Docstoc

1134.001 BSIM for 30nm MOSFETs

Document Sample
1134.001 BSIM for 30nm MOSFETs Powered By Docstoc
					           BSIM 4.5.0

                 BSIM Group

Prof. Chenming Hu, Prof. Ali M Niknejad,
University of California, Berkeley




          Compact Modeling Council May 8, 2006
                    Bug Report on BSIM4.5.0
       BSIM4 Manual Correction :

                 Suggested by Well Proximity Effect Sub-committee
                 Guidelines in BSIM manual for extracting the WPE instance
                  parameters “SCA, SCB and SCC” to be removed.
                 CMC maintained document on extraction to be referenced instead.
                 The change suggested by Mark Basel to be incorporated.



       BSIM4 Code Correction for Mobility Model:

                 Reported by ADI.
                 Two issues pointed out by ADI concerning the implementation of
                  Coulomb scattering effect in BSIM4.5.0
                 Derivative issue discovered by BSIM group.


CMC May 8, 2006                                                          UC Berkeley - 2
                  BSIM4.5.0 Mobility Model

           A new term was added in the mobility model to
            capture the Coulomb scattering effect.
           mobMod=0
                                                                         U 0  f ( Leff )
                      eff                                                                                                             2
                                                                  2V                      2Vth        V  TOXE                  
                                                                                                     2
                                                         V                     V
                               1  UA  UC Vbseff      gsteff th        UB  gsteff          UD  th                        
                                                          TOXE                  TOXE                 Vgsteff  2Vth            
                                                                                                                                   


        mobMod=1                                                       U 0  f ( Leff )
                      eff 
                                     Vgsteff  2Vth        Vgsteff  2Vth  
                                                                                                                                2
                                                                                                            V  TOXE 
                                                                                2

                               1  UA                  UB                      1  UC Vbseff   UD  th
                                     TOXE 
                                                      
                                                              TOXE  
                                                                                                           Vgsteff  2Vth 
                                                                                                                            
                                                                                                                         



        mobMod=2                                                       U 0  f ( Leff )
                      eff 
                                                                C0 VTHO  VFB   s  
                                                                                                EU                          2
                                                      V                                                   V  TOXE 
                               1  UA  UC Vbseff   gsteff                                        UD  th
                                                                      TOXE
                                                                                                          Vgsteff  2Vth 
                                                                                                                           
                                                                                                                        
CMC May 8, 2006                                                                                          UC Berkeley - 3
           Issue 1 : Backward Compatibility

          In BSIM4.5.0, the default value of the new parameter
           “UD” is 1e+14
          To have backward compatibility to BSIM 4.4.0 , UD
           should be explicitly specified to be zero in BSIM4.5.0
           modelcards.

          To fix the issue, the default value of the new
           parameter “UD” is now set to 0.

          For ex, for mobMod = 0,                                    NEW DEFAULT UD=0
                                                   U 0  f ( Leff )
      eff                                                                                        2
                                         Vgsteff  2Vth   Vgsteff  2Vth   Vth  TOXE     
                                                                        2

               1  UA  UC Vbseff     TOXE   UB  TOXE   UD  V  2V                   
                                                                           gsteff         
                                                                                         th   
CMC May 8, 2006                                                                    UC Berkeley - 4
           Issue 2 : Non-Monotonic Drain Current

          Due to the new “UD” term, drain current can be non-
           monotonic in certain circumstances.

          If UD is explicitly set to 0, the problem is absent in
           BSIM4.5.0.

          If Vth < 0 , Vgsteff + Vth can approach zero and mobility
           can go to zero.
                                                  500µ
                                                             UD = 0
                                                  400µ       UD = 1e14

                               Vx
                                                  300µ
                                        IVx (A)


                                                  200µ


                                                  100µ
                                                                   Coulomb Scattering
                                                     0
                                                     0.00   0.05       0.10      0.15      0.20
                                                                     Vx (V)
CMC May 8, 2006                                                                         UC Berkeley - 5
                    Issue 3 : Derivatives

          In BSIM4.5.0, the derivatives of the new mobility UD
           term with respect to Vg and Vb were coded with a
           bug

          The error stems only from the Coulomb scattering
           term.




