Subject: Re: Checking for dimensional consistency in VHDL-AMS models
From: Ernst Christen (christen@analogy.com)
Date: Tue Dec 21 1999 - 11:02:54 PST
I have followed this discussion with interest, but I haven't had a
chance to study the original work done at Los Alamos in detail.
However, I do agree with statements made by Mark Zwolinski, Dave
Barton, and others that the Los Alamos approach is not directly
applicable to VHDL-AMS becaus, among other things the C++ type system
is quite different from the VHDL type system. While there are some
similarities with the approach Dave Barton took with MHDL 8 or 9 years
ago, I did notice that fractional dimensions don't seem to be supported.
Such dimensions are used, for example, when dealing with noise
voltages whose unit is V/sqrt(Hz).
I can, however, add some historical perspective concerning dimensional
analysis etc. in the context of VHDL-AMS.
One of the requirements captured in the Design Objectives Document was
support for some dimensional analysis in VHDL-AMS. At some point this was
broken into two requirements:
- support for annotating units, so they could be picked up by a tool,
e.g. a waveform display tool
- full dimensional analysis
During the language design phase these requirements were studied, and
a white paper by J.J.Mayol and Serge Garcia-Sabiro documented the
findings and some possible approaches to support dimensional analysis
in VHDL-AMS. The conclusions were that dimensional analysis could not
be integrated into the VHDL type system without causing compatibility
problems with VHDL. At that time the Language Design Committee decided
that backward compatibility/interoperability with VHDL was the more
important requirement, and it postponed the requirement for a future
standardization phase. Because changes to the type system would be
required a request was submitted to the VASG to consider requirements
for dimensional analysis in future versions of the VHDL base language.
The issue has not been further considered till now. Note, however,
that annotation of units is possible with user-defined attributes.
Ernst Christen
This archive was generated by hypermail 2b28 : Wed Jan 26 2000 - 15:58:33 PST