[ath9k-devel] Difference 40MHz and 20MHz

[ath9k-devel] Difference 40MHz and 20MHz

[Kamran Nishat]

Hi,
FIRST, Will there be any difference in duration of PHY header for 20MHz and
40Mhz.
Do u think 40Mhz header will take lesser time?

Second, if my AP is connected to a client on 40MHz (Channel Bonding).
During Carrier Sense (CS) it sense ext channel of 20MHz is busy will it
automatically switch to 20MHz and start sending on primary channel. If yes
will it take some time to shift from 40MHz to 20MHz?

Kamran
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.ath9k.org/pipermail/ath9k-devel/attachments/20140923/a0def09f/attachment.htm 

[ath9k-devel] Difference 40MHz and 20MHz

[Adrian Chadd]

On 22 September 2014 14:14, Kamran Nishat <kamran.nishat@gmail.com> wrote:
> Hi,
> FIRST, Will there be any difference in duration of PHY header for 20MHz and
> 40Mhz.
> Do u think 40Mhz header will take lesser time?

Which bits of the header are you talking about? Just the initial PLCP
bits? I'm not sure how HT40 PLCP works.

> Second, if my AP is connected to a client on 40MHz (Channel Bonding). During
> Carrier Sense (CS) it sense ext channel of 20MHz is busy will it
> automatically switch to 20MHz and start sending on primary channel. If yes
> will it take some time to shift from 40MHz to 20MHz?

THat depends upon the AP model.

It can for sure just decide to monitor the air and if i's busy
downgrade to selecting HT20 rates to transmit on. There's also the
channel width update stuff that an AP can use to migrate a BSS between
20MHz and 40MHz.

Now, as for chip support with Atheros 11n parts - no, it doesn't
auto-detect 20 or 40MHz wide CCA/CS and select a rate. The 11ac chips
started doing that.

What they _do_ however is detect CCA/CS on both halves of the channel
and you can select whether to treat them separately or joint. (Ie,
whether to bother doing CCA on the full 40MHz for all transmissions or
just when it wants to do a HT40 transmission.)

So, you can do some tricks by watching the success rate of HT40
transmissions and the primary/extension MIB channel busy counters and
determine what's going on.


-a

[ath9k-devel] Difference 40MHz and 20MHz

[Kamran Nishat]

> Hi,
> FIRST, Will there be any difference in duration of PHY header for 20MHz
and
> 40Mhz.
> Do u think 40Mhz header will take lesser time?

Which bits of the header are you talking about? Just the initial PLCP
bits? I'm not sure how HT40 PLCP works.

I am not talking about the training (preamble) part. I was asking about MAC
header.

So, you can do some tricks by watching the success rate of HT40
transmissions and the primary/extension MIB channel busy counters and
determine what's going on.

Instead of monitoring MIB counters. If some controllers tells me its better
to switch to 20MHz do u have any idea what will be the delay and what rate
would be picked.

On Wed, Sep 24, 2014 at 9:52 PM, Adrian Chadd <adrian.chadd@gmail.com>
wrote:

> On 22 September 2014 14:14, Kamran Nishat <kamran.nishat@gmail.com> wrote:
> > Hi,
> > FIRST, Will there be any difference in duration of PHY header for 20MHz
> and
> > 40Mhz.
> > Do u think 40Mhz header will take lesser time?
>
> Which bits of the header are you talking about? Just the initial PLCP
> bits? I'm not sure how HT40 PLCP works.
>
> > Second, if my AP is connected to a client on 40MHz (Channel Bonding).
> During
> > Carrier Sense (CS) it sense ext channel of 20MHz is busy will it
> > automatically switch to 20MHz and start sending on primary channel. If
> yes
> > will it take some time to shift from 40MHz to 20MHz?
>
> THat depends upon the AP model.
>
> It can for sure just decide to monitor the air and if i's busy
> downgrade to selecting HT20 rates to transmit on. There's also the
> channel width update stuff that an AP can use to migrate a BSS between
> 20MHz and 40MHz.
>
> Now, as for chip support with Atheros 11n parts - no, it doesn't
> auto-detect 20 or 40MHz wide CCA/CS and select a rate. The 11ac chips
> started doing that.
>
> What they _do_ however is detect CCA/CS on both halves of the channel
> and you can select whether to treat them separately or joint. (Ie,
> whether to bother doing CCA on the full 40MHz for all transmissions or
> just when it wants to do a HT40 transmission.)
>
> So, you can do some tricks by watching the success rate of HT40
> transmissions and the primary/extension MIB channel busy counters and
> determine what's going on.
> -a
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.ath9k.org/pipermail/ath9k-devel/attachments/20140924/64136939/attachment.htm 

[ath9k-devel] Difference 40MHz and 20MHz

[Adrian Chadd]

On 24 September 2014 10:22, Kamran Nishat <kamran.nishat@gmail.com> wrote:
>
>
>
>> Hi,
>> FIRST, Will there be any difference in duration of PHY header for 20MHz
>> and
>> 40Mhz.
>> Do u think 40Mhz header will take lesser time?
>
> Which bits of the header are you talking about? Just the initial PLCP
> bits? I'm not sure how HT40 PLCP works.
>
> I am not talking about the training (preamble) part. I was asking about MAC
> header.

Ok, so the MAC header is already being transmitted at the HT40 rate.

So, take a look at the different symbol rate/timing for HT20 versus
HT40. Yes, they're slightly different.

> So, you can do some tricks by watching the success rate of HT40
> transmissions and the primary/extension MIB channel busy counters and
> determine what's going on.
>
> Instead of monitoring MIB counters. If some controllers tells me its better
> to switch to 20MHz do u have any idea what will be the delay and what rate
> would be picked.

Well, it doesn't work that way with the atheros MAC.

You have four rate table entries.

You get to select how many retries for each of them and whether to use
20 or 40MHz.

I don't know if there's a cute way to have it fast-fail short retries
(eg if it fails to get the RTS/ACK pair in time) in HT40 mode and fall
back to the HT20 rate.

So you could tag the last rate entry as being 20MHz regardless of channel width.

The chip isn't doing that dynamically. You can have it happen about as
fast as you can fail to transmit frames and get that feedback from the
MAC.


-a