Thermal Design Consideration For Surface Mount Layouts

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Thermal Considerations for Surface Mount

Layouts
Charles Mauney, Texas Instruments
ABSTRACT
The focus of this paper is to present the techniques that are effective in removing the heat from the
surface mounted power ICs and transferring it to the cooler surroundings. In the past the thermal design
was as simple as adding the theta values (C/W) of the power component, the pad and the heat sink
together and multiplying by the power dissipated in that component to arrive at the temperature rise
between the ambient and the junction of the IC. Today, with surface mount technology and ICs with
power pads the Printed Circuit Board (PCB) has become an integral part for removing the heat. Every
PCB is different and the task of mathematically deriving a temperature profile for the PCB and each
power component is beyond the scope of this paper. However the basic concepts of what makes a good
thermally conductive PCB, some first order modeling examples and evaluations of an existing EVM will
help the designer understand the techniques in producing a good thermally conductive PCB.
There are three basic methods to move heat away from the source, conduction, convection and radiation
heat transfer. Conduction is moving the heat through a material (PCB) typically from copper to FR4 to
copper and so on. Convection is for removing the heat off the surface material to the air. Radiation is
for moving the heat from one surface material to another through air.

TABLE 1 THERMAL CONDUCTIVITY FACTORS

I. HEAT TRANSFER BY CONDUCTION

Materials
Copper
Aluminum
FR4
Solder 63/67
Air

Conduction is important because it is the


surface area that eventually dissipates the heat
and only through conduction does the heat
spread out to the required surface area. The
trend is to migrate toward smaller packages
believing that the design area can be reduced.
The reality is that a smaller footprint, with the
same power dissipation, requires a better PCB
conduction design and a similar amount of
surface area, for a given temperature rise.
Heat conduction is better through some
materials (Copper) than others (FR4). Table 1
shows the thermal conductivity factor K for
different materials. These common materials
have significant different thermal conductivity
factors. The units are Watts per meter-Kelvin
(W/mK) or Watts per meter-Celcius (W/mC).
Since the temperature is a temperature rise (dT)
the 273.15 offset between K and C drops out.

K (W/mK)
355
175
0.25
39
0.0275

K (W/inC)
9
4.44
0.0064
1
0.0007

At first glance at the units, it appears that


one would multiply the Thermal Conductivity
Factor by the thermal path length to get W/C
(1/), but this is not correct since this would
indicate the longer the thermal path the larger
the heat transfer factor. The correct method is to
divide K by the thermal path length which
results in the units, W/in2C. Next multiplying
by the cross sectional surface area of the thermal
path yields W/C or 1/.

10-1

A. Formulas For Conduction

1. General Conduction Formula


The general conduction formula for thermal
impedance for a given material is:

L
K ACS

(4)

1
0.0064 W (in C) 0.060 in
= 2620C/W

sq 0.060 FR 4

(5)

It shows that the copper plane has much


better thermal impedance than FR4.
3.

(2)

Perpendicular Conduction Heat Flow


Formulas

where
L is the path length of the heat flow
W is the width and t is the thickness
Wt = ACS, area-cross section of heat flow,

1
9 W (in C ) 0.0014 in

1-oz. Cu is 1 ounce of copper spread over


1 square ft of area. (1 oz. is the equivalent to
1.14 mil thickness of Copper: which is also
eqivalent 310 grams of Copper per square
meter.
Thermal Impedance, , for lateral
conduction, for a square of 0.06 thick FR4
is:

2. Lateral Conduction Heat Flow Formula


Fig. 1 shows the lateral conduction heat flow
along a plane. The thermal impedance for the
lateral heat flow is given by

= 79.4C / W

A low theta is desired since this is the


temperature rise in C per Watt of heat
dissipated.

L
L
=
K W t K ACS

sq 1oz Cu

(1)

Where
K is the Thermal Conductivity Factor
L is the thermal path length
ACS is the cross sectional area where the
heat is being applied

Thermal Impedance, , for lateral


conduction, for a square of 1oz. Cu is:

1
K t

(3)

where
for a square area L=W

Fig. 2. Perpendicular conduction heat flow


through a plane.

