Power integrity engineers are running into a new reality: most modern voltage regulator modules (VRMs) control loops are no longer linear, time‑invariant systems, and the old “break the loop and run a Bode plot” workflow does not tell the whole story anymore. Keysight’s ADS offers a complete approach by letting you simulate the entire power‑delivery path end‑to‑end — from the VRM and PCB to the decoupling network and all the way into the ASIC. With the Sandler State‑Space Average Model at the core, ADS becomes a practical environment for closed‑loop stability assessment that reflects how today’s converters actually behave.

Two measurement‑driven methods fall naturally into this workflow. Non‑invasive stability measurement (NISM) provides a frequency‑domain view of stability directly from output‑impedance data, while scope embedded power integrity analysis (SEPIA) turns step‑load responses into a time‑domain stability model you can drop straight into ADS. Both approaches deliver quantitative stability metrics without breaking the loop, and both work cleanly with non‑LTI systems. Bode plots are still available in ADS for traditional analysis, but they don’t apply to most modern VRMs. These newer methods give engineers a realistic, simulation, and measurement-ready way to evaluate and optimize stability across the full power distribution network (PDN).

Control loop stability has always been of interest in power electronics because the control loop ultimately determines the circuit’s end performance. Poor stability shows up immediately as elevated noise. Steve Sandler first presented on this topic at CMSE 2011, where he showed the relationship between stability and a power supply’s closed‑loop functions. He continued presenting on this in the years that followed, including his APEC 2012 paper presentation titled, “High-Fidelity Power Supply Measurement.”  A key takeaway from that paper was that all closed-loop performance is affected by control loop stability, as summarized on slide four of that presentation, see in Figure 1.

Fig 1. A summary of all closed-loop power supply performance metrics and how they are impacted by control loop stability.¹

We do not always have the option of traditional Bode plot measurements, since in many cases open‑loop Bode plots are simply not possible. Voltage references and three‑terminal regulators do not expose the feedback loop. Op‑amps and RF amplifiers have bandwidths too high for practical signal injection. PMICs make this even more challenging by integrating multiple regulators into a single chip, with only the outputs available.

To overcome these limitations, Steve Sandler developed a closed‑loop stability tool, NISM, which mathematically relates the measured output impedance to the stability of the control loop. NISM provided a simple, quantitative assessment of stability without access to the loop. Adoption was slow at first as engineers were comfortable with Bode plots, and NISM did not always agree with traditional phase‑margin/gain‑margin results. Sandler repeatedly showed that NISM was correct and that the Bode plot, limited to phase and gain margins, does not always indicate relative stability. Engineers seemed to know this already: after running a Bode plot, they would generally check the step‑load response. If the step load was the trusted truth, why not just use that in the first place?

That led Sandler to develop SEPIA in 2013, which extracts quantitative stability directly from the step‑load response. Engineers trusted step loads, SEPIA agreed with NISM, and both methods worked non‑invasively. Still, adoption took time. It was not until the last few years — when measurement access became even more limited, noise limits tightened, dynamic currents increased, and VRMs shifted to non-linear, time‑variant control schemes — that interest accelerated. Suddenly, the industry needed methods that do not rely on assumptions of linearity or access to the control loop.

Simulation has always been part of the workflow, and this is where the Sandler State‑Space Average Model fits in. First published in 1995 and updated several times since, it provides a practical way to simulate closed‑loop behavior, including large‑signal effects, EMI, and other real‑world behaviors. Integrating NISM and SEPIA into Keysight ADS was the obvious next step. Ben Dannan and Heidi Barnes played a major role in getting these methods into ADS, and Masashi Nogawa — through his publications and continued advocacy — has been instrumental in driving renewed interest and adoption. These methods are now becoming even more accessible as they are integrated directly into measurement instruments. NISM has been available in the OMICRON Lab Bode Suite since 2012 and is now supported in a range of oscilloscopes, VNAs, and simulators. SEPIA is now being integrated into oscilloscopes as well. The introduction of Picotest high-speed step loads allows clean, high-bandwidth data to be acquired, furthering the benefits of SEPIA.

This article demonstrates a rigorous simulation-based approach to power integrity within Keysight ADS. We explore why traditional stability metrics often fail to capture the complexities of modern VRMs, particularly in non-linear or time-varying scenarios. By leveraging NISM and SEPIA directly within the ADS environment, designers can achieve a more realistic and accurate view of system stability. This unified simulation workflow ensures that the virtual model reflects the true behavioral dynamics of modern converters — principles that translate directly from the digital workspace to physical hardware validation.

Fig 3. Simple PDN with a Step Response showing SEPIA.

• Stability Metrics: Directly calculates phase margin and Q to ensure power supply reliability.
• Modeling: Converts physical measurements into actionable SPICE models.
• Optimization: Recommends specific capacitor solutions to hit your “Target Impedance” goals.
• Versatility: Functions across various system types, even those that are non-linear or time-varying.

Fig 4. LTM4637 stability frequency domain simulation example using NISM in ADS.

Fig 5. LTM4637 stability time domain simulation example using SEPIA in ADS.

Fig 6. LTM4637 stability simulation results - NISM (left) vs. SEPIA (right).

Fig 7. Simulation setup for extracted SPICE model for LTM4637 PDN.

Fig 8. Simulation result for LTM4637 state-space VRM model vs. extracted SPICE model for LTM4637 PDN using SEPIA.

• The Golden Rule: Always remember that noise follows impedance. To manage noise, you must manage your impedance profile.
• Design Strategy: Utilize SEPIA or NISM to aim for a flat impedance PDN. If a flat response is not achievable, ensure the system is designed so that Q < 2 to maintain stability.
• Comprehensive Modeling: Power integrity spans from the power supply to the load; achieving accuracy requires end-to-end modeling in ADS and validation through both large-signal and small-signal simulations.
• Efficiency: When paired with StepLoad Pro on a scope, SEPIA transforms complex characterization into a 60-second automated reality, offering a ubiquitous solution across the entire design cycle.