Calculating the worst-case performance for a circuit over its expected lifetime, including all specified operating conditions, can be difficult at best. It's not for the timid.
This first post in The Worst Case blog focuses on the skills, resources, and techniques needed to properly perform a worst-case circuit analysis and some of the reasons why worst-case circuit analysis (WCCA) is challenging to get right.
If you're thinking of performing a WCCA, you'll want to stay tuned to The Worst Case. You'll get a glimpse into what skills you need to properly perform WCCA and why WCCA needs to be an investigation that is separate from the design function. It's not a checkbox item, nor is it a task you can simply hand off to just any engineer. Furthermore, you can't use it as a learning task for the "junior" engineers.
Why engineers get WCCA wrong
The skills required to properly perform WCCA are diverse and yet each is critical to the outcome. In teaching WCCA techniques, I often find engineers so eager to tackle this important analysis on their own, even if it is their first time. Training a new engineer can take at least nine months to a year. Even then, there are multiple levels of review before an analysis is verified and complete. WCCA is that important. Often, it's the last line of defense against defects, recalls, hazards, and failure.
With those thoughts in mind, I compiled a list of challenges to having a successful outcome. Companies, program managers, and most of all engineers are eager to fix their circuit "leaks." Without the right tools, skills, and experience, your effort is bound to fall short. Think of WCCA the way you'd approach your first DIY plumbing job. Being your own plumber takes more than just a trip to the hardware store and a few YouTube videos.
The articles in The Worst Case blog will help you avoid the allure of saving a buck, enlisting professionals to support the process or perform the analysis, or at least realizing what they are getting into. While WCCA should be performed by someone outside the design team—be that someone internal or external to the company—you should understand the concepts, pitfalls, and goals of WCCA.
Here's a list of this and the next five topics we will cover, one per month. I expect to post six more.
- Going it Alone: Why worst-case circuit analysis is challenging to perform (this article)
- Failure to be rigorous
- Models, models, models
- Monte Carlo gone wrong
- There’s a hole in my data: Test data and datasheet issues
- Too many escapes and biases
What is WCCA?
A worst-case analysis is a quantitative assessment of a circuit or system's functional performance, accounting for manufacturing, environmental, aging, and, in the case of Space applications, radiation tolerances. This is contrasted by part-based analyses such as stress & derating, failure mode, effects and criticality analysis (FMECA), and mean-time between failures (MTBF). All have their place in reliability assessment and there is certainly overlap in the math, models, and resulting conclusions.
Through a worst-case analysis, you can compute many aspects of a circuit's performance and calculate the risks and margins of key metrics. WCCA examines the tolerance-induced effects on electronic circuits caused by potentially large and unknown variations of components beyond their initial nominal value. The variations can be the result of manufacturing, aging, or environmental influences, which can cause circuits to drift out of specification.
WCCA also lets you assess the mathematical sensitivity of circuit performance to these variations and provides both statistical and non-statistical methods for handling the variables that affect circuit performance. These results can be valuable in helping to improve product quality. Figure 1 shows the interrelationships between WCCA and various part-based analyses.
Figure 1 The relationship between various reliability assessments, both functional (WCCA) and part-based (Stress, FMECA, MTBF) is crucial in producing a reliable product.
Going it alone
The average age of aerospace engineers in the U.S. is 47. Jobs can't be filled by foreign workers, at least in some disciplines. Analysis, including both worst case stress and derating analysis and worst-case functional analysis require many different skills and a great deal of experience as we will learn throughout this blog set.
The skills and knowledge bases required to successfully perform a WCCA include:
- Knowledge of the design, its functional sub-blocks, and their requirements.
- In-depth experience of the related circuit application(s): What is important to get right about a power supply, an amplifier circuit, or a microprocessor interface.
- Knowledge of part characteristics, their performance nuances, and the related circuit sensitivities. See regulation example below.
- Circuit modeling: being able to generate a SPICE, IBIS, and other simulation models for parts and circuits with the appropriate fidelity; Knowledge of how to vet and adjust vendor provided models (rarely are they sufficiently accurate), ability to leverage an existing model library of known good models; ability to correlate a model to test data, and finally, knowledge of how to tolerance each model.
- Mathematics: being able to write a math equation for a circuit.
- Testing: the ability and proper equipment necessary to gather and interpret test data at the component, functional block, and system level. I'll cover this in greater depth in a later post.
- Tolerances: knowledge of the tolerance stack-up; environmental tolerances, historical database of aging tolerances and, for Space applications, radiation data.
- Tools: A circuit simulator, SPICE covers most applications but an EM/harmonic balance simulator capable of simulating both the circuit and the printed circuit board, such as Keysight's ADS or Cadence's Sigrity, and others are becoming more and more essential, digital analysis tools such as Mentor’s Hyperlynx, and, of course, Mathcad, and Excel (Figure 2).
Figure 2 Efficient circuit analysis requires a set of tools, experience, data, and application knowledge.
So, while the computational methods—parametric (extreme value analysis (EVA), sensitivity, and Monte Carlo analyses—are common, each analysis differs in scope and focus based on the features of each functional circuit block being analyzed.
Critical analyses should not be performed by junior engineers or hourly contractors who don't have the requisite experience and tools to spot and fix issues. If you need the results to be correct, WCCA is simply not a good training ground.
[Continue reading on EDN US: Designers are not analysts]
Charles Hymowitz is a technologist, marketer, and business executive with over 30 years of experience in the electrical engineering services and EDA software markets.