IPC-D-279 EN.pdf - 第78页

When variations of dissipated power are involved, the tem- perature swing related to the power cycling must be added to the variations of the device adiabatic temperature. When the dif ference in temperature between the …

system enclosure. It represents the temperature of the ulti-

mate heat sink towards which the entire energy of the sys-

tem is evacuated, and it represents the lowest temperature

of reference. The ambient temperature is different for dif-

ferent types of applications. It can be a near constant as for

implanted medical equipment, or it can vary over a wide

range of temperatures as for automotive, military or space

applications. For any speciﬁc case, this temperature is dic-

tated by the application and cannot be modiﬁed by design.

D-2.2 The Temperature Rise of the Cooling Agent at the

Device Level (T

) Except for some space applications,

the heat dissipated by device is transferred ﬁrst to a ﬂuid

cooling agent which transports the energy outside the sys-

tem to the ultimate heat sink. For most of the applications

the cooling agent is air which comes in direct contact with

the device. The mechanism of heat transfer is conjugate

convection conduction: part of the heat is transferred direct

to the ﬂuid, part is spread ﬁrst into the solids (board, enclo-

sure) and from there to the ﬂuid. In most cases the tem-

perature of the ﬂuid reaching the device is higher than the

ambient temperature (the temperature of the ultimate heat

sink) due to the fact that the ﬂuid already absorbed some

energy from other devices mounted upstream in the system.

The designer has the power to modify this term by modi-

fying the device position in the system or by introducing

additional cooling/heating elements in order to moderate

dangerous high/low absolute temperature values or strong

temperature swings. The temperature rise of the cooling

agent is also reduced if the mass (volume) ﬂow of the fluid

is increased.

D-2.3 The Temperature Rise Inside the Device Bound-

ary Layer (T

) The temperature rise inside the device

thermal boundary layer is related to the case-to-ambient

thermal resistance of the package. A certain gradient of

temperature is necessary in order to maintain the heat ﬂow

from the package into the ﬂuid acting as a cooling agent.

For devices dissipating less than 300 mW, this gradient of

temperature (the difference in temperature between the

package to the surrounding ﬂuid) is less than 5°C. This is

probably the reason that most designers have the tendency

of ignoring this term, focusing on the temperature rise of

the cooling agent at the device level as the critical factor of

ensuring the device functionability and reliability. As the

power per device increases, the temperature rise inside the

device thermal boundary layer becomes a critical factor.

For devices dissipating more than 1 W, this term alone

could be higher than 30°C and is certainly one of the domi-

nant terms in predicting the silicon junction temperature.

The designer has the capability of controlling the tempera-

ture rise inside the device thermal boundary layer by

increasing the local ﬂuid (air) velocity, by using heat

spreaders or increasing the substrate thermal conductivity.

D-2.4 The Temperature Rise Inside the Device Package

) Sometimes called junction-to-case temperature rise,

this term represents the temperature difference between the

silicon chip and the external temperature of the device

package. Being related to the thermal conductivity of the

materials used for packaging, this term can be controlled

by proper selection of materials and use of materials with

enhanced thermal properties. For devices dissipating high

powers (more than 0.5 W), the resistance of the interfaces

between different layers of materials (contact resistances)

becomes signiﬁcant and discontinuities, delaminations and

voids could add up to signiﬁcant thermal resistances.

D-2.5 Thermal Wake (T

) The thermal wake factor

should be considered when devices are mounted close

together and their thermal boundary layers intersect with

each other. This is particularly important when a sensitive

device is mounted downstream from a high power dissipa-

tor. Under such conditions the device is engulfed inside the

thermal plume of the upstream device. Even though this

term is generally low (lower than 5°C) it can produce sig-

niﬁcant problems if the layout of the board is not carefully

controlled.

D-3.0 DETERMINING THE SOLDER JOINTS TEMPERA-

TURE SWINGS

Considering the multitude of factors which determine the

device temperature, it can easily be seen that variation of

the ambient temperature alone is an indication of the actual

temperature variations the solder joint will see in the ﬁeld

conditions. Assuming the temperature of the solder joint is

almost identical with the external temperature of the pack-

age (case temperature), the temperature cycling is gener-

ated by two factors:

a) the variation of the local environmental conditions

for the considered device (sometimes called adiabatic

temperature or the temperature of the device with

power off)

b) the variation of the device power, which in turn pro-

duces variations of the temperature rise inside the

device thermal boundary layer.

The device adiabatic temperature is the equivalent of the

local ambient temperature for a speciﬁc device inside the

system. It is deﬁned as the temperature that an element

achieves when the convective heat from the element to the

ﬂuid goes to zero. When conduction and radiation from

other elements is negligible, the device adiabatic tempera-

ture is the temperature achieved when no power is applied

to the device and the rest of the system is activated.

The variation of the device adiabatic temperature is the real

temperature cycle that a device operating at a constant

power will see inside the system. The temperature cycle

could be very different for different devices inside the same

system, even though the external ambient is the same.