CMC May 8, 2006                                           UC Berkeley - 6
                       Model Correction

          Modify the coulomb scattering term for all mobMod’s.




          For ex. mobMod = 1




          The modification also fixes the UD term derivative bug
           of BSIM4.5.0.
CMC May 8, 2006                                           UC Berkeley - 7
                                     Model verification

                 Model implemented for all mobility models.
                 Derivatives ensured to be correct.


                              OLD                                                      NEW
           500µ                                                     500µ
                        UD = 0                                             New Coulomb scattering term
           400µ         UD = 1e14                                   400µ          UD = 0
                                                                                  UD = 1e14
                                                                                  UD = 1e17
           300µ                                                     300µ

                                                          IVx (A)
 IVx (A)




           200µ                                                     200µ


           100µ                                                     100µ
                              Coulomb Scattering
                0                                                      0
                0.00   0.05       0.10      0.15   0.20                0.00     0.05     0.10     0.15      0.20
                                Vx (V)                                                  Vx (V)
CMC May 8, 2006                                                                                  UC Berkeley - 8
                           Release Plan

          After further verification by member companies, the
           bugfix will be released as a part of BSIM4.5.1 release
           in June.




CMC May 8, 2006                                            UC Berkeley - 9
                    Acknowledgements


             Weidong Liu         Synopsys

             Geoffrey J. Coram   Analog Devices, Inc.

             Keith Green         Texas Instruments




CMC May 8, 2006                                 UC Berkeley - 10
Technology and Application
 Driven Enhancements for
          BSIM
Mohan Dunga, Wei-Hung Chen, Babak Heydari
          Darsen Lu, Hui Wan
      Chenming Hu, Ali M. Niknejad

 Department of Electrical Engineering and Computer Sciences,
             University of California, Berkeley




                        CMC May 8, 2006
           Emerging High Frequency Applications
                                                                                      10GbT
                               10G
                                                                    1000bT
                                1G
            Throughput (bps)


                                                       100bT                       802.15.3a
                               100M                                    802.11a/g
                                      10bT                                                              1Gbps: The next
                                                                    802.11b
                               10M                                                                     wireless challenge!
                                                             802.11
                                1M                                                              LAN
                                                                                                WLAN



                                        88   90   92    94     96      98     00     02    04     06

                                                                    Year
          High throughput MIMO based WiFi (support for video)
          WiMAX and multi-standard radios (2.4 – 5.75 GHz)
          UWB communication from 3-10 GHz
          Automotive radar (24 GHz, 77 GHz)
          High-speed short range communication (60 GHz)
          Satellite receivers for entertainment and data (11 GHz)


CMC May 8, 2006                                                                                                      UC Berkeley - 12
                  Extension of Portable Device
                                         Extended display for
                                          device
                                              PDA
                                              Digital camera
                                              Video camera

                                         Wireless USB
                                              Storage
                                              Printer

                                         Data transfer
                                              Digital Camera
                                              Video Camera
                                              Sync
                                              Music
                                              Movies

CMC May 8, 2006                                          UC Berkeley - 13
       Source: DC, Workshop IMS2002
                                          24/77 GHz Automotive Radar




                                     Safety, improved functionality, automatic cruise control …

CMC May 8, 2006                                                                                    UC Berkeley - 14
                    10 GHz Satellite LNB
      Exploit low cost CMOS technology to
       build an integrated 10 GHz low-noise
       block: LNA, mixer, and VCO
      Replace bulky microwave DRO LO
       with a frequency synthesizer
      Difficult Specs: 60 dB power gain,
       System NF < 1 dB!

