Question about causality in s-parameters

[yefj]

Hello, I am using VNA with ifbw=3KHz span 10MHz to 70GHZ step 10MHz.
I am measuring differential lines SDD11 for TDR uding IFFT.
If my IFFT TDR windows is 100ns then an event in 120ns will go to -20ns thus will create causality issue.
Anothe cause I know is deembeding creating ringing ,but i am not doing deembeding.
But this is all good in theory. In the lab what could cause causality issues in the s-params? how can I handle them?
Thanks.

 

[sophiecentaur]

f my IFFT TDR windows is 100ns then an event in 120ns will go to -20ns thus will create causality issue.

Nothing to worry about here, I think. Basically the processing takes time so you cannot 'know' about the future, you can only 'know' that Event B occurred a certain time after Event A. More formally, the integrations that the process uses will have an arbitrary constant and that constant would have to correspond to the answer coming out some time after the end of the set of samples used for the measurement.

 

[yefj]

Hello sophiecentaur,the main issue is the cables connected to the vnacould be making this causality issue.
I am using n2814a cable.
How can I test test them to see that they are properly functioning?
Thanks.

https://www.mouser.com/ProductDetai...wL09DEmHZq-VAJGHVgD7x8ZPrOI9cR2NK9ZIiOffq6QV3

 

[sophiecentaur]

How familiar are you with using Time Delay Reflecfometry? Interpreting what you see can be difficult; if you are trying to draw conclusions from what you see, you need to be familiar with the technique.
My experience of vector analysis and tdr is fairly dated but the basics have to be the same. The cables would be the most reliable parts of your setup unless you happen to have bad terminations; swapping cables and 'ends' sometimes reveals problems. Also a good 50Ohm load should reveal anything about the cable.

Why would you imagine that any signal passing through any circuit / cable would violate causality? A TDR trace will always have a delay relative to the input step. If you measure a signal at the input and at the output then how could you ever get a negative delay? It's normal to use a suitable delay in your input signal path to equalise the phase rotation (signal delay) so that the phase / frequency response is near constant over the frequency range. This link is a short summary of what you need to be considering. S parameters involve both amplitude and phase.

 

[yefj]

Hello , I did PCB tests in "oven" with long cable like 1m long cable , the causality was very bad like 30% instead of 80%.
So first of all I dont know if my calibration got ruined when the cables are at 85C temperature oven connected to DUT.

I am trying to think of a strategy.
few questions:
1.first is to connect a precition load 50 Omh and twist the cable and see if the S11 amplitude changes.
By how much I need to see a change in phase or amplitude to say that the cable is no good?

2.the cable inside the oven connected to DUT .How can I know that the temperature is the reason the cable is not calibrated anymore?

3.I tried to search the relevant data about N2814Ain google and they dont give you the information regarding what bending angle or temperature the cable can wistand.
So its just asking keysight directly or there is some cable library you could reccomend?
Thanks.
https://www.keysight.com/il/en/product/N2814A/ultra-high-performance-cable-65-ghz.html

 

[sophiecentaur]

How can I know that the temperature is the reason the cable is not calibrated anymore?

You can (as I already suggested) try swapping the cable (DUT and compensation cable). and see the effect before and after any disturbance (heating / twisting). If you bear in mind the speed of your signals will be something like 2c/3 then you can get an idea of what could be causing the delay changes you are getting.

Perhaps you could say exactly what you mean by a 'causality problem'. I have been assuming that your results tell you the output signal emerges before the input signal. Do you understand that is not possible unless you are wrongly calibrating the basic delay?

You should take the trouble to draw (for PF and yourself) a diagram of precisely what you are doing. It will get you better answers and help you to order your thoughts.

By how much I need to see a change in phase or amplitude to say that the cable is no good?

I think you need to do some more basics and see what the variation of S11 with frequency is telling you about the time delay. I am sure your cable will have a temperature range spec - you just need to look deeper in the manufacturer's information.

 

[Baluncore]

With TDR of any form, everything beyond the second step or kick, becomes indeterminate pinball insanity, in a hall of mirrors.

 

[sophiecentaur]

everything beyond the second step or kick, becomes indeterminate

An example of under-sampling and aliases, I think. Important to have good terminations to limit this.