(Axel Hammerschmidt) hath wroth:
>Spotty just means irregular. It doesn't have to be in physical "spots".
The neighbors dog is named Spot. The kids call him Spotty. Judging
by the mess he leaves behind my wood pile, he's not very regular.
>We don't know anything about the signal to noise ratio. The implication
>in the link is, that flow control "improves carrier sense information".
>> That was fairly straight forward with analog receivers. The
>> signal strength versus the remaining noise level between packets.
>> However, with the introduction of all digital receivers, there was a
>> change to measuring the SNR using the error rate. 100% reception
>> success, with no corrupted or trashed packets, was deemed perfect SNR.
>> If corrupted packets were decoded, the SNR value was reduced. The
>> correlation is marginal with the previous analog methods but good
>> enough for site surveys and antenna aiming.
>Hmm! Not sure how this relates to carrier sense?
In order to detect interference, the receiver needs to know what
constitutes interference and at what threshold to block transmission.
It makes no sense to have any and all detected RF be declared as
interference as BPSK and OFDM can extract data from rather low signal
to noise ratios. For example, 54Mbits/sec requires a 17dB minimum
SNR. I vaguely recall that it's about 6dB at 1Mbit/sec (not sure).
The device at one end of a link sends the receive SNR to the other end
to allow the sender to determine if it should be allowed to transmit.
Incidentally, much of the noise is contributed by internal digital
noise from other parts of the router or client radio, which is also
responsible for some of the weird variations in receiver sensitivity.
An ideal interference detection circuit would measure the receive SNR.
If it's over 17dB SNR for 54Mbits/sec, there is sufficient margin to
decode the data and transmissions are allowed. If the resultant SNR
is less than 17dB, then transmission is blocked under the assumption
that the interference will prevent reception. So far so good.
The problem is how to measure SNR. In the older analog devices, the
SNR was simply the peak RSSI during transmission divided by the
detected noise level between transmissions. This is adequate but
slow. As 802.11g speeds and short preambles made it more difficult to
use analog techniques, all digital methods were employed. The current
method is to use the detected data error rate as a measure of SNR. If
some packets arrived corrupted, it would be assumed to have been
perpetrated by "noise". The more corrupted packets, the more noise,
and the smaller the SNR per some conversion scheme. The effect is the
same. When the interference is sufficient to seriously interfere,
then transmission is blocked.
The term "carrier sense" is an analog term, that really only applies
to analog chipsets. The last of these were the Prism I chipsets with
it's mixers and analog front ends. Todays chips are all digital and
direct conversion (no mixers).
If one implements flow control (RTS/CTS) in such a system, the SNR
will improve simply because the system now has an additional method of
detecting interference and measuring SNR. Instead of relying on just
the SNR of the other end of the link to determine when to transmit,
the system now adds flow control handshake management packets that
determine when *BOTH* ends of the link have sufficient SNR to xmit.
This does very well for eliminating "hidden nodes" but also improves
the all digital version of SNR by simply improving the chances that a
packet will arrive intact.
As always, there's no free lunch. If you use flow control to improve
delivery reliability, the overhead will slow down the thruput. I
don't know the exact performance hit, but my testing showed that it's
substantial. It's as I hinted in a previous rant, 100% delivery at
1Mbit/sec is considerably slower than an error prone 30% delivery at
perhaps 12Mbits/sec. That same thing with flow control. If you
improve the delivery probability by blocking xmission during
interference, the system will by necessity slow down.
Jeff Liebermann email@example.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558