D2.2 Automatic adjustable reference path system Document Coordinator: Contributors: Dissemination: Keywords: Ger Folkersma (Demcon) Ger Folkersma, Kevin Voss, Marvin Klein (Demcon) Public Reference path, Optical Path Length, Reflectivity, Dispersion Date: 7-1-2016 Revision: 1.0
VERSION HISTORY VERSION DATE NOTES AND COMMENTS 0.1 13-11-2015 VERSION FOR REVIEW 0.2 24-11-2015 UPDATED PROJECT CONTEXT 0.3 7-12-2015 INCLUDED REVIEW COMMENTS FROM DATAPIXEL 0.4 7-1-2016 UPDATED DOCUMENT WITH TEST RESULTS AND CONCLUSION 1.0 8-1-2016 UPDATED AFTER INTERNAL REVIEW
Table of contents PROJECT CONTEXT 4 FUNCTIONS 5 SPECIFICATIONS 5 DYNAMIC OPL 5 REFLECTIVITY MATCHING 5 DISPERSION MATCHING 5 CONCEPTS AND DETAILED DESIGN 5 DYNAMIC OPL 6 REFLECTIVITY MATCHING 7 DISPERSION MATCHING 8 TEST SETUP 9 TESTS AND RESULTS 10 ABERRATIONS IN THE OPTICS 10 DYNAMIC OPL 10 REFLECTIVITY MATCHING 11 Controller accuracy 11 Controller speed 13 CONCLUSION 14
Project context This document details the concept choice, test setup and test results of task 2.4, Design and implementation of an automatic adjustable reference path system. The reference path is a part of the complete sensor system, see the encircled area in Figure 1. Note that the beam-splitter has been added to this for technical reasons, explained later in this document. The optical system of an ultra-short pulsed laser processing machine is optimized for the wavelength of the applied laser source, specially the internal dispersion properties of the f-theta scanning lens cause a wavelength and field angle dependent optical path change on the measurement beam. Additionally, adaptive optics are used in the beam coupling and sensor integration (task 2.5), which in general will also impact the required reference path. Furthermore the wide range over which the reflectivity of the measured area on the work-piece (or sample) can vary represents a challenge to the optical measurement system. In order to achieve a sufficient signal-to-noise ratio from the spectrometer, the light power feedback from the reference path has to be in balance with the light power returned from the work-piece. The measurement of polished or rough or oxidized surfaces therefore requires an adjustment of the light feedback levels from the reference path over a corresponding range. The above aspects require that the measurement system can be adjusted to compensate for changes to the optical characteristics of the measurement beam path including the work-piece itself. In the present document, the compensation of changes of the optical path length and of the local work-piece reflectivity will be addressed by the design and implementation of an adaptive reference path system. Figure 1: The reference path module as a part of the complete system is encircled in blue.
Functions The reference path module has three main functions 1. Dynamic adjustment of the optical path length (OPL) to match the OPL of the sample arm. 2. Dynamic adjustment of the reflectivity of the reference arm, to match that of the sample arm. 3. Adjustment and matching of the dispersion, introduced by optical elements in the sample path. Specifications For each task a set of specifications has been defined. The most important specifications are listed here per function. Dynamic OPL The main contributor to a difference in OPL of the sample arm is due to the beam-shaping module. However, it is expected that the beam-shaping module is not active during the measurement of one area on the work piece. The OPL is only adapted when the work-piece is out of the measurement range, either due to repositioning of the scan head with respect to the sample, or due to OPL changes in the sample arm introduced by corrections of the beam-shaping module. Therefore, the bandwidth of the OPL modulation can be much lower than the measurement speed of 1 khz. OPL adjustment range: 25mm OPL adjustment accuracy: ±50 µm OPL stability when measuring: ±0.1 µm Feedback power matching The reflectivity of the sample has to be matched to get the maximum interference modulation depth. Initial tests on samples of typical machined objects by Lightmotif have shown that the reflectivity can vary between mirror-like reflections, to matte black spots, depending on the machining parameters. Range: Accuracy: Speed: 0% to 100% with respect to the maximum feedback intensity from sample path. factor 0.5 to 2.0 (with respect to the actual feedback intensity from the sample path) 1 khz for entire dynamic range. Dispersion matching The dispersion in the sample path, which is mainly due to the f-theta lens, has to be matched closely by introducing a corresponding dispersion in the reference path. The amount of dispersion to be matched depends mostly on the given material