Simulating Complex Power- Ground Plane Shapes with Variable-Size Cell SPICE Grids Istvan Novak, Jason R. Miller, Eric Blomberg SUN Microsystems, Inc. One Network Drive, Burlington, MA 01803 Complex plane shapes 1 EPEP2002
Outline • Uniform, rectangular plane models • Need for adaptive, non-uniform grids • Impedance profiles with various cutouts • Hardware correlation with adaptive grid • Conclusions Complex plane shapes 2 EPEP2002
Conductive plane pair with dielectric separation: Simulation Upper conductor Dielectric material Model for Plane Pairs y Lower conductor x (Rectangular and Grid subcircuit model: Uniform) Subckt: L_line_e Y cells Subckt: L_line_g X cells Complex plane shapes 3 EPEP2002
Irregular Plane Shape with Cutouts Complex plane shapes 4 EPEP2002
Symmetrical Cutout in Middle (1) • 1/16” FR4 double-sided 4”x6” rectangular plane pair • Transfer impedance along shorter side • Removed copper – None – 0.5”x0.75” rectangular cutout – 1”x1.5” rectangular cutout – 2”x3” rectangular cutout – 3”x5” rectangular cutout Complex plane shapes 5 EPEP2002
Symmetrical Cutout in Middle (2) Impedance magnitude [ohm] 1.E+02 1.E+01 full 0.5x0.75 1.0x1.5 1.E+00 2.0x3.0 3.0x4.5 1.E-01 Frequency [Hz] 1.E-02 1.E+08 1.E+09 Complex plane shapes 6 EPEP2002
Symmetrical Cutout in Middle (3) Percentage frequency change over percentage copper removed 1.6 0.5”x0.75” 1.4 1.0”x1.5” 1.2 Frequency 2.0”x3.0” 1 of first modal 0.8 3.0”x5.0” resonance 0.6 peak 0.4 0.2 Relative copper area removed [%] 0 0 10 20 30 40 50 60 70 Complex plane shapes 7 EPEP2002
Symmetrical Slot in Middle (1) • 1/16” FR4 double-sided 4”x6” rectangular plane pair • Transfer impedance along shorter side • Slot in middle, 0.125” wide – None – 0.75” rectangular cutout – 1.5” rectangular cutout – 3” rectangular cutout – 4.5” rectangular cutout Complex plane shapes 8 EPEP2002
Symmetrical Slot in Middle (2) Impedance magnitude [ohm] 100 10 full 0.75 1 1.5 3 4.5 0.1 Frequency [Hz] 0.01 1.00E+08 1.00E+09 Complex plane shapes 9 EPEP2002
Symmetrical Slot in Middle (3) Percentage frequency change over percentage copper removed 1 0.75” 0.8 3.0” 1.5” Frequency 0.6 of first 0.4 modal resonance 0.2 peak 0 4.5” -0.2 Relative copper area removed [%] -0.4 0 0.5 1 1.5 2 2.5 Complex plane shapes 10 EPEP2002
Cut from Side (1) • 1/16” FR4 double-sided 3”x6” rectangular plane pair • Transfer impedance along 1” on side • Cut from side, 0.03” wide – None – 0.5” cut – 1” cut – 2” cut – 3” cut – 4” cut – 5” cut Complex plane shapes 11 EPEP2002
Cut from Side (2) Impedance magnitude [ohm] 1.E+03 1.E+02 full 0.5" cut 1" cut 1.E+01 2" cut 3" cut 1.E+00 4" cut 5" cut 1.E-01 Frequency [Hz] 1.E-02 1.E+08 1.E+09 Complex plane shapes 12 EPEP2002
Cut from Side (3) Percentage frequency change over percentage copper removed 60 50 Frequency 40 of first 30 modal resonance 20 peak 10 0 Relative copper area removed [%] -10 0 0.2 0.4 0.6 0.8 1 Complex plane shapes 13 EPEP2002
Limitations of • Many cells may fall Rectangular outside of shape • SPICE run-time grows Uniform Grids sharply with node numbers • Unnecessary nodes increase run time • Cant switch to fine mesh in sensitive areas • Modal resonances may not be captured correctly Complex plane shapes 14 EPEP2002
Cell and Diamond Node Definitions Diamond Cell Segment F upper Diamond1 F lower Diamond4 Diamond2 F pair Diamond3 Complex plane shapes 15 EPEP2002
Definitions of Full cell Example Cell Shape Diamond Partially empty cell Empty cell Complex plane shapes 16 EPEP2002
Grid with Adaptive Sub Gridding Complex plane shapes 17 EPEP2002
Correlation on Modal Resonances (1) Example shape from Slide 4: • Irregular outline • Cutouts Complex plane shapes 18 EPEP2002
Correlation on Modal Resonances (2) Self-impedance Impedance measured, simulated: adaptive at white arrow grid and simulated: fixed grid [ohm] Uniform grid: 1.E+01 Adaptive grid • Overestimates Measured static 1.E+00 capacitance • Overestimates 1.E-01 resonance frequencies Fixed uniform grid 1.E-02 Adaptive grid: Frequency [Hz] • Good 1.E-03 correlation 1.E+07 1.E+08 1.E+09 Complex plane shapes 19 EPEP2002
Acknowledgement Pre-processor SKILL script was written by Roger Cleghorn, Cadence SPICE equivalent circuits were created by perl code, written by Ken Laird, North Eastern University Further contributions and support: • Hemant Shah (Cadence) • Nick Laplaca (SUN) • Deborah Foltz (SUN) • Paul Baker (SUN) • Paul Sorkin (SUN) Complex plane shapes 20 EPEP2002
Conclusions • Odd shapes, cutouts and perforations change – Static capacitance – Modal resonances • Modal resonances do not scale with static capacitance • Adaptive, non-uniform plane models can – Allow for finer mesh in critical areas – Capture modal resonances of odd shapes – Capture signatures of perforated planes • Adaptive grid showed good hardware correlation Complex plane shapes 21 EPEP2002
Correlation on Perforated Plane (1) TDR and TDT response [V] 1.8”x1.6”x0.002” 5.00E-02 -2.34E-01 Measured in the 0.00E+00 -2.36E-01 middle, front/back -5.00E-02 -2.38E-01 Via pair: 20mil drill, 50-mil center-to- -1.00E-01 -2.40E-01 center -1.50E-01 -2.42E-01 TDR source: -2.00E-01 -2.44E-01 • 150psec -2.50E-01 -2.46E-01 • 50 ohm TDT input: -3.00E-01 -2.48E-01 4.50E-08 4.55E-08 4.60E-08 4.65E-08 4.70E-08 4.75E-08 • 50 ohm Time [sec] Complex plane shapes 22 EPEP2002
Correlation on Perforated Plane (2) Simulated and measured response [V] -2.42E-01 Adaptive grid -2.43E-01 captures accurately: -2.44E-01 • Plane -2.45E-01 perforations -2.46E-01 • Edge reflections -2.47E-01 -2.48E-01 4.55E-08 4.56E-08 4.57E-08 4.58E-08 4.59E-08 4.60E-08 4.61E-08 Time [sec] Complex plane shapes 23 EPEP2002
Recommend
More recommend
