Estimated reading time: 7 minutes

Elementary, Mr. Watson: Circuit Simulation, SPICE, and AI

According to the Federal Aviation Administration’s flight time limitations and rest requirements, a commercial pilot is restricted to 36 flight hours in a week, 100 hours in 28 days, and 1,000 hours in any calendar year¹. The cycle time depends heavily on the aircraft they fly and length of time in the air. It can often be long and grueling, so these limits are there to ensure the safety of the flight.

But before these pilots even start flying the friendly skies, they spend considerable time in a flight simulator. Airline pilots go through about a month of simulator training before they can fly on the line, and this comes after two to three weeks of systems and operations training.

Most airlines maintain a rigorous simulator training program of 18–24 simulated flights lasting four hours each. During these flights, they train on system reviews and procedures, including checklists and "flows"—special maneuvers, and real-world problem situations. All that can add up to 96 hours of simulator training before they walk aboard an actual plane. This level of training is done every six months to ensure that their regular and non-normal operation capabilities are up to standard.

Why? In the simulator, pilots have a controlled, safe environment to practice various flight maneuvers, emergency procedures, and critical decision-making without the risks associated with actual flight. That allows pilots to gain experience in handling challenging situations and improves their ability to respond appropriately in real-world emergencies without crashing a plane if they make a mistake. Furthermore, conducting flight training in a simulator is more cost-effective than training in an actual aircraft. Simulators reduce fuel, maintenance, and operational costs, allowing pilots to gain valuable training hours without the high expenses of flying real planes.

Take Your Design for a Test Flight
In the same way, PCB design and circuit simulations provide the same advantages: a controlled, safe environment for testing a circuit without the considerable overhead of fabricating and assembling the PCB. That is all done through SPICE simulations (Simulation Program with Integrated Circuit Emphasis). SPICE is a computer simulation and modeling program used to predict the behavior of electronic circuits mathematically.

Many PCB designers feel that circuit simulation is something they will never need to learn, and it is best handled by that infamous "someone else." But in our present business environment, with the lack of PCB design and engineering resources, many designers are pulling multiple duties throughout the design process. Having individuals understand various disciplines, including PCB simulation, increases your overall value to any company. I always recommend that new PCB designers learn anything and everything they can that would be an asset to the company.

Circuit simulations have a rich history that has mirrored the ever-changing electronic innovations. Circuit simulation must keep up with ever-increasing technology and more complex circuits. It all began in the early 1970s at the University of California, Berkeley, with Dr. Laurence W. Nagel and his team. Initially designed for simulating integrated circuits (and part of Nagel’s PhD dissertation), SPICE aimed to analyze and optimize the performance of electronic systems at the transistor level. It wasn't until the late 1990s to early 2000s that SPICE simulation became integral to comprehensive EDA software suites, integrated with schematic capture and PCB layout tools. This integration allowed designers to move between schematic design, layout, and simulation seamlessly.

The 2010s was the golden age of circuit simulations with significant advancements. With the advent of cloud computing, some EDA companies began offering cloud-based SPICE simulations, enabling scalable resources for more extensive simulations.

Furthermore, advancements in machine learning and AI techniques were integrated into SPICE simulators to optimize designs, perform parameter sweeps, and explore design spaces more efficiently, which we will discuss later in this column.

Altium’s SPICE Environment
Altium has what can be described as an ongoing development of its SPICE tool. Although we originally released a circuit simulation tool in the early 2000s, the release of Altium Designer 22 in 2022 featured a fully revamped simulation tool, with a focus on several key points.

Circuit simulation was fully and seamlessly integrated into the Altium Designer PCB design environment. This integration allows design engineers to move seamlessly between schematic capture, PCB layout, and simulation, streamlining the design process.

A completely redesigned advanced simulation engine that can handle complex circuits and mixed-signal designs drives the platform. That facilitates the co-simulation of analog and digital circuitry. It can quickly analyze the interaction between the two domains, ensuring proper integration and functionality. It supports various analysis types, including transient, AC, DC, and noise. Also, parameter sweeping is a remarkable advanced simulation technique for sweeping different circuit variables, enabling them to analyze the circuit's response to varying component values. This feature helps understand circuit sensitivity and optimize design parameters.