Fig. 1. Lateral conduction heat flow along the


length of a plane.
10-2

Fig. 2 shows the perpendicular conduction


heat flow through a plane. The thermal
impedance is given by:

t
t
=
K W L K ACS

L is the length of the via (the heat flows


through the length of the cylinder)
D02 D12
is the cross sectional area,

4
Substituting L=0.06 in., K=9 W/(in-C),
D0=0.016 in., and D1=0.013 in into the above
equation yields 97.6oC/W for a copper via.
For a solder filled via, the thermal
impedance for the solder plug is given

(6)

where
t is the path of heat flow (the heat flows
through the thickness of the material)
WL is the cross sectional area where the
heat is being applied, ACS
Thermal Impedance, , for perpendicular
conduction, for a 1 in2 of copper, 2 oz, is:

2oz _ Cu _ 1in =
2

0.0028 in
9W / (in C ) 1in 2

Solder via =

(7)

5.

= 0.0003 C / W
Thermal Impedance, , for perpendicular
conduction, for a 1 in2 of FR4, 60m, is:
0.060 in
0.0064 W / (in C ) 1 in 2
= 9.4C / W

0.06in FR 4 _ 1in =
2

L
=
K ACS

(10)

D12
K
4

Substituting L=0.06 in, K=1 W/(in-C), and


D1=0.013 in provides 452C/W.

Conduction Through Solder Mask


Formula

Solder Mask =

L
K ACS

(11)

Where K= 0.0075 W/in- C and L=1 mil.

(8)

The very thin sheet along with a large area


result in a 0.133 C/W rise for 1 in.2 of surface
area. This small temperature change is typically
ignored.

4. Heat Flow Through a Via Formula

B. Thermal Conduction Concepts


1. Concept 1: A Shorter Thermal Path Has
a Lower Theta
For a given material the shorter the thermal
path the better. It is more efficient to conduct
heat through the planes thickness rather than
along its length.
An example of short and long heat transfer
paths using FR4 is shown below.

Fig. 3. Vertical conduction heat flow through a


via.

Example: Heat
thickness 0.060 in:

Fig. 3 shows the vertical conduction heat


flow through a cylinder via. Its thermal
impedance is given by

Cu Via

L
=
=
K ACS

L
D02 D12
K
4

0.06in FR 4 Acs =

(9)

transfer

through

t
0.060 in
=
K 0.0064 W /(in C )

10 C in / W
o

This implies that a 10C drop over 1 in2


conducts 1 W)

where

10-3

FR4,

(12)

Example: Heat transfer along 1 in of 0.060in


FR4 layer:

This implies that for a 10C drop over 1-in.


wide and 0.0028-in thick conducts 25 mW.

2. Concept 2: A Larger Connection (Area /


Perimeter) Between IC & PCB Has a
Lower theta
The larger the perimeter of the power pad
(circumference), the greater the cross-sectional
thermal conduction area, the lower the PCBs
Theta and the cooler the PCB and IC.
As ACS (circumference around the power
pad times copper thickness) decreases the Theta
increases. Therefore selecting a smaller part
decreases the circumference and results in a
higher Theta. To compensate for this change
one can increase the copper thickness, which
increases the cross-sectional thermal conduction
area and reduces the Theta.
An analogy to heat flow is water flow.
Consider a pump which outputs a constant water
flow rate (5 gal/min), through a -in diameter,
1-ft pipe with a -in coupler on the end. A -in
diameter garden hose is connected to the -in
coupler and the pressure dropped is observed
across the coupler. Since two -in pipes are
connected together with a -in connection,
there is no measurable pressure drop observed.
The -in coupler is replaced with a -in
coupler and the pressure drop observed. As one
could imagine, the pump had to dramatically
increase its output pressure to get the same flow
through the smaller coupler connection. This
change in pressure is analogist to the
temperature change, and where the constant
flow rate is analogist to the constant power
dissipation.
One can see that a better connection (larger
ACS) to a source, the better the flow from the
source.

b.

1in FR 4 Acs =

L
1 in
=
K 0.0064W /(in C )

(13)

= 156C in / W
2

This implies that it conducts only 3.8 mW


with 10C rise for a 1 in. wide by 0.060 in. thick
FR4
a. Conclusion Short Thermal Path, FR4
FR4 is a fairly good thermal conductor
(relative to air) through thin layers
(perpendicular conduction) but not along the
length of the planes layer (lateral conduction).
An example of short and long heat transfer
paths using Copper is shown below.
Example: Heat transfer through 2 oz. Cu
(t=2.8 mils):

2 oz Cu Acs =

0.0028 in
t
=
K 9 W /(in C )

(14)

= 0.000311C in / W
2

can assume perfect heat transfer.


