The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design Languages >> Verilog-AMS >> problem with $realtime ???
https://designers-guide.org/forum/YaBB.pl?num=1149077059

Message started by rajdeep on May 31^st, 2006, 5:04am

Title: problem with $realtime ???
Post by rajdeep on May 31^st, 2006, 5:04am

Hi all, I am trying to model a sawtooth generator using verilog-A (not AMS, so I cannot use idtmod). The code looks like :

@(timer(0,period)) begin
offset = $realtime;
slope= 5e7;
end

V(ramp_out) <+ slope*($realtime - offset);

This works fine for a simulation time less than 10us. But once I set the simulation time somewhat higher than that, the output just goes bizzare!!! Number of triangles in the o/p starts to get flattened out .
I'm not sure why it is happenning??? Does $realtime have anything to do with the simulation time???

Plz help
Rajdeep

Title: Re: problem with $realtime ???
Post by rajdeep on May 31^st, 2006, 5:24am

Sorry folks!!! I got the solution. I have to increase the timestep of the simulation to the order of ns ,
But still can anybody suggest a way to circumvent this type of problem, where the inaccuracy is due to timestep of the simulation.
I forgot to mention , the period used in the code is 0.5us.....that's why probably a 1us simulation time step was not enough.

Rajdeep

Title: Re: problem with $realtime ???
Post by bernd on May 31^st, 2006, 6:11am

I assume you have set errpreset to liberal,
with modifies the parameter reltol and sets the
integration method to 'gear2' which damps the
signal.

Try with errpreset == moderate

Bernd

Title: Re: problem with $realtime ???
Post by rajdeep on May 31^st, 2006, 6:16am

It didn't even worked with conservative :-/..

Title: Re: problem with $realtime ???
Post by bernd on May 31^st, 2006, 6:26am

Why do you set the time step at all?
Spectre in general computes the best time step internally.

Bernd

Title: Re: problem with $realtime ???
Post by Ken Kundert on May 31^st, 2006, 7:19am

It seems to me that you should be using $abstime rather than $realtime. Also, idtmod() should be available in Verilog-A.

Title: Re: problem with $realtime ???
Post by rajdeep on May 31^st, 2006, 7:22am

Otherwise, it was not working. I don't think it can actually choose a good timestep by itself. Maybe it fails to do so n case of verilogA models.
A timestep of 10nS is giving fine result 8-)

Title: Re: problem with $realtime ???
Post by rajdeep on May 31^st, 2006, 7:28am

Is it the same Ken Kundert who wrote the book on Verilog-AMS???/
I have tried with abstime..didnot change. Infact, it gives a warning to change $realtime with $abstime!!!

Don't know!! I cudnot find idtmod in the whole Verilog-A LRM...i shud have tried though.

Thanks
Rajdeep

Title: Re: problem with $realtime ???
Post by Jess Chen on May 31^st, 2006, 8:27am

You could also try $bound_step. I am not sure I have the syntax correct but this VerilogA command limits the size of the time steps. You can use this command to link the maximum step size to a frequency or slope defined elsewhere in the code.

As a side note: if you are resetting your integral when some event occurs, you may want to use a @cross statement to make sure the reset occurs with the desired time resolution.

-Jess

Title: Re: problem with $realtime ???
Post by Geoffrey_Coram on May 31^st, 2006, 9:11am

Rather than @cross, you might want to investigate the optional 3rd argument -- timetol -- for the timer event.

Title: Re: problem with $realtime ???
Post by Geoffrey_Coram on May 31^st, 2006, 9:27am

rajdeep wrote on May 31^st, 2006, 7:28am:

1) Yes, it's the same Ken.

2) The simulator I'm using says that $abstime should be used instead of $realtime.

3) idtmod is in 4.4.6 of the Verilog-AMS LRM 2.2 (and the same section in 2.1), titled "Circular integrator operator"

Title: Re: problem with $realtime ???
Post by rajdeep on May 31^st, 2006, 12:59pm

Thanks a lot folks!! I will try all the options. Infact, was thinking of timetol option before leaving for home.

The cross option will make the code somewhat clumsy I guess...but then if it works fine.
Thanks,
Rajdeep