[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gnucap-devel] New Developer
From: |
al davis |
Subject: |
Re: [Gnucap-devel] New Developer |
Date: |
Tue, 20 Feb 2007 01:34:33 -0500 |
User-agent: |
KMail/1.9.5 |
On Monday 19 February 2007 19:49, David Cummings wrote:
> Hi all,
> I'm interested in helping out with the project, more
> specifically, i'm interested in an alternative to SPICE. The
> website seems to be full of old(er) information (just found
> the one on seul.org), maybe they should be connected? I'd be
> willing to help out with some web-devel, though it really
> grinds the nerves (bad memories of cross-browser
> compatibility).
That is obviously a place where help is desperately needed. The
latest is at gnucap.org, but even that needs lots of work. The
one at geda.seul.org is supposed to be a mirror of it. The one
at gnu.org is supposed to be a checkout of the CVS archive of
the one at gnucap.org.
> What's the status on IBIS? How hard would it
> be to complete?
IBIS-3 is about 90% complete, and on hold, but it is designed
around an older version of gnucap, assuming a Spice-style
netlist. It needs to be changed to use a Verilog style
netlist, and to work directly as a plug-in.
The important part that is still missing is part of the waveform
generator. I know how to do it, but have had other priorities.
It is conceptually difficult and requires expert knowledge.
> Also, has anyone done any work on harmonic
> balance simulation? That seems interesting, though probably
> fairly difficult.
No.
I have given some serious thought about periodic steady state by
a shooting method. This should be fairly easy. Most of the
infrastructure is in place. I have done some proof of concept
experiments but no real code. The proof of concept experiments
were successful. I was able to measure distortion of a high
quality oscillator by manually manipulating the shooting
parameters. The procedure is fairly easy. It just needs to
be coded. The most likely user interface is an option on the
Fourier command, which almost does all you need now.
> About me: I'm a senior at JHU studying computer
> engineering. I've worked on some midsize C++ projects (>20k
> lines) and am decent on teams. As a senior, my time is kinda
> limited, but I could try and help with the web while I spin
> up on the code. I'm taking a class on analog integrated
> circuits this semester, so some of this stuff is very current
> for me. I've taken classes on circuit analysis, parallel
> programming, etc., so I at least know the basic idea, which
> is to say I know very little in the grand scheme of things.
Johns Hopkins???
Try using gnucap for your homework. Your teacher probably told
you to use something else. Try both and compare. Gnucap will
give you a lot more information about your circuit than you can
get out of any Spice, but you need to know how to ask for it.
The main focus of current work is to move as much as possible
into plug-ins. That makes it much easier to extend and
experiment. You can experiment in a plug-in without the
baggage of dealing with the whole thing. You can distribute
experimental code without contaminating the core.
The other main focus, which needs plugins to work, is real
Verilog-AMS support.
> Anyway, sorry about the slow introduction, just wondering
> how I can help, and if there's a repository somewhere with
> the current development code.
There is no CVS yet. I have been trying to get the plugins and
at least part of Verilog-AMS working, which has been taking all
of my time.
The best code to work on is the latest development snapshot.
There will be a new one within a few days. I recommend that
you wait for the announcement, then get the new one. This one
will have 4 tarballs to get everything. That is one for the
simulator, and 3 more for models. The new one can use Spice C
models as plugins, with a wrapper.