Example: Heat transfer along 1 in of 2-oz.
Cu

1in Cu Acs =

1 in
L
=
K 9 W /(in C )

(15)

= 0.11C in / W
2

Conclusion Short Thermal Path, Cu


Copper is an excellent thermal conductor
through the 2-oz. thick layer and is a good
thermal conductor along its length for a large
cross-section, (Wt). At least one full copper
layer is needed to spread the heat.

L
L
=
K W t K ACS

(16)

3. Concept 3: Use Thick Copper and Thin


FR4 Layers
Heat conducts through a FR4 layer more
efficiently than through air, so FR4 is still
relatively a good heat conductor; therefore to
maximize heat flow through FR4, minimize the
thickness if possible.

10-4

Copper is 9/0.00641400 times the thermal


conductor as FR4. For copper, a thicker layer
helps the heat conduction along the length of the
plane but the does not significantly impede the
heat transferred through the thickness of the
copper layer. Therefore use a thick layer of
copper to maximize the heat flow along the
copper sheet (2 oz. is great for most power
designs).

Cu Solder =

(17)

1.

The temperature rise through convection


is given by

D0 D1
4
0.06 in
=
(0.016 in) 2 (0.013 in) 2
9 W /(in C )
4
= 97.6 C / W
(18)
K

dT =

Example: Heat transfer through solder


portion of via,

Solder =

P
P
= 133
h A
A

(21)

where
P is in Watts
h is 0.0075 W/(in2-C)
A is surface area, in2.
So the thermal impedance is given by
2. The Theta for Convection is given by

D2
K
4
0.06 in
=
= 452C / W
1W
0.0132 2

in
in C
4

Solder

A. Convection Formulas

L
2

(20)

Convection is the method of moving heat


from the surface of a material to the air. The
temperature rise is a function of the power
dissipated and inversely proportional to the
surface area and the heat transfer coefficient, h.
The heat transfer coefficient, h, is a function of
air speed and temperature difference. The range
is 7.8 (10C rise) to 9.1 mW/(in2-C) (60C rise),
for no air flow. A conservative value for h is
7.5 mW/(in2-C) for no air flow.

Example: Heat transfer through a Cu thermal


via, D=0.016, D0=0.013, L=0.06
L
=
K ACS

= 80.3C / W

II. HEAT TRANSFER BY CONVECTION

= 979C / W

Cu =

Combined they result in Cu/solder =


80.3C/W.
a. Conclusion Use Multiple Vias
Solder and copper greatly enhance thermal
performance of the FR4. Solder is less
conductive but often has a larger cross sectional
area so it can improve thermal conductivity
significantly. Multiple vias further lower the
Theta value.

9.4 C / W in 2
ACS

9.4 C / W in 2 mm

=
2 mm 3 mm 0.040 in

Cu

4. Concept 4: Use Multiple Vias to lower


the Theta through the PCB (FR4)
Copper is nine times better thermal
conductor than 63/37 solder, and solder is 157
times better than FR4. Consider the effects of
using multiple solder filled vias.
Example: Heat Transfer through 60m - FR4,
under IC (Pad Area, 0.08x0.12), with no Vias

60 m FR 4 IC =

Surface area
(19)

133 (C in 2 / W )
1
=
=
A h
A(in 2 )

(22)

3. Minimum PCB Surface Area need for a


given temperature rise, from TAMB to
TPCB-MIN.

A=

Example: Heat Transfer for Solder Filled


Vias

10-5

P
P
= 133
h dT
dT

(23)

The interior surface temperatures are:

4. Junction Temperature is given by

TINT Cu = TRISE Cu + TEXT SURFACE

TJ = T AMB + dTPCB Conv + dTPCB Cond + dTJC _ IC (24)

= 0.011C + 38.3C
= 38.311C

B. Convection Concepts
1. Concept 5: Convection is a Function of
T and Area, not the Material
The heat transfer off of a surface is related
to the temperature difference between the
ambient air and the surface of the material and
the air flow, but not the material itself. This at
first seems like an error because Copper is a
better conductor than FR4. Copper is a better
conductor than FR4, but both have the same
heat dissipation factor for Convection.

