Munaut Sylvain

@h_vogt : Ah no, you're right if I disable inp_fix_agauss_in_param then the example from 18.2 works but the IHP models don't work anymore. ( Although to be frank I don't exactly understand why ... If you decide not to modify ngspice, at least that behavior should be documented and the examples in the doc fixed to reflect that.

But why do the .parm to .func at all ? If I disable that function ( inp_fix_agauss_in_param ) it fixes the issue ...

No. I figured out the wrong apostrophe myself. This is the output : ( ngspice-43 but also tested git master ). v(1) = -1.20000e+00 v(2) = 1.200000e+00 v(3) = 1.200000e+00 v(11) = -8.80126e-02 v(12) = 1.521919e+00 v(13) = 1.089791e+00 v(14) = 5.222066e-01 v(15) = 5.640351e-01 v(16) = 8.792671e-01 v(17) = -3.34773e+00 v(18) = -1.20073e+00 v(1) and v(2) should be the same, they are not. v(14) and v(15) should be the same, they are not. v(16) should be twice v(14), it's not.

Hi, Yesterday while looking into the IHP SG13G2 PDK model behavior, we uncovered some unexpected (and undocumented AFAICT) behavior of ngspice. Specifically it seems that if a .param has a statistical distribution function ( like gauss) in it, instead of being evaluated once at startup like section 18.2 of the manual suggests, it will be evaluated every time it's being used. In the IHP PDK, you have lines like : .param sg13g2_hv_pmos_vfbo ='gauss(sg13g2_hv_pmos_vfbo_norm, 0.004, mc_ok)' that set...

I have a US keyboard so ... not really.

Segfault after calling XTextWidth with NULL font if Xft is enabled