I think that i have problem with the convergence that some times simulator fails to solve the dc point so it fails to do ac analysis that i need to measure psrr

You can try a DC sweep from the maximum voltage to the minimum voltage. Use fixed voltage source at 3 V and another between it and the real voltage source, so your simulation starts at 3 V for a a sweep beginnig from 0. You can try using .nodeset also. https://info.efabless.com/knowledgebase/overcoming-spice-convergence-issues/

Bandgaps are notorious for having multiple stable operating points. You may need additional circuitry to force the bandgap into the correct DC state.

@Luis Henrique Rodovalho The inversion of the sweep from maximum to minimum worked well but the problem is the wrong dc solution affects the ac solution so i got an unrealistic result

You should improve your start-up circuit. It also helps with the DC solution. My go to choice for start up circuit is this one, from this self-biased current source: https://www.researchgate.net/publication/224647766_Fast_startup_CMOS_current_references