River,

  Not sure that this is a good analogy since a square wave is a pathological case,
that is, it has frequency components at all frequencies and always suffers from
phase distortion. There is some discussion about this issue in Razavi's book,"RF
Microelectronics" in the intersymbol interference section(?). Don't confuse systematic
and random jitter. One way to look at this is to consider the following data patterns
01010101 and 00001111. Due to the finite bandwidth of the channel, the output signal
from the channel for these two signals will have different amplitudes and zero
crossings. When you plot the eye diagram, the difference in zero crossings
causes the "eye" to narrow and this is equivalent to jitter. That is for different
data patterns the zero crossings "jitters", changes. However, this is a determinstic
phenomena, that is, for the same data patterns you see the same "jitter".  You
have to account for the jitter due to the difference in the data pattern in the overall
system design. In addition to the systematic jitter there will also be random jitter
due to thermal, shot, 1/f, ... noise in the system.

                                                                        Best Regards,

                                                                            Sheldon

Neo_flash,

  You are correct that if the signal was periodic then there would be no pulse
distortion. However in general signals is not truly periodic, they are random.
This is why standards specify pseudo-random data, PRBS, data for testing
bit error rates.

   So if there is a system with a pulse input, the output pulse will be distorted
unless the group delay of the system is constant. The pulse distortion is dependent
on the data pattern, for example, the distortion for the data patterns 01010101
and 00001111 are different[if the group delay is not constant for two octaves].
When the outputs for the different data patterns are overlayed, the pulse
distortion appears as jitter. Even if the communication channel is perfect, systems
usually have internal filtering that causes pulse distortion, for example, the
channel filter in Bluetooth/GSM. The result is that systems have systematic
jitter that typically comes from the system specification, for example, 1 MHz
channel spacing, 20dB of attenuation in the next adjecent channel, Bit error
rate of 0.1%, ...

                                                                              Best Regards,

                                                                                   Sheldon