Quantitative Understanding of Induced Voltage Statistics for Electronics in Complex Enclosures COST MC Meeting, Toulouse, France

Transcription

1 Quantitative Understanding of Induced Voltage Statistics for Electronics in Complex Enclosures COST MC Meeting, Toulouse, France ACCREDIT ACTION IC 1407 IC1407 Advanced Characterisation and Classification of Radiated Emissions in Densely Integrated Technologies (ACCREDIT) MANAGEMENT COMMITTEE MEETING Toulouse, France February 2017 Research funded by ONR ONR/DURIP AFOSR / AFRL Center of Excellence 13 February, 2017 Bo Xiao 1 Zach Drikas 2 Jesus Gil Gil 2 T. D. Andreadis 2 Tom Antonsen 1 Ed Ott 1 Steven M. Anlage 1 1 University of Maryland, USA 2 US Naval Research Laboratory Random Coupling Model website: 1

2 Electromagnetic Interference and Electronics How to defend electronics from electromagnetic interference? R.F.I. International Electromagnetic Interference (EMI) Electromagnetic Compatibility (EMC) 2

3 Electromagnetic Coupling to Enclosures and Circuits A Complicated Problem Arbitrary Enclosure (electrically large, with losses 1/Q ) Coupling of external radiation to computer chips is a complex process: Apertures Resonant cavities Transmission Lines Circuit Elements Cross-Talk System Size >> Wavelength What can we say about the nature of fields and induced voltages inside such a cavity? Chaotic Ray Trajectories Statistical Description using Wave Chaos!! Rather than make predictions for a specific configuration, determine a probability distribution function (PDF) of the relevant quantities 3

4 Outline The Issue: Electromagnetic Interference Our Approach A Wave Chaos Statistical Description The Random Coupling Model (RCM) Example of the RCM in Practice Scaled measurement system for investigating new RCM predictions Extension of the RCM to Stochastic Sources Conclusions 4

5 Classical Chaos in Newtonian Billiards Best characterized as extreme sensitivity to initial conditions Regular system Chaotic system Newtonian particle trajectories x i, p i x i +Dx i, p i +Dp i x i, p i x i +Dx i, p i +Dp i 5 2-Dimensional billiard tables x 2 (0) D(0) x 1 (0) x 2 ( t) D(t) x 1 ( t) D( t) D(0) e 0 Lyapunov exponent t

6 Wave Chaos? 1) Waves do not have trajectories It makes no sense to talk about diverging trajectories for waves 2) Linear wave systems can t be chaotic Maxwell s equations, Schrödinger s equation are linear 3) However in the semiclassical limit, you can think about rays In the ray-limit it is possible to define chaos ray chaos Wave Chaos concerns solutions of linear wave equations which, in the semiclassical limit, can be described by chaotic ray trajectories 6

7 From Classical to Wave Chaos wave Quantum Semiclassical limit (quantum chaos) ray trajectory Classical (chaos) 7

8 How Common is Wave Chaos? Consider an infinite square-well potential (i.e. a billiard) that shows chaos in the classical limit: Sinai billiard Bunimovich Billiard Hard Walls Bow-tie Bunimovich stadium Solve the wave equation in the same potential well Examine the solutions in the semiclassical regime: 0 < << L YES L Some example physical systems: Nuclei, 2D electron gas billiards, acoustic waves in irregular blocks or rooms, electromagnetic waves in enclosures Will the chaos present in the classical limit have an affect on the wave properties? But how? 8

9 Random Matrix Theory (RMT) Wigner; Dyson; Mehta; Bohigas The RMT Approach: Complicated Hamiltonian: e.g. Nucleus: Solve H E Replace with a Hamiltonian with matrix elements chosen randomly from a Gaussian distribution Examine the statistical properties of the resulting Hamiltonians Universality Classes of RMT: Orthogonal (real matrix elements, b = 1) Unitary (complex matrix elements, b = 2) Symplectic (quaternion matrix elements, b = 4) H Hypothesis: Complicated Quantum/Wave systems that have chaotic classical/ray counterparts possess universal statistical properties described by Random Matrix Theory (RMT) BGS Conjecture Cassati, 1980 Bohigas, 1984 This hypothesis has been tested in many systems: Nuclei, atoms, molecules, quantum dots, acoustics (room, solid body, seismic), optical resonators, random lasers, 9 Some Questions to Investigate: Is this hypothesis supported by data in other systems? What new applications are enabled by wave chaos? Can losses / decoherence be included? What causes deviations from RMT predictions?

