Elegant users forum

Hi, Michael,
I need to split the CSBEND into many slices. However, I found that with the increase
of the number of slices, the vertical tune also increased. For example, if I split the
dipole into 1000 slices, nu_y increased about 0.0101%, for 10000 slices, nu_y increased
about 1.006%. I have tried splitting by hand and using the "divide_elements", this deviation
appeared in both situation. I also tested that in AT and nu_y stayed unchanged. I wonder how
to avoid this deviation in elegant. The attachment is an example.

Best regards.

Zhe,

I can confirm the problem but as far as I can tell, it is caused by limited numerical precision, since the tunes are computed by accumulating the phase advance element-by-element. I'm not sure how AT avoids this, but perhaps it uses the one-turn matrix instead; in that case, I'd expect the tunes to be inconsistent with psi(s).

Out of curiosity, why do you want to divide elements into 10k pieces?

--Michael

Hi, Michael,
I'm studying the fringe field effect of the longitudinal gradient magnets, so I want to use the
splitting method as a bench-mark model. I split the field 1 mm per slices, so the slice number
can reach about 2k (10k is just a test).

I'm also going to use the GGE and BRAT model, however, I’m confused about the coordinate of BRAT,
it is said “The coordinates of the field map are right-handed system (x, y, z), where z is along the
length of the magnet, x is to the right as viewed along the direction of beam propagation, and y is up.”
But I think that y should be down to be right-handed.

By the way, I'm Nan Li who temporarily using Zhe's account. Thank you for your reply!
Best regards.
Nan

Nan Li,

I can see how that would be confusing. It should say that (z, x, y) forms a right-handed system.

--Michael