© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 1
MOS-‐AK Mee;ng, Technical University of Munich Munich, April 11-‐12, 2013
Compact Modelling of HVMOSFETs
MaKhias Bucher12, Nikos Mavredakis1 Antonios Bazigos2, François Krummenacher2, Jean-‐Michel Sallese2
1Technical University of Crete 2École Polytechnique Fédérale de Lausanne
COMON project – HVMOS modelling WG
● Development / Implementa;on of a full compact model for High-‐Voltage MOSFET
● Key partners ○ EPFL: EPFL-‐HV MOSFET compact model
○ ams: technology supplier/model user
○ TUC: 1/f noise model development
○ AdMOS: 1/f noise measurement equipment
○ Dolphin Integra;on: Verilog-‐A code compiler, simulator vendor (SMASH)
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 2
COMON project – HVMOS publica;ons ● Modelling of the dria region of the HVMOSFET
○ A. Bazigos, F. Krummenacher, J.-‐M. Sallese, M. Bucher, E. Seebacher, W. Posch, K. Molnar, M. Tang, “A Physics-‐Based Analy1cal Compact Model for the Dri; Region of the HV-‐MOSFET”, IEEE Trans. on Electron Devices, Vol. 58, N° 6, pp. 1710-‐1721, June 2011.
● RF modelling of the HVMOSFET
○ A. Bazigos, F. Krummenacher, J.-‐M. Sallese, M. Bucher, E. Seebacher, W. Posch, K. Molnar, M. Tang, “RF Compact Modeling of High-‐voltage MOSFETs”, IETE Journal of Research, Vol. 58, N° 3, p. 214-‐221, May-‐June 2012.
● 1/f noise modelling of the HVMOSFET
○ N. Mavredakis, M. Bucher, R. Friedrich, A. Bazigos, F. Krummenacher, J.-‐M. Sallese, T. Gnei;ng, W. Pflanzl, E. Seebacher, “Measurement and Compact Modeling of 1/f Noise in HV-‐MOSFETs“, IEEE Trans. Electron Devices, Vol. 60, N° 2, pp. 670-‐676, Feb. 2013.
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 3
Outline
● Development of an HV-‐MOS compact model o Dria region accoun;ng for velocity satura;on
o TCAD verifica;ons
● Extensions for RF
● Extensions for 1/f noise o Verified the model against measurements on many HVMOS technologies from ams
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 4
high-‐voltage part dria region
n
low-‐voltage part inner MOS p
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich
● HV-‐MOS: 2 parts ○ Connec;ng point: ‘K’ ○ Doping changes type ○ Between channel and dria
● Low-‐voltage part [1] ○ inner MOSFET
● inner drain is ‘K’
● High-‐voltage part [2] ○ Dria region
● A physics-‐based model
Substrate
Source
Gate Drain K
HV-‐MOS macromodel
LDMOS TCAD: 2D doping profile
[1] Y. Chauhan, F. Krummenacher, R. Gillon, B. Bakeroot, M. Declercq, and A. Ionescu, “Compact Modeling of Lateral Nonuniform Doping in High-‐Voltage MOSFETs,” IEEE Trans. Electron Devices, vol. 54, No. 6, pp. 1527–1539, June 2007.
[2] A. Bazigos, F. Krummenacher, J.-‐M. Sallese, M. Bucher, E. Seebacher, W. Posch, K. Molnár and M. Tang, “A physics-‐based analy1cal compact model for the Dri; region of the HV-‐MOSFET” IEEE Trans. Electron Devices, Vol. 58, No. 6, pp. 1710-‐1721, June 2011.
