SPS Impedance Meeting

Minutes November 19th, 2008

Present:  Rama Calaga, Fritz Caspers, Elias Métral, Federico Roncarolo, Giovanni Rumolo, Benoit Salvant, Carlo Zannini, Bruno Zotter. 

   Agenda

 

- Elias Metral 

           "Procedures for frequency and time domain EM Simulations in asymmetric structures"  (pdf)
           (not discussed during this meeting)

 

-  Benoit Salvant 

"Trying to understand what Fourier Analysis Bruno Spataro is performing to get the impedance"  (slides, minutes)

 

Back to top

    Minutes (many thanks to Federico for his notes!)

Presentation by Benoit:

- The goal of this work is to understand the routine that is used by Bruno Spataro to get the impedance from the wake field, since it appears to be very different from a plain DFT. This FORTRAN routine is taken from Yong Ho Chin's ABCI older version of the code.

- After decomposing every step of the code, the routine indeed performs a DFT, but with the following additional steps:

1. The wake has to be deconvoluted from the gaussian bunch source (both in amplitude and time delay)
2. A 4-term-Blackmann-Harris-window (or gate) is applied on the wake
3. A zero-padding is applied (the length of the wake is multiplied by ~8)

- If it is clear that item 1 should be applied, as Headtail needs the point charge impedance, one may argue (and did argue...) whether using items 2 and 3 on the wake is legitimate.

- The general opinion of the colleagues present in the room was that

1. zero padding should never be applied to get the impedance spectrum (the effect on the impedance is rather small anyway, but it gives the fake impression of a much better resolution than it actually is)
2. if windowing is applied, it should be applied with great care (the effect of windowing is quite big on the signal).

- Ongoing work on simple models of impedance where both the wake and the impedance are known analytically shows that windowing widens the impedance resonance peak.

- Bruno Z. mentioned that he would prefer if Headtail used source slices that are not point charge, but gaussians or triangles, as it is done in TRISIM. This will be discussed at the next meeting in December.

- Fritz mentioned that all this work on Fourier Transforms can be performed directly on a spectrum analyzer.

- In the past, Federico applied some very basic zero padding to bpm data in order to improve the frequency resolution of the fft, and he was happy with the result. However, he said he was only interested in identifying the frequency of the main peak of the spectrum , disregarding the peak shape and the introduction of fake harmonics at other frequencies. Rama (and also Rogelio after the meeting) insisted that zero padding does not lead to any gain in information from the raw fft. Zero padding is equivalent to fitting the available fft points with a sinc function, giving the impression of a nice smooth result.

             

.

Back to top