IPC-D-279 EN.pdf - 第80页

heat exchanger , should be directed to cool the temperature- sensitive parts ﬁrst, and later be directed adjacent to the higher power heat producing parts. D-5.0 PRODUCT THERMAL DESIGN The data for thermal resistance of …

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

heat exchanger, should be directed to cool the temperature-

sensitive parts ﬁrst, and later be directed adjacent to the

higher power heat producing parts.

D-5.0 PRODUCT THERMAL DESIGN

The data for thermal resistance of junction-to-air, θ

, and

of junction-to-case, θ

should be available for all signiﬁ-

cant power dissipaters.

Time and money can be saved with thermal simulation,

analysis and prediction programs on printed board and

PWA when those prediction programs are veriﬁed by ther-

mal analysis data taken with IR imagers or thermometers.

Thermally sensitive components should be identiﬁed.

These would include technologies with maximum rated

junction temperature T

≤85°C, such as high speed CMOS

and gallium arsenide (GaAs) with aluminum metallization

electromigration limitations or time-dependent-dielectric

breakdown mechanisms.

Where there are thermally sensitive components, heat dis-

sipaters should generally be ‘‘downstream’’ in the air ﬂow;

under certain conditions of PWA geometry, component ori-

entation and relative component heights, turbulence may

result in ‘‘recirculation cells’’ conveying heat ‘‘upstream.’’

Where there are thermally massive heat dissipaters and

fans are required to cool the system, the design review

should include such items as component T

‘‘overshoot’’ on

fan turn-off.

Heat dissipaters should generally be ‘‘upstream’’ of tall

components to avoid recirculation cells.

Heat dissipater review should include those capacitors with

signiﬁcant ripple current and ripple voltage; the data should

include Equivalent Series Resistance (ESR) vs. ripple cur-

rent, temperature, and frequency.

Where lower T

is required, should be considered the fol-

lowing avenues particularly applicable to SMT: thermal

vias, thermal solder joints, thermally conductive adhesives

from component to printed board, and power and ground

plans included in the thermal design. In addition, consider

the following moves: sensitive components ‘‘upstream’’ of

power dissipaters, power dissipaters further apart to reduce

power density, power dissipaters closer to cold wall (edge

of card if card clamps are used), and power dissipaters

‘‘upstream’’ of tall components.

To allow computation of the reliability of solder attach-

ments (see Appendix A), component lead ﬂexural compli-

ance data should be available for the larger components or

for the components with stiffer leads; the critical package

dimensions and environmental temperature swings should

also be known.

Long, tall components such as connectors are ideally

placed parallel to the airﬂow. Placement of these compo-

nents perpendicular to the airﬂow results in the generation

of recirculation cells which reduce the heat transfer from

heat dissipaters or which increase heat transfer to heat sen-

sitive components.

Software for SM PWA and thermal design exist that allow

the designer to perform printed board thermal analysis

together with component junction temperature and reliabil-

ity prediction; some of the programs also address vibration,

fatigue, soldering, transmission line/parasitics aspects of

printed board design. Other programs perform thermal pre-

dictions on printed wiring boards, enclosures, heat sinks,

and plates, considering conduction within a printed board,

air temperature above components, and full convection and

radiation (steady-state and transient). Specialized programs

exist for computation of heat sink, heat ﬁn and cold plate

performance, under steady state and transient conditions.

Powerful computer simulation programs run on engineer-

ing computers or PCs for airﬂow thermal analysis of elec-

tronic equipment cabinets or mixed ﬂuid ﬂows.

D-5.1 Component Level Cooling SM components are

low proﬁle compared to through hole (TH) components

and are therefore more sensitive to the height of surround-

ing components and their airﬂow ‘‘shadowing’’ effects.

Also, the SM component has no airﬂow between the bot-

tom of the component and the printed board; the heat trans-

fer in this area is by conduction through the air.

