Study and Simulation of CMOS LC Oscillator Phase

Study and Simulation of CMOS LC Oscillator Phase Noise and Jitter Lecturer Michael S. Mc. Corquodale Authors Michael S. Mc. Corquodale, Mei Kim Ding, and Richard B. Brown Solid State Electronics Laboratory Center for Wireless Integrated Microsystems Department of Electrical Engineering and Computer Science University of Michigan Ann Arbor, MI USA 48109 -2122 International Symposium on Circuits and Systems, Bangkok, Thailand, May 2003

Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan Outline • • Motivation Phase Noise and Timing Jitter Expressions and Relationships Baseline Oscillator Topology Topological Enhancements Simulation Framework and Results Comparison of Results Conclusions 2 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Motivation • Review the relationship between phase noise and timing jitter which is particularly important when considering the overlap of time and frequency domain metrics in modern mixed-signal So. C development Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan • Study topological variations in a CMOS LC oscillator that reduce phase noise and jitter • Utilize both time and frequency domain simulation environments to measure phase noise and jitter • Demonstrate agreement between time and frequency domain simulation environments as well as between calculated and simulated phase noise and jitter • Draw conclusions regarding each approach 3 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Phase Noise and Timing Jitter Ideal Oscillator Output Noisy Oscillator Output Phase Noise Time domain uncertainty in period Power at frequency offset from fundamental Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan Timing Jitter Figure 1: Timing jitter illustrated Motivation P. Noise + Jitter Expressions Figure 2: Phase noise illustrated Baseline Top. Enhancements Simulation Comparison 4 Conclusions

Expressions and Relationships Short-Term Timing Jitter Expressions • N-cycle • Period (1 -cycle) Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan • Cycle-to-cycle Phase Noise • Power relative to fund. at offset fm Relationship Between Phase Noise and Jitter • Ignoring flicker noise 5 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Expressions and Relationships Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan Lesson’s Equation for Phase Noise Spectral Density (No/Po)fm = Phase noise density at offset fm F = Noise factor Fixed Parameters • k, T, fo Obvious Enhancements • Increase Q: Motivation for RF MEMS research • Increase C: Power is a constraint k. T = Thermal noise Another Approach • Reduce F, but how? C = Output power Q = Tank quality factor Topology fo = Fundamental frequency fm = Offset from fundamental 6 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan Baseline Oscillator Topology • 1. 8 GHz CMOS LC oscillator • Fully differential/symmetric • Cross-coupled negative resistance sustaining amplifier • NMOS tail • TSMC 0. 18 mm mixed-mode • BSIM RF noise models Figure 3: Baseline CMOS LC Oscillator Topology 7 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan Topological Enhancements • PMOS tail • Common mode capacitor • Cascode tail • Weak inversion tail Figure 4: Enhanced CMOS LC Oscillator Topology 8 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan Phase Noise Simulation Spectre. RF Pnoise Results • >10 d. B gain in tail current device polarity change • 1. 7 - 3. 4 d. B gain for each additional modification • ~20 d. B gain overall • Simulation convergence difficult but fast if achieved Figure 5: Phase noise performance for each studied topology 9 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan Time Domain Noise Simulation Figure 6: Collected period samples Figure 7: Distribution around mean ADS Time Domain Results • ~1000 periods collected with max step = 1/10 th expected jitter • Processor time: ~10 hrs • Data recorded: ~250 MB • Mean period determined • Period jitter determined • Gaussian distribution around mean period • Simulation convergence easy but time consuming 10 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan Comparison of Results Oscillator Configuration Spectre. RF Phase Noise at 600 k. Hz Offset (d. Bc/Hz) Calculated Period Jitter from Phase Noise (fs) Agilent ADS Simulated Period Jitter (fs) NMOS tail -86. 9 481 515 PMOS tail -99. 3 114 118 PMOS tail with capacitor -101. 0 94 93 PMOS cascode tail with capacitor -102. 8 76 83 PMOS weak inversion cascode tail with capacitor -106. 2 51 79 11 Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison Conclusions

Conclusions • Reviewed concepts of phase noise and jitter • Presented relationships for and between phase noise and jitter Michael S. Mc. Corquodale Center for Wireless Integrated Microsystems University of Michigan • Presented topological modifications that improve phase noise and jitter • Simulated both phase noise and jitter and showed good agreement between time and frequency domain results as well as calculated data • Time domain approach takes considerable time and data space but converges easily • Frequency domain approach is fast but it is often difficult to achieve convergence Motivation P. Noise + Jitter Expressions Baseline Top. Enhancements Simulation Comparison 12 Conclusions