Mextram 504 parameter extraction

Download Report

Transcript Mextram 504 parameter extraction

Mextram 504 parameter extraction
F. Yuan
Advisor: Prof. C. W. Liu
Graduate Institute of Electronics Engineering
and Department of Electrical Engineering,
National Taiwan University, Taipei, Taiwan
Extraction strategy
 SiGe parameters
 Low current parameters
 Temperature parameters
 S-parameter measurement (only for fT)
 High current parameters
 1/f noise parameters
Lay-out structure
 In S-parameter measurement, E-S are
connected and grounded to increase S/N ratio
 Early-measurement can be done on two
layout structures and used as a double check
normal DC
High frequency (S-parameters)
Initial guess
 We use the process parameters to set the
initial parameters
 All this guess is physically based
SiGe parameters
 Must be extracted first
 Linear graded Ge-profile
 Extract dEg
 Extract Xrec by IB-VCB plot at low VCB
dI
g

(No avalanche)
dV
C
out , f
VBE VF
CB

VtC 
I B1  I B 0 1  X rec 1  XI B1    I avl
Vef 

g out ,r 
dIE
dVEB
VBC VF
 g
I
C
dEg  kT  ln out ,r  2 ln E ,r  ln BE ,r
 g
IC, f
C BC , f
out , f





Low current parameters
 Depletion, overlap capacitances
 Early effect
 BC avalanche
 Transfer current
 BE base current
 series resistances
 PNP parasitic transistor
BE depletion capacitance
 LCR-meter or S-parameter
 VC=VE=VS=0, change VB but in low bias
 In low bias, we can assume the voltage drop
on all parasitic resistors are zero
 Smoothing factor ajE can not be extracted
C BE 
C jE
 vBE 
1 

 VdE 
pE
 C BEO
BC, SC depletion capacitance
 VB=VE=VS=0, change VC but in low bias
 Xp is reach-through parameter
 Cp,CS is the parasitic SC capacitance (not
modeled in Mextram 504)
CBC

1  X C

CSC 
p
 vBC 
1 

 VdC 
C jS
 vSC 
1 

 VdS 
jC
pC
pS
 X p C jC  CBCO
 C p ,CS
BC avalanche
 Iavl related to IC, VCB
 Measured at low VBE and high VCB
I B  I B 0  I avl
Wavl ,Vavl
Early effect
 Low bias, no BE tunneling breakdown
 Calculate Ver, then Vef, and repeat again
 Use the previous parameters in extraction
 Early voltage may be not normal if dEg is not
zero
V
1
V
1  tC
Vef
I E  IE0
V
V
1  tE  tC
Ver Vef
tE
IC  IC 0
Ver
 I avl
VtE VtC
1

Ver Vef
VtE VdE , pE ,VBE , a jE 
VtC VdC , pC ,VBC , a jC , X p 
Saturation current IS
 Measured in forward-Gummel (IC-VBE,
VBC=0V) at low VBE
IC 
VBE
VT
ISe
V
V
1  tE  tC
Ver Vef
Base current
 Measured in forward-Gummel (hfe-VBE,
VBC=0V) at low VBE
I B1 
IS
f
 B2E1
(e
VT
 1)
 B2E1
I B 2  I Bf (e
h fe 
mLf VT
IC
I B1  I B 2
 1)
series resistance
 For RE, measured VCE-IE at IC=0 and high VBE
 For RB, measured VBE-IB at IC=0 and high VBE
 Voff,Rb is only for optimization
 VB2E1 can be obtained by IB1
VCE  I E RE  vB2 E1  vB2C1
RE 
VCE
I E
VBE  vB2 E1  vB1B2  I B RBc  I C  I B RE  Voff , Rb
vB1B2
 RBv I C 1  XI B1  

 VT ln1 


 f VT


Substrate saturation current ISS
 Measured in reverse-Gummel (hfc,sub-VBC,
VBE=0V) at low VBC
 Iks is the point that hfc,sub becomes large
I sub  I Ss e
IE 
VBC
VT
VBC
VT
ISe
V
V
1  tE  tC
Ver Vef
h fc, sub 
IE

I sub
IS
 V
V 
I Ss 1  tE  tC 
 V
Vef 
er

series resistance
 Measured IX-VBE at VBC=0.6V
 From IC, IB, Isub, and RBc, the RCc is given
 B1C1
I sub 
2 I ss (e
VT
1 1 4
 1)
IS
e
I kS
 B1C1
VT
 B C  VBC  RCc I C  RBc I B
1 1
Parasitic PNP
 Measured in reverse-Gummel (hfc-VBC,
VBE=0V) at low VBC   V  V ln I 1  V
E
B1C1
I ex 
BC
IS
 ri
T
 IS 
 
