Re: [SI-LIST] : Differential Pair Theory

About this list Date view Thread view Subject view Author view

From: Dennis Tomlinson (det@tellabs.com)
Date: Mon Nov 15 1999 - 15:15:38 PST


Hi Chris,

My comments are interspersed.

Chris Bobek wrote:
>
> Hi,
>
> I am trying to understand how differential pairs behave and how you
> terminate them. I've read app notes and books that tell you how to
> terminate them, but I don't know how they derived the values and
> methodology. I would just like a general understanding of what's going
> on. I understand how and why termination works on normal, single-ended
> lines.

To understand differential lines, it is important to understand that
the pair of lines carry only one bit of information per bit time. This
results from the method of drive - which is, the two conductors are
normally driven out of phase. Restated: One conductor carries the
value of the signal, while the other carries it's compliment.
(Note: I have known engineers to drive a twisted pair with the signal
on one conductor, and earth (ground) on the other. This is, I guess, a
pseudo differential signalling scheme. Or maybe we could call it a -6dB
differential scheme. Nonetheless, all of those engineers have since
transferred to jobs in sales, or marketing;-).

>
> As an example, suppose you had a 100ohm, unshielded, twisted pair cable
> that ran for 5000 feet between two boards. Once the pair entered the
> board, it was routed over several inches to a differential receiver.
> The two boards do not share a ground (suppose they are battery powered).

This is not normally done using DC coupling. A difference in DC potential
between the two boards can exceed the compliance rating on either driver
or receiver, thus letting out some of the smoke;-).

AC coupling can be used in lieu of tying grounds together, but then you
face the mess of dealing with line coding schemes, code run lengths, etc.
This can all be alleviated with a good ground connection, if that's possible
over a 5000' run. Alternatively, a good, short reach fiber optic scheme
might be put to good use here.

>
> 1) How did I know to route the traces on the board as two 50 ohm
> traces?

The differential impedance, as seen from one conductor to the other,
is the sum of the individual trace impedances (modified to reflect any
significant coupling between them). 50 + 50 = 100 which matches the
differential impedance of the twisted pair.
The PCB trace impedances are chosen the same walue in order to preserve
signal balance.

> 2) How did I know to terminate the pair by placing a 100ohm resistor
> between the two traces at the receiver?

The simple answer is the 100 Ohm resistor provides a differential
termination to the differential signal.

> 3) Could I terminate each line separately as if they were two,
> single-ended lines (say, using a 50ohm resistor to ground on each line
> at the receiver?)

Yes, if the grounds are common, and the driver has the drive capability.
In ECL/PECL differential signaling schemes, it is common to terminate
each side of the differential signal to a termination voltage 2V below
the upper rail. This not only differentially terminates the signal,
but also completes the DC bias path for the driver's output transistor.
Again, grounds must be common.

> 4) Where does the return current flow as the signal is launched down
> the line from the driver, across the twisted pair, across the diff pair
> on the board, and to the receiver?

From driver to twisted pair interface: return currents flow on local PCB
                                       ground/power planes/busses.
On the twisted pair (no shield): return current flows in the other
                                       half of the twisted pair.
On the PCB to the receiver: Return current flows in the ground/
                                       power planes/busses of the receiver
                                       board.

>
> Thank you very much,
>
> Chris
>

Cheers,

Dennis

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:39:49 PST