Typical structure of HV-‐MOS
5
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich
● HV-‐MOS: 2D problem ○ Split into 2 axes
○ x-‐axis: inner MOS ● Source to ‘K’
○ y-‐axis: DRIFT region ● ‘K’ to Drain
● Simplified HV-‐MOS: two 1D problems ○ Solu;on possible
high-‐voltage part dria region
low-‐voltage part inner MOS
From “1×2D” to “2×1D”
6
© M. Bucher 3
MOS-‐AK, 11-‐12 April 2013, TU Munich
Model vs. TCAD simula;ons: ID vs. VG
Device: LDMOS L = 500nm W = 1um Tox = 15nm VDS = {0.1, 1.0, 5.0, 70.0}V VSB = 0V
7
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich
Model vs. TCAD simula;ons: ID vs. VD
Device: LDMOS L = 500nm W = 1um Tox = 15nm VGS = {1.0, 2.0, 5.0}V VSB = 0V
8
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich
DC measurements: ID vs. VG,VDS=0.1V
Device: LDMOS L = 400nm W = 40um Tox = 15nm VDS = 100mV VSB = {0.0, 0.4, 0.8, 1.2, 1.6}V
9
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich
DC measurements: ID vs. VG,VDS=35V
Device: LDMOS L = 400nm W = 40um Tox = 15nm VDS = 35V VSB = {0.0, 0.4, 0.8, 1.2, 1.6}V
10
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich
DC measurements: ID vs. VD
Device: LDMOS L = 400nm W = 40um Tox = 15nm VGS = {1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0, 5.5}V VSB = 0.0V
11
Dynamic MacroModel Extrinsic parasi;cs
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 12
Scalable Gate Resistance
Scalable, Bias-‐Independent Overlap Capacitances
Asymmetrical Junc;on Diodes
RF measurements Y-‐Parameters, real / imaginary part
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 13
Device: LDMOS L = 500nm W = 40um Tox = 15nm VDS = {0.0, 30.0}V VSB = 0.0V F = 1.2GHz
A. Bazigos, F. Krummenacher, J.-‐M. Sallese, M. Bucher, E. Seebacher, W. Posch, K. Molnar, M. Tang, “RF Compact Modeling of High-‐Voltage MOSFETs”, IETE Journal of Research, Volume 58, Nr. 3, pp. 214-‐221, 2012.
RF parameter extrac;on
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 14
Device: LDMOS L = 500nm W = 40um Tox = 15nm VDS = {0.0, 30.0}V VSB = 0.0V F = 1.2GHz
11G 2
11
Re(Y )RIm(Y )
=
RG on/off
11GG
Im(Y )C2 f
=π
12GD
Im(Y )C2 f
=π
21DG
Im(Y )C2 f
=π
22DD
Im(Y )C2 f
=π
● Low frequency noise under high-‐voltage condi;ons: largely unexplored. Does the dria region contribute to LFN? ○ Usual approach: use LFN models of ‘LV’ part of HVMOSFET models, adapted to LFN data under “LV” condi;ons.
● Difficulty of measuring LFN under high VDS – no adequate equipment (typically <5V)
● New measurement setup developed by AdMOS: can measure LFN of HVMOSFETs up to 200V!
1/f noise in HV-‐MOSFETs: what happens??
© M. Bucher 15 MOS-‐AK, 11-‐12 April 2013, TU Munich
○ N. Mavredakis, M. Bucher, R. Friedrich, A. Bazigos, F. Krummenacher, J.-M. Sallese, T. Gneiting, W. Pflanzl, E. Seebacher, “Measurement and Compact Modeling of 1/f Noise in HV-MOSFETs“, IEEE Trans. Electron Devices, Vol. 60, N° 2, pp. 670-676, Feb. 2013.
1/f Noise in HV-‐MOSFETs
• Noise spectra measured at high VDS = 50V o long/short, HVNMOS, HVPMOS – 20V, 50V HVMOS devices
• Noise in LV part is affected by Dria region: e.g. due to quasi-‐satura;on effect changing VK
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 16
● Total LFN is the sum of LV part LFN, plus Dria part LFN
● Combina;on of Number (ΔΝ) and mobility (Δμ) fluctua;on
Low frequency noise model in HV-‐MOS
© M. Bucher 17 MOS-‐AK, 11-‐12 April 2013, TU Munich
2 2 2 2
2 2 2 2
_ _ _
D D D DI I I I
D D D DLV LV DRIFT
S S S S
I I I Iµ
Δ Δ Δ Δ
ΔΝ Δ ΔΝ
= + +
○ N. Mavredakis, M. Bucher, R. Friedrich, A. Bazigos, F. Krummenacher, J.-M. Sallese, T. Gneiting, W. Pflanzl, E. Seebacher, “Measurement and Compact Modeling of 1/f Noise in HV-MOSFETs“, IEEE Trans. Electron Devices, Vol. 60, N° 2, pp. 670-676, Feb. 2013.
