Polytech`Montpellier – MEA M2 EEA – Systèmes
Transcription
Polytech`Montpellier – MEA M2 EEA – Systèmes
Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 Polytech’Montpellier – MEA M2 EEA – Systèmes Microélectroniques Advanced Analog IC Design Chapter I Introduction Pascal Nouet / 2013-2014 [email protected] http://www2.lirmm.fr/~nouet/homepage/lecture_ressources.html Outline of the complete course (four sessions – 3 hours each) 2 • Future of CMOS circuits: FinFET or FDSOI • Analog IC Design Flow • Advanced specifications • Advanced design techniques 1 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 1.1. History: today and future… • Today: 22nm node – CMOS bulk with Na=1018 atoms/cm3 – What is the number of dopant in a minimum size channel? – Cubic volume of 22nm of side 10-17 cm3 – 10 dopants d t !!! • Strong variability of the threshold voltage • Increase of leakage currents "Study of Random-Dopant-Fluctuation (RDF) Effects for the Trigate Bulk MOSFET« , IEEE Trans. on Electron Devices, 56(7):1538-1542, July 2009. 1.1. History: today and future… • Solutions for threshold voltage variability (and Ion/Ioff) – If substrate doping cannot be controlled then let’s work with undoped silicon – Solution 1: SOITEC-ST Microelectronics planar Fully-Depleted SOI http://hothardware.com/News/SoiTec-Announces-New-SOI-Roadmap--Industry-Uptake-Remains-Unclear/ 2 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 1.1. History: today and future… • Solutions for threshold voltage variability (and Ion/Ioff) – If substrate doping cannot be controlled then let’s work with undoped silicon – Solution 2: INTEL 3D FinFETs or tri-gate transistors (Berkeley, 1999) http://en.wikipedia.org/wiki/Multigate_device https://eda360insider.wordpress.com/2011/06/19/are-finfets-inevitable-at-20nm-“yes-no-maybe”-says-professor-chenming-hu/ FD-SOI Technology 6 http://www.youtube.com/watch?v=uvV7jcpQ7UY 3 Cours CIA Avancé - 2013/2014 - Séance 1 FinFET Technology 09/03/2014 7 http://www.youtube.com/watch?v=Jctk0DI7YP8 Outline of the complete course (four sessions – 3 hours each) 8 • Future of CMOS circuits: FinFET or FDSOI • Analog IC Design Flow • Advanced specifications • Advanced design techniques 4 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 9 Analog IC Design Flow Specifications Choice of the architecture From the Just designer to be sure point!!!of operating point, gain, bandwidth, noise, view: next THE step fabrication cost $ saturations, input number ofmore stages, slew-rate, stability, CMRR, (analog (10 layout -100k$) is and expert time and output ranges, AC small-signal output stage, input andto output dynamic (few than weeks digital layout) few months) MC Simulations gain, PSRR, output simulations folded cascode, ranges, output resistance, resistance differential and output output, t t current, t …… current, … output Initial sizing (1st (1st order order models) DC simulations Verify cell specifications vs stability, gain, technology spreadings bandwidth, CMRR, Verify system-level Application(variability) and mismatches PSRR, capacitive specifications load Temperature and Power Post-layout Specific effects, Layout … Supplies simulations Simulations (DC, AC, TRAN) 10 Analog IC Design Flow Specifications Choice of the architecture Initial sizing (1st order models) MC Simulations AC small-signal simulations (gain, fc) DC simulations (Op. point) ApplicationSpecific Simulations (DC, TRAN, AC) Layout Post-layout simulations 5 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 Outline of the complete course (four sessions – 3 hours each) 11 • Future of CMOS circuits: FinFET or FDSOI • Analog IC Design Flow • Advanced specifications – – – – – Offset considerations Common Mode Rejection Ratio Design for low mismatches Noise fundamentals Characterization • Advanced design techniques Offset considerations 12 • Definition: “input offset” of a differential amplifier is the differential input voltage that leads to a zero output voltage Symmetrical power supplies !!! 