Hi,

I have a question concerning measuring crosstalk between adjacent balanced pair channels. In particular, the DUT I am working with consists of 2 x 100 ohm twisted pair channels. I match the VNA to the DUT using baluns. All ends of both circuits are properly matched, for both common and differential mode, either by the test equipment or by passive terminations. I have access to both the common mode and differential mode ports on the baluns.

I measure crosstalk (near-end and far-end) by making the appropriate connections to the DUT and sweeping S21 across the frequency range of interest. This gives me a plot of crosstalk vs frequency.

Now my question is this - how can I determine where along the channel significant levels of crosstalk are introduced? Or put another way, can I somehow obtain a plot from the VNA which shows the level of crosstalk being induced vs distance from, say, the send end? I use an 8753A VNA which has option 010. I have more modern VNA's but these do not have the time domain option. Will a time domain plot of S21 give me this, or is there some other way?

Thanks in advance for any advice on this.

Regards,

George G

I have a question concerning measuring crosstalk between adjacent balanced pair channels. In particular, the DUT I am working with consists of 2 x 100 ohm twisted pair channels. I match the VNA to the DUT using baluns. All ends of both circuits are properly matched, for both common and differential mode, either by the test equipment or by passive terminations. I have access to both the common mode and differential mode ports on the baluns.

I measure crosstalk (near-end and far-end) by making the appropriate connections to the DUT and sweeping S21 across the frequency range of interest. This gives me a plot of crosstalk vs frequency.

Now my question is this - how can I determine where along the channel significant levels of crosstalk are introduced? Or put another way, can I somehow obtain a plot from the VNA which shows the level of crosstalk being induced vs distance from, say, the send end? I use an 8753A VNA which has option 010. I have more modern VNA's but these do not have the time domain option. Will a time domain plot of S21 give me this, or is there some other way?

Thanks in advance for any advice on this.

Regards,

George G

Thank you for your prompt reply.

In this case, the word "channel" has a rather specific meaning, and I should have made this clear at the start. It includes a length of cable, plus at least three connectors and a patch lead. I was hoping to see whether the connectors and patch lead make very significant contributions to crosstalk, and from what you say, I think I will be able to do that.

Regarding the influence of baluns, I do a full two port calibration which includes the baluns, so that the reference plane is at the 100 ohm terminals of each balun. In the frequency domain, this factors out the influence of the baluns. This must thus also be true for the time domain response, given that it is derived from the frequency domain response?

With regards to the near-end crosstalk (NEXT) and far-end crosstalk (FEXT) issue, in order to characterise a channel in the frequency domain, I make a set of measurements which includes NEXT, FEXT and insertion loss (IL). When I look at cable performance, I calculate a parameter known as Equal-level FEXT (ELFEXT), wherein I subtract the IL from the FEXT to give a true indication of FEXT over a long length of cable. From what you have said, I take it that when working in the time domain, I will need to do the same for NEXT, only the factor will need to be 2 x IL to account for the return trip losses. Have I got this right?

Also, I am trying to picture in my mind what the plot will mean when I view FEXT in the time domain. The X-axis represents time. So when I see a feature close to the origin of the plot, it will be near the receive port, which represents the far end of the cable. Have I got this right? Also, I will still need to subtract one lot of IL to account for cable losses, won't I? Consider the case where FEXT occurs near the send end. By the time it gets to the receiver, it will have suffered signal decrease due to the IL of the cable, so I need to account for this. Have I got this right?

Many thanks Dr Joel.

Regards,

George G

Regards,

George G

No, not at all in the case of Far End XT. Consider the case of the pair of lines, where you connect them together at the send side. The cross talk will be very large, and will occur at the position x=0. Consider the line length L, with effective total delay of D. The signal will reach the far end at time D, as measured at the the end of the supposed-to-be-isolated pair.

Now, consider the same scenario, but with the lines isolated at the source, and connected at the far end. Their will be a large signal, still at time D, because the signal will travel down the supposed-to-be-used pair to the end, then cross-talk over to the supposed-to-be-isolated pair at the far end. It will have similar strength and the same delay as the first case.

The reason near end works is the 2 x factor. If the pair is connected at the near end, you will see a peak in the time domain at T=0. If the line is connected in the middle, you will see the peak at T=D. If the lines are connected at the far end, you will see the peak at time T=2D.

I take it then that this means that a time domain channel FEXT measurement cannot be used to derive distance information?

If this is so, what information does the time domain plot of channel FEXT give - is it simply how FEXT is varying with time??

George G