spectre : variable defined as an explicit function of time

[fogh]

Hi All,

Is it possible to use time directly in a variable definition. in other
words can a variable be an explicit function of time ?

I have a RLC model of a cristal (piezo) that uses a scaling of the
quality-factor "scaf", so that the quality factor is decreased by a
factor scaf but the resonnant frequency stays the same. The lower Q
makes simulation possible, but I would like to make at least one
simulation (preferably a pnoise) with scaf=1, and thus a very high-Q.
The values of R,L and C in the cristal model are dependant on scaf.
I would like to let scaf decrease from 100 to 1 as an explicit
expression of time, for instance with
parameter _tmp1=0.5*( abs(time-25us) + time-25us )
parameter _tmp2= 1 - _tmp1/200us
parameter _tmp3= 0.5*(abs(_tmp2)+_tmp2)
scaf= 1 + 99*_tmp3

This way the Q of the cristal would increase slowly while the
oscillation regime has settled in low-Q. The problem is that spectre
doesn t let me use the variable "time". I expected a problem with pss ,
but it is not even allowed in transient.

I guess I could do this in verilog-AMS or with controlled sources, but
that seems a bit heavy. Does anyone know a more straightforward workaround ?

 

[Andrew Beckett]

In IC5033 (may be in IC5032 too), you can have resistors, capacitors, inductors and sources
controlled in terms of voltages and currents, and they can also reference the time via
$time. In IC5033 this is documented, and you'll be able to get this info by typing
"spectre -h bsource".

IC5033 is due out on 31st October.

Andrew.

On Thu, 16 Oct 2003 15:33:04 +0200, fogh <cad_support@skipthisandunderscores.catena.nl> wrote:

Hi All,

Is it possible to use time directly in a variable definition. in other
words can a variable be an explicit function of time ?

I have a RLC model of a cristal (piezo) that uses a scaling of the
quality-factor "scaf", so that the quality factor is decreased by a
factor scaf but the resonnant frequency stays the same. The lower Q
makes simulation possible, but I would like to make at least one
simulation (preferably a pnoise) with scaf=1, and thus a very high-Q.
The values of R,L and C in the cristal model are dependant on scaf.
I would like to let scaf decrease from 100 to 1 as an explicit
expression of time, for instance with
parameter _tmp1=0.5*( abs(time-25us) + time-25us )
parameter _tmp2= 1 - _tmp1/200us
parameter _tmp3= 0.5*(abs(_tmp2)+_tmp2)
scaf= 1 + 99*_tmp3

This way the Q of the cristal would increase slowly while the
oscillation regime has settled in low-Q. The problem is that spectre
doesn t let me use the variable "time". I expected a problem with pss ,
but it is not even allowed in transient.

I guess I could do this in verilog-AMS or with controlled sources, but
that seems a bit heavy. Does anyone know a more straightforward workaround ?

--
Andrew Beckett
Senior Technical Leader
Custom IC Solutions
Cadence Design Systems Ltd