[TowerTalk] 600 Ohm Line

[Jim Brown]

On Thu, 29 Jul 2004 13:15:34 -0700, Jim Lux wrote:

>Or, more properly, the adaptive equalizer in the modem had better 
>compensate for the eternally changing lumps and bumps in the circuit, and 
>not assume anything about the actual impedance (which probably varies 
>widely with frequency over the more than decade frequency range).

Yes. That's the modern equivalent of the passive equalizers both in the line driver and 
line receiver. But the modem MUST be doing its best to match the Z of the line -- the 
closer it gets to that ideal the less will be the response problems caused by mis-
termination. 

>
>   In the phone world, I don't think optimum power transfer (in the 
>Thevenin sense) is really the issue.  You've usually got as much signal 
>power as you need (although, this is why 56K modems don't really go 56K on 
>a POTS wire... The Part 68(?) limits on Tx power).  It's things like group 
>delay and differential effects that bite you.

Agreed. The impedance matching is ONLY to keep the transmission line behavior of 
the circuit under control.  

>That, and the 2wire to 4wire "hybrid" function (that modem has to separate 
>the Rx and Tx signals somehow, and it certainly can't depend on there being 
>a nice 600 ohm resistive impedance for a hybrid to look into).  At the CO 
>end, they have a similar problem related to echo suppression.

I haven't designed any hybrids, but Circuit Analysis 101 says that if you want the 
hybrid to work and be stable, the circuit gains have to be well behaved, which in turn 
says you want the impedance to be a well-behaved function. All of which points to 
getting as close as possible to a well matched and terminated line. 

>From my limited perspective, the biggest obstacles I see with this are 1) the 
impedance of the cables used was never a controlled parameter (or at least not to 
the extent needed for high speed data; 2) noise; and 3) losses in the cable above 
baseband were never a specified parameter. My hat is off to the guys who manage 
to make high speed data work on a typical subscriber loop! 

Jim Brown  K9YC