Silicon Characterization


For RF designs, EM simulations become indispensable. Silicon data are important reference data to ensure design closure. Silicon data-based EM device modeling can be quantified and enhanced by EM tools.

Typical Foundry Silicon Characterization has their limitations. From test key design, test key manufacturing, measurement to de-embedding, almost all manual work, long cycle, and hard to add new devices for different apps/frequencies. Uncertainty exists in the de-embedded results and de-embedding method. Limited silicon data reference leads to uncertainty in EM design.

Peakview-based Rapid Silicon Characterization approach provides a fast turnaround to support customer requirements, including software and simulation based flow. High-quality silicon data with a quantified error margin of inherent methods are used and compared. Reliable correlation with process information and EM simulations enables customer satisfactions.

Why PeakView-based silicon characterization?

Lorentz Solution’s Peakview EM platform support the automation of Silicon characterization.   It allows automatic synthesis of DUT (device under test). Typical de-embedding methods are supported, while new de-embedding methods can be developed and extended.

Different types of test keys can be developed, simulated and de-embedded:

  • DUT with Pad Structure
  • Shorted Pad Structure
  • Open Pad Structure 
  • Through Pad Structure

Before Tapeout b4TapeoutAfter tapeout course of action:



Silicon Data Correlation

PeakView LEM™ simplifies EM design, and comes close to sign-off quality with measurement data. Our products are seamlessly integrated into the IC design environment to support the advanced process nodes, thereby greatly facilitating the key steps for EM design. PeakView™ EM modeling solutions from DC to TeraHz frequency range demonstrate excellent correlation to silicon and golden simulation data. Our EM design flows have been demonstrated in TSMC RF Reference Design Kits and test-keys, simulation and measurement results.

Silicon correlation for PA and LNA designs at 60 GHz using PeakView™ Transmission Line Models for TSMC 65nm process is shown below. Then at TSMC 20nm process, PeakView™ simulations of inductors with a metal fill shapes and densities were compared to measurements. The result indicated a close match of silicon measurement when PeakView CMP was used for metal fill modeling.


Silicon Correlations for PeakView™ Transmission Line Models for (a) PA Module (b) low-noise amplifier (LNA) designs shows closely matched between circuit simulation and silicon measurements at 60GHz


                   Simulation vs. Measurement in TSMC 20nm Process

For silicon data correlation, 200μm and 400μm long slow-wave transmission lines and GSG Pads are also designed, simulated and fabricated in TSMC 65nm technology and are shown in the following figure. The profile is directly converted from iRCX file from the foundry.


(a) 200μm transmission line with GSG Pads (b) 400μm transmission line with GSG Pads


For 200μm transmission line and frequency up to 67GHz (a) S12, (b) S11. Blue lines are simulation data. (Silicon data courtesy of TSMC)


For 400μm transmission line and frequency up to 67GHz (a) S12, (b) S11. Blue lines are simulation data. (Silicon data courtesy of TSMC)

For more Si data correlation, please contact us at

Comments are closed.