Wireless Networking

Duncan Barclay dmlb at dmlb.org
Sat Jul 21 19:28:03 BST 2001


On 21-Jul-01 Andrew Boothman wrote:
> Howdy!
> 
> I would like to pick everyone's collective brains on the subject of 
> wireless networking.
> 
> I understand that most implementations have a maximum range of a couple 
> of hundred meters at best. But I have been reading some interesting web 
> sites including http://www.wlan.org.uk - which is full of quite useful 
> information, tempered by bad writing and site design :-/ and 
> http://www.seattlewireless.net/
> 
> Anyway, I was wondering what sort of outdoor range you could manage with 
> appropriately boosted connections? One problem I see is that you need to 
> boost the signal at both ends of the connection. It's all very well if 
> some central wireless ISP is using antennae costing hundreds, but 
> individual consumers won't be persuaded to spend that kind of cash on a 
> connection - cable modems and ADSL would be cheaper.
> 
> Can anyone see a way to overcome the above problem? Would it be possible 
> to still achieve larger external distances using a very powerful 
> transceiver at one end, and a less powerful one for the consumer at the 
> other? Would it be possible to set up community wireless ISPs using this 
> method?

It's not just a question of transmitted power (or even received power). A
digital RF system and 802.11 in particular does not simply scale range
with TX power. I'll treat the RF aspect first and then the 802.11 specifics.

On the RF side, there is a problem with reflections of the transmitted wave
entering the receiver antenna. This occurs even in line of sight links as the
earth's surface acts as a reflector. A problem called mutli-path occurs when
this happens. For short distances compared with the bit rate (assuming
300m/us propogation in air) the mutliple reflections can cause the drop out
of part of a bit. For longer distances where the time between the line of
sight path and the reflected path is the order of 1/2bit or more, you get
inter-symbol interference where one bit runs into another. Most 802.11b
chips sets use Rake receivers to migitate against these effects but there
is a limit to the mutlipath that they are designed to tolerate. Look at
Intersil's Prism II modem chip for details.

With respect to 802.11 (including 802.11b) there are a number of problems
that long range would generate. These all revolve around the timing aspects
of the on air packets. For example, most data packets are broadcast and within
in certain time a short ACK packet is expected. If the ACK packet is not seen
then the system will retry. Now with light travelling at 300m/us and about 10us
to 20us before the ACK is expected to be transmitted you will start to see
problems at a few km range. This problem also manifests itself in conjested 
networks becuase a particular station expects that it can transmit with
a low probability of collision if it sees no 802.11 transmissions within a
minimum of 15us (i.e. 4.5km) of the end of the last packet it saw (this is
the DIFs parameter in 802.11). Whilst things will still work at long
ranges through put will be reduced. Bear in mind that 802.11 is _designed_ to
have a range of about 200m. There is an assumption that the
propogation delay from one node to another is about 1us and the MAC
(medium access controller) timings reflect this. Extend the range too far
and the protocol will start to flounder.

The 802 specs. are now available to download at the IEEE. Have a look at
the 802.11 spec (and the 1999 suppliment for 802.11b, but the original 2Mb/s
DSSS system is the same). Chapters 8 and 9 deal with the MAC protocols and
explain how the DIFs, ACKs etc. are meant to work.

> Lots of questions and little information I know! :-)
> 
> I'd appreciate any and all thoughts.
> 
> Thanks all!

Duncan

> Andrew.
> 
> 
> ------ FreeBSD UK Users' Group  -  Mailing List ------
> http://listserver.uk.freebsd.org/mailman/listinfo/freebsd-users
> 

---
________________________________________________________________________
Duncan Barclay  | God smiles upon the little children,
dmlb at dmlb.org   | the alcoholics, and the permanently stoned.
dmlb at freebsd.org| Steven King




More information about the Ukfreebsd mailing list