Due to LV channel (ΔN and Δµ) Due to Drift region (ΔN)
Low frequency noise model for HV-‐MOS
© M. Bucher 18 MOS-‐AK, 11-‐12 April 2013, TU Munich
( )( )
24
2 2 2
_
2
220
0.51 2ln2 1 21 0.5
2
DI t
D oxLV
Cs d c
d c
Cd s kC s k k d c
S q NI kTWLn C f
q iii q qq q q i
λλ
λ
λ
λαµ αµ
λλ
Δ
ΔΝ
=
⎡ ⎤⎛ ⎞+ −⎢ ⎥⎜ ⎟ ⎛ ⎞+ +⎢ ⎥⎜ ⎟ ⎜ ⎟+ + ⎝ ⎠+ −⎢ ⎥⎜ ⎟+ −⎢ ⎥⎝ ⎠⎣ ⎦
( ) ( )2 2
0 0,
1 ( ) c
s s k kd c d d c
c s k
q q q qi i i
q qλ λ λλ =
+ − += =
+ −
( )2 T
CCN eff clm
UE L L
λ =−Δ
Carrier Number Fluctuations (LV)
( ) ( )( )
2
2
_
ln /1 1
2DI s kH
s kD ox s kLV
S q qk na q qI WLC f q q
µ
Δ
Δ
⎡ ⎤Τ= + + +⎢ ⎥
−⎣ ⎦Mobility
Fluctuations (LV)
(Dependent on qs, qk)
( )2
4_
2 2
_
1 ln 1 22
DI t DRIFTk
D OVD ox driftN DRIFT
S q Nq
I kTWL C f iλΔ
Δ
= ⋅ +
(Dependent only on qk)
Carrier Number Fluctuations (Drift)
1/f Noise in HV-‐MOS vs. bias – long channel
1/f noise vs. ID for 50V HVNMOS.
W=40um, L=10um.
VDS = 50mV, 50V.
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 19
SΔVG SΔID
G
S B
D
S B
D
G
SΔID = gm2 SΔVG
1/f Noise in HV-‐MOS vs. bias – short channel
1/f noise vs. ID for 50V HVNMOS.
W=40um, L=0.5um.
VDS = 50mV, 20V, 50V.
Single model @all geometries, bias.
Comparable results for HVPMOS. © M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 20
LFN model vs. bias and geometry
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 21
40um
/0.5
um
HV
NM
OS
(50V
)
40u
m/1
0um
40um/10um
HV
PM
OS
(50V) 40um
/1um
Analysis of different LFN contribu;ons • Different contribu;ons to LFN.
• Dominant LV channel noise:
o ΔN at high VG, for short devices (small geometry), and @ high VD (high gm)
o Δμ @ low VG
• Dria region LFN becomes significant:
o For long channel (large geometry channel)...
o .... and @ low VD, hence the dria region LFN becomes significant w.r.t. the channel noise.
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 22
● A complete HVMOS compact model has been developed. o Physics-‐based model for the dria region.
o Combined with a charge-‐model approach for the LV part of the device.
o Covers all essen;al effects in HVMOS.
● Extensions for RF have been demonstrated.
● LFN measurements under high-‐voltage condi;ons. o Dria region contributes to LFN: par;cularly long-‐channel, low VDS.
o First complete LFN compact model with contribu;on from dria region.
Conclusions
© M. Bucher 23 MOS-‐AK, 11-‐12 April 2013, TU Munich
Thank you for your aKen;on
● Special thanks to:
© M. Bucher MOS-‐AK, 11-‐12 April 2013, TU Munich 24