                                    Current
                                     130x130 mm2


                                          Final goal 
                                                 2mm2


CMC May 8, 2006                                          UC Berkeley - 15
                  Measured 10 GHz CMOS LNA




                            Complete model based LNA design
                            Layout parasitics substantial (~100pH)
                            Measured LNA is fully functional
                            Measured gain and NF worse

CMC May 8, 2006                                           UC Berkeley - 16
       SDR, Universal, Cognitive, Dynamic?
         Loose definitions:
                 SDR: Reprogram the baseband
                 Universal: Multi-standard
                 Multi-mode: short/long range, high/low data
                 Cognitive: Ability to sense spectrum and use it
                 Dynamic: Ability to alter bias currents to
                  tradeoff performance versus power
                  consumption
         Let’s enable CMOS to attack these new
          class of radios …
CMC May 8, 2006                                            UC Berkeley - 17
                  Too many external passives
                              Diode/Switch
                                       RX
                   DCS/PCS                        BPF: PCS

                                                  BPF: DCS
                   Diplexer
                                       RX         BPF: GSM
                    GSM
                              TX

                                                  LPF: PCS/DCS   Isolator   Coupler

                              Diode/Switch   TX   LPF: GSM       Isolator   Coupler


   Current trends in academia/industry have reduced component count
   The Low-IF/Direct-Conversion radio architectures eliminate (reduce) external IF
    filters
   Systems still heavily dependent on external components on the front end: SAW
    filters, switches, directional couplers, matching networks, pin diode, diplexers …
   Many of these components are expensive (high Q) and narrowband

CMC May 8, 2006                                                                       UC Berkeley - 18
                   Multiplicity of Standards

    Cellular voice: GSM, CDMA, W-
     CDMA, CDMA-2000, AMPS,
     TDMA…                                                 Image Reject                     Channel Select              IF IQ Mixers                    Baseband



    Same standard over multiple frequency             LNA 1             LNA 2
                                                                                 RF Mixer

                                                                                            IF         AGC
                                                                                                                            90
                                                                                                                                                              ADC   I




     bands (4-5 GSM bands exist today)                                                      IF Gain and AGC
                                                                                                                                                              ADC
                                                                                                                                                                    Q




                                                 PLL
                                                                VCO



    Data: 802.11x, Bluetooth, 3G,                             LC Tank
                                                                          RF Synthesizer                      IF Tank       IF PLL     IF Synthesizer



     WiMax…
    A typical handheld computer or laptop
     should be compatible with all of the
     above standards
    Today a typical cellular receiver has 3-4
     radio front-ends … this approach does
     not scale!



CMC May 8, 2006                                                                                                           UC Berkeley - 19
              Berkeley “COGUR” Transceiver

                                     Broadband dynamic
                                      LNA/mixer
                                     Wide tuning agile
                                      frequency synthesizer
                                     Dual-mode broadband
                                      PA with integrated
                                      power combining and
                                      control
                                     Linear VGA
                                     Programmable analog
                                      baseband filters
                                     High-speed
                                      background calibrated
                                      ADC/DAC

CMC May 8, 2006                                 UC Berkeley - 20
                      Multi Gain-Step Feedback LNA
                                                                                                              Vdd


                                                                                                                             PMOS




                                                                                       RF+            Vout-         Vout+           RF-




                                                                                                                            NMOS



                                                                                                   Bias_N




                                                                                                        Vss
                      Bruccoleri        This Work (simulated)
                      [1]
                      (Measured)     High Gain               Low Gain               Current reuse for higher gm
  Gain                   13.7 dB      13.1 dB                   3.18 dB

  -3 dB Bandwidth      2-1600 MHz               200-3500 MHz
                                                                                    Current-source noise is rejected
  |S12|, max             -36 dB                    -40 dB                           Resistive feedback creates self-biasing
  IIP3 (900M, 905M)      0 dBm       -11.7 dBm                  3.1 dBm
                                                                                    MOS switches/resistors in combinations
  NF (50 )              2.4 dB        2.4 dB                   4.7 dB

  Differential             No                       Yes
                                                                                     with passive resistors
  Gain Steps               No                       Yes

  Power                14mA x 2.5V              9.2 mA x 1.5V             [1] F. Bruccoleri, et al., “Wide-Band CMOS Low-Noise Amplifier Exploiting
                                                                          Thermal Noise Canceling,” IEEE Journal of Solid-State Circuits, Vol. 39, No. 2,
  Technology           0.25 CMOS                 0.13 CMOS                February 2004

CMC May 8, 2006                                                                                                                       UC Berkeley - 21
                  LNA with Noise Cancellation

                                                                                               R2
                                                                       R1
                                                                                              Vout
                                                                                Vsig (+)
                                                                                Vnoise (+)




