Simulated data explodes quickly

I’m using dynare to estimate a DSGE model. During the estimation, I use mode_compute=6, and I finally get some results. In order to verify whether the results reasonable, I want to simulate data using my model and estimate the model again with the simulated data. But the problem is, the simulated data explodes very quickly, e.g., from the 5th period, unless the standard error of innovation is very very small, e.g. 0.000001.
Anyone knows what’s wrong with the model?

Are you simulating at 1st or 2nd order ?

Explosive paths are frequent at 2nd order; you can use the new “pruning” option to circumvent that problem.

Using 1 order solves the problem!