Free Signal Integrity?

Cadence and Wild River Technology (WRT), a leading supplier of signal integrity measurement and optimization test fixtures for high-speed channels at data rates of up to 224G, collaborated on a compelling presentation at DesignCon 2025 describing how understanding anisotropic materials and tolerances could result in "free signal integrity" by increasing electronic product performance at 112/224Gbps and beyond. The paper was written by Dr. Eric Bogatin, University of Colorado; Alfred P. Neves, Wild River Technology; and Kristoffer Skytte, John Phillips, and Frank Zaosh, Cadence.

Driven by time-to-market requirements, there is constant pressure on the signal integrity (SI) engineer to sign off on a design without adequate knowledge about the vendor's PCB fabrication process and the laminate system. Consequently, assumptions are made about the PCB geometries using the layout as-designed data from ECAD and the PCB material properties based on material vendor data or qualified estimates (Figure 1) along with the PCB fabricator's confirmed stack up (as fabricated) (Figure 2).

Figure 1: As-designed typical assumptions

Figure 2: As fabricated PCB stackup variation

Goals of the Study

The paper outlines and describes the goals of the study as follows:

• Examine the difference between as designed and as fabricated
  • What is the impact on risk and cost of this approach
  • How much insertion loss margin is left on the table
• Examine influence of anisotropy
• Perform material identification using AI
  • Isotropic vs anisotropic materials—bandwidth limits?
• Examine variation of electrical performance with design tolerances
• Suggest ways to improve digital twin

Figure 3 is a sample correlation for a 70GHz platform. The simulation was done based on the best available information for signoff in a typical design process and shows a very good correlation on return loss. The insertion loss suffers due to uncertainty, mostly in surface roughness. How much insertion loss margin is the designer willing to allocate to such a gap in the system budget?

Figure 3: As-designed simulation vs. fabricated (measured) of 8-inch stripline, full path

With higher losses, either more complex equalization and/or forward error correction (FEC) is required, which translates into higher overall system cost and power consumption. With most current design approaches, a significant simulation margin of about 5-10dB of insertion loss at the Nyquist must be added (the exact figure depends on the workflow), which requires confidence in the simulation tools and simulation setup, the PCB vendor's capabilities, and the risk aversion of the designer. Reducing this simulation-to-measurement gap can recover design margin and reduce the cost of a system pushing the limits, leading to a more cost-effective system.

This presentation explores a process based on detailed simulation and measurements to examine some of the challenges in reducing the gap between simulation and the manufactured design to achieve a more robust system. In this process, the influence of material modeling approaches, such as the difference between bulk isotropy and anisotropic assumptions, and manufacturing tolerances on the modeling process and performance are examined.

Conclusions

Several important conclusions were reached in the presentation. Specifically, uncertainty in the PCB fabrication process and material properties poses significant risks when signing off on as-designed PCBs. On the bright side, automatic material identification of bulk-isotropic properties using an AI-enabled optimizer is feasible and corresponds well to measurement. However, further research is needed for anisotropic material identification that is decoupled from inhomogeneous materials and manufacturing tolerances. Finally, several digital twin recommendations were proffered:

• Control insertion loss margin using an early-design-phase calibration platform in the target laminate system
• Need to extract relevant material properties from the fabricated test platform
• Explore the sensitivity of metrics from each geometry and material term—we need macroscopic SI
• Bound expectations and challenge your assumptions

View this video of the presentation to learn more about understanding anisotropic materials and tolerances.