Is Momentum Aperture really deterministic?
Posted: 21 Feb 2019, 18:13
Hi,
This may seem like a philosophical question but I am having some substantive issues with calculating the MA for the ALS-U storage ring. This is a very nonlinear lattice with strong bends and sextupoles, not to mention that the fractional horizontal and vertical tunes are almost identical. I set the parameters, including correcting the tunes, before doing the momentum aperture calculation, sometimes using elegant and sometimes using Pelegant. As a result the last significant digit in the quads used for tuning can be different. I was shocked to find that only switching between these two sets of parameter files can give significant differences (as big as 10%) in the momentum aperture in certain locations. There don't seem to be any random number inputs, so I suppose this is just a consequence of tracking chaotic orbits.
I have been trying to make the calculations go quickly, so in momentum_aperture I usually use step_back=1 and splits=1. Is there a more reliable way you would recommend? Maybe exploring the boundary with at least a few extra passes would help make the result less jittery? It might help to allow for a more probabilistic definition of the boundary, although that would be a more expensive calculation. I am concerned about the size of the variation because I am using the MA as an input for calculating losses with touschek_scatter, and it makes the default value of 0.85 for Momentum_Aperture_scale seem risky.
I am attaching some files and results using the SPEAR lattice, it's not as dramatic as the ALS-U runs but I think it makes the point about how small changes can impact the momentum aperture. I added tiny (about 10^-9 T in one case, 10^-12 T in the last case) quadrupole errors and still see some impact on the MA. I did have to set a very large physical aperture to get that kind of impact.
Thank you,
Gregg
This may seem like a philosophical question but I am having some substantive issues with calculating the MA for the ALS-U storage ring. This is a very nonlinear lattice with strong bends and sextupoles, not to mention that the fractional horizontal and vertical tunes are almost identical. I set the parameters, including correcting the tunes, before doing the momentum aperture calculation, sometimes using elegant and sometimes using Pelegant. As a result the last significant digit in the quads used for tuning can be different. I was shocked to find that only switching between these two sets of parameter files can give significant differences (as big as 10%) in the momentum aperture in certain locations. There don't seem to be any random number inputs, so I suppose this is just a consequence of tracking chaotic orbits.
I have been trying to make the calculations go quickly, so in momentum_aperture I usually use step_back=1 and splits=1. Is there a more reliable way you would recommend? Maybe exploring the boundary with at least a few extra passes would help make the result less jittery? It might help to allow for a more probabilistic definition of the boundary, although that would be a more expensive calculation. I am concerned about the size of the variation because I am using the MA as an input for calculating losses with touschek_scatter, and it makes the default value of 0.85 for Momentum_Aperture_scale seem risky.
I am attaching some files and results using the SPEAR lattice, it's not as dramatic as the ALS-U runs but I think it makes the point about how small changes can impact the momentum aperture. I added tiny (about 10^-9 T in one case, 10^-12 T in the last case) quadrupole errors and still see some impact on the MA. I did have to set a very large physical aperture to get that kind of impact.
Thank you,
Gregg