10 Resistance (kw) Incoming Channel Chaos and Scattering Hypothesis: Random Matrix Theory quantitatively describes the statistical Nuclear scattering: Ericson fluctuations properties of all wave chaotic systems (closed and open) Outgoing Channel d d Compound nuclear reaction Proton energy Transport in 2D quantum dots: Universal Conductance Fluctuations Billiard mm 10 V V V Outgoing Voltage waves 1 2 N S matrix V V [ S ] V 1 2 N Incoming Channel Incoming Voltage waves Electromagnetic Cavities: Complicated S 11, S 22, S 21 versus frequency Outgoing Channel 1 2 S xx B (T) S 11 S 22 S 21 Frequency (GHz)

11 Universal Scattering Statistics Despite the very different physical circumstances, these measured scattering fluctuations have a common underlying origin! Universal Properties of the Scattering Matrix: φ n Im[S] S i S e Re[S] Unitary Case RMT prediction: Eigenphases of S uniformly distributed on the unit circle Eigenphase repulsion P,,... ) ~ ( 1 2 N nm e i n e i m b 1 b 2 4 GOE GUE GSE 11 Nuclear Scattering Cross Section d d 2D Electron Gas Quantum Dot Resistance R(B) Microwave Cavity Scattering Matrix, Impedance, Admittance, etc.

12 Outline The Issue: Electromagnetic Interference Our Approach A Wave Chaos Statistical Description The Random Coupling Model (RCM) Example of the RCM in Practice Scaled measurement system for investigating new RCM predictions Extension of the RCM to Stochastic Sources Conclusions 12

13 Statistical Model of Impedance (Z) Matrix S. Hemmady, et al., Phys. Rev. Lett. 94, (2005) L. K. Warne, et al., IEEE Trans. on Anten. and Prop. 51, 978 (2003) X. Zheng, et al., Electromagnetics 26, 3 (2006); Electromagnetics 26, 37 (2006) Port 1 R R1 () Port 1 Losses Other ports Port 2 R R2 () Enclosure Free-space radiation Resistance R R () Z R () = R R () + jx R () Z ij () j System parameters Statistical parameters Statistical Model Impedance R Ri modes n n 1/2 ( n ) Radiation Resistance R Ri () D 2 n - mean spectral spacing Q -quality factor n D 2 1/2 ( n ) n w in in w injn 2 (1 jq 1 2 ) n R Rj Statistics governed by a single parameter: - random spectrum (RMT) w in - Guassian Random variables = ω2 ω 2 Q 13

14 Probability Density Testing Insensitivity to System Port 1Details Cross Section View of Port CAVITY LID Z Cavity 2a=0.635mm 2a=1.27mm Radius (a) CAVITY BASE ix Rad R Rad RAW Impedance PDF Coaxial Cable z Freq. Range : 9 to 9.75 GHz Cavity Height : h= 7.87mm Statistics drawn from 100,125 pts. Metallic Perturbations z R R Cavity Rad i X Cavity Rad Rad 2a=0.635mm 2a=1.27mm R X NORMALIZED Impedance PDF Im( )( W) Z Cavity Im(z)

15 Universal Fluctuations are Usually Obscured by Non-Universal System-Specific Details Wave-Chaotic systems are sensitive to details The Most Common Non-Universal Effects: 1) Non-Ideal Coupling between external scattering states and internal modes (i.e. Antenna properties) 2) Short-Orbits between the antenna and fixed walls of the billiards Z-mismatch at interface of port and cavity. Ray-Chaotic Cavity Port Incoming wave Prompt Reflection due to Z-Mismatch between antenna and cavity Transmitted wave 15

