Additional Algorithms for Sensor Chip Alignment to Blind Datums

Alignment of the sensor focal plane array (FPA) to optical components is a critical design feature. Imaging system designs include reference datums that provide the basis for manufacturing alignment in each sub-assembly. Measurement of z and parallelism positioning can be problematic, since the relevant datum features are often beneath the mounting platform and are obscured to the measurement system. General algorithms for determining sensor chip alignment when datum features are inaccessible to the measurement system have been developed [5]. Precharacterization measurements of datum surfaces are stored for later use during alignment measurement to determine datum locations. The algorithms are useful for post-mounting alignment measurement, and can also be used for active manufacturing alignment. This paper presents additional algorithms that are useful for active alignment, including methods for determining rotation axes on the aligner-bonder system and for determining actuator motions to bring the FPA into alignment with datums. The algorithms have been successfully implemented for ultra-precision, active manufacturing alignment and post-alignment measurement of infrared imaging systems. INTRODUCTION For an imaging system, alignment of the Focal Plane Array (FPA) to optical components is a critical design feature. In particular, optics with shallow depth of field require tight control over parallelism alignment of the FPA image plane to the optical axis. Many applications, such as Telescopes and Satellite Imaging Systems, align optics to the FPA during system integration [e.g., 1,2]. Adjustment of FPA alignment at system integration is accomplished with test equipment and precision sources [3]. This process can overcome minor mis-alignments of the FPA to manufacturing datums that may occur when the FPA is bonded to its mounting platform. Nonetheless, minimizing the FPA mounting error is still important, and even more so for systems that do not have the opportunity to align at system integration, e.g., in volume production. Some systems, such as those containing multiple focal planes, are forced to pay more attention to mounting alignment when the FPA is bonded to its platform since there is less opportunity to overcome minor mis-alignments during system integration. In a manufacturing context, the sensor element (die) of an imaging system is typically mounted onto a platform using fiducial marks on the die and platform as alignment reference points. The mounting operation typically aims to align the two sets of fiducials. However, the true goal of the alignment should be to place the die in such a way as to position the active elements of the focal plane in the specified location of the datum coordinate system. That is, the active elements should be aligned to the datums. When using the die fiducials as alignment targets, some tolerance is lost when considering the position of the fiducials in relation to the active elements. That is, the active element is not always located at its nominal location with respect to the fiducial marks. A similar tolerance loss is incurred when aligning to a datum surface that is obscured to the alignment measurement system. Some assumptions are typically made regarding the location of a blind datum within the alignment system, e.g., that the datum is parallel to the measurement system x-y plane. Novel Optical Systems Design and Optimization IX, edited by José M. Sasian, Mary G. Turner, Proceedings of SPIE Vol. 6289, 62890K, (2006) · 0277-786X/06/$15 · doi: 10.1117/12.683300 Proc. of SPIE Vol. 6289 62890K-1 Accurate means are required for determining alignment of the FPA to its mounting platform based on measurements, for example from optical/encoder systems of an aligner-bonder. Modern methods have been established to determine the location of mechanical reference datums from optical measurements [4]. A common problem in this context is how to determine the datum locations on the mounting platform. Especially for parallelism alignment, reference datums are often obscured from the aligner-bonder optics because they are located on the bottom surface of the platform. A set of algorithms has been described for determining the location of alignment datums that are obscured from measurement during FPA mounting [5]. This method relies on precharacterization measurements of a mounting tool; the measurements are later used to establish the location of datum surfaces when the datums are obscured from the optical measurement system. This approach can be exploited to reduce the tolerance loss typically encountered when aligning fiducials on the die to non-datum features on the mounting platform. Pre-characterization measurements of both the sensor element and mounting platform are made. These measurements, with algorithms in [5], allow the active elements to be aligned directly to the datum coordinate system, even though neither set of features can be measured directly by the alignment system. As in [5], this paper assumes that the datum planes define a Euclidean coordinate system, and it is with respect to this datum coordinate system that the active elements of the focal plane must be aligned. The spatial relationships between the fiducials and the active elements and datum surfaces are calculated as ‘plane transfer values’. These values are scalars that denote the distance from the fiducials to the datum planes and active elements along vectors determined by the fiducial locations (Figure 1). The central idea is that the spatial relationship between the fiducials and datums does not change when the platform or die orientations change. The fixed relationship between fiducials, datums and active elements can be determined by pre-characterization measurements; the fixed relationships can then be utilized to determine datum and active element