The company I work for manufactures small, tightly toleranced components for the disk drive industry, and dimensional metrology/calibration is constantly being pushed to the bleeding edge in order to meet ever narrowing process and product specifications. I am facing requests for equipment that will have the accuracy and repeatability to measure product specifications of +/-.005mm within the next year and +/-.001mm within 3 years. We have equipment such as a Leitz UMS 432, a Leitz Libra, and various View Engineering platforms such as Pinnacle and Voyager. Most of these machines are capable of measuring a NIST certified artifact to an uncertainty of less than 1.5micron +/- (L/200mm) or so. This makes a target uncertainty of less than 20% of product specification virtually impossible with current equipment. One of the big problems with video inspection probes/systems in my mind is the diffraction limitation of light. With the lenses on equipment such as I listed, the resolution of features using white light is slightly larger than a micron. Equipment manufacturers have done things like sub-pixeling in order to obtain repeatabilities of 0.1 to 0.2 microns sigma(m), but I have yet to see that this is rigorously proven to be in fact, a relationship to the real measurand. I suspect that it is signal averaging after the fact, and indeed gives a fine fit to a theoretical best-fit model that may or may not have any true relation to physical fact. I have seen a paper published for a manufacturer of wafer inspection equipment that uses I-Line UV light and appropriate optics to take the theoretical resolution limit down to around 0.5 microns. This would obviously help but I doubt if it would even get the resolution and accuracy needed.

I am interested in learning:

1) if anybody else questions or has a definitive answer to the sub-pixeling as a valid and true measure issue,
and
2) if there are instruments with larger stages than wafer inspection systems, that are using I-Line UV optics.