16 The Random Coupling Model Divide and Conquer! Coupling Problem Solution: Radiation Impedance Matrix Z rad + Short Orbits Enclosure Problem Solution: Random Matrix Theory; Electromagnetic statistical properties are governed by Loss Parameter a k 2 /(Dk n2 Q) df 3dB /Df spacing 16 V V V 1 2 Z matrix I I [ ] N I N Z IEEE Trans. EMC 54, 758 (2012) 1 2 Mean part Port i Z Fluctuating Part (depends on a) Port 1 Port 4 ~ Z ix Port 5 ix R Rad Rad Port 3 <Imx> = 1 <Rex > = 0 Port j Port 2 Electromagnetics 26, 3 (2006) Electromagnetics 26, 37 (2006) Phys. Rev. Lett. 94, (2005)

17 Probability Density Statistical Properties of Scattering Systems Universal Z (reaction) and S statistics Inclusion of loss: P a (Z), P a (S) a = 3-dB bandwidth / mean-spacing Phys. Rev. Lett. 94, (2005) Phys. Rev. E 74, (2006) P(Re[ ˆ z ]) P(Im[ ˆ z ]) a 16 a 6.3 a Re[ ˆ z ] 2 1 a 16 a 6.3 a Im[ ˆ z ] 8 Removing Non-Universal Effects: Sensitivity to Details Coupling, Short Orbits RAW Impedance PDF 2a=0.635mm 2a=1.27mm NORMALIZED Impedance PDF 2a=1.27mm 2a=0.635mm Phys. Rev. E 80, (2009) Phys. Rev. E 81, (R) (2010) Im( )( W) Z Cav Im(z) 17

18 Outline The Issue: Electromagnetic Interference Our Approach A Wave Chaos Statistical Description The Random Coupling Model (RCM) Example of the RCM in Practice Scaled measurement system for investigating new RCM predictions Extension of the RCM to Stochastic Sources Conclusions 18

19 Test of the Random Coupling Model: Gigabox Experiment (NRL Collaboration) 1. Inject microwaves at port 1 and measure induced voltage at port 2 2. Rotate mode-stirrer and repeat 3. Plot the PDF of the induced voltage and compare with RCM prediction But this is at the limit of what we can test in our labs Z. B. Drikas, et al., IEEE EMC 56, 1480 (2014) J. Gil Gil, et al., IEEE EMC 58, 1535 (2016) US Naval Research Laboratory collaboration 19

20 Extensions to the Original Random Coupling Model Extension Short Orbits Multiple Coupled Enclosures Nonlinear Systems Coupling Through Apertures Mixed (regular + chaotic) systems Investigated Experimentally? Yes (Z avg, Fading, Time-reversal) In progress Yes (First results on billiard with nonlinear active circuit) Yes (Analyzing data) In progress Lossy Ports Yes (Radiation efficiency correction h) Integration with FEM codes and making connections to: Power Balance / SEA / DEA / Correlation Fcns. (U. Nottingham) Reverberation Chamber studies (NIST/Boulder) Electromagnetic Topology / BLT (U. New Mexico, ONERA) 20

21 Outline The Issue: Electromagnetic Interference Our Approach A Wave Chaos Statistical Description The Random Coupling Model (RCM) Example of the RCM in Practice Scaled measurement system for investigating new RCM predictions Extension of the RCM to Stochastic Sources Conclusions 21

22 Coupled Lossy Cavities RCM predicts the transmission through N cavities coupled through apertures Experiment requires a chain of Gigaboxes Pipes, cables, ducts Ability to predict coupled voltage and power along an arbitrary chain of ray-chaotic cavities/environments/compartments source G. Gradoni, T. M. Antonsen, and E. Ott, Phys. Rev. E 86, (2012). Enclosure 1 Loss Parameter a 1 Coupling Trans-Impedance Enclosure N Loss Parameter a N 22

