Re: Is there a range filter?


From: Jim Thompson (jim_at_musenki.com)
Date: 2002-03-23 00:44:33 UTC


Wim Lewis writes:
>
> On Friday, March 22, 2002, at 01:14 PM, Brad Colbert wrote:
> > Has anyone had experience using a prismII card with Jouni's
> > HostAP driver at a distance greater that 300meters?
> >
> > Is there a time(distance) filter? If so, can it be disabled?
>
> 802.11 uses link-layer acknowledgements --- that is, the receiver is
> expected to send an ACK packet whenever it receives a data packet. The
> timings of the CSMA algorithm take this into account: after a data
> packet goes by, there's a window in which only the receiver is allowed
> to transmit (so that it can send the ACK without worrying about
> contention with other nodes). So if the receiver is too far from the
> transmitter, the ACK may not get back within its allotted time slot.
>
> I don't know offhand what happens in this case --- I'm sure some people
> have tried this already. My guess is that your throughput will go way
> down (since the transmitter will be unnecessarily re-sending lots of
> packets) and the card will register lots of TX errors (since it thinks
> its packets aren't being received). But I'd guess that the data will in
> fact make it across the link, just not efficiently.
>
> You can probably tweak the MAC's timings to get around this, or maybe
> trick it into thinking all your traffic is multicast/broadcast
> traffic --- the ACKs are only used for unicast packets. (300m might not
> be far enough for this to matter anyway.

Some MACs have proprietary settings that eliminate the ACKs.

> I don't know offhand what 802.11's timings are. 300m is about 1
> microsecond at lightspeed, or 2 uS roundtrip.)

The ACK can't be transmitted before SIFS (28us), and has to be received before an additional DIFS (or PIFS) plus a virtual slot time for 128us.

Assuming 300m is 1ms, thats nearly 47km or 29 miles, unless I screwed the pooch.

-- 
"The Tivo is a wrapper around a bad interface" -- Bruce Sterling


This archive was generated by hypermail 2.1.4.