# Minimum time increment can't be lower than 1e-5

Hi,

I’ve noticed that if I set the minimum time increment for the *Static step lower than 1e-5 x time period of the step, the solver ignores this and doesn’t reduce the time increment below 1e-5 anyway. Because of that, it’s possible to end up with “increment size smaller than minimum” error message even when the current time increment is far from what was set as minimum. For example, in one analysis I set minimum time increment to 1e-9 (time period was 1) and I got the aforementioned error message when time increment reached only 0.329e-5.

Is that a known bug ? Can you confirm this ?

1 Like

have you checked the manual ccx?

Yes, here’s what it says about the minimum time increment:

Default is the initial time increment or 1.e-5 times the time period of the step, whichever is smaller.

No information about any limitations like that, the documentation only refers to the default value of minimum time increment.

I can confirm this behaviour for static steps.

1 Like

I’m getting smaller time steps with a period of 1 by also specifying a small maximum like this:

*STEP,NLGEOM=YES,INC=1000000000,AMPLITUDE=STEP
*STATIC
1E-09,1,1E-09,1E-08

But I guess that’s not really useful because it’ll be ridiculous to never let it exceed 1E-8.

Without a maximum the error message is different:

`````` *ERROR in nonlingeo
increment size smaller than one millionth of step size
increase increment size
``````

ending with:

`the increment size exceeds the remainder of the step and is decreased to 0.000000e+00`

Maybe it’s a limitation due to single precision which is about a million parts?

Dear CalculiX users,

my intention is that the lowest minimum time increment is 1.e-6 of the time step for a *static problem and 1.e-10 for a *dynamic problem. This was not very consistent so far (sometime 1.e-5 was used) but I changed that now in my development code. The reason for these limits is that if the calculation does not converge for a time increment that small, it usually means that something is wrong with the model and it will seldom converge for an even smaller value. Furthermore, the limits can be circumvented by subdividing the step into smaller steps.

Guido

1 Like