If you write good enough behavioral models, the modulator won't require spectreRF (however, some measurements might be faster with spectre RF).

That strongly depends on what you're trying to do.

If you want system simulations that can be simulated together with transistor level parts - use VerilogAMS.
For pure system level simulations with much IP reuse like modulated signals and so on - use Matlab.
I personally dislike ADS, but that's my own oppinion.

It all depends on how good you are writing models and how much control over your model's behaiour you desire...

spectreRF is not a language, it's a simulator.
"Hidden state" is  a phenomenom that occurs during periodic steady state analysis.
An accumulator behaves like a capacity ... that should be no problem to model.
However I thought we were talking about MIXED SIGNAL simulations.

And no, you can't use the Verilog-AMS language with the spectreRF simulator. (just verilogA with respect to hidden states)

module try(out_sin);

output out_sin;

electrical out_sin;

real count, period, cos_angle;

analog begin
 @(initial_step("static") or initial_step("pss") ) begin
   count = 0;
   period = 1 / 10M ;
 end  
@(timer(period/10, period)) begin
 count = count + 1;
end  
V(out_sin) <+ cos(2*3.14*1M*count);
end
endmodule

Geoffrey_Coram wrote on Feb 12^th, 2008, 6:41am:

It's (* instrument_module *) in front of the module name, see http://sourcelink.cadence.com/docs/db/kdb/2005/July/11187476.html.