When fitting a sincgauss model, if the broadening is not well measured (SNR is too low, initial guess value is too small, ... see #13 and #19) the returned uncertainty is a nan: e.g. [1.54e-17 +/- nan].
First thing to try is to change the initial guess on the broadening (by setting e.g. sigma_cov=20). If it does solve the problem then the SNR is good enough and the broadening is large enough to be measured.
If it does not solve the problem, using a non-broadened model like sinc is a very good alternative. If the broadening cannot be measured (see #13) the sincgauss model is an over-parametrized model and it just shouldn't be used.
When fitting a
sincgaussmodel, if the broadening is not well measured (SNR is too low, initial guess value is too small, ... see #13 and #19) the returned uncertainty is a nan: e.g.[1.54e-17 +/- nan].First thing to try is to change the initial guess on the broadening (by setting e.g.
sigma_cov=20). If it does solve the problem then the SNR is good enough and the broadening is large enough to be measured.If it does not solve the problem, using a non-broadened model like
sincis a very good alternative. If the broadening cannot be measured (see #13) thesincgaussmodel is an over-parametrized model and it just shouldn't be used.