23 Scaling Properties of Maxwell s Equations for Harmonic EM waves Maxwell s equations are left invariant upon scaling: r = r s ω = sω σ = sσ For example, scaling the GigaBox down by a factor of s = 20 requires: 1 m x 1 m x 1 m Box 5 cm x 5 cm x 5 cm Box 5 GHz measurement 100 GHz measurement Wall conductivity Wall conductivity s* 23

24 Scaled Coupled Cavities Simulation Circular aperture Dielectric sphere 24

25 Frequency Extender Scaled Structure mm-wave Horn antenna WR GHz WR GHz Frequency Extender Inside cryostat Teflon lens mm-wave windows ONR/DURIP funded experimental setup Coax ~10 GHz Reference plane Coax ~10 GHz 25

26 Experimental Setup Receiver Scaled Structure mm-wave Transmitter 26

27 Scaled Enclosure Direct Injection Measurement Scaled Enclosure (s = 40) mm-wave extender Scaled Enclosure (s = 20) with rotating perturber mm-wave extender 3 cm 27

28 Cryogenic Mode Stirrer Scaled enclosure wall Stirrer Use cryogenic stepper motor to rotate a magnetic strip below scaled cavity, causing the stirrer panel inside cavity to rotate. No holes in the cavity walls (no leakage) Bar Magnet Cryogenic stepper motor 28

29 Multiple Realizations of s = 20 Scaled Enclosure at Cryogenic Temperatures 29

30 Cryogenic Results for the s = 20 Scaled Enclosure Motion of the Perturber (Cu Sheet) Inside the s = 20 Scaled Enclosure a = 19 Rotating Perturber a = 19 Statistics with just 9 Realizations GHz 30

31 Problem: The ports are lossy The original RCM assumed loss-less ports Sample Equivalent Cryostat Lossy Free-space propagation A 1 2 Cavity C Lossy Free-space propagation B Measurement Ports A & B will influence RCM obtained α since they are lossy 31

32 Solution: Include the Radiation Efficiency of the Ports Scalar Form Lossy antenna Z cav = jim[z rad ] + Re[Z rad ]ξ Z cav = Z ant + ηre[z ant ](ξ 1) Matrix Form Z cav = jim[z rad ] + Re Z rad 1/2 ξ Re Z rad 1/2 Radiation efficiency Lossless port: η = I Lossy path Z cav = Z ant + R 1/2 ξ I R 1/2 R = η 1/2 Re Z ant η 1/2 B. D. Addissie, J. C. Rodgers and T. M. Antonsen, "Application of the random coupling model to lossy ports in complex enclosures," Metrology for Aerospace (MetroAeroSpace), 2015 IEEE, Benevento, 2015, pp

33 30.5 cm Scaled Enclosure Measurement Validation: Compare to NRL Full-Scale Experimental Setup cm a b 25 cm a = 27.9 cm b = 27.5 cm Stirrer (Motor Controlled) 2.1 cm 66 cm 25 cm Port 2 25 cm c c 2.5 cm cm We measure two scaled versions: c c = 25 cm 75 GHz 110 GHz (s = 20) 3.7 GHz 5.5 GHz WR-187 (G-Band) 220 GHz 330 GHz (s = 40) 5.5 GHz 8.25 GHz WR-137 (C-Band) 44.5 cm 90 Zach Drikas, Jesus Gil Gil, T. D. NRL 33

34 Single Cavity Experiment s = 20 scaled cavity, cryogenic Full-scale cavity, room temperature P Im z 12 Simulation with α = 2.6 P Im z 12 Simulation with α = 1.9 Im z 12 Im z 12 Our ongoing work is to increase full-scale cavity s α to make it within the tunable range of

35 Loss Parameter a Variation of Loss Parameter a With Temperature s = 20 scaled cavity α = k2 2 QΔk = k3 V n 2π 2 Q α=5.6 α=3.2