IPC-D-279 July 1996

When variations of dissipated power are involved, the tem-

perature swing related to the power cycling must be added

to the variations of the device adiabatic temperature. When

the difference in temperature between the device case and

the board underneath the device is small (not larger than

5°C), the simple superposition of the two cycles (local

environment and power cycling) is quite adequate for esti-

mating the real temperature cycle the solder joint of a spe-

ciﬁc device will see in different environments and under

different operational conditions.

D-4.0 COOLING OF ELECTRONIC EQUIPMENT

D-4.1 Radiation

Cooling PWAs by radiation is not prac-

tical as a primary heat transfer mechanism because of the

large temperature differences required for appreciable heat

transfer. In addition, control of heat paths is difficult due to

scattering of radiant energy. In fact, radiant heat transfer is

potentially objectionable because of the possible heating of

near-by temperature sensitive parts by hotter, less sensitive

components. The positive effects of radiation from SMT

printed boards can usually be neglected.

D-4.2 Free Convection There are some commercial

SMT applications which could effectively utilize free con-

vection, although even with free ﬂowing ambient air, the

packaging density would need to be decreased, with at least

5 mm, and preferably up to 20 mm spacings between ver-

tically oriented boards, with heat concentrations of no more

than 12,000 W/m

correspondingly, about 300 W/m

board area dissipation, based on the component side of the

board only; for boards with components on both sides, a

rough guideline would be 230 W/m

per side, maintaining

12,000 W/m

maximum.

These numbers for free convection printed board dissipa-

tions are relatively insensitive to the variations of printed

board size and calculated heat transfer coefficient and hold

for an air-to-board ∆T of 40°C. For sufficiently cooler air

or hotter board temperatures, greater dissipations can be

accommodated. For example, if a 60°C ∆T was acceptable,

then 60/40 = 1.5 times as much dissipation could be

allowed. A detailed design utilizing conduction spreading

of heat within the board should be performed around hot

spots in free convection designs to insure maximum junc-

tion temperature limits are maintained.

For SMT boards in military applications, free convection is

typically a second order effect in removing heat directly

from the board. This is especially true when, as would be

expected for SMT, that the boards are densely packaged in

a structure.

D-4.3 Direct Forced Convection Forced ambient air

directly over the board and components, is the most widely

utilized commercial method for cooling SMT (and high

density) printed boards. Forced conditioned or closed loop

air is also common in many military applications. The pen-

alties for this effective cooling mechanism is expenditure

of energy, added heat, weight, and volume, taken at the

enclosure or system level. The SMT printed board designer

will work to the requirements deﬁning air ﬂow rate and

temperature over the module, as well as possible variations

for altitude, in airborne applications.

Typically, 4 to 8 times as much heat can be removed with

forced convection compared to free convection. As a result,

heat dissipations in the range of 1500 to 3000 W/m

component side printed board area can be achieved with an

average module-to-air ∆T of 40°C.

D-4.4 Conduction Cooling SMT printed boards built

with conductive substrates or ‘‘cores’’ such as beryllium

oxide, aluminum, or one of several materials designed to

match the expansion rate of ceramic components (about 6

ppm/°C), can and should be conductively coupled to the

supporting structure by clamping the edges.

The allowable heat densities are signiﬁcantly greater than

those of free convection, and are a strong function of sub-

strate material, thickness, location of heat sensitive parts

and conduction paths of these parts to the substrate. The

boundary condition may be the temperature of the printed

board module substrate edge, or possibly the structural sur-

face to which this edge mates. In the latter case, the printed

board designer is dependent on the design of the structure

for the interface resistances, a key parameter in the conduc-

tion path to the ultimate heat sink. For a 40°C device

junction-to-module edge ∆T, dissipations on the order of

250,000 W/m

, or for a speciﬁc module, on the order of

4,500 W/m

can be maintained.

D-4.5 Heat Pipes The heat pipe is a thermal conductor

of very high conductivity. It is essentially a closed evacu-

ated chamber lined with a capillary structure or wick. Heat

is transported by evaporation of a suitable volatile ﬂuid,

which is condensed at the cold end and returned by capil-

lary force to the hot end. The vapor passes through the

cavity. Heat pipes can be constructed in practically an end-

less number of conﬁgurations, but always consist of three

zones or sections, namely: the evaporator, condenser, and

the adiabatic section connecting the other two. In some

designs, the adiabatic section may be very short.

Thermal resistances of 1°C/kW can be obtained and heat in

excess of 50 kilowatts can be transferred. Heat pipes offer

important advantages. First, a heat pipe has several thou-

sand times the heat transfer capacity of the best heat con-

ducting materials on a weight and size basis. Second, heat

pipes exhibit an essentially uniform temperature at the heat

input end. Third, the areas and conﬁgurations of contact at

each end of a heat pipe are independent and can be

designed separately to suit the application.

