CSR instability
Moderators: cyao, michael_borland
CSR instability
This is for Taiwan Photon Source 3 Gev ring. One superperiod is used for CSR tracking. It should be tracked for several turns, not just one superperiod.
-
- Posts: 1951
- Joined: 19 May 2008, 09:33
- Location: Argonne National Laboratory
- Contact:
Re: CSR instability
Ping,
Thanks for posting your files. One recommendation I have is to use the high_frequency_cutoff feature instead of the Savitzky-Golay smoothing. This gives better control over the noise cutoff frequency.
Here's an example of how to set the cutoff frequency:
Your bunch is gaussian with rms duration St and total duration 6*St. The characteristic frequency for a gaussian beam is fg=1/(2*pi*St) and the Nyquist frequency is fN=Nbins/(12*Sz). So the ratio of the characteristic frequency to the Nyquist frequency is approximately 2/Nbins. With Nbins=600 (as in your input files), this is 0.0033, so the dominant frequencies in the beam are well below the maximum frequency. (Which is good.)
Setting a cutoff of fc=fN/5 (high_frequency_cutoff0=0.2) is a good choice because it means that the highest frequency simulated is sampled with 10 time points. In this example, it also gives fg/fc=0.017, which means you will simulate frequencies that are up to 60 times fg. If you want to simulate higher frequencies, then you would increase Nbins (and the number of particles) but keep the cutoff fixed.
I've attached a PDF file that explains this a little more. I also modified your files. (I added the frequency cutoff. I also removed the rf cavity and radiation kick---these aren't needed when the beam only goes by once.)
--Michael
Thanks for posting your files. One recommendation I have is to use the high_frequency_cutoff feature instead of the Savitzky-Golay smoothing. This gives better control over the noise cutoff frequency.
Here's an example of how to set the cutoff frequency:
Your bunch is gaussian with rms duration St and total duration 6*St. The characteristic frequency for a gaussian beam is fg=1/(2*pi*St) and the Nyquist frequency is fN=Nbins/(12*Sz). So the ratio of the characteristic frequency to the Nyquist frequency is approximately 2/Nbins. With Nbins=600 (as in your input files), this is 0.0033, so the dominant frequencies in the beam are well below the maximum frequency. (Which is good.)
Setting a cutoff of fc=fN/5 (high_frequency_cutoff0=0.2) is a good choice because it means that the highest frequency simulated is sampled with 10 time points. In this example, it also gives fg/fc=0.017, which means you will simulate frequencies that are up to 60 times fg. If you want to simulate higher frequencies, then you would increase Nbins (and the number of particles) but keep the cutoff fixed.
I've attached a PDF file that explains this a little more. I also modified your files. (I added the frequency cutoff. I also removed the rf cavity and radiation kick---these aren't needed when the beam only goes by once.)
--Michael
- Attachments
-
- tnote2005-027.pdf
- Technical note describing setting of the binning and filtering parameters for CSR sims.
- (42.46 KiB) Downloaded 1240 times
-
- 79h2CSR4.txt
- Script file: I added making of a 2d histogram and contour plot.
- (848 Bytes) Downloaded 281 times
-
- 79h2CSR4.ele
- (954 Bytes) Downloaded 271 times
-
- 79h2CSR4.lte
- (5.99 KiB) Downloaded 263 times
-
- Posts: 1951
- Joined: 19 May 2008, 09:33
- Location: Argonne National Laboratory
- Contact:
Re: CSR instability
I used Pelegant on one of our computer clusters at APS to perform a simulation for ~100 turns (actually only 93 as I inadvertently stopped the job). This took several days on 32 processors (I didn't keep track of exactly how long it ran.) The results are pretty interesting, as the attached animated GIF shows. This is an unshielded case, which is why it looks so dramatic.
We have a program here at APS that calculates the shielded CSR impedance. (We'll be distributing this with the next release of elegant.) An initial check using this (via the ZLONGIT element) shows no issues.
--Michael
We have a program here at APS that calculates the shielded CSR impedance. (We'll be distributing this with the next release of elegant.) An initial check using this (via the ZLONGIT element) shows no issues.
--Michael
- Attachments
-
- 79h2CSR5.lte
- Lattice file
- (5.81 KiB) Downloaded 274 times
-
- 79h2CSR5.ele
- Command file
- (1.03 KiB) Downloaded 274 times
-
- Posts: 46
- Joined: 09 Jun 2008, 01:19
Re: CSR instability
Dear Michael,
I found your prescriptions on how to use HIGH_frequency_cutoff in CSRCSBEND in this old post.
I wonder if you have any similar guidance for the LOW_frequency_cutoff, in the same element. I could not find any discussion about this option in the forum.
Thank you in advance. Best,
Simone
I found your prescriptions on how to use HIGH_frequency_cutoff in CSRCSBEND in this old post.
I wonder if you have any similar guidance for the LOW_frequency_cutoff, in the same element. I could not find any discussion about this option in the forum.
Thank you in advance. Best,
Simone
-
- Posts: 1951
- Joined: 19 May 2008, 09:33
- Location: Argonne National Laboratory
- Contact:
Re: CSR instability
Simone,
I don't particularly recommend using the low frequency cutoff feature. Although it might in some sense imitate the effects of CSR shielding, I don't think it does so in a physical way. There is also a customized filter available with the WAKE_FILTER_FILE parameter, but my tests have indicated that this is also not a physical approach.
--Michael
I don't particularly recommend using the low frequency cutoff feature. Although it might in some sense imitate the effects of CSR shielding, I don't think it does so in a physical way. There is also a customized filter available with the WAKE_FILTER_FILE parameter, but my tests have indicated that this is also not a physical approach.
--Michael