36 Overview of the Scaled Cavity Measurement Project Test the Random Coupling Model in a set of increasingly complicated (and realistic) scenarios: o Multiple coupled cavities o Mixed (regular + chaotic) systems o Irradiation through irregular apertures o Evanescent coupling between enclosures o etc. Validation step: compare to full-scale measurements at NRL 36

37 Ports Multiple Cavity Experiment Under Development 3D Printed Scaled Structures: Multiple Connected Enclosures Backplane with Apertures Scaled Enclosures 37

38 Overall picture

39 3-cavity cascade Perpendicular port Lens for perpendicular port will be added later Parallel port

40 Outline The Issue: Electromagnetic Interference Our Approach A Wave Chaos Statistical Description The Random Coupling Model (RCM) Example of the RCM in Practice Scaled measurement system for investigating new RCM predictions Extension of the RCM to Stochastic Sources Conclusions

41 Extend RCM to Describe Stochastic Sources Located in Enclosures 1. Define a Radiation Impedance Z rad for a spatially-extended and time-dependent source 2. Determine the Radiation Impedance from measured fields near the source measurement plane H z ground plane trace printed circuit board u p x I p t 3. Create the RCM statistical Z cavity for the PCB in an enclosure H z x measurements determine u p x Calculate Z rad 4. Calculate the interaction of one stochastic source with another through an enclosure free space complex enclosure complex enclosure trace PCB Z rad Z cavity 41

42 Details of Extending RCM to Describe Stochastic Sources Located in Enclosures A port now becomes a current trace and it s image in the ground plane currents n z Ground plane I p t, u p x is the current trace profile function image currents In principle one can use measurements of H z x to deduce the trace profiles u p x The radiation impedance can be written in terms of the trace profiles u p x FT of the trace profile k k Z cav = iim Z rad ( ) + é ër rad ù 1/2 û x é ë R rad ù û 1/2 D 1 = 1k2 - k k k 2 The cavity impedance expression is the same as before, except for the updated Z rad and correlations between the random coupling variables + k k k 2 k (k k 2 ) 0 w n are zero mean, unit width, un-correlated Gaussian random variables All trace-to-trace correlations are built in to Z rad 42

43 The Maryland Wave Chaos Group Graduate Students (current + former) Jen-Hao Yeh LPS James Hart Lincoln Labs Biniyam Taddese FDA Bo Xiao Mark Herrera Heron Systems Ming-Jer Lee World Bank Trystan Koch Bisrat Addissie Also: Undergraduate Students Ali Gokirmak Eliot Bradshaw John Abrahams Gemstone Team TESLA Min Zhou Ziyuan Fu Faculty Not Pictured: Paul So Sameer Hemmady Xing (Henry) Zheng Jesse Bridgewater 43 Post-Docs Gabriele Gradoni Matthew Frazier Dong-Ho Wu John Rodgers NRL, Naval Academy, UMD Ed Ott Tom Antonsen NRL Collaborators: Tim Andreadis, Lou Pecora, Hai Tran, Sun Hong, Zach Drikas, Jesus Gil Gil Funding: ONR, AFOSR, DURIP Steve Anlage

44 Conclusions The Random Coupling Model constitutes a comprehensive (statistical) description of the wave properties of wave-chaotic systems in the short wavelength limit We believe the RCM is of value to the EMC / EMI community for predicting the statistics of induced voltages on objects in complex enclosures, for example. Extension to Stochastic Sources looks promising A new measurement system employing scaled structures enables new extensions of RCM: Multiple connected enclosures Irradiation through irregular apertures Systems with mixed (regular and chaotic) properties Systems with evanescent coupling Research funded by ONR ONR/DURIP AFOSR / AFRL Center of Excellence anlage@umd.edu RCM Review articles: G. Gradoni, et al., Wave Motion 51, 606 (2014) Z. Drikas, et al., IEEE Trans. EMC 56, 1480 (2014) 44

45 45 fin