The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Measurements >> Other Measurements >> extraction of poles and zeros from ac
https://designers-guide.org/forum/YaBB.pl?num=1119476961

Message started by ronny on Jun 22^nd, 2005, 2:49pm

Title: extraction of poles and zeros from ac
Post by ronny on Jun 22^nd, 2005, 2:49pm

Hi everyone,

does anybody know a possibility or tool which automatically extracts the poles and zeros from AC/XF analysis results? I'm using SpectreRF but the PZ analysis doesn't work right and isn't really powerful. So I'm looking for a workaround.

Thank you in advance!
Regards,
Ronny

Title: Re: extraction of poles and zeros from ac
Post by Andrew Beckett on Jun 23^rd, 2005, 10:38am

Ronny,

What do you mean it doesn't work right? That's not very descriptive. You need to make sure you've specified the input and output in order to capture the zeros - otherwise it can only compute poles (see spectre -h pz). Note also that IC5141 and later can do pole-zero cancellation for poles and zeros that are coincident (well, very closely spaced to allow for numerical inaccuracy).

Also, pz (like ac and xf) analyses are not SpectreRF analyses, but conventional spectre analyses. There is no periodic-pz analysis available.

Regards,

Andrew.

Title: Re: extraction of poles and zeros from ac
Post by ronny on Jun 23^rd, 2005, 2:00pm

Hi Andrew,

in fact the first problem with the PZ analysis is that no zeros are calculated. But I want to recheck the declaration of input and output! The second problem concerns the correctness of the results. We compared the results with the PZ analysis of another simulator (Eldo) and observed differences. After implementation in Verilog-A the results were verified. At the moment we can only use IC5.0. The third problem is in fact that we perspectively need a "periodic-PZ", because we analyse blocks of the RF frontend.

Regards,
Ronny

Title: Re: extraction of poles and zeros from ac
Post by Andrew Beckett on Jun 23^rd, 2005, 8:36pm

Hi Ronny,

Having done a number of comparisons myself between pz and other simulators, I've not seen problems with the accuracy of the results - but you do need to make sure you've specified the inputs and outputs correctly, of course.

What kind of blocks in the front-end are you analysing? Obviously the LNA wouldn't need a periodic analysis. Would you be analyzing the mixer, say? I'm asking so I can make an enhancement request for a ppz analysis based on a realistic example. I wondered a while back whether a periodic-pz analysis would be useful...

Regards,

Andrew.

Title: Re: extraction of poles and zeros from ac
Post by ronny on Jun 26^th, 2005, 1:50pm

Hi Andrew,

I investigated into the PZ analysis on friday again. The capture of the zeros is ok if a source is used instead of a port. Thank you for the note!

Now let me explain the remaining problems a little bit closer. The blocks of the frontend are for example a charge pump and a special power amp. The PZ analysis of a mixer is not a current task but maybe important in the future.

In the manual the PZ analysis is only recommended for small to medium circuits. This will be confirmed if I analyse the complex charge pump. The analysis dies after 3h and 272 poles with an arithmetic exception! The AC/XF analysis finishes after max. 30 seconds. There are a few poles which appear 20-30 times! Hmm?

A second problem I see is the following. Imagine, you have two or more blocks in your schematic and you want to capture the poles and zeros of the input-to-output path of only the first block. The poles of the complete schematic are calculated. Is it right? I cannot disassemble the whole schematic.

Because of these problems I think it would be a fast and efficient way to extract the poles and zeros from AC/XF results.

What do you mean, Andrew?

Thank you and regards,
Ronny

Title: Re: extraction of poles and zeros from ac
Post by Andrew Beckett on Jun 27^th, 2005, 11:18pm

Hi Ronny,

Are you using IC5141? As I mentioned, the pole-zero cancellation can help in this case - however, it may be that in your case the circuit is too big for the current algorithm. There were plans for a sparse solver as well as a detailed solver (we have detailed solver right now), but the sparse solver has not been implemented yet.

I don't know of a way of doing this from the ac results, sorry.

Regards,

Andrew.

Title: Re: extraction of poles and zeros from ac
Post by pancho_hideboo on Aug 30^th, 2005, 6:44pm

Hi.

> Because of these problems I think it would be a fast
> and efficient way to extract the poles and zeros from
> AC/XF results.

You can do this using Frequency Domain System Identification Toolbox in MATLAB.

http://elecwww.vub.ac.be/fdident/

Also PZ-Analysis for PSS/PAC/PXF is useful for RF behavioral modeling where power dependent frequency characteristics must be captured.