James Knott wrote:

glen herrmannsfeldt wrote:

It isn't so much signal to noise, but the signal couples to the
other wires. Depending on which wires, it can make a big
difference in the signal at the other end.

Anything other than the desired signal is noise. Therefor signal from the
other wires is noise. There's even a word for that type of noise. It's
called crosstalk.

Well, it is, but neither describes this case so well.

If you pair on (3,4) and (5,6) for 100baseTX the (3,6) signal
will couple to wires 4 and 5. Most likely the result is that
there isn't enough signal left at the end, even without any
noise. The receiver requires a certain amount of signal,
independent of the amount of noise actually present.

Depending on the arrangement of the pairs in the cable,
some will couple as common mode signals onto other pairs,
again lost to the intended receiver.

For a different example, if you manage to pair (1,3) and (2,6)
the signal will then couple between the active signaling
pairs. Though even in this case it should be absorbed by
the terminating resistors on the transmitter and not affect
the receiver much (for 100baseTX).

Now for gigabit, where all four pairs are used in both directions,
any coupling goes directly into the receiver of the wrong pair.
I don't believe that the OP is using gigabit, though.

-- glen