🚀 go-pugleaf

I have used Simulink/Simulab for the past 2 years, both for simulating
non-linear systems and for developing models for use in H-infinity
and mu-synthesis design work.

On Unix workstations Simulink is a fine product.  It is extremely easy to
build complex models.  It lets you fairly easily incorporate user defined
blocks and functions - either using the Matlab programming language or
using C.  I have not used the C option but have extensively used the 
Matlab programming language.

One valuable advantage to Simulink is that the graphical block diagram
is stored as an ASCII .m file.  Depending how you pass parameters into the
block diagram you can set it up so that ALL the info on the system you are 
simulating is contained in the file.  The .m file provides complete 
traceability for the simulation results.  The fact that the file is ASCII
just makes it easier to extract this info later if desired.

I heartily recommend Simulink, but am going to list what I know of
problems or areas I think need improvement:

--Occasionally (1 or 2 times/week) I get a segmentation violation and have to
exit Matlab & reenter.

--Within a function there are limits in what you can do to modify parameters
for a simulation run.  This is not a big deal running time domain simulations,
it is a big deal in developing functions to design controllers based on
linearizing a Simulink block.

--On the PC/Windows version the Simulink window editing is aggravatingly slow 
on a 25 MHz 386 machine.  It may be acceptable on a 486 machine.


I have only a little experience with ACSL, mostly with their first release on
the workstations.  I was not happy with it at all, it was very cumbersome to
use & used single precision variables.  I understand the product has changed
 substantially since then.

One big benefit I see with Matlab/Simulink is the availability of supplemental
toolboxes which can be used for associated design & analysis.

--
Mike Stoughton
stote@den.mmc.com