TINT FR 4 = TRISE FR 4 + TEXT SURFACE


= 15.6C + 38.3C
= 53.9C

C. PCB Area Estimate for Typical Design


How much board surface area is needed for
a design? A typical example is a power
component (JC of 7C/W, TJ-125C)
dissipating 2 W with a maximum ambient of
50C. How much area is needed to keep the
junction temperature from exceeding it
maximum temperature? Start with the maximum
junction temperature and subtract off the
temperature rise across the IC, PCB and
maximum ambient temperature, the remainder is
the allowable temperature rise allowed due to
convection.
The temperature rise across the IC-JC is

TEXT _ SURFACE
= 25C + 133C / W in 2 10 W 100 in 2 (25)
= 38.3C

Both external surfaces are the same


temperature and dissipate the same amount of
heat.
For conduction, the thermal conductivity
factor is different for different materials and the
temperature change for given shape and power
dissipation will be different.
The temperature change from the external
surface to internal surface, due to conduction, is:

TRise Cu

(26)

10 W
100 in 0.0064 W/in 2 - C
= 15.6C

TRise FR 4 =

TJC( max) = PDISS JC


= 2W 7C/W = 14C

(29)

The concept of having the same external


surface temperature, for Copper and FR4, is
more easily accepted knowing the internal
surface temperature of the FR4 is much higher
than that of the copper, for a given power
dissipation.
With 1.5 ft/sec of air flow, h can double and
lower the temperature rise due to convection by
half. Calculating h for the non linear air flow
patterns is complex and not explored in this
paper. For Portable Power, often the PCB are
enclosed and in an environment where there is
no air flow.

Example : Convection vs Conduction


Consider a sheet of copper and a sheet of
FR4 with the same dimensions (10 in 10 in
1 in). Each sheet is placed on top of a perfectly
insulated box (sides and bottom) with a 10W
heat source placed in each box. The ambient
temperature is 25C and the lid (top sheet) is
allowed to reach its steady state temperature.
The top surface temperature, by convection,
of each material (FR4 or Copper) is:

10W
=
2
100 in 9 W/in 2 - C
= 0.011C

(28)

(30)

A conservative estimate for theta, due to


conduction, across the PCB is 20C/W. The
temperature rise across the PCB is

TPCB( max ) = PDISS PCB Cond

(27)

= 2 W 20C/W = 40C
10-6

(31)

The dT between ambient and the PCB surface


is:

0.045

dTPCB Conv( max )

0.035

(32)

0.030

= 125C 14C 40C 50C = 21C

The surface area needed is:

133P
2W
A=
= 133 in 2 C/W
= 12.7 in 2 (33)
21C
dT

Heat Flux - W/cm2

= TJ( max ) TJC( max ) TPCB( max ) T AMB( max )

0.040

0.025
0.020
0.015
0.010
0.005

A 2.5 in. by 2.5 in. PCB gives 12.7 in2 of


surface area.

0.000
70

80

90

100

110

120

Board Temperature - C

III. HEAT TRANSFER BY THERMAL RADIATION

Thermal Radiation is the transfer of heat


from one surface to another. It is based off of
the Stefan-Boltzman Law of Radiation equation
and is given by

0.30

H = eAT1 T2

0.20

(34)

Heat Flux - W/in2

0.25

where
e = emissivity (0-1).
TK = Kelvin temperature
=Stefan-Boltzmann constant
(see Equation 17)

J
5.67 10
(s m 2 K 4 )
8

0.15

0.10

0.05

(35)
0.00
70

Plots were generated based on the thermal


radiation, in Watts/area, versus surface
temperature, of the radiator, and are shown in
Fig. 4.

80

90

100

110

120

Board Temperature - C

Fig. 4. Thermal Surface Radiation


The plots are fairly linear due to the small
change in surface temperature compared to the
large temperature values in K. It turns out that
the heat transfer due to radiation, assuming that
the radiating surface temperature is greater than
the ambient temperature, is approximately the
amount of heat transferred by convection. The
heat transfer contribution by Thermal Radiation
is often ignored in any analysis, due to its added
complexity, but may be considered as a margin
of safety in the thermal design.

10-7

This model has 20 horizontal resistors


representing 20, 0.1-in. circular rings (or a
radius of 2 in. from the heat source (power IC)).
The center row of vertical resistors (circular
rings) represent the thermal resistance through
the FR4 (PCB) from one Copper surface to the
other. The top and bottom row of vertical
resistor represent the theta values for the
convection off of the surface, of the PCB, into
the air. The ground potential represents the air
temperature. The voltage rises from the ground
toward the 3-A current source represent the
temperature rises above ambient. As mentioned
earlier, there is negligible temperature rise
through the thickness of the copper layer, so the
0- resistors are not shown. The horizontal
resistors for conduction in the FR4 plane are
very large and are in parallel with the small
copper horizontal resistors and thus are omitted.
Table 1 is a quick reference chart for the
theta value of a single via given its finished hole
diameter and the board thickness. The effective
theta value can be reduced by a factor of n by
adding n number of vias to the PCB.

