TDR math question

[joe smith]

I calibrated my VNA using SOL and left the load inserted.  I was expecting it to measure something very close to 50 ohms at the the reference plane (time=0).   I have attached the plots from my software along with  METAS VNA Tools.   Note how the METAS software at T0 is > than 50 ohms.  I wonder if this offset is due to how you mirror the data. 

"In the low pass mode, the data of the harmonic grid is mirrored (conjugate complex) to
the negative frequencies and the DC point is added."

I have included the Touchstone file as well.

[Michael Wollensack METAS]

Hi,

>  I was expecting it to measure something very close to 50 ohms at the the reference plane (time=0).

Wrong expectation. If you let the load connected and remeasure it you will see drift and noise effects. The drift can happen as well before your reference plane.

> I wonder if this offset is due to how you mirror the data.

No. By the way you can added the DC point to your Touchstone file yourself.

Regards

Michael

[joe smith]

I think there is some confusion.  The data was captured and saved to a Touchstone file.  It was then processed with METAS and my software.   Any drift or noise would not explain why a single file, processed with two different programs would have  such an offset at the reference plane.   Because the load standard that was used to calibrate the VNA was left in place and measured to create the Touchstone file, and because it was treated as a perfect 50 ohms, I would expect the reference plane (T0) to be very close to 50.  This is exactly what I see when I process the data.

[joe smith]

Let's try a different approach.  In this example, I have synthesized a very close to perfect load that I swept to 9GHz.  The Touchstone file was created without a VNA, removing your thoughts of drift.  Notice how at T=-20nS we see the 50 ohms.  No matter what I do, the negative side of the data set will always start at rather than T0 (the reference plane).    I downloaded CMT's software and tried it with the same settings and get the same results.  So, the problem is not unique to METAS.   Still it makes no sense from a physical standpoint.  If the VNA was cal'ed with a perfect 50 ohm load, the reference plane is not prior to T0.    I have included the Touchstone file.  

 

I have added was modeled.  

[joe smith]

Let's try locking the Mag Phase to -10db @ 0 deg and see what happens.   Notice how both programs will still not show a stable, correct impedance at the reference plane.  

Touchstone file is attached.   Hopefully this better explains my original question.

[Michael Wollensack METAS]

Let's assume you have an ideal open at the reference plane. So your S-Parameters are 1.0+0.0i.

The the Low Pass Step will go from 0 to 1. The value which you will have at T0 depends on your applied window function and if you provided the DC point or if it was extrapolated. It will depend as well on the used integral function. The last value of the integral of the time domain data will be exactly 1.0 independent of the used window function.

If I take a look at your -10dB data then I've the following comments:

1. Take a look at the Low Pass Impulse response with a rectangular window function. Only the value at T0 is 0.31623 all other are 0. That's what you and I would expect. Right?

2. In the Impedance world the T0 point is 92.248 Ohm and all other 50 Ohm.

3. The Low Pass Step response is not a simple sum up of the impulse response. I'm using cumulative trapezoidal numerical integration, e.g.:

>> d = [0 0 0.31623 0 0];
>> cumtrapz(d)

ans =

         0         0    0.1581    0.3162    0.3162

Hope this helps.

Regards

Michael

[joe smith]

We had a discussion about this topic on EEVBLOG as well where I further clarified my thoughts about it.

https://www.eevblog.com/forum/rf-microwave/vna-for-cable-characterization/msg5684457/#msg5684457