As mentioned earlier, SPICE circuit simulations mirror the advancements in our industry, which means the challenges are constantly shifting and changing. The primary area is the simulation of high-speed designs. Access to frequency domain analysis to analyze circuit behavior in the frequency domain is essential for RF and communication systems. It helps ensure proper frequency response, gain, and filter characteristics. Directly tied to that is transient response analysis, which assesses circuits' transient behavior during startup or switching events. This analysis is critical for understanding circuit stability, settling times, and response to input changes.

Shrinking the Simulation Learning Curve
SPICE was once seen as a complex tool that would take years to master, but today’s SPICE flavors, such as Altium SPICE, are a tad more user-friendly. Of course, based on what I've seen teaching this subject, the time necessary to learn it will vary from person to person. As a PCB design instructor, if I were to set up a curriculum for a circuit simulation course (hey, that’s a great idea), I would put the following time estimates for each objective of the course:

• Basic simulation set-up (two to three weeks). Users start working with the circuit simulation tool, learning to add and configure component simulation models. To understand component behavior, fundamental simulation analyses like DC and AC analysis are explored. Users begin by running simulations on simple circuits and verifying the results.
• Advanced analysis (two to four weeks). Users progress to more complex analysis types, such as transient analysis, noise analysis, and parameter sweeping. Convergence issues and adjusting simulation settings will be covered. Signal integrity analysis for high-speed designs may also be introduced.
• Mixed-signal simulation (two to three weeks). Mixed-signal simulations, including simulating the interaction between analog and digital components, as well as co-simulation techniques and troubleshooting mixed-signal designs are addressed.
• Design optimization and rules (two to three weeks). Students would use simulation results to optimize PCB designs iteratively and make design improvements based on simulations. Curriculum would also focus on implementing design rule checks to ensure compliance with specific design constraints and industry standards.

To become proficient with the Altium SPICE tool, I estimate roughly eight to 13 weeks.

Circuit Simulation for Beginners
I would start with simple circuits and progress to more complex designs. Master the fundamentals of circuit simulation, such as DC and AC analysis, before moving on to advanced techniques.

• Familiarize yourself with different simulation models (e.g., ideal, behavioral, and transistor-level models) and their limitations. Understanding the accuracy and applicability of these models will help you interpret simulation results effectively.
• Take advantage of tutorials and online resources provided by the simulation tool's documentation and the broader community. Work through examples to grasp simulation techniques and gain practical experience.
• Learn to perform parameter sweeps to explore how varying component values impact circuit performance. Parameter sweeping helps you understand the sensitivity of your circuit to different parameters.
• Familiarize yourself with component datasheets and their corresponding SPICE models. Accurate component models are crucial for reliable simulations. And remember that circuit simulation is an iterative process. Don't be discouraged by initial results; use them to refine and improve your design iteratively.

Simulation, AI, and the Future
With AI's development, I believe our world will never be the same. I would put this on the level of the discovery of fire or the invention of the wheel. AI is not going away, so we should embrace it. I am sure it will change how circuit simulations are performed. Here are some ways I see AI impacting this whole area in the future.

Simulation
AI-based algorithms can enhance simulation efficiency and accuracy by leveraging computational models and approximations. This can significantly reduce simulation times for large and complex circuits while maintaining acceptable accuracy. Furthermore, those AI algorithms can automatically optimize circuit designs based on specified objectives and constraints. By leveraging machine learning techniques, AI can explore vast design spaces to find optimal solutions for various performance metrics, such as power consumption, signal integrity, and thermal efficiency.

Predictive Analysis
With the ever-growing and knowledgeable platform of AI, predictive analysis will describe the behavior of complex circuits by learning from past simulation results and identifying patterns in circuit performance. Take this to its logical conclusion. There will be an incredible improvement in circuit failure analysis; AI will be able to analyze your circuit and, from the simulation results, predict potential points of failure in the circuit design, improving circuit reliability.

AI is an intriguing area of research, and we are barely in the infancy of this area. Fasten your seatbelts, because we haven't seen anything yet.

References

1. “How many hours and cycles do pilots amass in 12 months?” by Jack Herstam, simpleflying.com, Dec. 29, 2022.

This column originally appeared in the August 2023 issue of Design007 Magazine.