properties of the f-theta lens, but also to a small degree on the (static) alignment of the sample path. Any changes in the scan lens thickness (and thus dispersion modulation) over the scan range are neglected for now. This means the dispersion correction should only be needed when the dispersion in the measurement beam changes, e.g. exchanging or adding a lens, which will be done only very infrequently. For this only application scenario, the dispersion compensation need not be actuated automatically (neither fast nor slow), but needs only be manually adjustable, possibly including swapping of optical elements. Note that temperature changes of the optical elements in either beam path may introduce as a secondary effect a (relatively slow) net change of the dispersion. Whether any such change might become significant (thus requiring compensation) should be investigated. Note also that small amounts of uncompensated dispersion may also be corrected in the processing of the recorded interference spectra. Concepts and detailed design
For each function of the reference path module, numerous concepts have been formulated. The chosen concepts are detailed below. Dynamic OPL Concept studies have shown that a concept where a mirror is placed on a linear stage to adjust the OPL, the coupling efficiency of the beam back in the collimator is very sensitive to the pitch and yaw errors of such a stage over its travel range (see Figure 2). Calculations have shown that a maximum of 30 µrad pitch & yaw is allowable to prevent too much signal loss. Figure 2: Pitch and yaw errors of the stage (gray) influence the position of the reflected beam trough the collimator, on the fiber end face. This results in reflected power fluctuations during the motion of the stage. All (affordable) commercial stages have a pitch error that exceeds by far the requirement of ±30 µrad. The large pitch error will result in an excessive movement of the reflected beam on the fiber facet and thus a changing of the coupling loss. As a consequence, the settings of the reflectivity modulator that result in the desired attenuation effect may have to be re-calibrated every time the OPL is adapted. To solve this, the lens and mirror at the end of the reference path are replaced by a hollow retro-reflecting corner cube, see Figure 3. With such a retroreflector, the beam is always parallel with respect to the incoming beam in both planes, regardless of the orientation. Therefore the yaw / pitch errors of the stage are not relevant anymore. Furthermore, this retroreflector can replace the reflecting element (and lens) at the end of the reference path, which also avoids to the chromatic aberrations caused by this lens and mirror combination. The wave front distortion of the outgoing beam for such a reflector is 0.1λ, which is similar to the mirrors used before. Figure 3, hollow retroreflecting cornercube The reflector is mounted on a motorized linear stage with 25mm travel range and a maximum speed of 2.5mm/s, and the reference path is folded by a static mirror for compactness, see Figure 4. The fiber collimator is also reflective, which makes the reference path completely reflective, avoiding chromatic aberrations. Currently the setting of the OPL is a manual task, however the motorized stage allows for an automated range centering if this would be desirable later.
Figure 4: CAD image of the dynamic OPL module Feedback power matching The feedback power matching consists of two parts: The measurement of the feedback light power levels of both the sample arm and the reference arm. A modulating element in the reference arm that allows for a fast modulation of the laser power that is reflected back into the fiber. The difference of the two measured feedback levels will be the feedback signal for the control of the attenuation. Since the feedback light power of the sample arm needs to be measured at the sample arm port of the beamsplitter, for practical reasons the demonstration of the reference path module also includes the beamsplitter. The optical coupling to the rest of the sensor system will be realized by optical fibers, see Figure 1. The modulation of the open beam in the reference path will be realized by a dynamic aperture in the beam. Due to the required speed of 1 khz, a fast actuator is required. This is why a galvo scanner motor is used to drive a shutter rotor, see Figure 5. Such an actuator has a limited range (here 40 ) and requires a low rotor inertia to meet the speed requirements. Therefore, a lightweight rotor has been designed and manufactured by 3D plastic printing. As a bonus, 3D printing allows for a quick optimization of the aperture shape. Figure 5: Left: shutter rotor in the open beam. Right: Aperture shape.