                                                                            m
                                                                                      M3        M4
                                                                       M1
                   [Bruccoleri, et al., ISSCC 2002]
                                                              Vin               Vsig (+)
                                                                                Vnoise (-)


                                                                       Rs                     M2




                                                         Wideband and inductorless
                                                         Noise of M2-3 remains as residue. Can be
                                                          minimized free from S11 constraint
                                                         In power constraint design, the optimum
                                                          NF occurs at partial noise cancellation of
                                                          M1
CMC May 8, 2006                                                                              UC Berkeley - 22
                                NF Simulation




                  Technology     Bandwidth     S11     S21     NF

                  0.13um/1.5V   500M~2.5GHz   <-15dB   16dB   ~2.2dB

CMC May 8, 2006                                                     UC Berkeley - 23
                             Linearity Enhancement

    Device IIP3 determined by gm’’
    Composite transistor (M2/M2a) to                        g m’
     reduce IIP3 bias sensitivity                                            gm
    Choose M3 optimum bias

                                                                gm’’
                                         R2
                  R1
                                        Vout
                           Vsig (+)
                           Vnoise (+)




                       m
                                 M3      M4
                  M1

     Vin                   Vsig (+)
                           Vnoise (-)


                  Rs                    M2     M2a   Transistor IIP3 vs. bias point


CMC May 8, 2006                                                         UC Berkeley - 24
             Wideband Low 1/f Noise I/Q Mixer




                              0.8GHz    1.5 GHz    2.4GHz

   Conversion Voltage Gain    35.17dB   35.37dB    35.89 dB
                                                                      Complementary input for higher gm
          First Pole          269 kHz   268 kHz    264 kHz
                                                                       and linearity
      DSB NF @10kHz           10.8 dB   11.5 dB    12.4 dB

              @100kHz         8.51 dB   8.54 dB    8.84 dB
                                                                      Passive switching to get low 1/f noise
               @1MHz          8.23 dB   8.11 dB    8.27 dB            Implemented as a I/Q mixer with
          1/f Corner          20 kHz     30 kHz    40 kHz              integrated on-chip divider
   IIP3 1.9G, 10M and 20.1M             4-16 dBm                      1/f noise model and symmetry
                                                                       important!
                                                              Bias = 11.5 mA (for I+Q, excluding divider)
CMC May 8, 2006                                               Vdd = 1.5V                                    UC Berkeley - 25
              Linear CMOS Power Amplifiers
                                                                      35                                                       35

                                                                      30                                                       30




                                                                                                                                    Drain Efficiency (%)
                                                                      25                                                       25




                                                         Pout (dBm)
                                                                      20                                                       20
                                                                                                Pout
                                                                      15                                                       15

                                                                      10                                     Drain Efficiency 10
                                 50Ω
                        Output
                                                                             5                                                 5

                                                                             0                                                 0
                                                                             -20          -10              0        10        20
                                                                                                      Pin (dBm)


                  - +             -+    - +        -                         20
         +                                                                                       f1
                                                                             10




                                                                Pout (dBm)
                                                                                 0
                                                                                                IM3=-29dBc @ 18dBm
     Fully integrated CMOS PA (27dBm, 32%                              -10
      efficiency, no external passives)                                 -20                2f1-f2
     Linearity specs important for OFDM and other
                                                                        -30
      complex modulation schemes
     Power back-off possible with graceful efficiency                  -40
                                                                                     -5               0 P (dBm) 5
                                                                                                        in
                                                                                                                         10
      reduction
CMC May 8, 2006                                                                                              UC Berkeley - 26
                       Importance of Linearity

          Linearity over broad bias range increasingly important for
           multi-standard radios
                 Higher levels of TX leakage (e.g. WCDMA) mean higher linearity
                  specs
                 Reduced filtering and broadband operation
          Need predictive and bias dependent prediction of device
           non-linearity over entire range of operation (weak to strong
           inversion)
          Linearity especially important for LNA, driver amplifier
           and Power Amplifiers




CMC May 8, 2006                                                         UC Berkeley - 27
            Prediction of Spectral Regrowth




          Two-tone or three-tone test not sufficient to predict spectral regrowth
          Simulation of spectral regrowth requires long transient simulation to
           include many data cycles during RF operation.
          Need good model accuracy and speed!