IV. IS THERE AN EASY MODELING


APPROACH?

Modeling can be very complex since no two


PCB designs are exactly the same. The
temperature and power dissipation gradients are
non linear and is beyond the scope of this article
to create and integrate a formula to accurately
predict temperature rise, for power dissipation
of multiple ICs. However a generic model has
been created with a matrix of series and parallel
resistors to simulate temperature rise through
and along the PCB layers for various size boards
for a single power IC. The resistor values are
determined from the conduction and convection
formulas and represent the theta values for small
circular rings, 0.1 in. wide for both Copper and
FR4. The radius of each consecutive ring is 0.1
in. larger The larger the radius the longer the
circumference and the more 0.1 in. squares there
are per ring, thus lowering the theta.
The theta value for conduction through 2-oz.
thick copper is assumed zero where the theta
value along the length of the FR4 plane is
assumed infinite. The heat transfer is restricted
or limited by the PCB materials (theta values)
represented by resistors, the heat flow (Watts) is
modeled by the current and voltage drops
represent change in temperature (dT).
Fig. 5 is the thermal model for a 2-layer,
1-oz., 60-mil PCB with an IC that dissipates
3 W. The 3-A current source represents the 3 W
power dissipation of the IC. The two horizontal
rows of resistors represent the top and bottom
copper planes. Each resistor, from left to right,
represents the conduction horizontally along the
copper layer, starting at the ICs perimeter, for a
0.1 in. radius, and heading outward in a circular
pattern. As the 0.1-in, band, circular rings get
further from the IC, the circumference increases
and so does the number of squares of copper.
This explains the continual drop in theta values
as the radius increases from the IC.

10-8

TABLE 1. THETA VALUES

0.03

0.04.

