Sign In
Forgot Password?
Sign In | | Create Account

Analyzing the Big Picture

As engineers we often use “system” to describe different levels of design abstraction. Chip designers refer to integrated circuits as systems on chips. Audio designers refer to an amplifier as a system. Aircraft brake designers create systems for stopping aircraft. There truly are a variety of ways to define and implement a system. And most things we call a system are really just a subsystem of a larger system. I could get a little philosophical and propose that there isn’t a system that isn’t also a subsystem of a still larger system, but I’ll save that discussion for another time. For this blog post, I’ll stick with a general definition as suggested by a couple of examples: an aircraft is a system, but all the assemblies required to build it are subsystems; an automobile is a system, but all the assemblies used to create the car are subsystems. Get the picture? Good.

A colleague and I recently spent some time with an automotive customer. One of the promises of multi-domain simulators is the potential to do a full-system, not subsystem, simulation. But I’ve seldom seen a customer actually make a serious attempt to do so. Those that make the attempt usually use such high level models that the simulation results give a limited, and sometimes unrealistic, view of system performance. This customer, however, actually modeled the major subsystems in the target automotive platform, and then connected these subsystems together to model the dynamics of the car driving down the road. Subsystem models accounted for everything involved in moving the car, including engine, transmission, differential, wheels/tires, and brakes.

Along with developing models for vehicle subsystems, the company linked in several data files from actual vehicle test drives. Once simulation data matches test drive data, the complete system model will be qualified as a valid test bench for the vehicle. The company’s ultimate goal is to analyze, via simulation, the effects of driving conditions on vehicle performance. Think of the possibilities — with a running and validated system simulation, engineers can easily and accurately test multiple vehicle configurations and options, and make important automotive platform design decisions, before building a drivable prototype. Doing so is one of the many benefits of using virtual prototypes in a design flow.

Setting up a mixed-technology, full-system simulation for something as complex as an automobile takes a bit of thought and a lot of time. While the upfront resource investment required to build a complex system model is usually not trivial, the benefits measured in reduced prototyping costs and improved design quality easily justify the expense. Not many years ago, such a system simulation would have been impossible. But thanks to powerful modeling languages like VHDL-AMS, and robust simulators like SystemVision, full-system simulations are possible, practical, and quickly becoming a design flow necessity.

Modeling

More Blog Posts

About Mike Jensen

Mike JensenMost career paths rooted in high technology take many interesting (and often rewarding) twists and turns. Mine has certainly done just that. After graduating in electrical engineering from the University of Utah (go Utes!), I set off to explore the exciting, multi-faceted high tech industry. My career path since has wound its way from aircraft systems engineering for the United States Air Force, to over two decades in applications engineering and technical marketing for leading design automation software companies, working exclusively with mechatronic system modeling and analysis tools. Along the way, I’ve worked with customers in a broad range of industries and technologies including transportation, communications, automotive, aerospace, semiconductor, computers, and consumer electronics; all-in-all a very interesting, rewarding, and challenging ride. In my current gig, I work on technical marketing projects for Mentor Graphics' SystemVision product line. And in my spare time I dream up gadgets and gizmos, some even big enough to qualify as systems, that I hope someday to build -- providing I can find yet a little more of that increasingly elusive spare time. Visit Mike Jensen's Blog

More Posts by Mike Jensen

Comments 2

Post a Comment
Are you suggesting that with (traditional) HDL languages and robust simulators you are able to simulate an entire car? I agree the keyword here is "abstraction", but in such a way that important aspects are still visible at the system-level. This goes beyond abstraction of signals, structure and behavior, the elements which are normally handled well in traditional languages and tools. Relatively new in this abstraction game is how to deal with time. A good example is the transaction level modeling (TLM) standard which was introduced in the context of SystemC. This is introducing temporal abstraction, offering significant simulation speed-up, but still allowing refinement of timing properties where necessary. A nice feature for digital circuits, but how to do similar things in the analog and physical domain to tackle the multi-domain challenges we see in cars? Therefore I think we need to think a bit further than what we have today, and shape the design and verification environment of tomorrow, possibly with new system-level languages which can better address the multi-domain needs.

Martin
6:40 PM Dec 9, 2011

Thanks Martin. Not sure there will ever be a pure analog equivalent to TLM. Since analog behavior is based in the time or frequency domains, it's hard to visualize how either might be abstracted out. But there are a few very capable multi-domain modeling languages available today. For example, Synopsys offers MAST, which is proprietary and locked to their Saber simulator. Another option, which I think is one of the best solutions for modeling multi-domain behavior, is VHDL-AMS, an IEEE standard and supported by mutliple simulators, including Mentor Graphics SystemVision. Among it's other features, VHDL-AMS can model system behavior at multiple levels of abstraction.

Mike Jensen
10:26 PM Dec 16, 2011

Add Your Comment

Please complete the following information to comment or sign in.

(Your email will not be published)

Archives

 
Online Chat