Circuit Simulation Shapes Up for Energy-Conscious Design

Although simulation tools have evolved, the demands placed on engineers continue to intensify. System designs are becoming increasingly complex as the world looks to electronics and semiconductor innovators to solve today’s challenges. Performance expectations have increased, design margins have narrowed, and time-to-market pressure is perennially more critical.

In particular, today’s power designs are turning to more advanced architectures, including high-efficiency topologies and digital power. Hence, engineers want to perform more digital simulations with AI, driving increasingly complex designs alongside power and RF work.

 

Image used courtesy of Adobe Stock 

 

In 1970, SPICE (Simulation Program with Integrated Circuit Emphasis) was created at Berkeley and has become the reference for integrated circuit simulators, earning an IEEE Milestone in Electrical Engineering and Computing award in 2011. The ability to model components and simulate their behavior has empowered generations of engineers—not only chip designers but those developing all kinds of circuits and systems. They have been able to rapidly prototype and iterate their designs, quickly establish performance predictions under different operating conditions—optimizing before building the hardware, a costly and resource-intensive proposition—and generally analyze and optimize their designs more quickly and effectively to accelerate successful project completion. Simulation allows designers to design and optimize before building the hardware, a more costly and resource-intensive proposition.

 

Simulation Improvements Tackle Design Complexity

Some of the simulator improvements many engineers enjoy today are enabled by the vastly greater computing power on their desktops. Key developments include:

• solid-state drives that deliver high storage density and fast response
• faster processors 
• the adoption of GPUs

This move toward heterogeneous processing is a widespread trend throughout the industry. SPICE engines offer a close to 100,000-fold increase in the time taken to render data, enabling the tools to display the results quickly and accurately on screen. It’s now possible to display every data point in waveforms that would have been impossible to render at such a detailed level in the past.

 

The ability to simulate massive amounts of digital makes it possible to handle schematic capture and compile and run digital code. This is not usually possible with easily accessible and affordable tools, much less free ones. 

 

On the other hand, further enhancements have been intrinsic to the tools themselves, including Improvements to the SPICE modeling engine that ensure greater accuracy and increased power and analog capabilities. 

In addition, there are notable bug fixes. Some commercial tools have contained implementation errors that trace back to the original Berkeley SPICE code, allowing discontinuities in the I-V curves of semiconductor devices. Qorvo’s development of the QSPICE circuit simulator provided an opportunity to correct these errors and unleash the real power of the SPICE engine for analog circuit simulation. With these discontinuities now eliminated, simulations are completed more accurately and reliably—a significant step toward faster and more reliable simulation.

 

Figure 1. Improved current-voltage (I-V) curves in the QSPICE simulator. Removing discontinuities improves numerical convergence. Image used courtesy of Qorvo

 

Digital Capabilities

Simulating massive amounts of digital makes it possible to handle schematic capture and compile and run digital code. This is not usually possible with easily accessible and affordably priced tools, much less free tools. Historically, SPICE tools have been aimed purely at analog circuit simulation. Now, it is possible to click a button to compile Verilog or C++ code and quickly run high-speed digital and analog simulations.

This can help overcome challenges in developing applications, leveraging artificial intelligence (AI) and machine learning (ML). An engineer can take the code base for an open-source RISC-V processor, load it into a QSPICE code block, and compile it. They can then load in code for a TinyML algorithm that runs on RISC-V and simulate it next to their motor driver and model to run the analog simulation together. 

Another example involves developing systems to make batteries charge more effectively or to maximize the number of charge cycles. It’s now possible to put the battery model in and quickly start running machine-learning algorithms. So, power design challenges can be solved without needing to build a physical test bench or handle the additional concerns of large motors or batteries.

Moreover, the possibility of timestep control further enhances designers' ability to tackle power design challenges. Timestep control helps accurately and efficiently simulate power electronic circuits, facilitating insights into fast-switching events, nonlinearities, and dynamic behavior. 

With timestep control, the simulation can automatically adjust for smaller time intervals during critical events to capture dynamic response or noise and ripple effects. It can help analyze the system’s transient response or switching waveforms’ fast rise and fall times. While setting larger timesteps during slower variation periods can reduce computational costs, the flexibility to use smaller time steps during fast-switching events yields more accuracy.

 

Figure 2a. QSPICE simulation circuit, which captures the power dissipation and demonstrates automatic timestep control. Image used courtesy of Qorvo

 

Figures 2b and 2c. Instantaneous power dissipation in the circuit element is captured in the QSPICE simulation. The second plot shows how the dynamic timestep algorithm maintains accuracy and speed: When the circuit operating conditions change slowly, QSPICE takes longer time steps; when the operating conditions change rapidly, time steps are smaller. Image used courtesy of Qorvo

 

Accuracy in Power Analysis

While the systems to be simulated are becoming increasingly complex, the demand for accurate analysis is intensifying. Power analysis is a particularly pertinent example. Power consumption is critically important from the standpoint of product performance —influencing metrics such as electric vehicle range, runtime of cordless tools, renewable-energy cost per watt—and sustainability. Consumption of many devices is now in the order of nanoamps. Hence, in addition to accuracy, achieving a suitable resolution is extremely challenging.

 

Closer to Reality

While complexity and accuracy continue to become more challenging, some demands remain the same. Users want the results to be as close to real life as possible, and tool developers have the perennial ambition to achieve this. 

All this must be achieved without excessive demand for system resources such as storage and simulation execution time. The QSPICE simulator can perform digital simulations quickly because the Verilog or C++ source code that describes the digital logic is compiled into native desktop processor object code. 

Circuit simulation has come a long way but will always be chasing a target that is not only moving but accelerating. The journey will never end, and progress is self-fueling. As the tools enable technical development, the results of that development—especially faster and more efficient processors—can further improve the modeling process. By adding capabilities such as massive digital simulation and enhancing accuracy and general improvements to the SPICE engine, tools like the QSPICE simulator can support state-of-the-art projects in fields such as power conversion, e-mobility, and AI.

Qorvo is keynoting our inaugural EEPower Day virtual event. “QSPICE for Power Circuit Simulation: A Spicy Discussion with Mike Engelhardt” will feature tips and tricks for getting the most out of QSPICE, the hardware and software advancements that enabled QSPICE, and the lasting legacy of LTspice. Register now; this is an event you don’t want to miss. 

 

This article is co-authored by Jeff Strang, GM of Power Management, and  Surinder P. Singh, Sr. Manager of Customer Design Tools, with Qorvo, Inc.