Board Thickness
0.06
0.07

0.05

0.08

0.09

0.1

Filled Vias

0.01

61.5

53.1

82.0

70.7

102.5

88.4

123.0

106.1

143.5

123.8

164.0

141.5

184.5

159.2

205.0

176.8

0.013

48.8

40.2

65.0

53.6

81.3

67.0

97.6

80.4

113.8

93.8

130.1

107.2

146.3

120.6

162.6

134.0

0.016

40.4

31.9

53.9

42.5

67.4

53.1

80.8

63.7

94.3

74.4

107.8

85.0

121.3

95.6

134.7

106.2

0.019

34.5

26.1

46.0

34.8

57.5

43.5

69.0

52.2

80.5

60.9

92.0

69.6

103.5

78.3

115.0

86.9

0.022

30.1

21.8

40.1

29.1

50.2

36.4

60.2

43.7

70.2

51.0

80.3

58.3

90.3

65.5

100.3

72.8

0.025

26.7

18.6

35.6

24.8

44.5

31.1

53.4

37.3

62.3

43.5

71.2

49.7

80.1

55.9

89.0

62.1

0.028

24.0

16.1

32.0

21.5

40.0

26.9

48.0

32.2

55.9

37.6

63.9

43.0

71.9

48.4

79.9

53.7

0.031

21.8

14.1

29.0

18.8

36.3

23.5

43.5

28.2

50.8

32.9

58.0

37.6

65.3

42.3

72.5

47.0

0.034

19.9

12.5

26.6

16.6

33.2

20.8

39.9

25.0

46.5

29.1

53.1

33.3

59.8

37.4

66.4

41.6

0.037

18.4

11.1

24.5

14.8

30.6

18.5

36.7

22.2

42.9

26.0

49.0

29.7

55.1

33.4

61.2

37.1

0.04

17.0

10.0

22.7

13.3

28.4

16.6

34.1

20.0

39.8

23.3

45.5

26.6

51.1

30.0

56.8

33.3

0.043

15.9

9.0

21.2

12.0

26.5

15.0

31.8

18.0

37.1

21.1

42.4

24.1

47.7

27.1

53.0

30.1

0.046

14.9

8.2

19.9

10.9

24.8

13.7

29.8

16.4

34.7

19.1

39.7

21.9

44.7

24.6

49.6

27.3

0.049

14.0

7.5

18.7

10.0

23.3

12.5

28.0

15.0

32.7

17.5

37.4

20.0

42.0

22.5

46.7

24.9

0.052

13.2

6.9

17.6

9.2

22.0

11.4

26.4

13.7

30.9

16.0

35.3

18.3

39.7

20.6

44.1

22.9

0.055

12.5

6.3

16.7

8.4

20.9

10.5

25.0

12.6

29.2

14.7

33.4

16.8

37.6

19.0

41.7

21.1

0.058

11.9

5.8

15.9

7.8

19.8

9.7

23.8

11.7

27.7

13.6

31.7

15.6

35.7

17.5

39.6

19.5

0.061

11.3

5.4

15.1

7.2

18.9

9.0

22.6

10.8

26.4

12.6

30.2

14.4

34.0

16.2

37.7

18.0

0.064

10.8

5.0

14.4

6.7

18.0

8.4

21.6

10.1

25.2

11.7

28.8

13.4

32.4

15.1

36.0

16.8

0.067

10.3

4.7

13.8

6.3

17.2

7.8

20.7

9.4

24.1

10.9

27.5

12.5

31.0

14.1

34.4

15.6

0.07

9.9

4.4

13.2

5.8

16.5

7.3

19.8

8.8

23.1

10.2

26.4

11.7

29.7

13.2

33.0

14.6

0.073

9.5

4.1

12.7

5.5

15.8

6.8

19.0

8.2

22.2

9.6

25.3

11.0

28.5

12.3

31.6

13.7

Fig. 6 is the temperature RISE profile plot for


a 3-W, 2-layer, 1-oz. Cu, 60-mil FR4 PCB.
There are two sets of curves. The top set is for a
board that has a 1 in. radius around the part
similar to a 2 in 2 in PCB. The lower set of
curves is for a radius of 2 in. or a 4 in. 4 in.
PCB. The difference between the 1-in. and 2-in.
radius models is that for the 1-in. model the
horizontal resistors at the 1-in. radius are made
very large, effectively cutting off heat transfer to
the rest of the PWB. Note that for a given power
dissipation and board size that the PCBs
temperature tapers to a steady state base line
temperature toward the outer part of the PCB.
The smaller PCB reaches a higher base
temperature. Notice how the temperature rises

faster as the distance to the IC is reduced. The


sets of curves also fan out close to the part. The
curves that ramp up to a higher temperature are
the ones with isolated connections on the top
layer. For the totally isolated power pad the top
surface temperature drops rapidly toward the
average base line temperature of the PCB. This
fits well with the analogy about the larger water
pressure changes for smaller connections to the
source. The plot also shows how the top and
bottom surface layer temperatures merge as the
distance from the IC increases. This is the result
of the lower FR4 theta value, due to the
increased material area, as the distance from the
power IC is increased.

10-9

Fig. 5. Theta model for 3-W, 2-layer, 1-oz., Cu, 60 mil FR4 PCB (for Figure 2 plot )

10-10

Fig. 7 is similar to Fig. 6 except that it uses


2-oz. Copper. Note how the base line
temperature is the same verifying that the
temperature is a function of board area and not
Copper thickness. The temperature profile close
to the IC is much better for the 2-oz. case which
again supports the theory that a good connection
to the heat source is vital.

This model is a first order effect and does not


account for air flow or board orientation so in
reality, for a horizontal board the thetas would be
slightly lower making the top of the board 1 to
3C cooler. Also if the board is vertical, both
surfaces are a couple of degrees cooler. This
model is however fairly close to an enclosed
application where there is no air flow and
orientation becomes less of a factor.
Top Profile, Pad Isolated

Top Profile, Pad Isolated

Bottom Profile, Pad Isolated

Bottom Profile, Pad Isolated

Top Profile, Pad 50% Top

Top Profile, Pad 50% Top

Bottom Profile, Pad 50% Top

180

Bottom Profile, Pad 50% Top

140

R=1

R=1

160

3W

140

120

3W

100

2 oz. Cu

2 Layer

2 Layer
1 oz. Cu

120

60 mil FR4

100

60 mil FR4

80

Nine-13mil Vias

80

Nine-13mil Vias

60

60

40
T Temperature Rise - C

T Temperature Rise - C

40
20
0
0

0.2

0.4

0.6

0.8

1.0

140
R=2

120

3W

0.2

0.4

0.6

0.8

1.0

100
90

R=2
3W
2 Layer

70

1 oz. Cu
60 mil FR4

80

80

2 Layer

100

20

2 oz. Cu

60

60 mil FR4

50

Nine-13mil Vias

Nine-13mil Vias

60

40
40

30

20

20
10

0
0

0.4

0.8

1.2

1.6

2.0

0
0

R - Radius from PWR Device - Inches

Fig. 6. PCB temperature profile, 3-W, 2-layer,


1-oz., Cu, 60 mil FR4, nine-13 mil vias.

