Making Accurate RF On-Wafer Measurements for 5G and 6G
In the highly competitive RF semiconductor markets, companies strive to push the boundaries of power, efficiency, and linearity. Achieving even slight improvements — a few decibels (dB) in power or a few percentage points increase in efficiency — may catapult a product from “never was” to “best-in-class.” A crucial element in this pursuit is deploying an accurate on-wafer measurement strategy for device modeling and characterization. Only with accurate device models can design engineers simulate designs under different conditions, thus accelerating the development cycle of next-generation RF, millimeter-wave and terahertz products for 5G and 6G.
What is On-Wafer Measurement
On-wafer measurement, or wafer-level measurement, involves directly measuring the electrical performance of a device (i.e., transistor, inductor, and interconnect) on the wafer before dicing it. This step is crucial for understanding the device's behavior under various conditions and ensuring it meets performance specifications.
Emerging Challenges in RF Measurement
The RF design domain rapidly expands into higher frequency applications for 5G and 6G communications, automotive radars, and next-generation Wi-Fi standards.
One of the most significant benefits is the introduction of new frequency bands to alleviate spectrum scarcity within existing bands. Another advantage is the potential for wider bandwidths at higher operating frequencies.
However, the move to higher frequencies introduces significant challenges in ensuring measurement accuracy for device modeling.
First, greater throughput demands measuring an increased number of devices over multiple temperatures. The more devices measured, the better the statistical fit the modeling engineers will get, which is essential for better model quality. However, testing across temperatures requires continuous monitoring and time-consuming re-adjustment of the calibration.
Moreover, minimizing insertion loss becomes critical, as lower insertion loss translates to improved directivity, extended dynamic range, and reduced system drift. At frequencies like 330 gigahertz, even 1 micron's deviation in probe placement can introduce a degree of error. Thus, keeping a concise measurement channel is crucial.
In addition, software integration, critical for ensuring data correlation and measurement accuracy across different locales, can be time-consuming. Instruments and probers often run on proprietary software and firmware. Many R&D device characterization labs develop custom software to drive and control instruments, switch matrices, and probe stations using various languages such as Perl, MATLAB, LabVIEW, C#, or Python. While in-house software allows lab managers to control new features, custom drives, and data handling, the integration challenges are complex and often result in high hidden costs. Utilizing platforms like the Keysight WaferPro Express, which orchestrates the measurement process across various components, can improve efficiency and compatibility.
Key Considerations for On-Wafer Measurement Accuracy
Various factors can affect on-wafer measurement results, especially at millimeter-wave frequencies.
- Calibration method and reference plane choices: Calibration corrects systematic errors in measurement setups. However, the choice of method and reference plane must align with the specific requirements of the frequency range.
- Parasitic circuit elements: Parasitic elements arising from the interconnection between the probe and contact pads (like parasitic capacitance and inductance) can distort the measurements.
- Test environments: The material composition of the wafer chuck used during measurements (whether metallic, ceramic, or another material) can profoundly affect the results.
- Coupling between adjacent neighboring structures: In densely packed device layouts common to modern semiconductor wafers, electromagnetic coupling between adjacent devices or structures can lead to interference that affects measurement accuracy.
- Cross talk between probes: With multiple probes in proximity, often needed for complex measurements, electromagnetic crosstalk between the probes can occur. This phenomenon occurs when signals from one probe affect the signals of another, contaminating the measurement data with unwanted interference.
Understanding and controlling these factors is critical for engineers to ensure accurate device modeling and characterization.
The Path to Automated On-Wafer Measurement
To help companies automate RF on-wafer measurements for device modeling, Keysight has partnered with FormFactor to offer Integrated Measurement Systems (IMS), including integrated systems and application-specific hardware upgrades. IMS, designed to reduce the time to first measurement, can provide accurate and repeatable on-wafer device measurements for high-frequency applications.
A typical setup might feature Keysight's PNA Series microwave network analyzer and B1500A Semiconductor Device Parameter Analyzer, along with the WaferPro Express software, paired with FormFactor's semi or fully-automated wafer probe stations. Keysight’s WaferPro Express software takes the lead in automating control over instruments and prober software, including temperature management, simplifying test plan execution with advanced data handling and visualization capabilities.
To further improve modeling engineers’ productivity, the latest release has expanded WaferPro Express’ capabilities including:
- Simplify automated measurements with an easy-to-use modern interface.
- Control wafer temperature and prober chuck positioning automatically.
- Access an extensive library of configurations for semiconductor devices to quickly execute S-parameters, DC-IV, CV, noise figure, 1/f noise, gain compression, and more.
- Design and implement custom drivers and tests with Python or PEL and save measured data to ASCII files or SQL database format.
- Visualize data across wafers using a wafer data mapping viewer.
- Easily export data to modeling platforms such as Keysight’s Device Modeling Platform (ICCAP) and Model Builder (MBP).
- Enable instrument links, customized measurement algorithms, and data processing via Python 3.
Request a free trial today to see how it can help your lab accelerate on-wafer measurements and create highly accurate PDK models at a lower cost.