[mesa-users] Timestep control

[Josiah Schwab]

In-reply-to: <3A079F20-5FCB-4940-8B87-5996B3E5C42C at kitp.ucsb.edu>

Hi Bill,

> I stated it for mass loss since Mathieu had asked for a limit rated to
> mass loss timescale.

> For your case, I think we are already covered with "delta_lg_star_mass_limit",
> although perhaps the name isn't the best.   Note that we also have limits for
> the change in Mdot ("delta_mdot_limit").

Thanks for reminding me of this control.

Mathieu's case is also covered by this control, which enforces

    abs(log10_cr(s% mstar/s% mstar_old)) < delta_lg_star_mass_limit

which for small mass changes is 

    abs(dt*Mdot) / (M*ln(10)) < delta_lg_star_mass_limit

and so is equivalent (up to a dimensionless constant)

    dt / (M / abs(Mdot) < ln(10) * delta_lg_star_mass_limit

to the new control.


As part of my campaign for fewer MESA controls,  I say either:

  (1) eliminate delta_lg_star_mass_limit in favor of
  dt_div_dt_mass_change

  (2) just use the existing control


I don't like adding a semi-redundant control.

Josiah

P.S.  I vote* (1), but maybe I'm biased, because every time I use
delta_lg_star_mass_limit, I always screw up that ln(10) :)

* Not that this is a democracy :)