CMC May 8, 2006                                                            UC Berkeley - 28
                          Small Signal AC Models
    • Applicable for amplifier
      design

    • Useful for device
      optimization

    • Predictive for higher
      frequencies and noise                Add Physical                Initial value
                                            Parasitics                 for the core




      Core
                                                          Use ICCAP for
                                                           Optimization
                       Substrate network
CMC May 8, 2006                                                      UC Berkeley - 29
                                                                                                                    AC Modeling Result
                                       1.0
                                                                                                                                                                                                                      4



                                       0.8                                                                                                                                                                            3



                                       0.6                                                                                                                                                                            2


                                       0.4
                                                                                                                                                                                                                      1
               imag(SP1.SP.S(3,3))
               imag(SP1.SP.S(1,1))




                                                                                                                                                                                             imag(SP1.SP.S(4,3))
                                                                                                                                                                                             imag(SP1.SP.S(2,1))
                real(SP1.SP.S(3,3))
                real(SP1.SP.S(1,1))




                                                                                                                                                                                              real(SP1.SP.S(4,3))
                                                                                                                                                                                              real(SP1.SP.S(2,1))
                                       0.2
                                                                                                                                                                                                                      0


                                       0.0
                                                                                                                                                                                                                     -1

                                       -0.2
                                                                                                                                                                                                                     -2

                                       -0.4
                                                                                                                                                                                                                     -3

                                       -0.6

                                                                                                                                                                                                                     -4

                                       -0.8

                                                                                                                                                                                                                     -5
                                       -1.0
                                              0   5   10    15   20    25   30        35   40        45   50    55   60    65   70    75   80   85   90   95   100   105   110
                                                                                                                                                                                                                     -6
                                                                                                           f req, GHz
                                                                                                                                                                                                                          0   10   20   30   40   50     60    70    80    90    100    110

                                                                                                                                                                                                                                                  freq, GHz




                                                  Measured and modeled s11                                                                                                                                           Measured and moded s12

                                                                                                                                                                                                                    0.4
                                      0.20
                                                                                                                                                                                                                    0.3
                                      0.18

                                      0.16                                                                                                                                                                          0.2

                                      0.14                                                                                                                                                                          0.1

                                      0.12
                                                                                                                                                                                                                    0.0
         imag(SP1.SP.S(3,4))
         imag(SP1.SP.S(1,2))
          real(SP1.SP.S(3,4))
          real(SP1.SP.S(1,2))




                                                                                                                                                                                       imag(SP1.SP.S(4,4))
                                                                                                                                                                                       imag(SP1.SP.S(2,2))
                                      0.10




                                                                                                                                                                                        real(SP1.SP.S(4,4))

                                                                                                                                                                                        real(SP1.SP.S(2,2))
                                                                                                                                                                                                               -0.1
                                      0.08
                                                                                                                                                                                                               -0.2
                                      0.06

                                      0.04                                                                                                                                                                     -0.3


                                      0.02                                                                                                                                                                     -0.4

                                      -0.00
                                                                                                                                                                                                               -0.5

                                      -0.02
                                                                                                                                                                                                               -0.6
                                      -0.04
                                                                                                                                                                                                               -0.7
                                      -0.06

                                      -0.08                                                                                                                                                                    -0.8
                                                                                                                                                                                                                          0   10   20   30   40   50      60    70    80    90    100     110
                                              0        10         20             30             40             50         60         70         80        90         100         110
                                                                                                                                                                                                                                                   freq, GHz
                                                                                                               freq, GHz




                           Measured and modeled s22                                                                                                                                                           Measured and modeled s21
CMC May 8, 2006                                                                                                                                                                                                                                                           UC Berkeley - 30
                        Sensitivity analysis

                                                   • Make Cgd as small as possible
      • Analyzing the effect of different            (MSG and Fmax)
        parasitics on device performances          • Make Rs as small as possible
        using small-signal model                     (MSG and Fmax)
                                                   • Reduce Rs/Rd to decrease
      • Analyzing trade-offs to optimize             noise and increase Fmax
        the layout
                                    |S |
                            MSG  21                        In conditionally
                                         | S12 |
                                                             stable region
                                                          Unilateral Gain
                      Rgate                                           Cgd



