[mesa-users] Hook in HR when hitting max_age

[Josiah Schwab]

paxton at kitp.ucsb.edu writes:

> So smaller max timesteps make the problem worse!!!!
>
> smaller dt => worse approx for eps_grav?

I think this depends on how one defines "the problem" :)

The way I interpreted my result was that the blue line (Radek's inlist),
where the final step timestep was "suddenly" logdt ~= 6.5, agrees with
the purple line, where the model has run at a timestep of logdt ~=6.5
for the previous 50 timesteps.  So that's reassuring: there's no
"transient" problem associated with hitting a final (shorter) timestep.

As I decrease the max_timestep, all the curves dstart to end up in the
same place.  So eps_grav is converging (at the 1e-4 level), at
max_timestep of 1e5 yr or so.  That's also reassuring.

A big reduction in timestep gives a bigger "hook", but the hook size
  a) converges to a constant value as dt shrinks
  b) is consistent with a model using that reduced timestep

So I think this is saying smaller dt => better approx for eps_grav, and
is consistent with your previous messages.

Josiah