Re: analysis names

Hi, Paul -
I agree that analysis_list needs to be defined; in LRM 2.2,
initial_step and final_step also took analysis_list as an
argument, and the syntax was defined (in Syntax 6-11).
I don't remember why the change was made.

My understanding of Table 5-1 is that
   initial_step
without any arguments will return 1 for the initial step of any
analysis, and it's up to the simulator vendor to document what
additional analyses are supported and what the "initial step" of
such analyses are, so that
   initial_step("hb_noise")
could return 1 for the initial step of that analysis, if it is
known to the simulator running the simulation, whereas
   initial_step("completelyunknownname")
will definitely return 0 for all the phases of the analyses
listed in Table 5-1.

I believe it would be allowed for a simulator to document
an analysis name such as "smallsignal" that would return 1
for initial_step("ac") and for initial_step("noise").
I made a request once to a vendor to support something like
    analysis("any_noise_analysis")
so that I could tell a simulator to run or bypass noise code
without having to explicitly list "noise", "hb_noise",
"pnoise", "transient_noise", ... and whatever other names
for noise analyses were used in other simulators that I
hadn't come across yet.

-Geoffrey



Paul Floyd wrote:
> Hi
> 
> initial_step and final_step [5.10.2] can take a list of 
> "analysis_identifier"s from "ac", "noise", "tran", "dc" and unknown, and 
> there is a reference to analysis [4.6.1]. 4.6.1 adds "ic", "static" and 
> "nodeset", and it has an optional "analysis_list" as argument - which is 
> not defined anywhere else.
> 
> It's not clear to me if "Additional analysis names can also be used as 
> necessary for specific implementations" means that the extra names can 
> also cause initial/final_step to return 1, or if table 5-1 is exhaustive 
> and any further names will fall into the unknown category.
> 
> It would look better if initial/final_step and analysis used the same 
> style for the syntax of their argument(s).
> 
> ac_stim [4.6.3] can also take an analysis_name, but there is no 
> reference to analysis [4.6.1]. I think that this is an ommission.
> 
> Regards
> Paul Floyd

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.