 B1C1
e
VT
 B1C1
I B 3  I Br
e
VT
 B1C1
1
VLr
e 2VT  e 2VT
IE
IE
h fc 

I B  I sub
I ex  I B 3
VtC  

Ver Vef  
tE
Temperature parameters
 We don’t extract AE, AB, Aepi, Aex, AC, AS
 The other 8 parameters should be extracted
  T A
AQB 0
Ver ,Vef
Vg B
IS
Vg C
I Br , C jC
Vg j
I Bf
Vg S
I SS
dVg  f
f
dVg 
 ri
dVg
r
E
E
Temperature parameters
 Extract all low current parameters at
reference temperature
 Extract temperature parameters at elevated
temperature (temperature independence
parameters will set the same as ref temp)
 Extract high current parameters
High current parameters
 Self-heating
 Output-characteristics
 Cutoff frequency
 Quasi-saturation
Self-heating
 IB fixed, increase VCE to self-heating
 Then IS will increase with temperature
 VBE will now decrease by fixed IB
 No avalanche (Iavl << IB)
T  Rth I BVBE  I CVCE 
 No hard saturation
v
I S (T ) V
IB 
e
(Iex, Isub << IB)
 f (T )
B2E1
T
VBE  vB2 E1  I C RE (T )
Output characteristics
I B  I B1 (vB2 E1 )  I B 2 (vB2 E1 )
vB2C1  vB2 E1  I C RCcT  I B  I C RET
v*B2C2
v B2E1
I C1C2  I N  I ST
e
VT
e
qB
vB2 E1 , vB2C1 , I C1C2 , vB* 2C2
VT
 IB fixed,
VCE=0…VCB,max+1
 Solve the voltage of all
 V internal nodes
 No hard saturation
(Iex, Isub << IB)
CE
Cutoff frequency
 S-parameters are measured
 fT measurement is very easy
 The fT value from extracted parameters
should be compared to the measured value
h fe  h21 
Y21
Y11
Cutoff frequency
 VB2E1, VB2C1, IC1C2
 Small signal variable (VCE constant)
v
v
dv

dv

dv
 Solve two equation
v
v
 Calculate dQ and fT dI  I dv  I dv
CE
CE
CE
B2 E1
B2 E1
B2C1
N
N
dQ 
vB2 E1
Q
Q
Q
dvB2 E1 
dvB2C1 
dIC1C2
vB2 E1
vB2C1
I C1C2
1
dQ
 T 
2fT
dIC
VCE 
B2C1

N
B2 E1
vB2C1
dvB2 E1 dvB2C1
,
dIC1C2 dIC1C2
Q dvB2 E1
Q dvB2C1
Q


vB2 E1 dIC1C2 vB2C1 dIC1C2 I C1C2
B2C1

vCE
dIC1C2  0
I C1C2
I N
dIC1C2  dIC1C2
I C1C2
Cutoff frequency
 For some parameters in Q , we use the initial
guess only
 Only  E left for extraction
Ver 1  XC jE C jE
 No avalanche (Iavl << IB)  B 
Ik
V
2
 No hard saturation
I I R
 epi   B S k 2Cv e V
4VT
(Iex, Isub << IB)
dC
T
1  XC jC
 R   B   epi 
XC jC
Quasi-saturation
 External VCB > 0, but Internal VCB < 0
VCB,in  VCB,out  IC RC
Quasi-saturation
 Extract Ik (hfe-VBE at high VCE)
 Extract Rcv (forward-Gummel)
 Repeat first two procedures
 Extract  E at maximum fT and high VCB
 Extract  epi , I hc , SCRcv at fT roll-off
 axi , m ,VdC can be used as fitting parameters
 Repeat the whole loop again (and again)
Avalanche at high current
 Extract SFh


I B  I B1 (vB2 E1 )  I B 2 (vB2 E1 )  Gem vB2 E1 , I C1C2  I C1C2


I N  I C1C2 1  Gem vB2 E1 , I C1C2

Xext
vB2 E1  VBE  I N RET  I ex  I B 3  I sub RBCT
v B2C2
I N  I ST


e
VT
qB vB2 E1 , vB2C2


I C1C2 vB2C1 , vB2C2  I N
vB2C1  vB1C1
VBC  vB1C1  I ex  I B 3  I sub RBCT
 I ex  I B 3  I N  XIex RCCT
IE  IN
I B  I ex  XIex  I sub  XIsub  I B 3
I sub ,ext   I sub  XIsub
 We can take IB,IE-VBC
(reverse-Gummel)
 VB2E1, VB2C2, VB2C1,
VB1C1, Xext
 Check IE, IB, Isub,ext
Y-parameters
Y21
Y11
 Only for fT measuring
 We can also use it to double check the
extracted parameters (at small current)
h fe  h21 
g m rB
 1
Y11 
gm


 j Cex  Cin  C BE 
Y21  g m  j Cex  Cin 1  g m rB   C BE g m rB 
Y12   2 rB Cin Cin  C BE   j Cex  Cin 
Y22  g out  j Cex  Cin 1  g m rB   CSC 
1/f noise parameters
 SIB is base current noise spectral density
S IB ,1Hz  S IB  f
S IB 
Af
2
i
I
 K f b , b 1
f
f
S IB ,1Hz  K f I B f
A
logS IB ,1Hz   logK f   A f logI B 
Thermal capacitance
 Can be measured in time-domain
 RthCth will be the order of 1us
Geometric scaling
 There is a primitive scaling rules in Mextram
504 parameter extraction
 If the devices have different B, E fingers or
different contacts, the model may inaccurate
 High current parameters can be more easily to
extract