July 1996 IPC-D-279

Fourth, a heat pipe can transfer heat up to several feet with

a very small temperature drop. Fifth, a heat pipe can be

made entirely of insulating material to provide electrical

isolation for high voltage equipment. Sixth, heat pipes

require no power for their operation, since capillary pump-

ing is provided by the heat being transferred.

Heat pipes can be designed for operation at any tempera-

ture pertinent to cooling electronic equipment. In general,

they must be designed specially for each application. One

of the major difficulties in heat pipe application is in

achieving adequately low thermal resistances at the inter-

faces at the heat pipe ends. Also, the capillary pumping in

a heat pipe is inﬂuenced by gravity and external forces.

Some heat pipes have exhibited as much as an order of

magnitude change in internal thermal resistance due to

gravity (i.e., transferring heat downward vs. upward). Fur-

ther heat pipes exhibit startup difficulties because enough

condensate must collect to saturate the wick before pump-

ing can be initiated. If a heat pipe is overloaded, it can

rupture.

D-4.6 Direct Liquid Cooling In utilizing direct liquid

cooling as an alternative cooling technique, much more is

involved than merely replacing the internal air with a liq-

uid coolant. The full effects of the coolant intimate contact-

thermal, electrical, chemical, and mechanical must be

evaluated, both from a circuit and from a component view-

point.

The primary effect from a circuit viewpoint is, of course,

electrical, and involves the breakdown voltage (dielectric

strength) of the coolant in high power applications, and the

dielectric constant of the coolant in higher frequency appli-

cations. In addition, the ﬂuid selected must not be

adversely affected by the electrical ﬁelds or voltages nor-

mally produced in the equipment.

Components must be carefully screened to insure compat-

ibility with the coolant. The electrical effect on normal air-

dielectric components such as trimmer capacitors is obvi-

ous. Chemical compatibility with component cases, circuit

board material, rosins, and potting compounds, and any

other internal module material must be veriﬁed. Plastic

encased semiconductors are susceptible to leakage in some

ﬂuids when the parts are immersed directly in the liquid.

D-4.6.1 Direct Natural Convection Liquid Cooling In

direct natural convection liquid cooling systems, it is nec-

essary to expose a maximum of the surface of the heat-

producing parts to the coolant and to direct the free con-

vection currents of the ﬂuid around these parts. Thus, in

liquids, as in air, parts should be mounted to promote con-

vective cooling. Metallic conduction paths of low thermal

resistance from the heat production parts to the surface of

the case are not as important in direct liquid-cooled assem-

blies as in most other types of assembly since adequate

cooling is usually obtained by convection. The parts may

be supported by solid insulating materials so long as the

ﬂuid is permitted to freely circulate around the parts.

Where applicable, the part should be mounted such that the

longer part dimension is vertical. The construction of this

type of equipment must be given special consideration.

With viscous coolants, a slight mechanical advantage is

gained by the immersion of the electronic parts in the fluid,

since the ﬂuid tends to lend support to parts. Also, it can

provide a damping action which assists in resisting vibra-

tion and shock, dependent upon the viscosity of the ﬂuid.

Generally, the lowest thermal resistances from the parts to

the coolant are obtained with direct liquid cooling.

An accurate quantitative analysis of the temperature distri-

bution within a sealed direct liquid cooled system is usu-

ally very difficult to obtain, due to the indeterminate over-

all nature of the circulating convection currents.

Nevertheless, at least a qualitative inspection should be

made with regard to parts placement. Thus, temperature-

sensitive parts should not be placed where they would be

in the convection current generated by a high dissipation

part. Usually, a temperature gradient will exist in a direct

cooled module, with the lowest temperature at the bottom.

Therefore, the more sensitive parts should be placed in the

lowermost locations. Holes can be provided in subchassis

to direct the ﬂow of the free convection currents around the

heat producing parts. The use of direct liquid cooling

should not be considered a panacea for thermal problems.

The effectiveness of the contained coolant depends prima-

rily on the generation of auto-convection currents. Parts

placement should recognize this requirement and not inter-

fere with circulation. Printed wiring boards should be

mounted vertically, since a horizontal orientation would

restrict the convection currents.

While direct liquid cooling generally allows an increase in

parts density in a module because of improved thermal

paths, the increase in packaging density is limited by the

space requirement for allowing adequate circulating con-

vection currents to develop. Sufficient gaps must be pro-

vided between adjacent vertical plates to insure liquid flow,

i.e., between a circuit board and a structural wall. Liquids

may be as much as 10 times more effective than natural

convection in air, allowing up to 3000 W/m

board dissipa-

tions at a 40°C component to bulk ﬂuid temperature.

D-4.6.2 Direct Forced Liquid Cooling Just as increased

cooling effectiveness is obtained with forced air as com-

pared to free or natural air convection, forced circulation of

the coolant greatly increases the cooling rate when liquid

cooling is used.

In direct forced liquid cooling, part layout and location are

most important. The cooled liquid, or that coming from the

IPC-D-279 July 1996