Channel interference scan and manual selection

Adrian Chadd adrian at freebsd.org
Wed Dec 4 04:14:07 UTC 2013


No, I get this bit. Totally. I had to deal with this with HT40
channels where .. the rejection isn't as great. Adjacent channel
interference into the secondary channel just killed things.

I'm right now more interested in making sure we can gather the data we
need first. The channel selection bit is orthogonal.

I completely agree with you. We're better off with completely
overlapping channels where the competing access points have a chance
of correctly detecting each others signals and backing off without
raising the channel noise floor.

There's also the HT20/HT40 coexistence stuff where we should actually
disable doing HT40 on 2GHz channels if there's adjacent channel
traffic.



-adrian

On 3 December 2013 19:59, Jim Thompson <jim at netgate.com> wrote:
> There is no "1, 6, 11" anymore.  You're not listening.
>
> Back in the day, the adjacent  (1 on 6, 6 on 1, 11 on 6, 6 on 11)
> channel rejection spec for 802.11b was 35dBm at 11Mbps modulation
> There never was an alternate (1 on 11, 11 on 1) channel rejection spec
> in the days of 802.11b.
>
> Most decent super-het designs back then used a SAW filter in the
> middle of the downconversion chain, and got around 41dBm of adjacent
> channel rejection.
>
> But we all moved to direct-conversion receivers ... because: cost.  We
> all loves us some cheap 802.11 gear.
>
> The adjacent channel rejection spec for 802.11g (and 802.11a) is -1dBm
> @ 54Mbps modulation.
> The alternate channel rejection spec for 802.11g/a is 15dBm @ 54Mbps modulation.
> Just FYI, minimum ACR for 802.11g/a at 6Mbps is 16dB (alternate is
> 16dB more, for 32dB)
>
> If you're underwhelmed by the difference between 41dBm and -1dBm, then
> I can't help you.
>
> Free Space Path Loss = 20 log(4*p*r/λ) dB, where
> r = distance between transmitter and receiver
> λ = wavelength
>
> Path loss in the first meter @ 2.4GHz is 41dB. At 10m it's 60dB.
>
> Lets say you've got a garden-variety radio that puts up 32mW (15dBm)
> of tx power, and ignore antenna gain for now (so 0 dBi antennas on
> both radios).
>
> Old 802.11b (super-het receivers) world:
>
> 15dBm - 60dB - 41dB = -86dBm This is the in-channel 'noise power' of
> the adjacent channel radio.
> Notice that it is at least 15dB above the thermal noise floor.
> Translated: you've lowered your SINR.
>
> New 802.11g/a (direct conversion receivers) world:
> 15dBm - 60dB - -1 dBm = -44dBm.   This is about 20dB higher than what
> is necessary to recover a 54Mbps receiver, but remember, it's the
> *noise power* of a radio operated on an adjacent channel.   (Most
> 802.11 OFDM receivers are EVM-limited at 48Mbps and higher, but I
> digress.)
>
> Note as well that we're 57dB above the thermal noise floor, at 10m (33')
>
> Even if you back off to alternate channels (1 and 11), you're still at -60dBm.
>
>
>
> On Tue, Dec 3, 2013 at 9:40 PM, Adrian Chadd <adrian at freebsd.org> wrote:
>> On 3 December 2013 19:32, Tek Wiz <tx2023 at gmail.com> wrote:
>>> Yes, the interference can be calculated by averaging the beacon RSSI (dB) of
>>> the overlapping channels, e.g. for channel 6, the beacon RSSI of channels 4,
>>> 5, 6, 7, 8 are averaged. Normally channels 1, 6, and 11 are chosen, although
>>> other channels may have less interference, in order to be a 'good neighbor'.
>>
>> Sure. But we can also look at what kinds of frames we actually receive
>> during a sample window.
>>
>> What I'd also like to do is finally add receive power histogram
>> support. Ie, look at the RSSI of all the frames you receive, stick
>> them in buckets, respond to the power histogram action request frames,
>> use them for this.. that doesn't require any driver support. Just
>> net80211 work in the RX path.
>>
>>
>>
>> -adrian


More information about the freebsd-wireless mailing list