Enhanced Machine-Learning Based Probe Alignment for On-Wafer RF Measurements

D. Vitali^#, A. Chillico^#, B. Puri^*$, R. Heldmaier^*, R.P. Klausen^*, W. Samek^*$, O. Bengtsson^#

Published in:

Proc. 55th European Microwave Conference (EuMC 2025), Utrecht, Netherlands, Sep. 23-25, ISBN: 978-2-87487-081-1, pp. 580-583 (2025).

Abstract:

In this paper an automatic device under test (DUT) to probe-tip alignment technique for on-wafer measurements is presented. The method uses a machine learning (ML) model to identify the boundaries of the DUT and correct the position of the probes, with a 95% confidence level of 1.2 μm. A comparison of wafer-wide measurements with and without the correction show an average deviation due to the operator’s manual positioning of X=-13 μm and Y=-1.5 μm from the correct position, with a variation due to mechanical tolerances from the probestation of 2σ ≈ 6 μm. S-parameter measurements of a RF transistor show that the correction can increase the accuracy in the variation of the measurements up to 10%. Additional measurements on a 200 μm coplanar waveguide line, with intentionally introduced offsets, have been conducted to further quantify these effects. These measurements show a considerable variation when more than 5 μm misplacement from the correct position is present.

Keywords:

Automation, GaN, HEMT, machine learning, microwave measurements, on-wafer measurements.

Copyright © 2025 EuMA. All rights reserved. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the EuMA.

Full version in pdf-format.