Staying vigilant of circuit and system model flaws

Article By : Bill Schweber

Electrical, thermal, and mechanical models are essential design tools, but they are not always good enough, as a new book explains.

Engineers of all disciplines rely on models and subsequent simulations to assess likely performance of components, circuits, and systems. Whether it’s an electrical model of an RF component (Figure 1), a component’s thermal model (Figure 2), or a mechanical model of a car’s suspension (Figure 3), engineers need these to determine likely nominal performance. These models would also be used to understand the impact of expected tolerances, temperature-based effects, and long-term changes. In nearly all cases, it would be prudent to use models to have some level of confidence in the design.

Staying vigilant of circuit and system model flaws

Figure 1 This is just one small-signal equivalent model of a 60-GHz RF CMOS transistor. Source: Semantic Scholar/Allen Institute for AI

Staying vigilant of circuit and system model flaws

Figure 2 This thermal model shows the heat-transfer paths and impedances from a hot die to the ambient environment, including the critical PCB. Source: Richtek Technology

Staying vigilant of circuit and system model flaws

Figure 3 The four suspension subsystems of a vehicle to ground through the shock-absorber system and tires must also be modeled. Source: ResearchGate

Of course, these models do not exist in silos where they can be fully modeled, such as by using Spice or Ansys for electrical performance. Thermal factors will affect electronic and mechanical performance, and vice versa. It’s a complicated cross-linking, and there are modeling and simulation applications such as COMSOL, which do a fairly good job of bridging the different perspectives, but with inevitable limitations.

Models usually begin as simple generalizations, but they soon get enhanced, refined, and modified based on knowledge and experience. After all, a model of a water-based system may soon need to accommodate the dramatic inflection points of 0°C and 100°C, and the fact that those two critical points are actually a function of pressure may also be critical (or maybe not at all).

However, as every engineer knows or quickly learns, no model is perfect. So, the issue is how imperfect it is and in what ways. That’s a question to which there is no single or simple answer since it’s highly dependent on how and for what the model is being used.

But beyond that obvious acknowledgement, what do we really divine about the broad underpinnings of these models on which we rely so heavily? This is an issue about which we may not be adequately aware of or prefer to not think, because it’s a tangled discussion which cannot be fully unraveled and placed into neatly stated summaries.

Nonetheless, these are important model-related issues, and an excellent place to start is with the recently published book “Escape from Model Land: How Mathematical Models Can Lead Us Astray and What We Can Do About It” by Erica Thompson (Figure 4). Don’t worry, this is not a rant against models, despite the “escape” in the title. Instead, it’s a carefully thought-out discussion of the many issues surrounding models.

Staying vigilant of circuit and system model flaws

Figure 4 The equation-free yet sophisticated book provides multiple perspectives on the many implications and realities about models. Source: Basic Books

Not that the author is not a science writer or journalist who has latched onto a sensationalistic hot topic—usually nothing wrong with that, but sometimes there is—but is a Ph.D. in Economics with a highly mathematical and model-focused background. Her goal is not to get rid of models, but to help us understand how they are created and when to not be overly constrained or misled by them.

If you worry this book is yet another academic, math- and jargon-laden treatise on models, it’s far from it. You’ll have no concerns about getting bogged down in deep math since there are no equations at all. Yet, at the same time, it doesn’t “dumb down” the topic for non-technical audiences; instead, it’s a sophisticated discussion of this important topic. Also surprisingly, it’s not encumbered with graphs, block diagrams, charts and tables, as there are only three figures in the book, and they are not technical but just provide a brief visual relief.

This very timely book looks at issues with an emphasis on financial, virus-related, and weather/climate models, as they generally try to predict the future trends based on past data and in-depth analysis. And there is some discussion related to astrophysics and atomic-level models. It forces us to think about what it means with respect to various terms and types of assumptions, and how models can both illuminate and obscure reality.

It also poses some interesting but hard-to-answer questions. What do we even mean by a good model? If we model the weather with absolute accuracy 364 days each year but fail miserably one or two days each year, how does it compare to a model which is fairly close every day but never way off? What are the implications if tweaking a constant such as π to 3.5 makes the model very good, while using 3.14 makes it just mediocre? How do we reconcile that “experts” have created so many widely different models of the same situation, using the same data and even the same or similar equations?

Even if the model is not good enough for the situation, the act of making a model forces you to think about what is happening, what is important, what you know, what you don’t know, and maybe even things that you don’t know that you don’t know. The author makes the point from various perspectives that “all models are wrong, but some are useful” with the emphasis on “some.” The book explains why even apparently good models can limit thinking and clarity, and confine our creative, non-conforming assessment to what the model will allow.

In many ways, engineers have it easier than those who model economics, viruses, or climate. How so? In many cases, we can build a real testbed and check it versus the model via controlled experiments to see if the model is working well enough for the application. We can model the theoretical performance of a power or RF device, build one, start with known conditions, and test the model versus reality. We can set up multiple controlled experiments and vary key parameters to see if the sensitivity and correlations built into the model are in line with observable reality. That’s a luxury that economists, doctors, and climatologists don’t have.

To paraphrase one of the many insightful paragraphs in the book, the advantage of a model is that it lets you neglect irrelevant details, so you can see the essence of the problem, linkages between parts of the situation, and get insight into the behavior of a system. That’s the best-case scenario. But don’t delude yourself: a model can also lead you to neglect factors that are not immediately relevant, tractable or understood, and thus remove the needed context and development of genuine insight—and that’s the worst-case scenario.

Have you ever been misled by an over-simplified or inadequate model? Have you ever had to create a more meaningful one even if there’s a penalty in accuracy in order to better capture the relevant perspective?


This article was originally published on Planet Analog.

Bill Schweber is an electronics engineer who has written three textbooks on electronic communications systems, as well as hundreds of technical articles, opinion columns, and product features. In past roles, he worked as a technical website manager for multiple EE Times sites and as both Executive Editor and Analog Editor at EDN. At Analog Devices, he was in marketing communications; as a result, he has been on both sides of the technical PR function, presenting company products, stories, and messages to the media and also as the recipient of these. Prior to the marcom role at Analog, Bill was Associate Editor of its respected technical journal, and also worked in its product marketing and applications engineering groups. Before those roles, he was at Instron Corp., doing hands-on analog- and power-circuit design and systems integration for materials-testing machine controls. He has a BSEE from Columbia University and an MSEE from the University of Massachusetts, is a Registered Professional Engineer, and holds an Advanced Class amateur radio license. He has also planned, written, and presented online courses on a variety of engineering topics, including MOSFET basics, ADC selection, and driving LEDs.


Leave a comment