6 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 13 Offset considerations • Impact of offset on a simple design: high for low-input levels iin vin vos vout vos RF .iin 596mV 1k • The gain is 60 instead of 100 14 Offset considerations • Offset is a random phenomenon due to: – Technology gy spreading p g low “frequency” q y variations (die to die ; wafer to wafer ; run to run) – Mismatches high “frequency” variations (device to device) – Variability hot topic covering both previous origins • affecting technology parameters and dimensions • generally following a Gaussian distribution. distribution • Propagation to circuit behavior – Example: incertitude on saturation current – µCox, W/L and Vt are affected µCox W I dsat 2 L V gs Vt 2 7 Cours CIA Avancé - 2013/2014 - Séance 1 Offset considerations 09/03/2014 15 • Gaussian distribution basics – For a large g number of identical devices the distribution of actual Vt (µCox, W/L) follows a Gaussian distribution – 0.5% of the values are more than ±3 away from the average value – 6 designs g are then the standard in industry 99.5% of yield in absence of defects – Run to run (technology spreading) is much higher than device to device (mismatches) MC simulations Offset considerations 16 • Standard deviations are related to design !!! – Set of equations q can be A Vt Vt found in the literature WL – Vt spreading increases with the AVt tox 4 NB (mV.µm) substrate doping and the oxide thickness – Vt spreading decreases with the area of the transistor – PMOS fabricated in a N-well exhibits more Vt spreading (ND>>NA) – Overall, Vt spreading reduces with modern technologies (seems to saturate at around 3mV.µm) 8 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 17 Offset considerations µCox • µCox mismatches µCox – Similar expressions p than for Vt AµCox W / L • W/L mismatches t Vt 1,2mV 0,2% 600mV µC ox µCox AµCox WL 0,0056 µm 1 1 W/L W 2 L2 AW / L 0,02µm • Example: 100µm/1µm NMOS in a 0,6 µm tech. V AW / L W / L 0,00056 0,056% W/L 2% – 50% more for a PMOS – 10 times less for MOST on the same die (mismatches) 18 Offset considerations • Random offset in a current mirror Iin T1 I S I dsat IS T2 µCox W Vgs Vt 2 2 L Vs • Design tips Large area and Veff, long I S IS 2.Vt Vgs Vt µC ox W / L µCox W / L • W=100µm ; L=1µm ; Veff=0,1V 0,24% 56 ppm 0,2% • W=10µm ; L=10µm ; Veff=1V 0,024% 56 ppm 0,028% 9 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 19 We’ve already studied… Somedesign variability • Analog flowparameters (wafer to wafer) AV g must fit specifications in the typical yp case µCoxp AµCox V – A design W / L 1 1 AWpossible 2 – A WL design must also fit specifications forall /L 2 µCox WL / W L W L of: process (variability in process AV combinations 12mV.µm AW / L 0,and 02µm 0,0056µm parameters andAµC dimensions), temperature power ox t t t AVt supply 8mV.µm 6 designs • Advanced specifications and variability – Uncertainties U t i ti ttranslate l t iin a Idsat standard t d dd deviation i ti I dsat I dsat 2.Vt Vgs Vt µC ox W / L µCox W / L ... – Random offset in a simple current mirror… • Large transistors and large Veff 20 Offset considerations • Random offset in a differential pair with resistive load and symmetrical supply voltages vod RL I B vos g m RL Veff • Spreading in load resistance vod RL IB RL Veff vos 2 RL 2 • Other spreading vod RL .I dsat vos I dsat Veff I dsat 2 I dsat 2.Vt µCox W / L I dsat Veff µCox W/L vos Vt Veff RL µCox W / L 2 RL µCox W / L 10 Cours CIA Avancé - 2013/2014 - Séance 