An interferometric surface profiler for high-magnification fixed through glass measurement (patent pending), such as for a packaged MEMS or MOEMS measurement, is described in this paper. Three techniques are introduced into the profiler, including aberration correction and long working distance for the objective, substantially shaped illumination and dispersive compensation. Measurement results illuminate that the data of the through glass objective are very close to those of the standard objective.
Optical profilers that employ white-light interferometry (WLI) techniques are widely used in industry and research to map surface structures with precise sub-nanometer resolution. The accuracy, repeatability and consistency of these instruments depend mainly on the proper calibration of two parameters: the mean wavelength and the scanner speed. Equally important in industrial applications is high throughput of measuring devices; high throughput can be achieved by increasing the speed of the measurement, which in turn reduces the sampling rate of the interference signal. Again, exact calibration of the system is the key to obtaining accurate measurements. Our solution for calibrating the system involves inserting into the optical profiler a second, laser-based, reference signal interferometer that creates a primary length standard for calibration. This innovation allows for substantially improved performance and repeatability at different measurement speeds up to 100 microns per second and over an 8 millimeter scanning range.
One of the primary challenges in MEMS metrology is the large variety of shapes, lateral feature sizes, and vertical steps on MEMS devices. This paper describes a software approach by which ideal surface templates are generated for each MEMS device from the design files or prior measurements. These templates may contain multiple sub-regions, or data islands, each of which is generally characterized in a different manner.
Surface measurements from a white-light interference microscope are matched with the ideal MEMS template using a variety of techniques and threshold criteria. The template-based technique is tolerant of errors both rotation and translation, allowing accurate characterization of each data island and their relative positions.
This paper will explore the concepts used to generate templates, align actual data with the original template, and sources of error and robustness of each technique on several datasets. The effect of measurement and positional errors on both the overall match and on the sub-region analyses will be explored for characterization of datasets.