Transcript (poster)
An Equivalent Circuit Model for a Faraday Cage Substrate Crosstalk Isolation Structure Joyce H. Wu ([email protected]) and Jesús A. del Alamo Massachusetts Institute of Technology Technology Motivation Faraday Cage Isolation Structure Substrate-Via Technology Key Features Noisy or sensitive devices/circuits System-on-Chip Al A/D Al Digital Logic A/D Si substrate Faraday cage Analog/RF A/D Silicon nitride 75-100 µm Silicon nitride Noisy Substrate Si Substrate crosstalk is considered one of the biggest problems in mixed-signal circuits Ta-Ti-Cu seed Cu Grounded via • Deep reactive ion etch (DRIE) • Silicon nitride barrier liner • Electroplated Cu fills via Cu ground plane Fabrication (4) e-beam deposition of Ta-Ti-Cu seed on backside (1) photoresist mask patterning (7) Cu CMP frontside for a flush surface Si (5) Cu electroplating to close bottom of via (2) DRIE through the substrate (3) photoresist strip, silicon nitride deposition from front and backside (8) Al e-beam deposition and patterning to form contact to via (6) Cu electroplating to fill via Cu 12 µm 12 µm x 100 µm vias before Cu CMP step (aspect ratio = 8) Measurements Test Structures Substrate-Via Impedance Reference 100 Faraday Cage Test Structure Substrate Via Model: Faraday cage Ground 100 µm Signal 100 µm 200 µm 70 µm Z11 (Ω) 10 Im(Z11) Re(Z11) Rv 1 Z11 Ground Lv 0.1 Transmitter Receiver Simulation Measured Substrate via 0.01 0.1 1 Faraday Cage Substrate Noise Isolation 80 Reference At 100 µm distance, on average: 32 dB Faraday Cage 42 dB -60 -70 Substrate thickness=77 µm Separation dist.=100 µm Via separation=10 µm Via diameter=10 µm -80 -90 Air -100 1 GHz: 41 dB improvement 10 GHz: 30 dB improvement 50 GHz: 16 dB improvement 60 L (pH) |S 21| (dB) -50 Frequency=10 GHz Substrate thickness=77 µm 70 18 dB -40 100 Frequency (GHz) -20 -30 10 50 40 30 20 Measurement Theory (Goldfarb) 10 0 0 10 20 30 Frequency (GHz) 40 50 0 10 20 Nominal aspect ratio 30 Equivalent Circuit Model Reference Structure Reference Structure with center split Rr Cpad R2 R3 Cpad R2 Cr C3 R3 Cpad Rr R1 = 2 C1 = 2Cr C3 R1 R2 R3 Cpad R2 C1 C3 Rr = 5 k Cr = 3 fF C1 R3 C3 R1 = 2.5 k C1 = 6 fF -20 -20 100-µm transmitter-receiver separation 100-µm transmitter-receiver separation Model unchanged by split -30 |S 21| (dB) |S 21| (dB) R1 -40 Add series Rv and Lv of via -50 0.1 1 10 100 -30 -40 -50 0.1 1 Frequency (GHz) 10 100 Frequency (GHz) 100-µm transmitter-receiver separation Faraday Cage Structure -20 Reference Cpad R1 R2 R2 C1 R3 Cpad Rv C3 |S 21| (dB) R1 -40 C1 R3 Rv=1 kΩ Lv=500 pH Rv=250 Ω Lv=200 pH -60 Rv=70 Ω Lv=70 pH -80 C3 Change only Rv and Lv to evolve from reference to Faraday cage structure Faraday Cage Rv=25 Ω Lv=50 pH Simulation Measured Lv -100 R1 = 2.5 k C1 = 6 fF 0.1 1 10 100 Frequency (GHz) Simulations Comparison of Measurement and Simulation Equivalent circuit lumped-element values Imag -20 Ref. Cage Ref. Cage Cage 100 µm 100 µm 200 µm 200 µm 200 µm 10 µm 10 µm 10 µm 10 µm 70 µm Cr Rr C1 3 fF 5 kΩ 6 fF 2 fF 6.4 kΩ 5 fF 5 fF R1 2.5 kΩ 6 kΩ 6 kΩ R2 200 Ω 200 Ω 250 Ω 250 Ω 250 Ω C3 17 fF 17 fF 17 fF 17 fF 17 fF R3 260 Ω 260 Ω 260 Ω 260 Ω 260 Ω Rv 25-50 Ω 20 Ω 45-85 Ω Lv 10-130 30-50 30 pH Tx-Rx separation Reference S21 S21 Measured -60 Simulation Faraday Cage -80 100-µm pad separation Simulation Measured freq (540.0MHz to 49.05GHz) -100 0.1 1 10 100 Frequency (GHz) • Simple model matches data well (including real and imaginary S21) • Range of Rv and Lv consistent with measured values • Spread of Rv and Lv of substrate via accounts for spread in S21 of Faraday cage Increase Tx-Rx separation distance Increase via spacing -20 -30 100-µm Reference -60 100-µm Faraday Cage 200-µm Simulation Measured Cpad 1.4 pF 10-µm via spacing -90 -100 -100 0.1 1 10 Frequency (GHz) • Increasing pad separation reduces substrate noise • R1 and C1 to account for greater pad separation 100 0.1 1 1.4 pF 1.4 pF 1.4 pF • Developed a simple, lumpedelement equivalent circuit model Rv=20 Ω Lv=30 pH 70-µm via spacing -70 1.4 pF pH Conclusions -60 -80 -80 pH Rv=45 Ω Lv=30 pH -50 200-µm spacing Simulation Measured -40 |S 21| (dB) -40 |S 21| (dB) Real 0.012 0.010 0.008 0.006 0.004 0.002 0.000 -0.002 -0.004 -0.006 -0.008 -0.010 -0.012 |S 21| (dB) -40 Via 10 100 Frequency (GHz) • Increasing via spacing reduces substrate noise isolation • Effectiveness of Rv-Lv shunt is reduced due to fewer vias • Only need to increase Rv-Lv values for larger via spacing • Model matches experimental data into mm-wave regime • Model will be useful to evaluate substrate noise isolation schemes in actual circuits