.
> > > The ath9k codebase doesn't - it calibrates all three chains, but the
> > > ADC calibration results in legacy mode give all 0x0's for chain 1 and 2.
> > > (And seemingly invalid results for chain 0 in 11g mode, but that's another
> > > issue which I am hoping someone will comment on.)
> > > 
> > > Is there any benefit or detriment to leaving RX chains 1 and 2 enabled
> > > when in 11a/11bg mode? Does the unit do any kind of multipath
> > > discrimination?
> Leaving all RX chains enabled will give a huge gain for all OFDM rates,
> they get MRC combined
> (http://en.wikipedia.org/wiki/Maximal-ratio_combining) in the frequency
> domain(after the FFT).

Interesting, thanks Felix, I suppose this may be applicable to
carl9170 as well then, think it uses the same radio? Anyway
patches welcomed.

  Luis

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Felix]

On Mon, 2010-09-27 at 12:33 -0700, Luis R. Rodriguez wrote:
> On Mon, Sep 27, 2010 at 12:25:36PM -0700, Felix wrote:
> > On Mon, 2010-09-27 at 11:19 -0700, Luis R. Rodriguez wrote:
> > > On Sat, Sep 25, 2010 at 08:14:43PM -0700, Adrian Chadd wrote:
> > > > Hi,
> > > > 
> > > > FreeBSD's AR9160 support enables all three RX chains in legacy mode.
> > > > The ath9k codebase doesn't - it calibrates all three chains, but the
> > > > ADC calibration results in legacy mode give all 0x0's for chain 1 and 2.
> > > > (And seemingly invalid results for chain 0 in 11g mode, but that's another
> > > > issue which I am hoping someone will comment on.)
> > > > 
> > > > Is there any benefit or detriment to leaving RX chains 1 and 2 enabled
> > > > when in 11a/11bg mode? Does the unit do any kind of multipath
> > > > discrimination?
> > Leaving all RX chains enabled will give a huge gain for all OFDM rates,
> > they get MRC combined
> > (http://en.wikipedia.org/wiki/Maximal-ratio_combining) in the frequency
> > domain(after the FFT).
> 
> Interesting, thanks Felix, I suppose this may be applicable to
> carl9170 as well then, think it uses the same radio? Anyway
> patches welcomed.

Correct.
  Felix

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Daniel Halperin]

On Mon, 27 Sep 2010, Felix wrote:

> On Mon, 2010-09-27 at 11:19 -0700, Luis R. Rodriguez wrote:
>> On Sat, Sep 25, 2010 at 08:14:43PM -0700, Adrian Chadd wrote:
>>> Hi,
>>>
>>> FreeBSD's AR9160 support enables all three RX chains in legacy mode.
>>> The ath9k codebase doesn't - it calibrates all three chains, but the
>>> ADC calibration results in legacy mode give all 0x0's for chain 1 and 2.
>>> (And seemingly invalid results for chain 0 in 11g mode, but that's another
>>> issue which I am hoping someone will comment on.)
>>>
>>> Is there any benefit or detriment to leaving RX chains 1 and 2 enabled
>>> when in 11a/11bg mode? Does the unit do any kind of multipath
>>> discrimination?
> Leaving all RX chains enabled will give a huge gain for all OFDM rates,
> they get MRC combined
> (http://en.wikipedia.org/wiki/Maximal-ratio_combining) in the frequency
> domain(after the FFT).
>
>  Felix
>

Strongly seconded.  With 3 antennas on and using MRC, you get something 
like an average of 7--10 dB (i.e. a factor of 4 to 10) more useful signal 
power in indoor environments, precisely because of the impact of diversity 
on multipath fading.

Dan

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Luis R. Rodriguez]

On Mon, Sep 27, 2010 at 01:41:02PM -0700, Daniel Halperin wrote:
> 
> On Mon, 27 Sep 2010, Felix wrote:
> 
> > On Mon, 2010-09-27 at 11:19 -0700, Luis R. Rodriguez wrote:
> >> On Sat, Sep 25, 2010 at 08:14:43PM -0700, Adrian Chadd wrote:
> >>> Hi,
> >>>
> >>> FreeBSD's AR9160 support enables all three RX chains in legacy mode.
> >>> The ath9k codebase doesn't - it calibrates all three chains, but the
> >>> ADC calibration results in legacy mode give all 0x0's for chain 1 and 2.
> >>> (And seemingly invalid results for chain 0 in 11g mode, but that's another
> >>> issue which I am hoping someone will comment on.)
> >>>
> >>> Is there any benefit or detriment to leaving RX chains 1 and 2 enabled
> >>> when in 11a/11bg mode? Does the unit do any kind of multipath
> >>> discrimination?
> > Leaving all RX chains enabled will give a huge gain for all OFDM rates,
> > they get MRC combined
> > (http://en.wikipedia.org/wiki/Maximal-ratio_combining) in the frequency
> > domain(after the FFT).
> >
> >  Felix
> >
> 
> Strongly seconded.  With 3 antennas on and using MRC, you get something 
> like an average of 7--10 dB (i.e. a factor of 4 to 10) more useful signal 
> power in indoor environments, precisely because of the impact of diversity 
> on multipath fading.

Patches welcomed :D

  Luis

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Adrian Chadd]

On 28 September 2010 04:53, Luis R. Rodriguez <lrodriguez@atheros.com> wrote:

>> Strongly seconded. ?With 3 antennas on and using MRC, you get something
>> like an average of 7--10 dB (i.e. a factor of 4 to 10) more useful signal
>> power in indoor environments, precisely because of the impact of diversity
>> on multipath fading.
>
> Patches welcomed :D

I'll see what I can do, unless felix beats me to it. :)


adrian

[ath9k-devel] [AR9160]: 3 RX chain operation in 11abg legacy-only mode?

[Adrian Chadd]

On 28 September 2010 03:25, Felix <ic.felix@gmail.com> wrote:

>> > Is there any benefit or detriment to leaving RX chains 1 and 2 enabled
>> > when in 11a/11bg mode? Does the unit do any kind of multipath
>> > discrimination?
> Leaving all RX chains enabled will give a huge gain for all OFDM rates,
> they get MRC combined
> (http://en.wikipedia.org/wiki/Maximal-ratio_combining) in the frequency
> domain(after the FFT).

On the AR9160 too?

Is there anything that can be monitored to verify this is actually
occuring? Or just watching the retry/TX rate information for each
client?


Adrian