gnucap-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gnucap-devel] "transient" branch


From: al davis
Subject: Re: [Gnucap-devel] "transient" branch
Date: Thu, 22 Sep 2016 12:45:52 -0400

On Fri, 12 Aug 2016 19:48:35 -0400
al davis <address@hidden> wrote:
> The main thing is to remove the requirement to specify a step size for
> transient analysis, and if you don't specify a step size, it shows
> all steps, as if "trace all".

I think what is now in the transient-3 branch (transient, rebased a few
times). is almost ready.

The latest commit has changes to time step control related to the first
few steps.  It now makes a real attempt to have even the first step
having a proper step size, with intent that it is not necessary to
specify a step size for accuracy, including accuracy of downstream
applications.

An oscillator circuit (Thanks to Richard Gipps) has been very helpful
in making this progress.  It now gets stepping correct for the
oscillator, a fully autonomous circuit, including startup, even if the
user strobe (printed step interval) is unreasonably large, such as
200x the oscillator period, even with relaxed tolerances, then with
tightened tolerances seems accurate enough for distortion
and noise measurements.

In the existing transient-3, I did not update the test references
(tests/==out) so many regressions appear to fail, in the sense that a
text comparison shows a difference.  New results are temporarily there
as "tests/0922".  In this case, it is necessary to manually look at each
one to determine if the new output is good or not.  I have done this, as
much as I could, and the new results look good.

I need others to check it too.  Is new (0922) close enough to old
(==out)? When you run the test, does it match new (0922) as it should?

al.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]