`Sunday Times' article on GSM changes

[Jay Prime Positive <jpp@markv.com>]

Brad Huntting <huntting@glarp.com> asks:
>[...]this would mean the sender and reciever would need to be synced
>to well within 500ns of each other.  Isn't this a bit difficult?  How
>do they do it?

  You are right.  This is perhaps *the* central design issue of this
kind of spread spectrum system.

  One standard solution is to use "gold codes".  Gold codes are
special in that they are very self dissilimar.  That is they look very
unlike any shifted version of themselves.  So you can build a very
simple corelator which tries all the possible shiftings of a code to
the signal, until one pops up with "low frequency" data rather than
"high frequency" noise.

  Another is to begin a transmission with a special sync header (and
concievably intersperse additional ones bassed on the expected
frequency of loss of lock).  Currently available PLL's working at
900MHz have very low phase noise, and I can imagine the construction
of fixed frequency PLL's with even lower phase noise.

  A third is to transmit BOTH the spreading code, and the data.  You
can think of this technique as sending two channels of data, one which
is all 1's (or 0's), the other which is a little more interesting.
The two channels are then combined at the reciever to yield the data.

  A fourth is to use an externaly generated sync signal -- for example
a radio transmission that both sender and reciever can hear.  (For
this aplication, I don't see how this would be used...)

  Aditional solutions are possible.  (What is this S.A.W. thing I read
about??)

j'