gEDA-dev: New diagram (attempt at UML)

[John Doty]

On Apr 10, 2007, at 11:18 PM, al davis wrote:

> On Wednesday 11 April 2007 00:01, John Doty wrote:
>> We used HSpice for interchange. I used a modified spicepp.pl
>> to convert to ngspice for my simulations. Not without bumps,
>> but it got the job done without too much wasted time or
>> money. A part-timer like me can't afford expensive commercial
>> tools. I was spending more time doing spacecraft operations
>> and gamma ray astronomy than IC design.
>
> Do you want a real up to date simulator or don't you?

I don't know. I want to do astrophysics. For that I need data. For  
that I need instrumentation. For that I need electronic technology.  
For that I need design tools. But "up to date" isn't always the same  
as "effective".

> The purpose of gnucap is be a real up to date simulator with all
> the stuff you see only in the expensive ones, and more.  I am
> not interested in "me-too".

I understand that: no point in duplication. I'm actually pretty happy  
with ngspice, but of course I always want more ;-). However, it  
sounds to me that right now, gnucap would give me less, since  
sensitive instrumentation design begins and ends with understanding  
the noise. Maybe I can find some time to look at gnucap: if the code  
is straightforward enough for a greybeard physicist, I might try  
putting in some noise stuff.

>
>>
>>> I don't know what the priorities of the other developers
>>> are, but for me the people doing free/open-source hardware
>>> development are the top priority. We are the enabler for
>>> that, in the same sense that gcc was the enabler for the
>>> whole free/open-source software movement.  Another priority
>>> for me is a base for researchers, so they can build on free
>>> tools rather than proprietary ones.
>>
>> Well, I am a researcher. One of my collaborators is the
>> publisher of OpenIP (http://research.kek.jp/people/ikeda/).
>> But those netlists are all SPICE.
>
> Because it is all they can get.  I am looking forward.  When we
> develop new stuff, we need to think of leading the way into the
> future, not following everyone's mistakes of the past.

As far as I can tell, all the designs and models one can presently  
get are SPICE (or some more proprietary code), and I can't penetrate  
the Veriog-AMS rhetoric to see how it can help in this world.

>
> Gnucap will continue to support the Spice formats.  (notice the
> plural).  With plugins, it should be able to get exact
> compatibility.  Previously, compatibility and moving forward
> were mutually exclusive.

So what you might be trying to tell me is that I can netlist my  
designs in Verilog-AMS and import SPICE models for the underlying  
stuff. That might be cool, but I'm having trouble with the sales  
pitch here.

>
>> Even researchers need to make money. I provide services to
>> universities doing sponsored research. I've also done work
>> for NASA.
>
> Even developers of free software need to make money.....

Indeed.

>
>> I have no axe to grind here, but I've little clue as to what
>> Verilog- AMS is and what it can do. Since vendors provide
>> SPICE models, it's hard to understand how I could actually
>> use it.
>
> Actually now a lot of them are done first in Verilog-AMS,
> because it is easier.  Then they make a Spice model for
> everyone else.

Where are they published in Verilog-AMS, then? Googling finds much  
Verilog-AMS sales rhetoric, but no models.

>
> The simplest way to put it is that Verilog-AMS is a better
> Spice.  It is really much more than that.

PL/I was once touted as a great improvement over Fortran, but it  
never really made it. I'd love a better simulator than SPICE, but  
"better" means it helps get the work done faster and more accurately,  
not some abstraction. A lot of the the attraction of SPICE is  
community leverage (things like OpenIP and manufacturers' model  
libraries). I can't do *everything* myself, although for my work IC  
design is a long way down that road ;-)

John Doty              Noqsi Aerospace, Ltd.
jpd at noqsi.com