0.4
0.8
1.2
1.6
R - Radius from PWR Device - Inches

2.0

Fig 7. PCB temperature profile, 3-W, 2-layer,


2-oz., Cu, 60 mil FR4, nine-13mil vias.
10-11

The difference between Fig. 8 and Fig. 7 is


that the FR4 is half as thick. Once again the base
temperatures are the same and the connection
between the IC and board has improved since the
vias and FR4 are half the thickness and half the
theta value. The result is half the temperature rise
perpendicular through the PCB, near the IC.
Fig. 9 is a model of a 4 layer PCB and
therefore has two extra copper layers and two
extra FR4 layers. Note that when a Copper layer
is isolated, from the IC, or the board radius is
reduced the theta value at that location is
effectively removed by placing a series of 9s in
front of the value. This allows quick changes
between simulations.

Top Profile, Pad Isolated


Bottom Profile, Pad Isolated
Top Profile, Pad 50% Top
Bottom Profile, Pad 50% Top

120

R=1
3W

100

2 Layer
2 oz. Cu

80

30 mil FR4
Nine-13mil Vias

60

40

20

0
0

0.2

0.4

0.6

0.8

1.0

80
R=2

T Temperature Rise - C

70

3W
2 Layer

60

2 oz. Cu
30mil FR4

50

Nine-13mil Vias

40
30
20
10
0
0

0.4

0.8

1.2

1.6

R - Radius from PWR Device - Inches

Fig. 8. PCB temperature profile-3W, 2-layer,


2-oz. Cu, 30 mil FR4, nine-13 mil vias.

10-12

2.0

Fig. 9 Theta model for 3-W, 4-layer, 2-oz. Cu, 30 mil FR4 PCB (for Fig. 10 plot)
10-13

The difference between Fig. 10 and Fig. 8 is


that Fig. 10 has four layers. Once again, the base
temperatures are the same for a give surface
area. Also there is at least twice the copper
connection near the IC reducing the temperature
rise. The extra curve to the right end of traces
represents the base line temperature for the PCB
ranging from approximately 1.4 in. per side to
4 in. (2 in. radius) per side. For a 3-W
dissipation, a PCB smaller than 2-in. per side (1in. radius) has a difficult time dissipating its
heat and is very unlikely to keep the junction
temperatures, of the IC, under its maximum
rating.

110
Bottom EVM
Top EVM
Bottom Model
Top Model

T Temperature Rise - C

100

90

80

70

60

100

50
0

0.4

0.6

0.8

1.0

1.2

R - Radius from PWR Device - Inches

4 Layer

80
T Temperature Rise - C

0.2

3W

90

Fig. 11. Temperature profile of bq24032 EVM RHL package dissipating 3W vs. Model Data.

2 oz. Cu

1-in Radius

70

30 mil FR4
Nine-13mil Vias

60
50

Typical PCB
Edge

40

Temperature

30

2-in Radius

20
10
0
0

0.5

1.0

1.5

2.0

R - Radius from PWR Device - Inches

Fig. 10. . PCB temperature profile-3W, 4-layer,


2-oz. Cu, 30 mil FR4, nine-13 mil vias
Fig. 11 shows the temperature profile of an
EVM and its Model Data. The ambient
temperature of 25C was added to the model
data which is in C temperature rise above
ambient. The PCB of the EVM is approximately
2.1 in. 2.3 in., has four 2-oz. copper layers,
two of which connect to the IC through 6-13 m
vias, one routing layer with no copper
connection to the IC and the top layer that has a
60% connection to the ICs power pad.

10-14

The model in Fig. 9 was adjusted for these


differences. The power IC is mounted
approximately 0.9 in from the south and east
sides of the board. The board is 0.031 in thick.
The measured data was taken from the center of
the IC toward the west side of the PCB.
The Model Data is higher, by about 20C,
than the measure data and is explained by
absence of the Radiation effect in the model. Is
is estimated that up to 35% of the power
dissipation is due the heat transfer by radiation.
Notice how both plots have a similar
temperature curve profile. The measured EVM
temperature data ramps up sooner which implies
a poorer thermal path than the model. This is
true since the model did not take into account all
the signal vias that break up the copper planes
(similar to swiss cheese). The separation in
temperatures between the top and bottom
surfaces also occurs further out form the IC than
the model indicates. This is explained by the
signal vias reducing the FR4 effectiveness and
other surface mounted components that change
the thermal characteristics. The temperature
spread from the top and bottom surfaces, at the
ICs power pad, is similar for the model and
measured data and indicates that the model and
PCB are in agreement.

