Re: [ng-spice-devel] New data file format


To ng-spice-devel@ieee.ing.uniroma1.it
From Steve Hamm <Steve.Hamm@motorola.com>
Date Thu, 7 Dec 2000 09:30:49 -0600 (CST)
Delivered-To mailing list ng-spice-devel@ieee.ing.uniroma1.it
In-Reply-To <CF654D964573D311A1CA0090278A36FF4CBCF2@EDIN_EXM1 >
Mailing-List contact ng-spice-devel-help@ieee.ing.uniroma1.it; run by ezmlm
References <CF654D964573D311A1CA0090278A36FF4CBCF2@EDIN_EXM1 >
Reply-To ng-spice-devel@ieee.ing.uniroma1.it

---"AG" == Gillespie, Alan <Alan.Gillespie@analog.com> writes:

AG> I think the standard binary format outputs double precision
AG> data, which is overkill in my opinion, although I know some
AG> people disagree. But single precision halves the file size,
AG> and therefore doubles the access speed. Single precision has
AG> plenty resolution, way more than any reasonable reltol
AG> setting. At least, the precision could be optional.

Generally, any situation where the dynamic range exceeds 140db would
need double precision. I can think of a few examples, off the top of
my head: Looking at settling time of a precision amplifier, microvolts
on a 1v signal may be significant.  To accurately resolve those
microvolts, 1e-7 or 1e-8 reltol would be useful. Or, transient
analysis followed by an FFT, where you want the noise floor more than
140db down, would also fit.  Also, AC simulation of very high-Q
circuits may also need results of more than 7 digits.

I've thought more than once about doing single precision output, but
we've never done it because of all the reasons... Even for "hack"
digital design, cell timing measurements, if they are to give
monotonic results for small changes in cell size and loading, require
very tight tolerances. So you're not off the hook even there...

AG> But it's access speed that can be dramatically improved with
AG> a wee trick which I'll try to explain :-

Interesting!

--Steve

Partial thread listing: