TL;DR
This paper presents an optimized on-wafer calibration kit using coplanar waveguide (CPW) transmission lines for characterizing Indium-Phosphide (InP) devices up to 500 GHz. This advancement enables more accurate measurements of high-frequency transistors, pushing the boundaries of sub-millimeter-wave electronics.
What Problem Does This Paper Address?
As electronic devices push into higher frequency ranges (sub-millimeter-wave and terahertz), accurately measuring their performance becomes increasingly challenging. Traditional calibration methods, like SOLT (Short-Open-Load-Thru), struggle at these frequencies due to limitations in modeling lumped elements and probe-to-substrate coupling. Specifically, for advanced Indium-Phosphide (InP) Double Heterojunction Bipolar Transistors (DHBTs), which are crucial for high-speed communication and sensing, characterizing their behavior beyond 110 GHz requires more sophisticated on-wafer calibration techniques. Existing on-wafer calibration methods often rely on specific substrate types or have limitations in achieving the required accuracy at very high frequencies. This paper aims to bridge this gap by developing and validating an optimized on-wafer calibration kit for InP technologies that can reliably perform measurements up to 500 GHz.
Methods
The core of this work involves the design and optimization of on-wafer Thru-Reflect-Line (TRL) calibration structures. The authors first analyzed the limitations of their previous generation (G0) test structures through electromagnetic (EM) simulations and measurements, identifying issues with pad parasitics and band discontinuities at higher frequencies.
Based on this analysis, they developed an improved generation (G1) of test structures. Key improvements include:
- Optimized Pad Layout: The signal pad size was reduced, and ground pads were connected to better confine RF signals and reduce parasitic effects.
- Coplanar Waveguide (CPW) TRL Kit: A TRL calibration kit was designed using CPW transmission lines, which are well-suited for InP processes that may not have a ground plane on the bottom metal layer. This kit includes Thru, Pad-Open (reflect), and two different Line standards to cover the desired frequency range.
- EM Simulation-Assisted Design: Extensive EM simulations using Ansys HFSS were employed throughout the design process. This included simulating the intrinsic device, the complete test structures with RF pads and probes, and the interaction between neighboring structures. This allowed for the prediction and mitigation of measurement artifacts.
- On-Wafer Calibration: The developed TRL kit was used for on-wafer calibration, setting the reference plane at the edge of the Pad-Open. A subsequent Short-Open de-embedding step was then applied to extract the intrinsic transistor's parameters.
Datasets
Not discussed in the paper. The paper focuses on the design and characterization of test structures and devices, rather than training on large datasets.
Key Insights and Ideas
- CPW TRL is Crucial for InP: The use of CPW transmission lines for TRL calibration is essential for InP technologies, especially those lacking a bottom ground plane, enabling accurate on-wafer measurements.
- Pad Parasitics are a Major Hurdle: The size and design of the RF pads on test structures significantly impact high-frequency measurements. Reducing pad parasitics is critical for extending accurate characterization to higher frequencies.
- EM Simulation is Indispensable: EM simulations are not just for verifying designs but are crucial for understanding and predicting measurement artifacts like coupling and crosstalk, which become dominant at sub-THz frequencies.
- Neighboring Structure Coupling is a Significant Issue: The proximity of test structures can introduce coupling and crosstalk, leading to inaccuracies in measurements, particularly for parameters like fmax. This highlights the need for careful layout design and potentially new calibration strategies for future generations.
- On-Wafer TRL Outperforms Off-Wafer SOLT at High Frequencies: The paper demonstrates that on-wafer TRL calibration provides significantly better results and a flatter frequency response for passive structures compared to previous off-wafer SOLT calibrations.
Key Results
- Characterization up to 500 GHz: The developed on-wafer TRL calibration kit enabled accurate measurements of passive devices up to 500 GHz.
- Transistor Characterization up to 330 GHz: InP DHBTs were reliably characterized up to 330 GHz using the new calibration approach.
- Improved De-embedding: The optimized G1 test structures showed a significant reduction in parasitic effects (e.g., ~10 fF in open capacitances and ~50 pH in short inductances) compared to the G0 generation.
- Accurate Compact Model Validation: The HiCuM compact model for InP DHBTs was validated against measurements up to 330 GHz, demonstrating satisfactory agreement in both magnitude and phase of S-parameters.
- Identification of Crosstalk Effects: EM simulations and measurements revealed significant crosstalk effects impacting measurements beyond 220 GHz, particularly affecting fmax.
Limitations and Future Work
- Measurement Uncertainties Beyond 220 GHz: While transistor characterization was achieved up to 330 GHz, some measurement uncertainties, particularly for fmax, were observed beyond 220 GHz. This is attributed to crosstalk and parasitic modes.
- Future Work on Crosstalk Reduction: The authors propose further layout improvements for future generations (G2) to reduce crosstalk and coupling. This includes:
- Continuous Ground Plane: Implementing a continuous ground plane around test structures.
- CPW Scaling: Further reducing CPW line dimensions.
- Modified Back-End Structures: Exploring options like benzocyclobutene (BCB) dielectric and microstrip line topologies.
- Extended Thru Length: Increasing the Thru line length in the TRL kit to 100 µm to mitigate crosstalk.
- Extending to 500 GHz for Transistors: The ultimate goal is to extend reliable transistor characterization to 500 GHz, which will require addressing the identified crosstalk and parasitic issues.
Summary updated