This is an author-created, un-copyedited version of the article L. Galatro and M. Spirito, "Fused silica based RSOL calibration substrate for improved probelevel calibration accuracy,"which is published in the digest of the Microwave Measurement Conference (ARFTG), 2016 88th ARFTG, Dec. 2016. IEEE is not responsible for any errors or omissions in this version of the manuscript or any version derived from it. The definitive publisher-authenticated version is available online at http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7839721 Fused Silica based RSOL calibration substrate for improved probe- level calibration accuracy Luca Galatro, Marco Spirito Electronics Research Laboratory (ERL), Delft University of Technology, Delft, The Netherlands, Abstract — In this contribution we present a calibration substrate manufactured with integrated circuit technology on fused silica, used for reciprocal SOL calibration. The fabrication technology is described together with the standard layout, and the precise control of the geometrical properties is exploited to create accurate pitch-dependent standard model to be used during the calibration procedure. The calibration accuracy is benchmarked with the conventional alumina impedance standard substrates, using the provided (polynomial fit) standard definitions, giving an estimate of the accuracy improvement that the proposed calibration substrate can provide. Index Terms — VNA, calibration, on-wafer, probe-level, RSOL, fused silica. I. INTRODUCTION The accuracy of S-parameter measurements of any device under test (DUT) is set, at the first order, by the quality of the VNA calibration. This is typically performed by measuring a certain number of known devices (i.e., the calibration standards). Depending on the specific calibration technique employed, the quality of the calibration is directly dependent on the accuracy with which the calibration standards are known/modeled [1][2]. When considering planar devices, for which wafer-probes need to be employed, it is common practice to perform a probe-level calibration (first-tier) using a low-loss substrate (i.e., alumina or fused silica), which can then be transferred to the environment where the DUT is embedded. In planar environments, the accurate modeling of the calibration standards can become cumbersome, due to ambiguity in the definition of the calibration reference plane [3] and the limited control on the accuracy with which the standards are manufactured. For this reason, calibration techniques in which little knowledge of the standards is required, like TRL [4] and LRM [5], might be preferred. However, TRL calibration results to be impractical for probe level calibrations performed on general purpose substrates when a broadband range of frequencies is considered, due to the large number of long transmission lines that would be required. Also, both TRL and LRM calibration define the calibration reference plane at the center of the (non-zero) thru standard, and not at the probe tips. While moving the reference plane to the probe tips is possible by considering single mode propagation in the thru line, this assumption is generally violated in the close proximity of the probe tips, due to higher order modes generated at the non-ideal probe-to-line transition [6]. On the other end, SOL calibration using an unknown thru (RSOL [7]) has been proven to be as accurate as TRL calibration, when a sufficiently accurate model of the standards is provided [8]-[10]. Also in this case, as the authors described in [3] and [10], the calibration reference plane can be univocally defined using EM simulations for the standard modeling. This technique also allows to create probe- independent calibration kits, as opposed to the conventional probe-paired calibration substrates [11], where the imperfections of the probe-to-line transition are embedded in the DUT instead than in the calibration error terms, leading to measurement error [3]. In order to achieve accurate calibration standard modeling, additional care should be dedicated to the manufacturing of the calibration standards. The process presented in [10], based on integrated circuit technology, and the use of fused silica as substrate, had the goal of improving the quality of the standard manufacturing with a special focus on the load, in order to increase repeatability and avoid the common practice of resistance tuning via laser trimming. In this contribution we present a RSOL calibration kit manufactured on fused silica with the process described in [10] and its performances, in terms of measurement accuracy, when employed for VNA calibration in the frequency range from 10 MHz to 50 GHz. The paper is structured as follows, first the process technique and the layout of the calibration kit are described. Then, the modeling of the calibration standards is discussed. Finally, the proposed calibration kit is used for VNA calibration and measurements of both one port and two ports passive devices are employed to compare it towards a commercially available alumina calibration kit. II. FABRICATION TECHNOLOGY For RSOL as well as for LRM calibration, the variation of the load performances, both in terms of DC resistance and electromagnetic behavior, constitute one of the biggest sources of uncertainty [9][10]. In commercially available calibration kits, the load standard is typically defined by means of a purely inductive model [12]. In this, while the DC resistance is very well controlled by means of laser trimming [11], the purely inductive model results to be inaccurate since the large capacitive loading provided by the contacting metal stripes is neglected [10]. At the same time, the laser trimming procedure, while keeping the resistance value highly repeatable, poses a limit in the EM modeling of the load