locations from fiducial measurements during alignment or post-mounting measurement of die placement. Algorithms required for post-mounting measurement are given explicitly in [5]. This paper extends those algorithms for use during active alignment of imaging elements on a sensor die to blind datums on the mounting platform. Pre-characterization measurements should provide means to determine two spatial relationships: (1) the location of the active elements with respect to the die fiducials; and (2) the location of the (blind) datum coordinate system with respect to features (often fiducials placed on a mounting tool) that will be accessible to the aligner-bonder measurement system during alignment. These two requirements are first considered separately, then combined to provide a seamless method for active FPA alignment. DIE PRE-CHARACTERIZATION Hybridization slip and gap measurements can be utilized to provide plane transfer values from the die fiducials to the probable locations of the active elements. Slip and gap data are typically recorded for feedback to the hybridization process, but these values can also be useful to the FPA alignment process. The alignment operation often aims to place the FPA so that Pixel (1, 1) is at a specified (x, y, z) location within the datum coordinate system established by mechanical datums in the mounting platform. For clocking alignment, the (x, y, z) location of Pixel (1, n) is also required. For parallelism alignment, the (x, y, z) location of Pixel (m, n) is also required. A common approach to alignment uses the nominal locations of Pixel (1, 1), Pixel (1, n) and Pixel (m, n) as ‘reference pixels’. The uncertainty of the reference pixel locations in the hybridized FPA is typically Proc. of SPIE Vol. 6289 62890K-2 established by the maximum allowable hybridization slip with maximum allowable hybridization gap wedge. This uncertainty could be reduced for the FPA alignment process by using hybridization slip and gap data to estimate the (x, y, z) locations of all reference pixels after hybridization. It is possible to estimate the (x, y, z) locations of reference pixels after hybridization using hybridization slip and gap data, along with algorithms from [5]. These measurements provide important feedback to the hybridization process, but they can also be used to improve the FPA alignment process. Figure 2 illustrates the transformations that are used to determine the slipped location of reference pixels. For FPA alignment, the hybridization slip is characterized as the movement of the Detector relative to the Read-Out Integrated Circuit (ROIC) where the FPA fiducials are located. The relative motions are applied to the nominal locations of the detector and ROIC targets to give the absolute positions of the slipped detector targets in the ROIC coordinate system. The hybrid slip values are not always self-consistent; any inconsistencies can be averaged by determining the detector target midpoint after hybridization. If the hybridization slip measurements at both sets of targets give the same xand y-values, then only a translation in x and y (∆x, ∆y) is required to describe the offset of the detector with respect to the ROIC. However, if any of the xor y-values differ between the two sets of targets, then a translation (∆x, ∆y) and a rotation (through some angle θ about some rotation center (xc, yc)) is required to describe the offset. For example, suppose that three of the four hybridization slip measurements were zero, and the fourth was 3 microns. There is no single translation (∆x, ∆y) that could account for these measurements. A rotation is required. The ‘probable’ hybridization slip movements can be summarized by a translation (∆x, ∆y) and a rotation through an angle θ about the midpoint of the line segment joining the detector targets. The ‘probable’ resulting position of Pixel (1, 1) is estimated by applying the translation and rotation to the nominal location of Pixel (1, 1) in the ROIC coordinate system. Use of the midpoint for slip calculations provides a ‘best-fit’ approximation to account for the observed slips. Hybrid gap measurements are then used to determine the plane equation of the bottom of the detector. The x-y location of Pixel (1, 1) and the image plane equation (from hybrid gap measurements) can then be used to determine the most probable (x, y, z) location of Pixel (1, 1). Some care must be made when interpreting hybridization slip and gap data. The values are recorded in local coordinate systems, and translation to the ROIC coordinate system is critical. The example in Figure 1 shows measurements of individual Detector Slip at targets near ROIC F2 and ROIC F4. These values can be used to determine individual hybridization slip values in ROIC fiducial coordinates: (x, y) slip of Detector target near ROIC F2 after hybridization: (xROICt arg et F2 + x2 , yROICt arg et F2 + y2 )= (xnom + x2 , ynom + y2 ) (x, y) slip of Detector target near ROIC F4 after hybridization: (xROICt arg et F4 + x4 , yROICt arg et F4 + y4 )= (xnom + x4 , ynom + y4 ) From the midpoint formula, and the locations of the individual slipped detector targets, the probable location of the detector target midpoint after hybridization is given by: ⎛ xROICt arg et F2 + xROICt arg et F4 + x2 + x4 yROICt arg et F2 + y arg F4 + y + y ⎞ ⎜ , ROICt et 2 4 ⎟ ⎜ ⎝ 2 2 ⎠ The difference between the nominal midpoint and the slipped midpoint indicates the overall (average) slip of the detector, i.e., the ‘average’ xand y-displacements of the two detector targets. Proc. of SPIE Vol. 6289 62890K-3 This can be calculated by using the distance between the ROIC midpoint and the Detector midpoint after hybridization: x + x + x + x x + x x + x ∆x = ROICt arg et F2 ROICt arg et F4 2 4 − ROICt arg et F2 ROICt arg et F4 = 2 4 2 2 2 y ∆y = ROICt arg et F2 + yROICt arg et F4 + y2 + y4 yROICt arg et F2 + yROICt arg et F4 y2 + y − = 4