Rod, at 08:28 on Mar 01 1996, you wrote to Bill Grimsley...

BG> Sure, but even fibre-optic has inherent line loss
BG> (attenuation), which must be reamplified every so often.

RS> Not in the sense of affecting the calls call by call. Once its digital,
RS> that that, what goes in one end, comes out perfect at the other end.

Theoretically.  However, I just called Alan Kennedy, my senior engineer
mate with Telstra, and the reality can be rather different (although for
the most part, there should be no appreciable loss along the entire length
of FO).  The actual FO design specs call for an exchange transmit level of
-10dBm, with automatic regeneration once the level drops to -47dBm. 
However, Telstra have set the system to regenerate if the signal reaches
-40dBm.  This rarely, if ever happens, although the digital signal does
degrade into sinusoidal waves after a time, and regeneration restores the
square wave, as well as the -10dBm amplitude.  Without any regeneration
along the FO, the signal will attenuate by 30dB (to the aforementioned
-40dBm) after just 150km.  That said, it is fairly safe to say that if
handled correctly, the signal will arrive at the other end in much the same
condition as it was sent.

Now comes the tricky bit.  The maximum allowable inter-exchange losses are
12dB on a normal copper pair, with a further allowable loss of up to 7dB to
each subscriber.  This means that a total loss of 26dB is deemed to be
acceptable, but when added to the original -10dBm Tx level, that translates
to a totally unacceptable line level (AFA V.34 modems are concerned) of
-36dBm (and this is at voice frequencies of around 1400Hz, not data
frequencies of 3KHz or higher, upon which the loss is logarithmically
greater.

RS> With the sole exception of a completely different problem called
RS> clock slip, but thats a fault, a different matter entirely.

I know nothing of cock slip, but I strongly suspect that Bob does.  :)

Incidentally, clock slip is a non-issue, as clock synchronisation at each
end of the FO cable is performed automatically.

BG> Something is not working properly one that particular
BG> path which is used for calls to Paul's board,

RS> Nope, its just a combination of YOUR indifferent line from your
RS> exchange to your house and PAULS indifferent line from his exchange
RS> to his house. Thats VERY clearly visible on the freq response graph.

Agreed, and Alan concurs.

RS> Nothing to do with whats between the two exchanges
RS> once the signal is digital in the fibre optic network.

In general, that's also true.

BG> Still subject to signal amps and shapers every so often though.

RS> Nope, digital doesnt work like that. Thats WHY you do
RS> digital in the first place. You can regenerate as often
RS> as you like with no added degradation each time you do it.

Correct, but without regeneration, there is still an average loss of 30dB
every 150km or so with FO cable.

RS> BUT, when you call each other, the COMBINATION of that degradation at
RS> EACH end of the session produces an pretty obscene result. Which the
RS> M34F really cant handle very well at all when called by a Courier.

BG> It's a logical assumption I suppose,

RS> Its not an assumption, the result of careful analysis.

After speaking with Alan, I'd have to agree with that.

BG> and may have been masked by our respective USRs'
BG> superior ability to handle marginal line conditions,

RS> Yep, no doubt about that whatever. And we have other examples
RS> like when Steve Anderson found that his problems went away when he
RS> replaced his M34F with a Courier, from memory calling Poes Courier.
RS> AND his M34F works fine for the fellow in Sydney he flogged it to.

Yeah, the USRs really do allow a higher bit rate for a given carrier freq.

RS> Bet that bloke has a rather better freq response curve when calling
RS> say Poe than Steve does.

Sure looks that way now.

BG> but now that Paul is using Rockwells, the problem has become more obvious.

RS> Well, I think its better to say that its other than Courier
RS> that matters, coz clearly the Spirit wasnt terrific either.

True, the M34F hasn't failed to connect at V.32bis once, unlike the S2.

BG> And if little Johhny sells Telstra, it'll likely only get worse.  :(

RS> Dunno, that can see operations like that trying a lot harder to please.
RS> The problem up till now has been that they had a monopoly on the local
RS> lines and it may well be the threat of those customers who are pissed
RS> off with their line performance pissing off to Optus that may well shake
RS> the shit out of Telstra, whatever they do on ownership. Corse at a vast
RS> subsidy from the people who just use their line to yack on by voice.

Looks like we'll be finding out soon enough, as the Libs appear to be a
shoe-in at this stage.  Oh well, people get the government they deserve, I
guess.  :(

Regards, Bill

--- Msgedsq/2 3.20