RF Noise Simulation for Submicron MOSFET s Based on Hydrodynamic Model Jung-Suk Goo, Chang-Hoon Choi, Eiji Morifuji, Hisayo Sasaki Momose, Zhiping Yu, Hiroshi Iwai, Thomas H. Lee, and Robert W. Dutton, ULSI Engineering Laboratory, Toshiba Corporation 1
Outline Motivation Simulation Method Validity Examination Drift-Diffusion Model vs. Hydrodynamic Model Simulation Results for.25µm nmosfet Conclusions 2
Motivation (RF CMOS) Rapid f t increase of MOSFETs, driven by the microprocessor industry, attracts RF designers. Promise of integrating whole systems on a single chip. Noise behavior in short channel MOSFETs is not well understood yet. 3
Motivation (Continue) (MOSFET Noise) Flicker (1/f) Noise Dominant up to few MHz range Significant in mixer circuits (Up-conversion Error) Shot Noise Dominant in the subthreshold region Thermal Noise (Velocity Fluctuation Noise) Dominant in high frequencies 4
Motivation (Continue) (HF MOSFET Noise - Thermal) Excess drain noise in short channel MOSFETs caused by carrier heating near drain junction. - A. A. Abidi (IEEE TED, 1986) - B. Wang et al. (IEEE JSSC, 1994) Induced gate noise due to the distributed nature of MOS devices. - Introduced by A. van der Ziel in 1976 - NO QUANTITATIVE report to date - D. K. Shaeffer et al. (IEEE JSSC, 1997) 5
Simulation Method (1D vs. 2D/3D) 1D Approach - Using transmission line analogy - Reasonable comutational cost - Poor accuracy 2D/3D Approach - Impedance Field Method + Adjoint analysis - Better accuracy, incorporating 2nd order effects - Expensive in implementation and simulation (no HD to date) 6
Cgd gm Vgs;local Q inv V gs local I DS V gs local Simulation Method (Continue) (Hybrid Approach) Gate (V GS ) Source Gate Drain Oxide V gs,local Drain (V DS ) V ds,local I DS Silicon x Cgs C gs +C gd = ro o = V ds local r DS I g m = in S in = 4 kt n I DS V ds local 7
Validity Check - I (First-Order Network Parameters) Gate Source Drain Gate Gate Y 11 + Y 21 Y 12 + Y 22 Y 12 Source (Y 12 Y 21 )v GS Drain 8
Validity Check - II (Non-segmented Uniform Transmission Line) Drain Current Noise [A 2 /Hz] 1 22 1 23 1 24 1 25.1 µm 1. µm 1 µm 1 6 1 8 1 1 1 12 Gate Current Noise [A 2 /Hz] 1 2 1 25 1 3 1 35 1 4 1 µm 1. µm.1 µm 1 6 1 8 1 1 1 12 Frequency [Hz] Frequency [Hz] for V DS =V 9
Validity Check - II (Continue) (Segmentation Error for Uniform Transmission Line) Error in Drain Current Noise [%] 1 5 1 1 5 1 1 1 15.1 µm 1. µm 1 µm 1 15 1 1 1 5 1 Segment Length Device Length Wave Length [µm] Error in Gate Current Noise [%] 1 3 1 2 1 1 1 1 1.1 µm 1. µm 1 µm 1 15 1 1 1 5 1 Segment Length Device Length Wave Length [µm] x L/λ=1-5 µm corresponds to 28GHz for L=.25µm divided into 2 segments 1
Validity Check - III (Long Channel MOSFET Case) Noise Parameters 2 1.5 1.5 5µm nmosfet V GS =.9V f = 1MHz δ γ I[c] Classical Values γ = 1. (Linear) = 2/3 (Saturation) δ = 4/3 c = j.395 (Saturation) (Saturation) = i 2 d 4 kt f gd.5 1 1.5 2 Drain to Source Voltage [V] = i 2 g 4 kt f <[YGS] c = i gi d q 2 g i2 d i 11
Short Channel Effect (Drift-Diffusion vs. Hydrodynamic).25µm nmosfet Noise Parameters 4 3 2 1 V GS =.9V f = 1MHz δ HD γ HD δ DD γ DD.5 1 1.5 2 2.5 Drain to Source Voltage [V] 12
n = jbj 2 i 2 d R kt f 4 Comparison to Measured Data (γ δ c vs. F min -R n -Y opt ) Intrinsic Drain Noise Gate Noise Correlation Extrinsic Interlayer Capacitance Gate Resistance Parasitic Pad Loss Routing Inductance F min = 1+2R n (G opt + G c ) G opt = B opt = ;B c Y c = D B ; c B 1+2R n G opt G r u + G 2 c r G u n R i2 g vut i 2 d D B i 2 g u = (1 ;jcj 2 ) G ktf 4 R n 1 B = Y 21 + Y 22 = 11 + Y 12 D +1 Y + Y 22 21 Y 13
Four Noise Parameters I (Intrinsic).25µm nmosfet Minimum Noise Figure [db] 3 2.5 2 1.5 1.5 V DS =2.5V f = 4GHz.5 1 1.5 2 2.5 Noise Resistance [Ω] 2 15 1 5 V DS =2.5V f = 4GHz.5 1 1.5 2 2.5 Gate to Source Voltage [V] Gate to Source Voltage [V] Red Blue Green : Measured Data : Hydrodynamic Simulation : Drift-Diffusion Simulation 14
Four Noise Parameters II (Intrinsic).25µm nmosfet Optimum Conductance [ms] 6 5 4 3 2 1 V DS =2.5V f = 4GHz.5 1 1.5 2 2.5 Gate to Source Voltage [V] Optimum Susceptance [ms] 5 1 15 V DS =2.5V f = 4GHz.5 1 1.5 2 2.5 Gate to Source Voltage [V] Red Blue Green : Measured Data : Hydrodynamic Simulation : Drift-Diffusion Simulation 15
Four Noise Parameters I (After correction for the loss due to G opt ).25µm nmosfet Minimum Noise Figure [db] 3 2.5 2 1.5 1.5 V DS =2.5V f = 4GHz.5 1 1.5 2 2.5 Noise Resistance [Ω] 2 15 1 5 V DS =2.5V f = 4GHz.5 1 1.5 2 2.5 Gate to Source Voltage [V] Gate to Source Voltage [V] Red Blue : Measured Data : Hydrodynamic Simulation 16
Four Noise Parameters II (After correction for the loss due to G opt ).25µm nmosfet Optimum Conductance [ms] 6 5 4 3 2 1 V DS =2.5V f = 4GHz.5 1 1.5 2 2.5 Gate to Source Voltage [V] Minimum Noise Figure [db] 2.5 2 1.5 1.5 V GS =1.1V V DS =2.5V 2 4 6 8 1 Frequency [GHz] Red Blue : Measured Data : Hydrodynamic Simulation 17
Conclusions Accurate and efficient noise simulation technique : 1D active transmission line + 2D device simulation First known attempt to use an advanced (HD) transport model for 2D noise analysis First successful noise simulation results for deep submicron MOSFETs 18