Since the PCB model for Fig. 8 is similar to


the one in our example, we can use that data and
estimate the maximum ambient temperature
using the formula:

IV. APPLYING THE BASIC CONCEPTS


IN A PCB DESIGN

Taking what was learned about thermal


convection design, the goal is to spread the heat
out and at the same time get it to both top and
bottom surfaces. We know that copper does the
best job conducting heat, followed by solder,
and then FR4. Therefore the best approach is to
have a good thermal path between the power
pad of the device to at least one copper plane,
either directly connected to the copper plane or
by vias and then let the copper plane spread the
heat out. The better the copper connection to the
heat source, the smaller the temperature change
(rise) across the connection. As the heat, from
the source, is spread out, it becomes more
efficient to transfer the heat through the FR4 to
the adjacent surface and then be dissipated into
the air.
Consider a design using an IC, as shown in
Fig. 12, that dissipates 1.5 W, has a theta of JC=
10C/W and will be mounted on a PCB that is
2-layer, 2-oz. Cu., with up to 9-13 mil vias, in
the power pad. Assume the ICs power pad is
isolated from the top copper layer.
What is the highest ambient temperature the
PCB assembly should see if the TJ(MAX) =125C
is not to be exceeded, for a 2 in. 2 in. PCB (1in. radius from IC)?

TJ-MAX = tAMB+dtPWB+dtIC
or
tAMB-MAX = TJ-MAX - dtIC - dtPWB
The power dissipation in this example is half
the referenced model data so the temperature
changes should be half. The temperature rise
from ambient to the power pad, for the isolated
1 radius curve is 110C. The PCB rise for this
example is half of this or 55C. The rise from
the case to junction of the IC is 1.5 W10C/W
= 15C. The maximum ambient temperature
should not exceed 55C (125C-15C-55C).
What about for a 3 in. 3 in. (R=1.5 in.)
board? Fig. 10 shows how the base line
temperature of the board changes for the size of
the PCB for a 3-W dissipation. The change
between R=1 and R=1.5 is 32C (57C-25C).
For this example at half the power, the
temperature change would be half of 32C or
16C. It is estimated that the ambient can be
16C higher on the R=1.5 in. PCB (3 in. 3 in.)
for a maximum ambient of 71C.
The convection temperature rise above
ambient for the R=1 example above is
55C=27.5C. The temperature margin due
to the Thermal Radiation contribution in parallel
with the convection would result 1/3 of the
calculated convection temperature rise or 9.2C
(1/327.5). This means that the assembly should
be 9.2C lower in temperature than estimated
due to the contribution of the Thermal Radiation
losses.
Understanding the electrical equivalent
concept of the current (heat) flowing through
the resistors (thetas) from the source (IC) to
ground (ambient temperature) resulting in
voltage (temperature) drops is helpful in
understanding what PCB parameters one has to
modify to achieve the thermal results desired.

Fig. 12. Bottom view of IC showing power pad


restriction by signal pins.

10-15

V. CONCLUSION

VI. REFERENCES

Thermal PCB design is a critical component


design that is often over looked or not
understood. This article tries to break down the
components of thermal design, analyze them
and give examples to show what makes an
effective design. The ultimate goal is to keep the
junction temperature of the ICs below their
maximum temperature values. The IC typically
has a theta value which accounts for the
temperature rise from the Power pad to the ICs
junction. The board area and power dissipation
determines the base line temperature of the
board, assuming there is at least one copper
plane connected, by copper, to the IC. The
critical area of the design is the connection
method close by the power IC. Multiple planes
(thicker copper) with multiple ground vias helps
reduce the theta values. A larger package for a
given heat dissipation provides a larger initial
path and less temperature rise. The maximum
junction temperature can be estimated by adding
the temperature rise of the PCB and the ICs
temperature rise to the maximum ambient
temperature. The heat loss due to Thermal
Radiation is typically not considered in the
analysis, but its contribution is used as the
margin of safety in the design.

Kollman, Robert, Constructing Your


Power Supply - Layout Considerations ,
Texas Instruments Power Supply Design
Seminar SEM-1700, 2004 (SLUP100).

10-16

You might also like