Fwd: Re: [MLB-WIRELESS] ./: Can 802.11 Become A Viable Last-Mile Alte rnative?

[Jason Hecker]

> But both types essentially rapidly change frequencies to avoid
> interference.

No, to be pedantic, if you take a successive snapshots of a FH hopped signal
and display the spectrum, you will observe each hopped transmission as a peak
moving about the frequency band.

With direct sequence you will just see the whole band of interest (22MHz)
taken up with energy, the spectrum envelope taking a distinct lumpy shape
with a large main lobe at the centre frequency and successive smaller lobes
to the band edge.

By examples below will apply to something like the older WaveLAN cards, whose
parameters differ from 802.11b.  They only operated at 2Mbit.

FH and DS tackle noise in different ways as well.  Where you might have a
strong peak, such as a narrowband transmitter (microwave oven in our case) FH
will only occasionally encounter this frequency.  Because it will transmit
the same bit 11 times at 11 different frequencies you will end up with 10
good bits and one bit swamped by the interference (a bit simplistic but it
will illustrate the point).  Never mind, 10 out of 11 bits is plenty good
enough to detertmine what the actual bit transmitted was.

DS takes your original bit and XORs it with a psuedorandom sequence whose bit
rate is 11 times the data rate.  Doing this smears your original 2Mbit stream
with it's 4MHz bandwidth to the 11MBit stream's 22MHz bandwidth which is what
you transmit.  The other end has the same 11 Mbit random sequence in sync and
XORs the received signal with it, inverting the original process.  Any
received peaks like an oven or other transmitter will be smeared to 22MHz but
your original signal will pop out above this noise.

CDMA and DS are much the same thing.  On the CDMA phone network different
spreading codes are used so phones can occupy the same channel.  You will
only get a usable signal back if you correlate the signal sent to you with
the right sequence.  All other sequences will end up looking like noise.
BUT, if you have too many CDMA transmitters in one place this noise level
after decorrelation may well be too high to pick a usable signal out of.

If you have too many FH transmitters in one place, you will get too many
transmissions colliding with each other, maybe giving you less than 5 or 6
good bits out of 11 for example which isn't good enough to determine what the
original bit was.

> What does "It uses direct sequence which involves modulating the signal
> with a high speed "chipping" code to spread it across a wider bandwidth"
> mean?

See above.

> > Putting a signal on picosecond pulses will _certainly_ cause its
> > spectrum to spread.
>
> Why? I don't see how the one implies the other.

By virtue of the fact that the shorter your transmission pulse time, there is
a tradeoff with the bandwidth it needs.  A 1 picosecond (1E-12) pulse will
require an inordinately large bandwith 1/1E-12 Hz.  Huge.  That's assuming
your pulse is a square one.  You can reduce the bandwith requirements by
pulse shaping with gaussian curves or raised cosines or whatever.

Are you still awake?


To unsubscribe: send mail to majordomo at wireless.org.au
with "unsubscribe melbwireless" in the body of the message