Stack-up and routing optimization by understanding micro-scale PCB effects Authors: G. Romo, CST of America Chudy Nwachukwu, Isola Group Reydezel Torres-Torres, INAOE Seung-Won Baek, CST of America Martin Schauer, CST of America 1
Outline Introduction Fiber Weave Effects Analytical Model Simulations and Measurements Summary and Conclusions 2
Introduction As communication frequencies continue to increase, new (microscopic) effects arise We present an experimental, numerical, and analytical investigation of the role that Fiber weave effects play in high speed interconnects 3
Fiber Weave Effect FW effects originate from the inhomogeneous properties of PCB laminates 4
Laminate Properties Electrical properties of prepreg material Electrical properties of cured laminate material 5
Fiber Weave Effects Fiber weave used in this work 1080 Warp & Weft Count: 60 x 47 Sparse material (inhomogeneous) Thickness: 0.0025 /0.065mm 2116 Warp & Weft Count: 60 x 58 Dense material (homogeneous) Thickness: 0.00387 /0.097mm 6
Fiber Weave Effects (FWE) Effects due to location of trace w.r.t. fiber weave bundles High Er; Low Zo Low Er; High Zo Effects due to periodic loadingof trace by fiber weave bundles Resonance Ф 7
FWE: Resonance For incident broadband electromagnetic radiation on a periodic structure: λ' λ d The wavelength satisfying the in-phase constructive interference will be reflected and the rest transmitted. Resonance d = λ' 2 λ' Reflected λ λ' Transmitted Size or resonance depends on materialcontrastand length of propagation λ 8
Special cases: Ф=0, 45, 90 FWE: Periodic loading For the above angles: f c d ~ 20mils ~ f ~ 150GHz 2d ε r 9
Arbitrary Angle: Ф=0 > 45 FWE: Periodic loading Challenge is to find/define the spatial period Separate the Weft and Warp loading Warp loading is in pitch scale high frequencies neglected Weft loading occurs in a larger scale lower frequencies key role Weft spatial period is obtained from trigonometric expressions: c 1 [ ] fres = 2 weft period = pitch + 1 2 2 1 tan( Φ) 2 ε eff pitch 2 + 1 [ tan( Φ) ] 10
Resonance Frequency and Spatial Period Angles Ф<20 are potentially the most critical Large angles Ф>5 More periods per length strong resonances 11
FWE: Simulation results 3D parameterized model created in CST MICROWAVE STUDIO 1080 2116 12
FWE: Simulation results Position- Simulation of 1080 laminate at Ф=0, for different trace positions Positions 1 and 9: Trace over bundles Predicted impedance variations is very small (~1Ω) 13
FWE: Simulation results Position- Simulation of 1080 laminate at Ф=0, for different trace positions Positions 1 and 9: Trace over bundles Predicted Effective relative permittivity is also small (~0.13) 14
FWE: Simulation results Angle- Simulation of 1080 laminate at Ф=0, Ф=7, Ф=10 andф=15, 7in Insertion Loss Return Loss Excellent agreement with analytical equation Resonance strength is also consistent: the larger the angle more periods per length Stronger resonance Angle Equation 2 Simulated (GHz) [Deg] (GHz) 7 19.6 19.67 10 26.76 27.0 15 41.39 41.8 15
FWE: Simulation results Angle- Simulation of 2116 laminate at Ф=7, Ф=10 andф=15, 4in Insertion Loss Return Loss Excellent agreement with analytical equation Resonance strength is weaker compared with that from 1080 due to the more homogeneous laminate 16
Experimental Verification: Board Stack-Up I-Tera, 1080 and 2116 weave based resin-system 50 ohm singleended microstrip configuration 17
Experimental Verification: Test Structures 2 sets of 3-in long microstrips. 2 sets of 4-in and 8- in microstrips with solder mask. 2 sets of 4-in and 8- in long microstrips without solder mask. Same number of sets of striplines (work in progess) 18
Measurement Setup Two-port S-parameter measurements were performed to the fabricated transmission lines Board Probe-2 Camera view Probe-1 VNA 19
VNA Calibration An off-board calibration was applied to correct the measurement system from systematic errors Probe-1 Probe-2 For that matter, the line-reflect-match (LRM) algorithm and an impedance standard substrate (ISS) were used Calibration substrate (ISS) 20
Fiber Weave Effect: Position The characteristic impedance was obtained for lines at different locations on the board 3 in Position 1 Position 7 Re(Z Z c ) (Ω) 70 68 66 64 62 Top (1080) Bottom (2116) 1 2 3 4 5 6 7 Position (#) The observed changes in Zc are within the tolerances typically specified by the PCB shop (±5%) 21
Fiber Weave Effect: Angle (Top) The same experiment was carried out but now varying Φ Re(Z Z c ) (Ω) 52 50 48 Φ = 0 Φ = 7 Φ = 10 Φ = 15 Top (1080) 46 0 10 20 30 40 50 f (GHz) Relatively small changes are observed in Zc 22
Fiber Weave Effect: Angle (Bottom) Similar variations are observed on the bottom side of the board ) (Ω) Re(Z c ) 53 52 51 Φ = 0 Φ = 7 Φ = 10 50 Φ = 15 Bottom (2116) 49 48 0 10 20 30 40 50 f (GHz) Zc remains relatively low sensitive to changes in Φ 23
Fiber Weave Effect: Angle, ε Within the measured frequency range, the maximum variation of the permittivity is less than 2% effective relativ ve permittivity 2.9 Φ = 0 Φ = 7 2.8 Φ = 10 Bottom (2116) 2.7 Φ = 15 2.6 2.5 Top (1080) 2.4 0 10 20 30 40 50 f (GHz) effective tive permittivity relat 2.7 2.6 2.5 2.4 0 determined @20 GHz Top (1080) Bottom (2116) 7 10 Φ ( ) 15 The small variations of Zc and Eeff are consistent with the predicted simulation results 24
Fiber Weave Effect: Angle (1080) Insertion and Return Loss, 4-in traces Insertion Loss Return Loss Excellent agreement with analytical equation Strength of resonance increases with angle Angle [Deg] Equation 2 (GHz) Measured (GHz) 7 19.6 19.17 10 26.76 28.2 15 41.39 42.86 25
Fiber Weave Effect: Angle (1080) Insertion and Return Loss, 8-in traces Insertion Loss Return Loss Excellent agreement with analytical equation Resonances spread over frequency due to fiber weave imperfections 26
Fiber Weave Effect: Angle (2116) Insertion and Return Loss, 4- and 8-in traces Insertion Loss Return Loss (8-in) No resonances were detected from measurement This is due to the highly homogeneous material 27
α (Np/m) Fiber Weave Effect: Angle, attenuation The effect of the resonances clearly impacts the propagation characteristics 8 6 4 Top (1080) Φ = 0 Φ = 7 Φ = 10 Φ = 15 2 0 0 10 20 30 f (GHz) 40 50 α (Np/m) 8 Top (1080) Bottom (2116) 6 Φ = 15 4 2 0 0 10 20 30 f (GHz) 40 50 This is more accentuated on the 1080 side of the board due to the sparser weave This, as well as other discussed results, points out that the more accentuated fiber weave effect is that associated with the resonances 28
Conclusions Most important FWE is that associated with resonances An analytical model, verified by modeling and measurements was proposed Small angles (5-15 ) are potentially the most critical to signal integrity at bandwidths below 50 GHz The use of homogenous spread fiber weave has been shown to mitigate these effects 29