Access Official Technical PDF

Share Summary:LinkedInEmail

On-wafer VNA testing is essential to characterizing semiconductor devices across the broad frequency ranges used in modern millimeter-wave applications. As measurement requirements extend from near DC to frequencies above 100 GHz, test teams must maintain accuracy and stability while managing the time spent calibrating vector network analyzers (VNAs). More stable broadband measurement systems can extend the time between calibrations, making more time available for device measurements and production testing.

Why calibration frequency matters in on-wafer VNA testing

VNA calibration establishes the measurement reference needed to obtain reliable device data. At millimeter-wave frequencies, however, maintaining that reference over time can be challenging. If system drift affects measurements, engineers may have difficulty distinguishing actual device-to-device variation from changes in the test system. Frequent recalibration helps manage that risk, but it also interrupts measurement work.

The impact depends on the manufacturing stage. Device characterization involves collecting detailed data to inform product specifications. Evaluation uses that data to make engineering decisions, while production testing checks devices against established specifications. Calibration downtime affects the first and third stages directly, and the quality of characterization data influences decisions made during evaluation.

Calibration tradeoffs across the semiconductor workflow

Device characterization

During initial characterization, engineers collect full S-parameter data from multiple devices across a wafer. Measurements may cover a wide frequency range and vary by bias condition, power level and, in some cases, temperature. Because probing is often performed manually, engineers also need to manage probe movement and contact issues such as oxidation.

Calibration standards may be placed on the wafer, but doing so uses wafer area. A separate calibration substrate may therefore be used, adding further steps to the process. In the example described in the source, calibration can take 20 minutes and may be required every hour. Across a four-hour session, four such calibrations account for 80 minutes, leaving 160 minutes for measurement. That corresponds to a measurement efficiency of about 67% of the session.

Reducing calibration frequency can return a substantial portion of that session to measurement. With one 20-minute calibration instead of four, the example leaves 220 minutes for measurement, or about 92% of the four-hour session. That is a 38% increase in available measurement time compared with the hourly-calibration case. More measurement time can support broader data collection without changing the characterization plan.

Evaluation and production test

Engineers use characterization data during the evaluation stage to assess device performance and set product specifications. Calibration time may be less significant during evaluation itself, but confidence in the underlying data matters. Measurement uncertainty influences how confidently teams can make tradeoffs between tighter specifications and manufacturing yield.

In production, automated test equipment collects a reduced data set to determine whether each device meets its specifications. Individual tests may take only seconds, so calibration pauses can have a meaningful effect on station availability. The source compares eight hourly calibrations of five minutes each over an eight-hour period with one five-minute calibration. Available measurement time rises from 440 to 475 minutes, an increase of about 8%.

In the example, at a three-second test time per device, that additional time could allow a station to test about 700 more devices per day. The source also notes that for a 10-second test time, the corresponding increase would be about 210 devices. These figures are tied to the stated test times and calibration assumptions; actual output depends on the test process.

How broadband VNA design supports stability

Earlier broadband systems commonly used large test heads assembled from waveguide and coaxial components. Such systems provided a route to frequencies up to 110 GHz, but could be less stable and offer reduced performance. More recent solutions use integrated millimeter-wave integrated circuit (MMIC) technologies. Their compact test heads can be mounted on a probe station and, in some cases, positioned directly on the probe.

The capabilities described for modern broadband VNAs include single-sweep frequency coverage from 40 kHz to 125 GHz, a stated dynamic range of 107 dB at 110 GHz, and positive raw directivity. The latter contributes to measurement stability and can reduce the need for frequent recalibration. Real-time power-level control supports more accurate linear-gain and 1 dB compression measurements, while compact modules can simplify positioning on the wafer probe station.

Stability over an extended period is central to the efficiency gains. The cited measurement example shows magnitude and phase stability 24 hours after calibration. When a system holds its measurement reference for longer, teams may be able to calibrate once per characterization session or once per day, rather than hourly. The appropriate interval still depends on the measurement requirements and the stability of the specific test setup.

Practical implications for test teams

For engineers and operations leaders, calibration frequency is not just an instrument setting: it affects how test capacity is divided between setup and data collection. When assessing an on-wafer measurement process, teams can consider:

  • How much of each characterization session is consumed by calibration.
  • Whether measurement drift could be mistaken for variation between devices.
  • How probe contact, calibration standards and substrate handling affect workflow.
  • How calibration pauses influence automated station availability and daily test output.
  • Whether measurement stability supports the required frequency range and data quality.

Calibration intervals should be matched to the accuracy and stability required for the work. The opportunity is not to treat calibration as unnecessary, but to reduce avoidable interruptions when the measurement system can maintain reliable performance for longer.

Conclusion

Stable broadband VNA measurements can improve on-wafer test efficiency at both the device characterization and production stages. Longer intervals between calibrations give engineers more time to collect data and give automated stations more time to test devices. For semiconductor teams, the practical benefit is a more productive use of test capacity, supported by measurements that remain accurate and dependable across the required frequency range.

Access Official Technical PDF

About The Author