Possible low-pass filter

[michael_borland]

Alex,

Regarding the high-frequency cutoff for LSC, there is one imposed by your choice of the number of bins, just from Nyquist. There are also parameters in the element definition to allow you to vary the cutoff, but by default there isn't one.

--Michael

[headdoggy64]

Michael,

Regarding to the number of bins in LSC. I tried N=1024,2048,4096 but it seems it doesn't affect the cutoff value.
Just now I reran the simulation with the additional parameters you suggested, but the cutoff hasn't shifted. (at least not that I can notice on the plot)

Thanks,
Alex

[headdoggy64]

Michael,

By the way, have you used some SDDS tool to calculate the bunching factor?

Thanks,
Alex

[michael_borland]

Alex,

I'm not trying LSC at this point, just looking at the initial particle distribution. I'll take up LSC tomorrow.

I wrote a C program to perform the bunching factor computations. It's much faster than using a script and will be included in the next distribution of elegant. Code is attached in case you build from source.

--Michael

[michael_borland]

Alex,

Here are updated plots for random and quiet-start distributions, extending the number of particles another order of magnitude. While both improve, the Halton-sequence-based distribution clearly performs better, both in terms of the noise level for a given number of particles and in terms of reliable improvement in the noise level as the number of particles increases.

I don't think a revision of the random number generator is going to improve the distributions enough to equal the Halton sequence result, which is after all constructed specifically to remove noise.

Bunching factor for random-number-based distributions.

Bunching factor for Halton-sequence-based distributions.

--Michael

[michael_borland]

Alex,

I did some runs with LSC on (1024 bins) and 1M particles, scanning p from 100 to 1000 in steps of 5. The results seem regular and very different from yours. The trend seems to make sense, with LSC basically becoming less and less influential as the momentum increases. I didn't perform any analysis to make this quantitative.

--Michael

Bunching factor with LSC as p is varied.

[headdoggy64]

Michael,

Thank you very much for your replies!

I compiled the new sddsbunchingfactor tool and compared it to my script output. For some reason they do not match very well. I am still looking into this.
But! With the new tool I reran my scan and I can confirm the LSC plot that you got! Here is the slice of gamma 850-900 with the step of 1. I will post more here when I make further progress.

Thanks again!

P.S. Could you please show me how to make an image plot above within SDDS toolkit?
Alex

[michael_borland]

Alex,

Here are the files and scripts I used to perform the runs and analysis. Script "process2" shows how to make the color map using sddscontour.

--Michael

[headdoggy64]

Hi Michael,

Thanks a lot. I believe this topic is obsolete since there will be a built-in SDDS tool for computing the bunching factor
https://www.aps.anl.gov/Accelerator_Sys ... f=17&t=567

Alex

[michael_borland]

Alex,

Thanks, I added a comment on the referenced post.

--Michael