1. Simulation and Experimental Verification of Model Based Opto-Electronic Automation Drexel University Department of Electrical and Computer Engineering skb25@drexel.edu, kurzweg@ece.drexel.edu, guezal@drexel.edu Shubham K. Bhat, Timothy P. Kurzweg, and Allon Guez

2. Motivation Current State-of-the-Art Photonic Automation Our Technique: Model Based Control Optical Modeling Techniques Learning Model Identification Technique Conclusion and Future Work Overview

3. No standard for OE packaging and assembly automation. Misalignment between optical and geometric axes Packaging is critical to success or failure of optical microsystems 60-80 % cost is in packaging Automation is the key to high volume, low cost, and high consistency manufacturing ensuring performance, reliability, and quality. Motivation

4. Current State-of-the-Art LIMITATIONS: Multi-modal Functions Multi-Axes convergence Slow, expensive “Hill-Climbing” Technique Visual Inspect and Manual Alignment Initialization Loop Move to set point (X o ) Measure Power (P o ) Stop motion Fix Alignment Approximate Set Point=X o Assembly Alignment Task Parameters Off the shelf Motion Control (PID) (Servo Loop) Stop

5. Model Based Control ADVANTAGES: Support for Multi-modal Functions Technique is fast Cost-efficient Visual Inspect and Manual Alignment Initialization Loop Move to set point (X o ) Measure Power (P o ) Stop motion Fix Alignment Set Point=X o Learning Algorithm Model Parameter Adjustment Optical Power Propagation Model Correction to Model Parameter {X k }, {P k } FEED - FORWARD Off the shelf Motion Control (PID) (Servo Loop) Assembly Alignment Task Parameters

6. Model Based Control Theory KpKp KpKp P d (s) P r (s) + + + - R(s) E(s) If = P,

7. Optical Modeling Technique Use the Rayleigh-Sommerfeld Formulation to find a Power Distribution model at attachment point Solve using Angular Spectrum Technique – Accurate for optical Microsystems – Efficient for on-line computation Spatial DomainFourier Domain Spatial Domain

8. Inverse Model For Model Based control, we require an accurate inverse model of the power However, most transfer functions are not invertible Zeros at the right half plane Unstable systems Excess of poles over zeros of P Power distribution is non- monotonic (no 1-1 mapping) Find “equivalent” set of monotonic functions

9. Inverse Model: Our Approach Decompose complex waveform into Piece- Wise Linear (PWL) Segments Each segment valid in specified region Find an inverse model for each segment

10. The structure of the system and all of its parameter values are often not available. Noise, an external disturbance, or inaccurate modeling could lead to deviation from the actual values. Adjust the accuracy on the basis of experience. Need for Learning Model Real System Adjustment Scheme Model + - Input Output Error Estimated model

11. Learning Model Identification Algorithm It follows thatand Step 1: Assume system to be described as, where y is the output, u is the input and is the vector of all unknown parameters. Step 2: A mathematical model with the same form, with different parameter values, is used as a learning model such that Step 3: The output error vector, e, is defined as Step 4: Manipulate such that the output is equal to zero. Step 5:

12. Real System Model Sensitivity equations + - Learning Model Identification Technique Output Input Output Error Estimated model of unknown parameters

13. We present a two unknown system having input-output differential equation ( a and K are unknown ) The variables u, y, and are to be measured Step 1: and { } Step 2: { Assume estimated model and } The Sensitivity coefficients are contained in Step 3: where, and Learning Model Identification example

14. Learning Loop of PWL Segment Power Displacement + - K Input. (x) (P) (u) For each PWL Segment: L -1

15. Learning for 2 Unknown Variables (PWL Segment) a and K have initial estimates of 0.1 and 4 Actual values of a and K are 1.44 and 5.23 S: Sensitivity matrix : Updated Model e: error matrix Q: weighting matrix e: tracking parameter

16. Distance = 10um No. of. Peaks = 10 Edge Emitting Laser Coupled To a Fiber Aperture = 20um x 20um Fiber Core = 4 um Prop. Distance = 10 um Example: Laser Diode Coupling NEAR FIELD COUPLING

17. Nominal Model - K K + Proportional Gain Motor DynamicsPlant Model Derivative Desired Power Time Taken = 7 seconds Model Based Control System (1.41) + Inverse Model + + 20 18 16 14 12 Fiber Position (12.6 um) 1.5 1 0.5 0 1.5 1.3 1.1 0.9 0.7 Received Power (1.41 )

18. Experimental Setup of Laser- diode example

19. Test bed for Verification Optical Power Sensor Optical Source X-Y Stage Motion Control Card X Amplifier Y Amplifier Laser Diode Driver Pre-amplifiers, encoders We acknowledge Kulicke and Soffa, Inc. for the donation of the XY Table

20. Test bed for Verification Optical Power Sensor X-Y Stage Laser Diode Driver

21. Model based control leads to better system performance Inverse model determined with PWL segments Learning loop can increase accuracy of model Shown increased performance in simulated systems Hardware implementation Evaluate other learning techniques Error prediction in models Conclusions and Future Work