                                                   Stable Gain



CMC May 8, 2006                                                         UC Berkeley - 31
                  First and Second generation Devices

100 fingers, 1um/finger
  Version 1

                                                                    Uv2
                                                                          Uv2_predicted
           gate           drain
                                                        Uv1


                                                       MSGv2
                                  Version 2
                                                         MSGv2_predicted       MSGv1
                                  bridges

                                                  Ft~100 GHz (both)
                  gate    drain                   Fmax (1) ~ 145 GHz
                                                  Fmax (2) ~ 175 GHz
                                                  MSG ~ 7.6 dB (both)
                                              Modification of first generation transistor
                                                    using the sensitivity analysis
            small taper
CMC May 8, 2006                                                                     UC Berkeley - 32
                  3rd Generation ( Round-Table)
          Double contact
          Multi-path connections
          MSG 60GHz = 8.3dB
          fmax= 300 GHz (extrapolated), fT= 100 GHz
          Highest reported fmax/fT=3 ratio for CMOS!




CMC May 8, 2006                                         UC Berkeley - 33
                          Large Signal Modeling

   • Non-linear Circuits

   • Distortion analysis for amplifiers



           Extraction of            Modification of
          BSIM modelcard                BSIM




             DC Fitting             Adding parasitics




                                                        BSIM3v3 Modified Model
                          S-Param Fitting



CMC May 8, 2006                                                         UC Berkeley - 34
             BSIM Large Signal Modeling Results




                  Some BSIM3 parameters modified to get the fit
CMC May 8, 2006                                                   UC Berkeley - 35
                  Measurements above 40 GHz




         Measurement above 40 GHz difficult due to de-embedding
          issues (open/short does not work)
         TRL not good on Si since Z0 is not known
         Model increasingly important for predictive design above
          40 GHz
         Berkeley in process of upgrading measurement capability
          to 110 GHz
CMC May 8, 2006                                             UC Berkeley - 36
           Technology Driven Enhancements

          Body dependent GIDL model
                 Body node used for power control in digital circuits
                 GIDL leakage an important component of standby power
                 Important to model it correctly as a function of body-bias


          Flicker noise for small geometry devices
                 1/f dependence not observed (Telegrapher’s noise)
                 Need a model that can fit worse case noise from measurements


          High-K gate dielectrics/ metal gate transistors
                 Gate leakage, noise, and C-V



CMC May 8, 2006                                                                UC Berkeley - 37
                              TI’s Suggestions

          Fully Asymmetric Leakage Models
                 Implement independent source-side model
                 Add gate to source leakage model
                    Source-side equivalent of DLCIG, AIGSD, BIGSD, CIGSD

                 Add Gate Induced Source Leakage (GISL) model
                    Source-side equivalent of AIGDL, BIGDL, CIGDL, EGIDL

                 Improve Gate Edge Source Leakage (GESL) model
                    Drain/Source separation for NJTS[SW,SWG], TNJTS[SW,

                     SWG]
          TI can provide code support

          Impact Ionization Model
                 Re-open action item toward an improved II model


CMC May 8, 2006                                                     UC Berkeley - 38
                             Your Input?

          As always, the BSIM team is open to suggestions for
           future model enhancements
          Current SRC supported research will end this summer
          With the continued importance of CMOS technology,
           BSIM will continue to play an important role in the next
           three years

          Email: niknejad@berkeley.edu




CMC May 8, 2006                                               UC Berkeley - 39
                      Acknowledgements

          Weidong Liu (Synopsys)
          Keith Green, Claude Cirba (Texas Instruments)
          Josef Watts, Dawn Wang, Jack Pekarik (IBM)
          Geoffrey J. Coram (Analog Devices, Inc.)

          BWRC Member Companies
          Broadcom and Analog Devices for support
          ST Microelectronics, Infineon, Jazz for fabrication




CMC May 8, 2006                                                  UC Berkeley - 40

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:21
posted:9/29/2011
language:English
pages:40