Transcript www.nxp.com

NXP Power MOSFET spice models
Quick introduction
Phil Ellis
April 2015
Introduction
NXP power MOSFET spice “models” attempt to give an accurate
representation of a typical device at 25°C for key static and dynamic
characteristics.
The “models” are actually SPICE subcircuits and contain the Berkeley
SPICE level 3 semi-empirical .MOSFET model and the SPICE .DIODE
model. The subcircuit contains additional SPICE components to
represent device non linear capacitances and package parasitics.
2
Key points
Power MOSFET models attempt to accurately simulate
–
–
–
–
–
Rdson considering Id, Vgs and also global .TEMP (default is 25°C)
Transfer characteristic (Id vs Vgs characteristic) @ 25°C
Diode forward characteristic @ 25°C
Capacitances (Ciss, Coss, Crss) @ 25°C
Gate charge characteristic (Qg vs Vgs considering Id, Vds) @25°C
NXP also create RC thermal models (in Foster network format) to
simulate the relationship between Tj and Tmb
NXP also provides some LTspice VDMOS models although these don’t
model package parasitics but can run faster if a small loss in accuracy
can be tolerated
The following limitations apply
– We generally don’t have thermally corrected models (Tj affects Vgs
threshold, Rdson, diode characteristics, transfer characteristic)
– Qrr is poorly modelled by the standard spice diode model
– The spice mosfet model is only valid at 25°C
– Applies to a typical device (not worst case)
3
Gate charge characteristics
Vgs vs Qg simulated compared to measured for a given sample
Modelled using capacitance measurements
4
Transfer characteristics
Transfer curve: simulated vs measured
Output characteristics follow from this
5
Other characteristics modelled
Rdson vs Vgs
Diode forward characteristics
Note that curves are from a single typical device measurement.
Datasheet values may be from averages of large batches
There can be significant variation between parts due to manufacturing
variation, this is indicated in the datasheet.
The devices used to make the datasheet can be different from the
devices used to make the spice model, so the mosfet model might not
exactly match the datasheet
6
Switching characteristics
Example of real device switching vs spice simulation. Switch off
simulation agrees with real circuit quite well but switch on is not so
good due to the poor modelling of reverse recovery in the diode
(SPICE program doesn’t account for this properly). There are
modifications to the subcircuit to more accurately model this behaviour.
Switching characteristics don’t change too much with temperature
(increase by approx. 15% at 150 for power MOSFETs since the
capacitances are dominated by gate oxide thickness however the
depeletion (drift) region behaviour is temperature dependant.
Vgsth is strongly temperature dependant and this will affect
performance (Miller plateau voltage will influence gate drive current,
this tracks Vgsth)
7
Switching: simulation compared to actual
Actual: MOSFET switch off
Salmon = High side Vds 10V/div
Green= Low side Vds 10V/div
Blue = Low side Id 20A/ div
Horizontal 50ns/division
Simulation
Pink = High side Vds 10V/div
Green= Low side Vds 10V/div
Blue = Low side Id 20A/ div
Horizontal 50ns/division
Note Id inverted
MOSFET Switch on is less accurate but still useful
8
Usage of NXP power MOSFET subcircuits
Can be used to calculate switching and conduction loss in pwm
applications, particularly if the load current varies such as in motor
control.
Temperature compensation must be applied in order to give best
accuracy at other temperatures
Particularly useful in determining optimum performance when trying to
decide which device to use rather than determining the exact losses
For ultimate accuracy, the simulation must be benchmarked against
actual device operation
SPICE simulation is easier to use and more accurate than using
calculation tools in spreadsheets, maths programs etc. Easier to apply
to any topology, can include circuit parasitics easily.
9
10
Subject / Department / Author -
July 18, 2015