1 Offset considerations 09/03/2014 21 • But offset can be also systematic – due to the chosen architecture, the bias p point, a wrong layout (systematic mismatch)… – Must be fixed by designer !!! • Examples – Related to design (layout) Veff 3 Veff 1 RS iout 2 iout 2 iin g m RS iout 2 – Related to usage • If Vds2>Vds1 (Vt+Veff1) iout 1 iin g out Vds 2 Vds1 Laboratory work 22 • Random offset in a current mirror I – Using g T1 and T2 at minimum size I and Iin=7µA, let’s record Is=f(Vs) T1 T2 Vs – Let’s identify Vs0 such that Is=Iin. – Let’s run MC simulations (process only then mismatch only then all) and analyze Is=f(Vs) and Is(Vs0). – Do the same simulation for W/L multiplied by ten and then divided by ten – Let’s compare the obtained for the three simulations – Conclusions in S 11 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 23 Laboratory work • Fully differential pair – Let’s simulate a fully y differential pair p with IB=14µA and W/L=10 using ideal resistors of 200k. One input is fixed to 1.65V while the other one is swept. – Using the calculator functions compare and cross determine input voltage that leads to Vo1=Vo2. – Run a MC analysis to determine typical offset – Do the same analysis for different W/L, real resistors and finally using the previous current mirror. Let’s conclude. Connexion serveur Linux MEARH14 • mearh14.pedagogie-virtuel.polytech.univ-montp2.fr • mkdir AMS035 mkdir = make directory (dir name can be choosen freely) • cd AMS035 cd = change directory • source /soft_eii/Cadence/.config_AMS410 • ams_cds -64 -tech c35b4 -mode virt & will prepare your folder (directory) to design in AMS C35B4 technology • Next runs : ams_cds -64 12 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 Outline of the complete course (four sessions – 3 hours each) 25 • Future of CMOS circuits: FinFET or FDSOI • Analog IC Design Flow • Advanced specifications – – – – – Offset considerations Common Mode Rejection Ratio Design for low mismatches Noise fundamentals Characterization • Advanced design techniques Common Mode Rejection Ratio, CMRR 26 • Definition: CMRR characterizes the ability of a differential amplifier to reject the common mode V VMC Vd 2 + Vd Vs Ad Vd AMC VMC - V VMC CMRR Ad AMC Vd 2 A CMRRdB Ad ,dB AMC ,dB 20 log d AMC 13 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 Common Mode Rejection Ratio, CMRR 27 • Random CMRR in a differential pair Ad vod vid Amc g m RL vinc 0 vod vinc RL I B Veff 0 CMRR vid 0 • Impact of spreading spread ng in n RL – vinc vinc/RB in the current source output resistance vod RL vinc v Amc od 2 RB vinc vid 0 2 g m RB RL CMRR 2 RB RL RL Common Mode Rejection Ratio, CMRR 28 • Without RL spreading, vinc/(2.RB) in each load resistance • Impact of spreading in MOST vod RL I ( RL ) RL vinc I dsat 2 RB I dsat 2.Vt µCox W L Veff W L µC ox g m RL 2 g m RB CMRR RL 2.Vt µCox W L Amc RL Veff µCox W L Amc vod R L vinc 2 RB 14 Cours CIA Avancé - 2013/2014 - Séance 1 09/03/2014 29 Random CMRR and offset trade-off • Design for low offset and high CMRR vos Vt Veff RL µCox W L 2 g m RB CMRR W L 2 RL µCox RL 2.Vt µC ox W L RL Veff µC ox W L vos CMRR g mVeff RB I B RB The lower the offset, the higher the CMRR !!! 1. Optimize for low offset – Low Veff (0,1V), large transistors, matched resistors reduce Vt spreading and current mismatch 2. Optimize for large CMRR High gm(IB) & RB 30 Systematic CMRR • CMRR can be also systematic: example with a differential pair with referenced output – Current source output resistance: RB – Common Mode VMC change bias current VMC/RB T3 – VMC/RB does not equally share between T1 and T2 I – Small-signal Small signal analysis to calculate V+ induced output voltage Vdd T4 Id4 Id2 d1 CMRR 2 g m 2 g m 3 rds 1 R B T1 T2 Vout V- Ibias 15