Part of the  

Chip Design Magazine

  Network

About  |  Contact

Headlines

Headlines

Accessibility Trumps Accuracy In Today’s Hardware Models

By John Blyler
Shrinking time-to-market windows and competition to keep end-product costs low have resulted in a growing usage of hardware prototypes and software algorithm models.

This trend is fairly well known. But what many designers don’t appreciate is that accessibility of hardware models in software is sometimes more important than the accuracy of the hardware being modeled. Another trend – really, a re-emerging trend – is the use of hardware prototypes as the initial product offering. Both of these trends are occurring at the chip, board, module and end-user product versions of the electronics market.

“The ability of companies to better understand product design alternatives based on variations of software and hardware modules is vital as time-to-market pressure continues to increase,” says Greg Sikes, director of enterprise architecture and systems modeling at IBM Rational. He sees more companies spending time and money to model complex systems that contain both hardware and software before “bending metal” – that is, implementation of the final design in hardware and software.

In an ideal environment, both hardware and software designers would share a unified modeling environment. Such an approach would be very useful for embedded developers. According to Pranav Mehta, senior principal engineer and CTO of Intel’s Embedded and Communications Group (ESG), such a modeling environment would let chip designers perform architecture modeling of the silicon before implementation; validation and test engineers test the implementation against a range of “real” application usages; firmware developers complete the initialization code before silicon is available, and a system/application developer performs necessary functional and performance checks to not only speed up the time-to-market, but also to have high confidence in the qualitative aspect of the end platform.”

Beyond the detailed world of chip and embedded development, many designers must cope with cross-disciplinary models that include electronics, mechanics and control systems. “Many designers who have traditionally used The Mathworks Simulink to model algorithms now realize that they can do more if they tie that model to models from other disciplines,” says Bill Chown, director of marketing for the model driven engineering group at Mentor Graphics.

Traditionally, many designers believed that the accuracy of the models is of primary importance. But in many cases, accessibility trumps accuracy. How can this be true? Thanks to increasing software complexity and decreasing time-to-market product windows, more software developers are needed for a given project. More software developers mean that they need greater accessibility to hardware prototypes on which to test out their applications.

Naturally, designers will eventually need to use hardware models to gain confidence in the accuracy of their models. But the cost to build or acquire additional hardware prototypes to accommodate the extra software development teams is often to expensive and time consuming, especially in the automotive, aerospace and upcoming medical industries. Hardware prototypes in these industries typically requires lots of electronic subsystems and the associated cables/peripherals – the cost of which is high.

“Designers realize that they must try out their software but cannot afford to make or buy extra hardware prototypes,” explains Chown. “Instead, they try to do the next best thing, namely, make a model of the hardware.”

Models of software-intensive hardware electronics are nothing new. Such virtual hardware prototypes have been around for many years, certainly in the EDA and semiconductor industry. What is relatively new is the closer linkage between the virtual prototypes – which model the chip hardware for software developers – and the FPGA-base prototypes used for hardware verification.

The tighter integration has led to an interesting side effect, namely that many designers and manufacturing engineers are opting to use the prototype as the initial version of a product. This is particularly appealing when being the first to market is more important than the initial cost, notes Frank Schirrmeister, director of product management for system-level solutions at Synopsys. “Designers that have used a hardware prototype to do architecture exploration will increasingly use that prototype for the initial production run to capture market share. They will optimize that design in the next version.”

Modeling hardware in software and manufacturing prototype models for initial product runs are trends finding favor with many designers. It’s a balancing act between accessibility and accuracy, first to market and cost priorities. Time will tell which companies can maintain this balance.

Tags: , , , , , , , ,

Leave a Reply