Tunes from tracking vs twiss params

Moderators: cyao, michael_borland

Post Reply
twatanabe
Posts: 22
Joined: 15 Jan 2009, 20:30

Tunes from tracking vs twiss params

Post by twatanabe » 12 Jul 2010, 03:39

Hi Michael,

I successfully got almost perfect consistent between tunes from tracking and those from twiss params for some lattice, thanks to your help before. Now I just did the same thing for another lattice (with combined mags and other changes in magnetic strengths etc), then got completely different tunes between the two; tracking and twiss params. As far as I checked, the tunes given by twiss params are right numbers, but also the Fourier analysis using 'sddsnaff' works right (i mean tracked particles saved in the .twi file really oscillates as the sddsnaff gives).
Could you please check the attached file and see why particles in the lattice seem to oscillate at wrong tunes?

Thank you and Best,

Takahiro
Attachments
sp8Track.tar
(10 KiB) Downloaded 833 times

michael_borland
Posts: 1927
Joined: 19 May 2008, 09:33
Location: Argonne National Laboratory
Contact:

Re: Tunes from tracking vs twiss params

Post by michael_borland » 12 Jul 2010, 12:51

Takahiro,

One problem is that the N_KICKS parameter for the CSBEND elements is left at the default value. You should set it to something like 20 to 40 to get good results.

The second problem is more difficult. In particular, even when I set N_KICKS to a sufficient value, I don't see good agreement between tracking and the twiss parameter result for the vertical plane. The agreement is very good if I change all the CSBEND elements to SBEND elements, and for lattices that have no gradient in the dipole. It seems to be a bug but I'm not sure for which element. At this point, I would tend to trust the tracking results and not the Twiss output results.

The result is puzzling because the transport properties of the SBEN and CSBEND elements agree quite well. I'm working to diagnose the problem.

--Michael

michael_borland
Posts: 1927
Joined: 19 May 2008, 09:33
Location: Argonne National Laboratory
Contact:

Re: Tunes from tracking vs twiss params

Post by michael_borland » 19 Jul 2010, 08:25

Takahiro,

I've come to the conclusion that the reason for the discrepancy in the vertical plane is the lack of third-order terms in the dipole matrix. When K1 is non-zero, there are many such terms, which elegant doesn't include in the analysis. I'm guessing that they affect the vertical plane and not the horizontal because the vertical beta function is much larger in the dipole than the horizontal beta function.

I'm looking into adding the third-order dipole matrix. Meanwhile, tracking with CSBEND will include these terms.

--Michael

twatanabe
Posts: 22
Joined: 15 Jan 2009, 20:30

Re: Tunes from tracking vs twiss params

Post by twatanabe » 19 Jul 2010, 23:53

Michael,

Thanks for your quick investigation. Yes, as you mentioned, I forgot to optimize N_KICKS when I switched my lattice from SBEN to CSBEND. Now the horizontal nonlinear chromaticities almost agree between tracking and twiss params by using N_KICKS=40.
Regarding the vertical plane, I look forward to the modified CSBEND you are working on, and also hope that the horizontal chromaticities would match even better like they do in usual elegant runs without combined mags :-)

Thank you,

Takahiro

Post Reply