Transient simulation with noise does not converge

Hello,

I am running transient noise simulation on my circuit (even a boostrapped switch)

i am using moderate errpreset, reltol, vabstol and iabstol and gmin are all the default values.

the transient noise simulation does not converge and I get the following errors: 

Notice from spectre at time = 156.261 us during transient analysis `tran'.
Newton iteration fails to converge at time = 156.261 us step = 2.00577 as.
Disaster recovery algorithm is enabled to search for a converged solution.
Error found by spectre at time = 180.834 us during transient analysis `tran'.
ERROR (SPECTRE-16192): No convergence achieved with the minimum time step specified.

Meanwhile I found that there are many warning about mos occurred after noise simulation selected. Without noise there are no such warning and simulation is ok.

WARNING (CMI-2682): I18.I21.I11.M3: The bulk-drain junction forward bias voltage (3.3185 V) exceeds `VjdmFwd' = 938.719 mV. The results are now incorrect because the junction current model has been linearized
WARNING (CMI-2682): I157.I30.I23.MU1_17-M_u2_turbo_m24: The bulk-drain junction forward bias voltage (1.17157 V) exceeds `VjdmFwd' = 963.055 mV. The results are now incorrect because the junction current model has been linearized

Do you have any advice on how to achieve convergence? I do not want to use a liberal setting as that gives unreliable results for the amount of precision we need in capturing voltage noise values. 

I have tried different integration methods (gear2, trapgear, euler) as well as larger gmin but neither works.

thanks,

Guozhu