1. Full Wave Simulation and Validation of a Simple Via Structure Bruce Archambeault, Samuel Connor, Daniel N. de Araujo, C. Schuster, A.Ruehli, IBM barch@us.ibm.combarch@us.ibm.com, sconnor@us.ibm.com, dearaujo@us.ibm.com, cschuste@us.ibm.com, ruehli@us.ibm.comsconnor@us.ibm.com dearaujo@us.ibm.comcschuste@us.ibm.com ruehli@us.ibm.com M. R. Has hemi, R. Mistral, Pennsylvania State University mmh244 @engr.psu.edu, mittra@engr.psu.edumittra@engr.psu.edu

2. PCBs and Very High Frequency Data Rates Gaga-bit signal rates in common practice today 10-12 Gb/s plan in near future Require at least 5 th harmonic for reasonable signal quality –12 Gb/s  30 GHz Both measurements and simulations require extreme care at these frequencies!

3. Measurements @ 10-50 GHz + Calibration of VNA requires special calibration fixtures Special probing techniques Extremely expensive equipment Requires costly PCBs with various structures to be studied Difficult but possible

4. Full Wave Electromagnetic Simulations Can save costs associated with test equipment and test PCBs Eliminate calibration concerns Probes within models can be ‘perfect’ –No impact on results from the act of measuring Software tools are mature and reasonably priced (sometimes)

5. The High Speed PCB Problem Many layers Entry/exit on different layers Inexpensive dielectrics have more loss at high frequencies Need to be able to properly analyze all effects Quasi-static models not accurate at high frequencies

6. Initial Geometry Single Plane for Initial Models 50 mil Load 10 mil Source 100 mil Round Geometries! Dielectric Constant = 3.8Trace Thickness = 1 mil Via Barrel Diameter = 20 mil Plane thickness = 1 mil Via Antipad Diameter = 44 milTrace terminated in characteristic impedance = 80 Ohm Via Pad Diameter = 34 mil PCB size = 300 mil x 300 mil Trace width = 7 milMetal = perfect electric conductor Loss tan = 0Absorbing boundary conditions!Background DC = 1

7. Validation of Simulation Results Extremely Important Measurement data not available Assumptions must be known –By user –Built into the software –Inherent in simulation technique Different techniques have different assumptions! Simulation validation by using multiple modeling techniques

8. Initial Models Used  r =1  Good Agreement

9. How Good is “Good”? “I know it when I see it” Feature Selective Validation (FSV) technique –Allows a numerical comparison that agrees with ‘experts’ –Used in IEEE Standard on Model Validation –Performs an FFT, separates low frequency data and high frequency data Low frequency data provides indication of overall amplitude agreement (ADM) High frequency data provides indication of rapidly changing feature agreement (FDM)

10. FSV Results for ADM Example (One Plot for Each Pair of Techniques) cFDTD vs FIT PEEC vs FDTD

11. FSV Grade and Spread Easier to quickly compare plots Grade  Number of categories required to get 85% Confidence starting at highest (excellent agreement) –Quality of agreement Spread  Number of categories required to get 85% Confidence starting with the largest category –Consistency of the agreement

12. Summary of FSV Grade/Spread for Initial Agreement Between Techniques TechniquesGradeSpread cFDTD-FIT22 cFDTD-FDTD44 cFDTD-FEM11 cFDTD-PEEC33 FIT-FDTD44 FIT-FEM22 FIT-PEEC33 FTDT-FEM54 FDTD-PEEC44 FEM-PEEC33 Average 3.1 (Excellent-Good)3.0 (Excellent-Good)

13. Final Models Used  r =3.8  Good Agreement

14. FSV Results  Excellent-to-Very Good GradeSpread FDTD-CST44 FDTD-PEEC33 CST-PEEC33 FDTD-FIT PEEC-FIT FDTD-PEEC

15. Traces often Enter/Exit in different Directions Will This Make a Difference? Zero Degrees between entry/exit traces 90 Degrees between entry/exit traces 180 Degrees between entry/exit traces

16. Direction of Entry/Exit for Traces

17. Two Plane Geometry Dielectric Constant = 3.8Trace Thickness = 1 mil Via Barrel Diameter = 20 mil Plane thickness = 1 mil Via Antipad Diameter = 44 milTrace terminated in characteristic impedance = 80 Ohm Via Pad Diameter = 34 mil PCB size = 300 mil x 300 mil Trace width = 7 milMetal = perfect electric conductor Loss tan = 0Absorbing boundary conditions!Background DC = 1

18. Validation Using Different Techniques GradeSpread FDTD-MWS33 FDTD-HFSS33 MWS-HFSS33

19. Direction of Entry/Exit for Traces

20. Ground-Return Via between Planes

21. Real-World PCBs have many Layers/Planes 6 Plane Example10 Plane Example

22. Comparison of 2/6/10 Planes Two Techniques for Validation

23. Model Detail Subtleties Assumptions may be hidden –Simulation techniques –Software simulation tool –User

24. Model Detail Subtleties Examples Dielectric slab vs background dielectric –Volume based techniques can handle dielectric slabs without significant increase in computer resources FDTD, FIT, FEM –Surface based techniques require significant increase in computer resources for dielectric slab Background dielectric easy to simulate MoM, PEEC Does it matter?

25. Dielectric Model Using Background vs Slab Dielectric

26. Model Detail Subtleties Examples Round vs Square objects –Via, Via pad, Via antipad Some techniques have non-rectangular grids to make closer approximation to round objects –FEM Some software tools handle round objects with special internal calculations ‘Regular’ rectangular grids require stairstepping for round objects Does it matter?

27. Square vs. Stairstep Via Structure Example

28. Models Can Provide Information that is Impossible to Measure Examples –Current through ground return via –Displacement return current mapping

29. G round-Return Via Current from FDTD Simulation (Time Domain)

30. G round-Return Via Current from FDTD Simulation (Frequency Domain)

31. V iew of Displacement Current Through Dielectric (Two Ground-Return Vias)

32. Summary Model validation is IMPORTANT –Do NOT simply believe simulation results because a technique or tool was accurate on a different model!!! FSV technique allows user to compare overall quality of agreement –Similar to expert opinion Modeling allows the user to view voltage/currents/fields in ways not possible in the laboratory –Helps understand underlying physics –Helps with validation Via configuration changes transmission at high frequencies significantly

33. FSV Follow up FSV Web site –http://www.eng.dmu.ac.uk/~apd/FSV/FSV%20web/http://www.eng.dmu.ac.uk/~apd/FSV/FSV%20web/ On-line Survey – http://orlandi.ing.univaq.it/test/default.asphttp://orlandi.ing.univaq.it/test/default.asp FSV Program free download – http://ing.univaq.it/uaqemc/public_html/http://ing.univaq.it/uaqemc/public_html/