As I expect most people reading this forum know, option 010 on most HP/Agilent VNAs allows one to perform time domain measurements. But the temporal resolution is limited by the frequency range of the VNA, so it is impossible to isolate two reflections close in time, unless one has a VNA going to a very high frequency.

This paper:

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?reload=true&arnumber=75298&contentType=Journals+%26+Magazines

(there was a copy here http://www.cis.syr.edu/~tksarkar/pdf/1991_Mar.pdf when I first wronte this post, but now it seems to have gone. A bit odd)

*"Time-domain measurements with the Hewlett-Packard network analyzer HP 8510 using the matrix pencil method"*,

which was published in 1991 in an IEEE journal, describes a method which appears on the face of it to give far better resolution than the inverse Fourier Transform (IFT) method implemented on the HP/Agilent VNAs as option 010. The results from a stepped impedance airline terminated in 50 Ohms look pretty impressive, with far better resolution than using the normal IFT for the same bandwidth.

The method was suggested to the authors by a Dr. Rytting at Hewlett-Packard, and the paper concludes the method could be implemented in the firmware of a VNA. But I've never seen the method impemented, so I'm guessing there are some problems with it. I'm just interested in what those problems are.

Dave

Edited by: drkirkby on Feb 10, 2013 2:50 PM

Edited by: drkirkby on Feb 11, 2013 12:20 AM

This paper:

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?reload=true&arnumber=75298&contentType=Journals+%26+Magazines

(there was a copy here http://www.cis.syr.edu/~tksarkar/pdf/1991_Mar.pdf when I first wronte this post, but now it seems to have gone. A bit odd)

*"Time-domain measurements with the Hewlett-Packard network analyzer HP 8510 using the matrix pencil method"*,

which was published in 1991 in an IEEE journal, describes a method which appears on the face of it to give far better resolution than the inverse Fourier Transform (IFT) method implemented on the HP/Agilent VNAs as option 010. The results from a stepped impedance airline terminated in 50 Ohms look pretty impressive, with far better resolution than using the normal IFT for the same bandwidth.

The method was suggested to the authors by a Dr. Rytting at Hewlett-Packard, and the paper concludes the method could be implemented in the firmware of a VNA. But I've never seen the method impemented, so I'm guessing there are some problems with it. I'm just interested in what those problems are.

Dave

Edited by: drkirkby on Feb 10, 2013 2:50 PM

Edited by: drkirkby on Feb 11, 2013 12:20 AM

It does provide an interesting way of getting time domain from an VNA.

As I see it the solving for eigenvalues is what takes time, but by using better algoritms (Jacobi-Davidson) and having Moores law applied to that 22 year old article it should be possible to do the conversion fast (for reasonable size matrixes) on an computer.

Being able to do this on S-Parameter files (Touchstone) would be an nice pice of software, or even better to be able to use this method for tuning of filters with an VNA would be great.

The big question: Is this method implemented in any of the VNAs avaible today or do they all use IFT?