Has anyone looked at using hdf5 as a file format for storing simulation
data? I'm thinking that this might be a good way to go for
spice/gnucap/etc results. One thing on all of these tools which I think
is most useful is the ability to run a simulation where you basiclly
save all node voltages and optionally all currents. Then you can pick
out what to plot later.
Why yet another file format? It seems that hdf5 is gaining popularity.
You can read it from octave and from matlab. It seems it wouldn't be
too hard to extend gwave to read it too. It is less clear that the
nutmeg raw file format is as well represented in some of the various
plotting tools around. ASCII files are inefficient for lots of data and
much more restrictive in what you can put in them if you want to be able
to directly load into octave/matlab/scilab/grace/gwave/whatever.
I haven't had a chance to dig through all of the documentation yet, but
it does seem that hdf5 will let you grow a data array in either
dimension. This is critical for simulation data and is why matlab
binary .mat files (besides being proprietary) are a bad choice. The
problem with a .mat file is the file format is basically:
header
all of matrix/vector #1
all of matrix/vector #2
...
where something like spice wants to write more in the format
header
time/frequency/whatever point #1 for all signals
time/frequency/whatever point #2 for all signals
Anyway, before I wasted time on this I wanted to see if anyone else has
looked into it.
-Dan