Figure 6 shows a schematic of the complete test setup. The reflectivity of both the sample and reference arm is measured by four photo detectors denoted D 1 to D 4 which are connected to the PLC system. The ratio between detectors D 1 and D 2 is the effective reflection of the reference arm, which has to be matched to the ratio between D 4 and D 3, the effective reflection of the sample arm. The light to those detectors is tapped with fiber-optic splitters with a small tap ratio of 90:10, to prevent excessive light losses. To get the correct reference path reflection (R r) and sample path reflection (R s), we have to correct for the splitter losses: R r = P out,6 P in,6 = D 1 0.9 D 2 R s = P out,3 P in,3 = D 4 0.9 D 3 The control signal for the shutter is then computed from the mismatch between R r and R s. This controller includes a feed-forward term that is calibrated beforehand. This feed-forward term compensates for any imbalances due to e.g. the beam splitter and connector losses as well as imperfections of the photodiode gains. Feed-back control using a proportional and an integral term is added to correct for the sample reflectivity and also for disturbances in the system, such as the influence of the moving retro-reflector. Figure 6: Schematic overview of the reference path test setup Dispersion matching As explained in the requirements, a static, but interchangeable dispersive element suffices. Therefore, a glass rod is placed into the beam, which can be interchanged with different lengths. This is not detailed further as it involves off-the-shelf components.
Test setup To test these concepts, a stand-alone test setup has been designed, see Figure 7 and Figure 8. The setup consist of the OPL and reflectivity modulators, the fiber splitters and photodetectors, a 532 nm diode laser source, and a dummy sample path. Most components are off-the-shelf, which reduces the cost of the test system. A spectrometer is not included, as it is not required to do inteferometric measurements with this setup. Figure 7: Photo of the reference path test setup. Figure 8, Photo of the reference path The surrogate sample path consists of a fiber collimator and a mirror, with a motorized optical beam chopper in between. This allows for testing of the speed with which the sample and reference arm feedback signals can be matched.
All signals are processed by a Beckhoff PLC system, that allows for real-time control, and easy integration in a stand-alone PLC system later for integration at Lightmotif. Tests and results For each function a series of tests are required to verify its performance and check for unwanted sideeffects. Aberrations in the optics One of the goals of this task, was to identify the optical aberrations in the reference path, which can be introduced by imperfect optics. The first tests showed a very low reflectivity in the reference arm (R r) of 5%, even with the shutter rotor removed. It was identified that the coupling efficiency of the used reflective collimators (Thorlabs RC08APC-P01) is very low when the collimated beam is reflected back into the collimator. An M 2 measurement of the beam revealed that the collimator introduces a significant amount of astigmatism. The reflective collimator was therefore replaced by a microscope objective, which increased the maximum reflectivity to 32%. The additional dispersion introduced by this objective is compensated when the same objective is used in the sample path. Dynamic OPL First, the dynamic OPL was tested for stability of the total reference path reflection. That means, the stage is moved over its full range, and reference path reflection R r is measured. This reflection should not change significantly. However some deviation (<5%) is allowed, since the reflectivity will be controlled in real time by the shutter. Figure 9 shows this ratio for 25mm movement at a constant speed of 1mm/s of the OPL stage and retroreflector (which corresponds to 50mm OPL change). The maximum difference was measured to be about 3%. This (slow) variation van be corrected for by the shutter, without sacrificing too much of its dynamic range. Note that the noise in this figure is mainly due to the instability of the testing laser, in combination with noise in the electronics. 33 32.8 32.6 32.4 Reflection (%) 32.2 32 31.8 31.6 31.4 31.2 31 0 5 10 15 20 25 OPL stage position (mm) Figure 9, Reflection stability for the full stroke of the OPL stage. The maximum deviation is about 3%.
