Final note on XTK ( Cross Talk)
I think I see where the confusion about cross talk is coming from in this thread. Once again, I think it stems from the words we are using. In this thread, we have been discussing three different topics, yet, using the same terms of “cross talk”.
The actual cross talk voltage noise at the near and far end of an adjacent quiet line is one set of terms. These voltage signatures depend on the intrinsic coupling between the lines, and the extrinsic line features of length, the terminations (due to the impact of the reflected noise) (which Scott McMorrow pointed out off line) and signal magnitude and rise time, and as Yuri and other have pointed out, the losses.
Then there is the near end cross talk coefficient (NEXT) and the far end cross talk coefficient (FEXT). These are generally defined in a transmission line environment where the ends of all the lines are terminated, so as to not compound the complexity of the problem with the impact from the reflections of both the signals and the noise. The NEXT depends on the intrinsic line parameters, and provided the coupling length is longer than the saturation length, not on the rise time or length. The FEXT depends on intrinsic line parameters and the line length and the rise time.
Finally, there are the intrinsic coupling coefficients, related to the relative distributed coupling capacitance per length between the two transmission lines, to the capacitance per length of the signal line to its return and the distributed loop mutual inductance per length to the loop self inductance per length of the signal line. Granted, this is as per the circuit’s model, as an approximation to the more accurate field’s model.
It is absolutely correct that the NEXT is related to the relative capacitive coupling + the relative inductive coupling, while the FEXT is related to the difference between the capacitive and inductive coupling. This might lead one to believe that NEXT will always be > FEXT. However, the FEXT also scales with the coupling length and inverse with the rise time, while the NEXT does not. This means that for long lines, and short rise times, FEXT can be >> NEXT.
How long and how short..it depends. Knowing the coupling terms, it is easy to calculate how long and how short to have FEXT > NEXT. If you want the details, see the chapter in my book on cross talk.
Hope this helps.
–eric
*******************************************************
Dr. Eric Bogatin, Signal Integrity Evangelist
Bogatin Enterprises
Eric and All,
Excellent summary on crosstalk.
Just one minor clarification to the statement:
” However, the FEXT also scales with the coupling length and inverse with the rise time, while the NEXT does not.”
In the time domain, both NEXT and FEXT have saturated and unsaturated regions. Both NEXT and FEXT scale with length and (inversely) with aggressor rise-time in their unsaturated region. NEXT saturates at the backward crosstalk coefficient, whereas FEXT saturates practically at full coupling. Full coupling finds very good applications in microwave and optical applications, but in high speed digital circuits, where crosstalk is mostly an unwanted bad side effect, we want to stay clear from the saturated far end crosstalk region.
Regards,
Istvan