Constantly, my colleagues and I are asked the same questions about PCB effects such as:
When are PCB effects required for simulation?
Why are PCB effects required for time domain simulation?
Does the frequency matter when PCB effects must be included in your simulation?
How much should designers care about including the PCB effects with the VRM model?
These are all valid questions, but my answer is always the same. If you want the right answer, PCB effects are ALWAYS required regardless of the frequency. In other words, whether you are simulating a 1 Gbps SerDes, 56 Gbps SerDes, DDR5-5600 bus, DDR4, power supply, voltage regulator module (VRM), or power distribution network (PDN), you must always include your printed circuit board effects to get the right answer! There are multiple inductive and capacitive parasitics on the PCB that can completely change the dynamics of a circuit's performance.
Why PCB Effects Are Essential for Accurate Simulation
I think we can all agree that the Signal Integrity community mostly takes this "school of thought" very seriously by generally including the PCB effects. However, there are still outliers, which is not always true. The largest gap of not including the PCB effects seems to exist within the power electronics community.
When moving beyond the schematic, it is crucial to understand the impacts of the PCB effects in your simulation.
The Impact of PCB Effects on Simulation Accuracy
For example, let's use the Texas Instrument TPS7H4003 evaluation board. The TPS7H4003 is an 18A DC/DC converter. With reference to Figure 1, the left waveform shows the time domain voltage response output from the Sandler State-Space Average VRM model without PCB effects included. The right waveform shows the measurement result of the output voltage response which clearly indicates an additional 1MHz oscillation not seen in the left plot. The center waveform depicts output voltage response when the PCB effects are included in the Sandler State-Space Average VRM model. As we can see, the center waveform not only now includes the 1MHz oscillation but also matches the measured waveform. This result would not have been achieved unless the PCB effects were included in the simulation result using Keysight ADS.
Figure 1 - Time Domain Simulation with and without PCB effects vs. Measurement [1]
Impedance and Noise Spectrum Analysis
We can look at another example showing the importance of PCB effects in your simulation by referring to an excerpt from a recent DesignCon 2024 paper done with some colleagues, "Design, Simulation, and Validation Challenges of 2000Amp Core Rail." With reference to Figure 2, when comparing the simulation result without PCB effects and with PCB effects, there are two very quick observations.
The first is a 22,000% increase in the path resistance (2.455 uOhm to 57.441 uOhm) indicated by marker m1 on each plot. In this case, the target impedance for a 2000-Amp sub-1V PDN is around 40 uOhm. Without PCB effects as designers, we may think we have plenty of margin, when in fact, you're actual resistance needs to be lowered by adding more copper to the design to meet the desired 40 uOhm impedance target.
The second observation is a 76% increase in our VRM control loop inductance (242.425 pH to 426.882 pH), observed when including the PCB effects.
With reference to EQ(1), this 76% increase in inductance means that this design will have a 76% increase in capacitance to meet the 40 uOhm impedance target. Again, this would not have been observed or caught without PCB effects. Ultimately, this would have led to a PCB re-spin, which would have cost more time and money in the schedule.
EQ(1)
Figure 2 - Impedance Simulation of a 2000-Amp Core Power Rail PDN without and with PCB Effects [2]
As a final justification for why the PCB effects are important to include in the design of a 2000A VRM and PDN, Figure 3 compares the noise spectrums with and without PCB effects. At the VRM’s switching frequency, the noise spectrum with PCB effects is 3 dB higher [2].
Figure 3 - 2000 Amp VRM and PDN Noise Spectrum with and without PCB Effects [2]
The Cost of Overlooking PCB Effects
In short, being right matters! Unplanned PCB re-spins can destroy a project development schedule. In addition to the additional costs necessary for a PCB re-spin, To avoid this PCB effects must also be included in your simulation. The parasitics that make up our PCB, whether it is for a transmission line or power plane can greatly impact the overall design if not properly accounted for before design sign-off.
Whether you need help with:
Generating PCB (SnP) models or package (SnP) models for your design sign-off
Want to ensure the signal integrity of your interconnects on your design with accurate electromagnetic analysis
Want to analyze the power integrity of the PDNs in your design with accurate electromagnetic analysis
Support end-to-end simulation efforts for power integrity or signal integrity
DC drop analysis on your power planes
Electrothermal Analysis
Generate EM extractions for differential pairs, DDR4, DDR5, or high-speed SerDes nets
Or even just ensure S-parameter (SnP) model quality Signal Edge Solutions is more than happy to work with you.
Signal Edge Solutions uses 3D and 2.5D field solvers to provide electromagnetic (EM) extraction services for PCB, package, substrate, or interposer designs.
To learn more about what Signal Edge Solutions can do to support you, please check out our services page:
In addition, we can be reached by email at info@signaledgesolutions.com
References: