l1 mos model

Upload: tarang-mogre

Post on 06-Apr-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/2/2019 L1 MOS Model

    1/34

    MOS MODEL

  • 8/2/2019 L1 MOS Model

    2/34

    In this lecture, you will learn about

    Basic components that form MOS model.

    Basic equations that describe these functions andcomponents.

    Small signal model useful for analog design.

    Contact Information: A/P Ganesh Samudra, Tel: 6516-2293, email:

    [email protected] Room : E5 03-13.

    mailto:[email protected]:[email protected]
  • 8/2/2019 L1 MOS Model

    3/34

    S G

    N+

    D

    N+

    Schematic representation of MOSFET is shown below.The main purpose of MOS model to represent thecomplex MOS device with simple circuit elements likedependent sources, capacitors and resistors.

    Although we would like to see MOSFET as a gatevoltage controlled current source like below, unwantedparasitic devices/elements are always present as aconsequence of actual physical structure of MOSFET.

    MOS Model-Dependent Source

    Id Source

    We will now develop each of these elements.

  • 8/2/2019 L1 MOS Model

    4/34

    S G

    N+

    D

    N+

    The primary device functions through the gate

    capacitance. As gate overlaps with the bulk (also calledbody or substrate), the source and drain, this capacitancecan be distributed between all these 3 terminals givingfollowing representation for the gate capacitance.

    MOS Model-Gate Capacitance

    CGB

    CGDCGS

  • 8/2/2019 L1 MOS Model

    5/34

    S G

    N+

    D

    N+

    Along the width direction, S/D doped regions normally

    diffuse laterally under the gate which gives this overlapcapacitance. When positive voltage is applied to the gate,these N+ regions pileup (accumulate) negative chargedelectrons below the gate oxide giving overlap parasiticcapacitances as shown below.

    MOS Model-Overlap Capacitance

    C GSO C GSO CGDO

    Gate-S and D Overlap Regions

  • 8/2/2019 L1 MOS Model

    6/34

    S G

    N+

    D

    N+

    There is parasitic resistance from terminals of S/D

    (Source/Drain) to the inversion layer as current has toflow through doped regions. This is modeled as a normalresistor.

    MOS Model-S/D Resistance

    RD

    RS

  • 8/2/2019 L1 MOS Model

    7/34

    S G

    N+

    D

    N+

    N+ S/D regions form N+P junction with substrate due to

    vertical implant and diffusion of dopants. They form a junction diode and also have junction capacitancesassociated with them as shown below. The junctioncapacitance and diode current values are proportional toarea of the junction.

    MOS Model- Bottom Junction

    S-Bulkbottom

    JunctionCapacitance

    S-Bulkbottom

    JunctionDiode

    D-Bulkbottom

    JunctionCapacitance

    D-Bulkbottom

    JunctionDiode

    Vertical Implant/Diffusion

  • 8/2/2019 L1 MOS Model

    8/34

    S G

    N+

    D

    N+

    N+ S/D regions form N+P sidewall junction with substrate

    due to lateral encroachment of implant and diffusion ofdopants under the gate. They form a junction diode andalso have junction capacitances associated with them asshown below. The junction capacitance and diode currentvalues are proportional to perimeter of the junction.

    MOS Model-Sidewall Junction

    D SidewallJunction

    Diode and

    capacitance

    S SidewallJunction

    Diode and

    capacitance

  • 8/2/2019 L1 MOS Model

    9/34

    SPICE MODEL for MOSFET

    C GBO is in missing dimension shown on layout

    width edges due to birds beak encroachment.

    S G

    D

    C GSO C GDO

    N + N +

    SBdiode(genreversebiased)

    SBSide wallCap &diode

    DBSidewall

    diode andcapacitance

    C GB

    R S R D

    C GD C G S

    Noise Source

    I d Source

    Source toBulk (SB)bottomcapacitance

    DBdiode&bottomcap

    B

    Overall MOS Model

    Combining all aboveelements and adding anoise source in

    parallel, we have theintegrated model.

  • 8/2/2019 L1 MOS Model

    10/34

    S G

    N+

    D

    N+

    Now we will look at sample of detailed models for each of

    the elements described in the previous slides. The first element is the dependent current source

    between S and D. This models how I d depends on V DS ,VGS , VBS and the dependence is modified with short

    channel effects. All these parameters are grouped in thispart.

    MOS Model-Dependent Source

    Id Source

  • 8/2/2019 L1 MOS Model

    11/34

    * The following basic set of 6 parameters determines theproperties of the current source, I

    d.

    VTO : Zero bias, long channel threshold voltage.

    KP: called as transconductance parameter so that KP = COX(see below) and hence has units of A/V 2.

    GAMMA (): Bulk threshold parameter which specifiesdependence of V T on V BS . Called body effect.

    PHI : Surface potential. Mainly determined by substrate

    doping. LAMBADA (): Channel length modulation parameter.

    Models increase in drain current with increase in V DS in thesaturation region.

    MOS Model-Dependent Source

    f )(

  • 8/2/2019 L1 MOS Model

    12/34

    In this simple level 1 model that we will use,

    Subthreshold Region.

    Linear region

    saturation region.

    MOS Model-Dependent Source

    TGS D VVfor0I

    DS2DSDSTGSox D V1V21 -VV-VC

    LW

    I

    'DSDSTGS VV,VVfor

    DS 2TGSox D V1V-VC2LW

    I

    'DSDSTGS VV,VVfor

    .VV-VVwhere DSATTGS'DS

  • 8/2/2019 L1 MOS Model

    13/34

    Normally, a compact notation is used to express the currentwith .

    The actual threshold voltage is found with the followingexpressions.

    where D is drain induced barrier lowering parameter thatreduces V T at higher V DS for short channel devices and

    Vfb is the flat band voltage. V TO is the Zero bias, longchannel threshold voltage and PHI is the surfacepotential, mainly determined by substrate doping as onthe next slide.

    MOS Model-Dependent Source

    ox CL

    W

    V - V +2+2+V =V DS DSB f f fbT

    2+2+V =V f f fbT 0

    f )(

  • 8/2/2019 L1 MOS Model

    14/34

    The other quantities in the equation are VT = threshold voltage = Surface mobility of the carriers W = Width of the MOS device L = Length of the MOS device ND NA = Net doping concentration in the channel and ni = the intrinsic carrier concentration.

    MOS Model-Dependent Source

    and N-N

    lnq

    kT

    i

    AD

    n f

    ,C

    N-Nq2 )(

    OX

    ADsi parameter effect body

    .thicknessoxidebeingd ,d

    C oxox

    oxOX

  • 8/2/2019 L1 MOS Model

    15/34

    Clearly, change in the threshold voltage due to bodyeffect or substrate bias effect is given by

    Note that the p-substrate of n-channel devices is alwaysgrounded. The n-substrate of p-channel devices isgenerally tied to the power supply voltage V DD or couldalso be shorted to the source to eliminate body effectcompletely.

    If any of these parameters are not specified, they arecalculated with device physics equations and some otherparameters related to process such as d OX, oxidethickness for gate, N Sub , substrate doping, NSS surfacestate density, NFS fast surface state density, etc.

    MOS Model-Dependent Source

    ][ 2 V +2 =V f SB f T

  • 8/2/2019 L1 MOS Model

    16/34

    The characteristics for N channel device using equationsabove are as shown. They help determine by extendingthe saturation region I

    Dto negative V

    DSaxis.

    MOS Model-Dependent Source

    In addition, the velocity saturation effect can be includedby adding a critical field parameter or reducing saturationcurrent exponent below 2 and reducing the saturationvoltage V DS . Higher level models in circuit simulators likeSPICE have much larger set of parameters for accuracy.

    VGS = V T + 1

    VGS = V T + 2

    VGS = V T + 3

    VDS

    ID

    1/ Figure : I D vs V DS characteristics

  • 8/2/2019 L1 MOS Model

    17/34

    S G

    N+

    D

    N+

    These are modeled through 3 non-linear bias dependent

    capacitances C GS , C GD, C GB defined by piecewise linearMeyers model. There are other advanced modelsavailable for more accuracy and at higher frequency.

    MOS Model-Gate Capacitance

    CGB

    CGDCGS

  • 8/2/2019 L1 MOS Model

    18/34

    Meyer model for gate capacitance.

    Linear Region - strong inversion

    MOS Model-Gate Capacitance

    0=C GB

    ])2V T 2-V GS+V GD(

    )2V T -V GS( -[1WLC ox32 =C GD

    ])2V T 2-V GS+V GD(

    )2V T -V GD( -[1WLC ox32

    GS =C

    Saturation Region -strong inversion

    C GB=0=C GD

    WLC ox32

    GS =C

    Weak inversion Ignored in the simple model.

    Why these both are zero?

  • 8/2/2019 L1 MOS Model

    19/34

    S G

    N+

    D

    N+

    there are three overlap capacitances.

    CGSO: Gate to source overlap capacitance per meter ofchannel width. This and drain overlap capacitance(CGDO) are independent of length L as they areassociated with edge effects along the width.

    MOS Model-Overlap Capacitance

    C GSO C GSO CGDO

    Gate-S and D Overlap Regions

  • 8/2/2019 L1 MOS Model

    20/34

    CGBO: Gate to bulk overlap capacitance per meter ofchannel length and is associated with edge effects at thewidth ends of the gate area. This capacitance is not seenin the slides as it is in the third dimension missing in ourfigures.

    MOS Model-Overlap Capacitance

  • 8/2/2019 L1 MOS Model

    21/34

    S G

    N+

    D

    N+

    RS and R D are the source and drain resistances due towhich inside source and drain voltages applied to actual

    inversion layer will change as there will be voltage drops.They are specified normally as an effective sheetresistance of S/D regions and found using number ofsquares in S/D regions.

    MOS Model-S/D Resistance

    RDRS

    How to find number of squares in S/D regions?

  • 8/2/2019 L1 MOS Model

    22/34

    S G

    N+

    D

    N+

    The source to substrate and drain to substrate bottom junction parasitic diodes are modeled with ( Expectation?)

    MOS Model- Bottom Junction

    S-Bulkbottom

    JunctionCapacitance

    S-Bulkbottom

    JunctionDiode

    D-Bulkbottom

    JunctionCapacitance

    D-Bulkbottom

    JunctionDiode

    1eAII kTdiodeqV

    sdiode

    where I S is reverse saturationcurrent per unit area modelparameter and A is the junctionarea. Since junctions are reversedbiased in MOS, I diode -AI s.

  • 8/2/2019 L1 MOS Model

    23/34

    Although these S-D region doping profiles are quite

    complex, the junction capacitance is calculated usingsimple abrupt junction capacitance formula fromEE2004/EE2021 for simplicity. Since source draindoping profiles are normally the same, the bottom

    capacitances are modeled with CJ: Zero bias bulk junction bottom capacitance per unit

    area

    MJ: Bulk junction bottom grading coefficient

    PB: Bulk junction built in potential

    MOS Model- Bottom Junction

  • 8/2/2019 L1 MOS Model

    24/34

    The model is based on the same equation with step junction and depletion approximation in EE2004/EE2021so that

    MOS Model- Bottom Junction

    MJ

    BSBD

    SD

    PB)V(V

    -1

    CJ)(AA(Source)drainateCapacitanc

    layout.fromce)drain(Sourtheof areatheis)(AAwhere SD

    Here, the expressions in parenthesis () apply to sourceand the expression replaces the drain term just precedingthe parenthesis.

  • 8/2/2019 L1 MOS Model

    25/34

    S G

    N+

    D

    N+

    The source to substrate and drain to substrate sidewall junction parasitic diodes are modeled with

    MOS Model-Sidewall Junction

    D SidewallJunction

    Diode and

    capacitance

    S SidewallJunction

    Diode and

    capacitance

    1ePII kTdiodeqV

    sdiode

    where I S is reverse saturation currentper unit length model parameter and Pis the junction perimeter. Since

    junctions are reversed biased in MOS,Idiode -PI s.

  • 8/2/2019 L1 MOS Model

    26/34

    Although these S-D region sidewall (lateral) doping

    profiles are quite complex, the junction capacitance iscalculated using simple abrupt junction capacitanceformula from EE2004/EE2021 for simplicity. Since source

    drain sidewall doping profiles are normally the same,

    the sidewall capacitances are modeled with CJSW: Zero bias bulk junction bottom capacitance per

    unit length

    MJSW: Bulk junction bottom grading coefficient

    PB: Bulk junction built in potential taken same as that ofthe bottom junction.

    MOS Model- Sidewall Junction

  • 8/2/2019 L1 MOS Model

    27/34

    The model is based on the same equation with step junction and depletion approximation in EE2004/EE2021so that

    MOS Model- Sidewall Junction

    MJSWBSBD

    SD

    PB

    )V(V -1

    CJSW)(PP(Source)drainateCapacitancSidewall

    layout.fromce)drain(Sourtheof perimetertheis)(PPwhere SD

    Here, the expressions in parenthesis () apply to sourceand the expression replaces the drain term just precedingthe parenthesis.

  • 8/2/2019 L1 MOS Model

    28/34

    In addition, there are noise coefficients which may addsome noise to I D. As one can see, this model ignoresmany parasitic resistances, bipolar devices and PNPNdevices.

    This completes the discussion on accurate modeling ofMOSFET.

    MOS Model

  • 8/2/2019 L1 MOS Model

    29/34

    The model shown earlier is good for circuit simulation, buttoo complex for simple hand calculations of gain, cornerfrequency, output resistance, etc. For this purpose,simpler small signal model is developed in this part.

    In this analysis, a small AC signal v is superimposed onnormal DC biases. In the notations, AC signals willnormally use all small letters, DC all capitals and totalvoltage with AC and DC components added will usemixed case. For example, V OUT is a DC voltage, v out ACvoltage and V out is the total voltage (this is normally usedin large signal response which is determined by the totalvoltage).

    Lumped Small Signal (AC) Model for MOSFET

  • 8/2/2019 L1 MOS Model

    30/34

    The effect of gate voltage is modeled usingtransconductance

    Hence this gm

    models how drain current will change withchange in gate voltage (which is the AC voltage applied tothe gate). Normally, the input voltage applied to the gatechanges I D and hence changes output voltage, normallytaken at the drain.

    Lumped Small Signal (AC) Model for MOSFET

    .GSVDI mg

    This effect is captured as adependent current sourcecircuit element between drainand source.

    gmVgs

    D

    S

  • 8/2/2019 L1 MOS Model

    31/34

    In the linear region,

    by taking partial derivative of the model equation in thelinear region. Hence there is a drop with decreasing drainto source voltage.

    In the saturation region where many analog amplifiersoperate,

    Lumped Small Signal (AC) Model for MOSFET

    .V1VCLWg DSDSox m

    .V-V

    I2V1V-VCLWg

    TGS

    DDS TGSox m

  • 8/2/2019 L1 MOS Model

    32/34

    The effect of the drain voltage is modeled using the drainconductance

    This models how I D changes when the output voltagenormally taken at the drain changes. Since we want onlyinput voltage to affect I D and not the output voltage,ideally, this should be zero.

    Hence in saturation region using I D expression in the

    model

    This effect is captured as a resistor circuit elementbetween drain and source.

    Lumped Small Signal (AC) Model for MOSFET

    .DSVDI dg

    D2TGSox d IV-VC2LW

    g

  • 8/2/2019 L1 MOS Model

    33/34

    The effect of body bias or substrate voltage is modeledusing transconductance

    by taking partial derivative of the model equation.

    Hence this value is mainly determined by as thedependence is through changes in V T. This effect iscaptured as a dependent current source circuit element

    between drain and source.

    Lumped Small Signal (AC) Model for MOSFET

    ,BSVDI mbg

    All these three componentsare heavily dependent onthe models used.

    gmVsb

    D

    S

  • 8/2/2019 L1 MOS Model

    34/34

    The complete AC representation for nMOS device iscaptured in the following figure. We add all capacitors thatare in parallel to lump them together as though they areconstant mainly determined by DC voltages (Which onceare combined in each one?). For analytic calculations,this is adequate. Circuit simulators like SPICE can take

    care of actual voltage dependencies.

    Lumped Small Signal (AC) Model for MOSFET

    G

    C gb

    C gs

    C sb

    S

    g mv gs g mbv sbr 0 = . 1

    g d

    D

    C db

    C gd