Different modeemittances as calculated with atx.m for the same ring with different starting points! about at HOT 4 CLOSED

There is some typo in the matlab file I posted above, please see the attached file, the suffix of the file needs to be changed to .m.
test_emit_esrf_1.txt

from at.

Sorry, I found out the numerical difference stepsize need to be set properly, since the default value 6.055454452393343e-006 could be too large for the concerned calculation, for example
global NUMDIFPARAMS;
NUMDIFPARAMS.XYStep = 1e-8;
NUMDIFPARAMS.DPStep = 1e-8;

from at.

@z-duan : Thanks for giving the solution so rapidly! According to your findings, I think we should reduce the default step size to avoid falling again in that trap. Furthermore, I could not find any justification for the very strange default values. I'll prepare a pull request to clarify this.

from at.

I encountered a similar problem with atx on our lattice (attached):

load('ring_atx_wrong.mat')
[lo,pa]=atx(ring,0);
 pa.modemittance

gives 7.8269e-06, while:

ats=atsummary(ring,'NoDisplay');
ats.naturalEmittance

gives 4.5276e-09, which is much more reasonable. As @z-duan said it can be solved by just:

global NUMDIFPARAMS;
NUMDIFPARAMS.XYStep = 1e-8;
NUMDIFPARAMS.DPStep = 1e-8;

I wonder if maybe the default should be 1e-8?

ring_atx_wrong.txt
Zeus

from at.