Thermal Modeling Using Analog Power Thermal Spice Models

Thermal Modeling using SPICE
April 2015
Analog Power Confidential
Concept
•
Use up to 5 Uniform Resistor Capacitor (URC) elements to
model RTH:
1.
2.
3.
4.
Die ( vertical heat flow)
Die attach
Lead Frame vertical heat flow ( copper under die)
Lead Frame horizontal heat flow (not included in RTHJC for DFN
DPak, included in RTH JL for SO-8, includes some lateral flow for
DFN & DPAK on PC Board as leadframe works as a heat spreader)
5. PC Board (somewhat fixed but small footprint devices have a higher
resistance due to small periphery near foot)
Analog Power Confidential
Advantages of Model
By using URC elements with
defined applicability, this
method allows:
• Re-use of some elements,
eg leads, PC board
• Use of the same model for
RTH JC and RTHJA
• PC Board thermal models
can be added
Short here for RTHJC
• URC Model is more
accurate than 4 element
lumped value models
Short here for RTHJL
Analog Power Confidential
Netlist (export from LT Spice)
1 0 N001 1
U1 N001 N002 0 URC_DIE L=1 N=50
U2 N002 N003 0 URC_DA L=1 N=10
U3 N003 N004 0 URC_LF_Vert L=1 N=10
U4 N004 N005 0 URC_LF_Horiz L=1 N=5
U5 N005 0 0 URC_PCB L=1 N=5
• Model values stored in ascii file of model for easy editing
• Force 1A DC and voltage is temp rise for 1W heat = thermal
impedance °C/W
• N can be decreased to improve speed, reduce accuracy
Analog Power Confidential
Accuracy based on 2.2mm2
AM4874N R5
Time
RTH
SPICE
°C/W
RTH DS LIMIT RTH Old
(RTHJA ∞ =
Model °C/W
80°C/W+
1ms
1.15
1.2
1.17
10ms
3.2
3.6
3.12
100ms
6.4
10.4
8
1s
16.2
25.6
17.8
10s
25.8
40
24
100s
37.5
64
41.5
500s
48
80
49
Notes
Low, C_LF_H high?
Analog Power Confidential
Values Used for Prior Page
Die size mm2
2.2
RDIE
2.27
CDIE m
2.2
RDA
0.681818
CDA m
0.22
R_LF_Vert
1.818
C_LF_Vert
0.02
R_LF_Horiz
16
C_LF_Horiz
0.08
RPCB
CPCB
28
Analog Power Confidential
10
Maxes and Aggressiveness
• Based on typical SO-8 RTHJA on 2 x 25mm = 46.2°C/W,
so model of 49 °C/W is about right for a small die
• However data sheet gives max of 80 °C/W as it is typ for
short pulses and then max for long pulses
• Suggest use model + 15% as limit
• 56 x 1.15 = 65 °C/W a reasonable limit
Package
Observed RTHJA
10 mm x 15 mm board
25 mm x 25 mm board
51.3 C/W
47.7 C/W
SO-8
°
46.2°C/W
38.8°C/W
52.4°C/W
DPAK
-
DFN2x3
DFN3x3
DFN5x6
°
39.5°C/W
32.8°C/W
46.2°C/W
28.6°C/W
Analog Power Confidential
DC RTH JA, JL Sanity Checks
DFN56
Die
size
mm2
RDIE
CDIE m
RDA
CDA m
R_LF_V
C_LF_V
R_LF_H
C_LF_H
RPCB
CPCB
RTOT JA
RTOT JC
2.2
2.27
2.2
0.681
0.22
1.8181
0.02
5
0.2
26
10
35.7
4.77
10
0.5
10
0.15
1
0.4
0.02
5
0.2
26
10
32.05
1.05
CDA m
R_LF_V
R_LF_H
C_LF_H
SO-8
Die size
mm2
RDIE
CDIE m
RDA
C_LF_V
RPCB
CPCB
RTOT JA
RTOT JL
1
5
1
1.5
0.1
4
0.02
16
0.08
28
10
54.5
26.5
2.2
2.272
2.2
0.681
0.22
1.8181
0.02
16
0.08
28
10
48.77
20.77
8
0.625
8
0.187
0.8
0.5
0.02
16
0.08
28
10
45.31
17.31
Assume FDMS85200 is approx 10 mm2, and add ~20% for limit fits with Fairchild spec (and SiRa00 )
for RTH JC. Experimental Data has shown that RTHJA for medium die DFN 56 is ~ 33°C/W
Analog Power Confidential
Conclusion
• This new model allows calculation of RTHJC, JL and JA and
transient curves, even JL curves. And allows non linear input
• By separating die from package performance and allowing plotting
of each element, thermal performance can be better understood
• Same model used for RTH JL and JA
• By separating each element, one element may be changed to
determine effect, eg model improved performance PC Board
• Thermal Spice models available from Analog Power on a
experimental individual basis to ensure accuracy and
improvement over time
• Spice can be run at different time steps depending on the
granularity needed. Ideally run at step size of 1/10000 of the limit
being determined (ie 100n for 1ms)
Analog Power Confidential