Viewpoint

Analog Language Standards Need Working Models There is a conflict between the immediate need for analog HDLs, and the need for a few years to work out a standard. Is there no easy solution? by Larry Moore

Much attention has recently been placed on the controversy surrounding commercial availability of analog hardware description language (HDL) products prior to the ratification of the 1076.1 standard, due at the end of 1997 or early 1998. Critics claim that such products impede the standardization process and are not in the interests of the user community. They argue that a de facto standard could be established, thereby derailing the IEEE balloting process. They also argue that users of these products will have to rewrite their models to comply with the IEEE standard when it eventually takes effect.

Companies operating in highly competitive and cost-driven industries such as semiconductors will suffer if the true merger of analog and digital HDLs, as now foreseen for Verilog as well as VHDL, will not be widely available to them for two or more years. These languages are important because they enable new, more productive methodologies and support the ongoing globalization and transformation of our high-tech industrial landscape.

Proponents of pre-standard analog language products­a growing constituency of EDA suppliers and users­hold that the use of these products helps and improves the standardization process by bringing practical aspects into full view. They argue that the availability of products today addresses real and imminent needs of a user community that can no longer tolerate being tied to proprietary analog HDLs.

Semantics, rather than the difference between keywords such as "terminal" and "node," are what is important to a user. While the discussions in the 1076.1 Working Group often revolve around more fundamental issues than just keywords, it is also true that the definition of the VHDL language itself constrains the 1076.1 work to a significant degree. A new language is not being designed, but rather an existing one is being extended. The extension, by definition, must conform to the rules of the existing language (VHDL-93).

The EDA community would be doing a disservice to customers and ultimately, itself, if it did not find ways of passing on the benefits of all available technology. The thought of the EDA community "holding back" for the good of users is both patronizing and out of touch with reality. Perhaps for this reason, all EDA companies currently engaged in the 1076.1 standardization process either have VHDL-based, pre-standard analog HDL products on the market or have announced plans to introduce one by the end of 1995.

Some EDA vendors have taken care to provide their products with a wide range of supporting tools, preparing the way to automatic migration paths as the standard evolves. Without exception, those EDA vendors that already have made the courageous effort to advance their thoughts on how to merge practical with theoretical requirements are aiding the whole industry.

Sturdier products will follow from early user experiences and criticisms of these first attempts. Standards are always most useful when they are built on practical foundations. By putting working products in users' hands, these EDA vendors may have gained some market advantage. More importantly, they have responded to the critical needs of companies that simply cannot afford to wait.

Larry Moore is president of Anacad Electrical Engineering Software (Milpitas, CA).

integrated system design August 1995