tran infotimes via ocean?

[danmc]

Anyone know how to specify the "info" and "infotimes" parameters for a
spectre transient analysis via ocean? Along with this, how do you
specify an "info" analysis in ocean since I'll need a named info
analysis for the transient "info" parameter.

Maybe the fact that

ocnDisplay('tran)

doesn't show anything that sounds right means I'm just stuck.

As an aside, it seems to me that ocean is lagging behind spectre and
that not all spectre capabilities are supported via ocean. Does anyone
know if cadence is planning on trying to fix this or are they planning
on letting ocean die out? I really hope they'll fix it instead of
killing it as I have a huge investment in ocean/skill code that I'd
really hate to have to port over to some other system.


Thanks
-Dan

 

[Svenn Bjerkem]

danmc wrote:
[snip]

As an aside, it seems to me that ocean is lagging behind spectre and
that not all spectre capabilities are supported via ocean. Does anyone
know if cadence is planning on trying to fix this or are they planning
on letting ocean die out? I really hope they'll fix it instead of
killing it as I have a huge investment in ocean/skill code that I'd
really hate to have to port over to some other system.

I have about the same experience when trying to utilize spectre beyond
the capabilities of Analog Artist. I wanted to do some measurement and
characterization work on standard cells, and found that this is either
done in extremely long time manually with analog artist, semilong time
with ocean and awkwardly complex with spectremdl.

Analog artist have limited the engineer to one simulation of each type
in each state. There was no real advances in analog artist front end
development from Cadence bought it up to now. I can't judge 6.0 because
we are not open access yet. There is no possibility to automate the
work in analog artist with ocean. It is only possible to write out the
current setup to a script file and then do command line work from there
on.

SpectreMDL is completely command line based. My work so far on the
command line has been different variations of spice. The syntax is well
known and the measurements are sort of well implemented. Cadence chose
with mdl to break totally with the spice concept to give the designer
full freedom by dividing test code and netlist. This concept takes way
to long to learn, and even if there is a spice to mdl translator the
documentation must be improved a lot before I will call it acceptable.
There are also nobody pushing mdl, it is just there.

I have been positively impressed by some of the features available in
the new wavescan. I am not impressed by the waveforms themselves, but
the new CIW for skill mode, which I expect to be available in the other
applications also.

My advice would be to focus on pure skill implementations and drop the
use of the word "ocean". Ocean is nothing but skill anyway and the
postings on this group the last years show that Cadence confuse their
customers more than they help them by this.

Still the simplest simulation automation is perl and (h)spice, that
Cadence will not be able to change.
--
Svenn

 

[danmc]

setting up the infotimes bit via the gui and then writing out an ocean
script showed that the magic words to add to the analysis() arguments
are:

?infotimes list("10n" "20n" "30n")
?oppoint "rawfile"

-Dan