Analog Power Proprietary Fast, Accurate Spice Model

Analog Power Inc.
SPICE MODELING Revised August 2015
Analog Power Spice Model Goals
•
To provide simple, fast, accurate SPICE model to be used predominantly for
switching speed and loss calculations
•
To be able to support model and its use
•
To allow temperature modeling, but not at the expense of speed and simplicity
•
To focus on the useful modeling parameters, and do not include breakdown
voltage etc as stresses can be checked via model probing
•
To support on low cost universal platforms, WinSpice, NGSPICE, LT Spice
•
To be able to use SPICE to select the optimum part for PWM applications
Not Intended for (significant added complexity/run time):
•
Modeling of breakdown voltage, high temperature leakage
•
Thermal modeling including RDS variation with temperature
– Model can be adjusted for temperature
•
Qrr modeling
Analog Power Confidential
Analog Power SPICE Subcircuit
Analog Power model R1.0:
•
Simple Level 1 model with 8
external components
•
Excellent fitting to data sheet
curves
•
All parameters are devicederived not process derived
•
Very simple to see effects of
each parameter
Analog Power Confidential
Analog Power SPICE Subcircuit 2.0
Improved implementation:
Non-Linear dependent voltage
sources used to set dv/dt over
diode and fixed capacitor =
zero appropriately , and
therefore isolate them when
not in circuit.
No switches used, but
identical concept
Analog Power Confidential
Analog Power Spice Example (switch, L1)
.SUBCKT AM90N08-04B 1 2 3
*Nom Temp=25 deg C
Mos1 4 5 6 6 APLMOS w=1 l=1
.MODEL APLMOS NMOS (Level=1 VTO=1.94 KP=116 CGDO = 600p CGSO = 9100p)
Rdrain 1 4 4.7e-3
Dbody 6 1 BodyDiode
.MODEL BodyDiode D (RS=2e-3 IS=2e-12 vj=0.6 RS=2m CJO=1000p M=0.67)
Lsource 3 6 3e-9
Rgate 2 5 1.8
Dcgd 5 7 Crss
.MODEL Crss D (RS=1e-3 CJO=10.9n M=1.18)
Sdiode 7 1 1 5 Sfb1
Scap 8 1 5 1 Sfb2
.MODEL Sfb1 VSWITCH (RON=100m ROFF=10e11 VON=0 VOFF=-0.01)
.MODEL Sfb2 VSWITCH (RON=100m ROFF=10e11 VON=0.01 VOFF=0)
C2 8 5 10.9n IC=0
R2 8 5 1e8
.ENDS
NOTE some Spice versions (E.G. NGSPICE) use different syntax for the switches:
Sdiode 7 1 1 5 Sfb1 ON
Scap 8 1 5 1 Sfb2 Off
.MODEL Sfb1 SW (RON=0.1 ROFF=10e11 VT=0 VH=-0.01)
.MODEL Sfb2 SW (RON=0.1 ROFF=10e11 VT=0.01 VH=0)
Analog Power Confidential
Analog Power Spice Example (no switch, L3)
* Analog Power Spice Level 3 Model - fixed temperature
* Rev 4/14/2014 PD
* Level 3 model is experimental, sub-threshold leakage is overestimated, which will affect VSD diode
modeling
.SUBCKT AM3446N 1 2 3
*Nom Temp=25 deg C
Mos1 4 5 6 6 APLMOS l=0.25u w=1
.MODEL APLMOS NMOS (LEVEL = 3 vto=0.91 KP =10u NSUB=2e+17 Kappa = 0.12 CGDO = 55p
CGSO = 336p)
Rdrain 1 4 28m
Dbody 6 1 BodyD
.MODEL BodyD D (IS=0.5e-12 RS=30e-3 CJO=10p M=0.2)
Lsource 3 6 0.3e-9
Rgate 2 5 3
Dcgd 5 7 Crss
.MODEL Crss D (RS=1e-3 CJO=400p M=0.91)
B1 7 1 V= (abs(v(5)-V(1))+(v(5)-V(1)))/2
C2 8 5 400p IC=0
R2 8 5 1e9
B2 1 8 V= (abs(v(1)-V(5))+(v(1)-V(5)))/2
.ENDS
Analog Power Confidential
Level 1 and Level 3 Models
• Many vendors add many external components and use more
complicated BSIM models, but do they work any better?
• Parameter-derived level 1 model is recognized as suitable for
discrete devices for modeling switching time etc
– more complicated models are intended for use in IC modeling
and are process-derived
• However level 1 model does not model RDS variation with VGS well
for modern devices and switching characteristics may suffer
– Use Analog Power Level 3 model instead
• Analog Power models are device derived, as opposed to process
and theory derived and therefore more accurate
• Rather than include temperature dependencies, our simple model
can be accurately tweaked for any temperate and this provides a
faster iteration to modeling at higher temperatures
Analog Power Confidential
Transfer Characteristics
•
Level 1 model uses simple equation:
ID α (VGS – VGS(TH))2
•
This model holds true for state of the art
Trench MOSFETs, especially surface mount
where current density is quite low
•
KP (Spice model parameter) easy to derive
from characterization data. L1 model gives
accurate transfer function, but inaccurate
RDS VS VGS plot
•
Level 3 model allows KP to reflect accurate
RDS VDS VGS yet still provides accurate
Transfer characteristics
Analog Power Confidential
CRSS – The key to accuracy
Three components:
•
CGDO – part of level 1 model
–
•
•
does not vary with VDS
Reverse biased diode
–
gives variation with VDS
–
removed when VGD>0
Fixed CDG2 capacitor
–
switched in when VGD>0
–
1.2 X value of diode at VDG = 0 to reflect
higher CRSS for +ve VGD
•
The combination of a diode with fixed
parallel capacitance gives fit with data
•
Alternate implementation with voltage
sources works in identical manner, but
use of switches is avoided – simplifies
syntax, and avoids SPICE transients
Analog Power Confidential
Comparison Data Sheet Vs. Model
25000
Capacitance (pF)
20000
15000
10000
5000
0
0
5
10
CISS
Data sheet
COSS
15
CRSS
20
VDS (V)
Model
Analog Power Confidential
Conclusion
Analog Power proposed SPICE subcircuit:
• Simple and therefore fast, no switches, no extra active components
• Device derived. Accurate in operation (switching, RDS)
• L3: Developed predominantly for optimum part selection in PWM
applications
– L1 most accurate for linear regulator type aplications
• Device-derived and therefore contains accurate parameters
• Easily adjusted to any operating temperature
• Allows what-if experiments, parameters are easily identifiable and editable
• L3 model includes accurate transfer AND RDS Vs VGS modeling
Analog Power Confidential