symbol + schematic pcells

Hi all,

We've designed a schematic pcell which includes a variable number of
instances. Also, the I/O pins are busses whose size varies according to
some "width" parameter.

We created also a symbol pcell with the varying pin width so that the
terminals on the symbol and the schematic match.

When we instanciate the symbol, the pin width come right. When we
instanciante the *schematic* we also see the pcell evaluated correctly.

What we would like is to be able to use the symbol in a schematic, and
expect the netlister to switch into the evaluated submaster of the
schematic pcell. This doesn't work, that is, the port list appears
correctly in the netlist but the inside of the subcircuit is always
netlisted the same regardless of the parameter values.

Are we missing something here ?

thanks in advance

stéphane.

 

I have developed pcells that have similar functionality for current
mirrors and diff pairs, where both the schematic and symbol are
evaluated dynamically based on a "numberOfOutputs" parameter.

In doing this, I ran into a possibly related bug in schematic pcells -
they are not always evaluated unless a certain events are triggered.

I never did figure this out completely, but the pcell was always
evaluated successfully if it involved a change in the number of pins,
number of instances, or other major structural changes. However, if I
tried to perform calculations and update fields on the children
instances, that would not work. It seems that a new sub-master was not
created unless there was a required _structural_ change. If _only_
changes to properties were required, a new submaster was not created,
and the default values were used. However, if the new sub-master would
require changes to the number of instances, pins, etc., then the
properties would have the correct values.

It seemed like a bug to me, but submitting a SR rolled off my TODO list,
because I was pushing for a deadline. So, I calculated the necessary
data in the form callback and passed the data through pPar's for hidden
parms.

Hope this helps...

 

Are you both using IC50 or later? Schematic pcells did not netlist properly
until then - it did not correctly handle the variant generation at netlisting
time.

These days it seems to work OK - I've done a couple of these in the last year or
so, and they netlisted fine. One which had a bus with variable width, for
example.

Regards,

Andrew.

 

Are you both using IC50 or later?

Yes, IC5033USR3 exactly.

If you couldn't spot any obvious mistake in the explanation, then it should work...
I will try it again.

thanks,

stéphane

 

If you couldn't spot any obvious mistake in the explanation, then it

Finally, I sorted it out. Not sure exactly how, but recompiling the pcell
from scratch helped.

Thanks for your input,

stéphane

 

I was using IC5.1.41, IC5.0.33 at the latest (or earliest

 

Well, I'm surprised. Can you log this with customer support?

Regards,

Andrew.

 

It's on my TODO list.