I developed this code to generate a random noise. However, it seems it will keep constant after initial change, why?

module rj_src (rj_out);
output rj_out;
electrical rj_out,rj_int;
parameter real rj_amp = 1 from (0:inf);
parameter real rj_tau = 10m from (1p:inf);
parameter integer seed = 10 from (1:inf);
real next;
real noise;

analog begin
 
  // this rj_src will generate a random noise voltage and sum with sin_clk_src's output;
  // Thus, jitter is modeled as additive noise;
  @(initial_step) next = $abstime + rj_tau;

  @(timer(next))
     begin
        noise = rj_amp*$rdist_normal(seed,0,1);
        next = next + rj_tau;
     end

  V(rj_int) <+ noise;
  V(rj_out) <+ V(rj_int) + I(rj_out)*50;

end
endmodule

The model would be a bit more efficient if you make rj_int a local real variable rather than a node.

-Ken

Why would you, does the REAL noise nicely prevent double crossing?

one way would be to include a delay in the detector, so that ONCE a crossing has happend, a certain amount of time must pass before another one.. that might be more realistic..
for the stats you want to measure it might be more interesting to report any times when there would have been double crossing.. even if the detector masks it in the output.
have fun