The next test will be the stability of the OPL over time. This is to be tested with the interference signal of the complete system. To prevent influence of other optical components as much as possible, the simple surrogate sample path of this test setup can be used to verify this measurement. At this moment, such a measurement is not yet possible, since this reference path is not integrated in an interferometric setup. However, it is expected to be very stable, when the DC motor controller of the OPL stage is disabled as soon as the desired position is reached. Reflectivity matching Unfortunately, the scanner motor for the shutter for which the setup was designed, was not available in time for these tests. However, a similar model with a 4 times smaller rotational stroke and 4 times lower speed was available, but could only be controlled via USB (instead of directly from the PLC), which results in a slow reaction time for the controller of 60 ms. In the results below, this delay was removed in post processing to evaluate the controller. As soon as the scanner motor becomes available, these tests will be revisited to include the full stroke and speed. Open-loop response Figure 10 shows the result of measuring R r while slowly rotating the shutter. Note that this scanner has a rotation range of 10, where 40 is required to achieve a fully open shutter, resulting in the maximum of 32% reflection. As designed, the response follows a second order curve so that at small angles Φ, corresponding to low feedback signals, dr r/dφ is smaller than at large angles. A very small feedback level from the sample are can be matched more accurately with this response than would be possible with a linear response. A second order fit of this curve is used as the feed-forward term of the shutter controller. 7 6 Measured 2nd order fit 5 R r (%) 4 3 2 1 0 0 2 4 6 8 10 12 Scanner angle (deg) Figure 10, Measured reflection (Rr) versus the scanner angle. The curve is fitted by a second order polynomial fit. Controller accuracy With the photodiodes calibrated, R r must be equal to R s when the shutter rotor is controlling the reflection level. First, this is tested for accuracy, by modulating the signal from the surrogate sample
path by means of the beam chopper, while measuring all signals in real-time. Figure 11 shows an excerpt of the sample and reference path reflectivity while actively controlling the shutter rotor position. For such a case of relatively slow modulation, the error stays well within the specification of a factor 0.5 to 2.0. 5.2 5.1 R r R s 5 4.9 Reflectivity (%) 4.8 4.7 4.6 4.5 4.4 4.3 4.2 5.5 6 6.5 7 7.5 8 Time (s) Figure 11, Excerpt of the reflectivity of the sample and reference path, with active controlled shutter. The shutter follows the noise that is present in the (stationary) dummy sample path. Figure 12 shows a similar measurement where the sample reflectivity is modulated by manually disturbing its path. The reflectivity error factor stays well within the specifications.
6 Reflectivity (%) 4 2 R r R s 0 0 5 10 15 20 Time (s) Reflectivity factor R r / R s (-) 2 1.5 1 0.5 0 5 10 15 20 Time (s) Figure 12, Measurement of a slowly changing sample reflectivity, while actively controlling the shutter. The reflectivity error factor stays well within the bounds of 0.5 to 2.0, shown by the red lines. Controller speed Due to the delay in the control of the scanner, a characterization of the control performance at high frequencies is not useful. However, the maximum attenuation speed can be identified by oscillating the scanner at full speed while measuring the reflection. Figure 13 shows 6 of these oscillations, in 48ms, which means that it takes 4ms for the shutter to go from 0% to fully open (in this case limited to 5.5% reference signal feedback). Note that the scanner angle follows a saw-tooth profile, but due to the nonlinear response of the reflection (see Figure 10), this figure shows a somewhat asymmetric shape. The performance of the new scanner is estimated to be similar, since the maximum rotational speed is 4 times larger, but the required rotation angle is also 4 times larger. This means that the specification of 1kHz for the entire dynamic range will most likely not be met. However, such extreme variations in sample reflection, where the system has to wait for the reflectivity to be matched, are unlikely to occur frequently. This means that the additional processing time of measuring the depth profile of an entire sample due to the limitation of the reference feedback signal adaptation is very small.
6 5 4 Reflectivity (%) 3 2 1 Figure 13, Reference path reflectivity for an oscillation of the scanner at full speed. Conclusion 0 0 10 20 30 40 50 Time (ms) For each function of the reference path module a concept has been chosen and detailed. First, it is expected that a semi-static dispersion correction is sufficient, which means a dispersive element is replaced by hand if for example lenses are added in the sample path. Second, a motorized variation of the optical path length has been realized by a linear stage in combination with a retro-reflector. It has been shown that this results in a stable signal, with minimal cross-talk to the reference path reflection. Finally, an actively controlled matching of the sample reflectivity was realized for the reference path, by moving a shutter rotor in the collimated beam. The feedback signals of both the sample path and the reference path are measured by photodiodes. The difference between photodiode signals is the error signal for the control loop to the shutter. With the currently available scanner motor, this error stays within the specifications for slow disturbances. It is expected that with the correct scanner motor, the full-scale reaction time will be about 4 ms.