Guidelines on placement of components based on two

dimensional airﬂow simulations and models do not apply

to typical SM PWAs. Under natural or low air ﬂow forced

convection conditions, heat dissipaters should be ‘‘down-

stream’’ from thermally sensitive components. Taller pas-

sive, temperature-insensitive components should be

‘‘downstream’’ of heat dissipaters, and heat dissipaters

should not be in the airﬂow ‘‘shadow’’ (recirculation

region) of taller ‘‘upstream’’ components. Recirculation

regions result in a decrease in the heat transfer coefficient

of the heat dissipater and an increase in temperature of the

other component(s) sharing the heated recirculated airﬂow.

Under high ﬂow forced convection conditions, recircula-

tion cells can form upstream of a taller, wider component

and reduce its heat transfer coefficient or unexpectedly

increase the temperature of upstream parts. It helps to turn

the long axis of the component parallel to the air ﬂow, to

avoid stacking such components end to end perpendicular

to the air ﬂow, and to increase the distance to upstream

parts. Low frequency oscillations in air ﬂow and subse-

quent variations in component temperature may result

under some circumstances of parts orientation, heat dissi-

pation and air velocity.

D-5.2 Hot Parts (Thermal Considerations) Common

methods for controlling thermal rise of our electronic

assemblies is forced air, liquid cooling systems or natural

July 1996 IPC-D-279

convection. For most products using low power devices,

strategic placement of a few ICs and careful venting in a

housing or enclosure is adequate. For the products used in

a friendly environment, such as the typical air conditioned

office, one board assembly will survive indeﬁnitely. But as

the complexity and component density increases, creative

thermal management techniques must be adopted.

Transferring heat from the component body of ICs and

power regulators usually requires some form of physical

contact to a mass of material attached to the device. Using

hardware or thermally conductive adhesives, as shown in

Figure D-1 is a common practice for thermal management.

The thermal rise can be distributed to a larger surface area,

thereby keeping the component within a recommended

operating temperature range. The efficiency of this dissipa-

tion and distribution is reliant on the area of the heatsink

mass and the loss expected through the interface media.

When attachment hardware is used, a thermal compound is

applied to prevent air, moisture and contaminants from

forming between surfaces. When epoxy attachment is nec-

essary, a thermally conductive material is chosen. The most

efficient thermal conductive epoxy usually contains an

electrically conductive ﬁller. When maximum insulation is

also critical, the heat transfer efficiency from the compo-

nent to the heatsink will be reduced. Researchers are work-

ing on new compounds each year to further improve these

thermal transfer characteristics, and heat transfer tech-

niques for through hole devices are as varied as the engi-

neer’s imagination.

Care must be taken that the thermal planes do not cause

tensile loading of the solder joints.

Typically signal layers on a printed board board are etched

⁄

ounce or 1 ounce copper clad dielectric. But 2 and 3

ounce copper is also available for use on power and ground

layers. Of course, if the heat transfer is a mechanical

attachment to the chassis, the ground plane or planes would

provide a thermal conduit to transfer the heat from the

component’s body into the internal copper layer of this

substrate. When specifying the copper thickness on each

side of one dielectric layer, the layers and copper thickness

should be distributed evenly as the example in Figure D-1.

A non-symmetrical lamination may warp excessively dur-

ing reﬂow solder process.

The thermal rise expected from the components must be

determined, and if this collective temperature rise is

beyond the recommended operating temperature limits of

the device; a careful evaluation of the component grouping

is in order. Distribution of higher power devices on the

outer edge of the substrate panels will provide the neces-

sary correction to control and distribute thermal rise away

from the components. The more direct the thermal path is

away from these components the better; don’t locate high

power devices in the center of the substrate. Rather, place

them closer to the heat transfer edge leaving low power

devices in the center area. See ﬁgure D-2.

Figure D-3 describes techniques used to transfer thermal

rise of a SMT device through the outer dielectric layer of

the printed board to an internal ground plane of metal core.

Multilayer printed boards may have power and ground

planes internally layered and intermixed with signal layers.

This mass of copper planes can be an efficient thermal

conductor to transfer heat to the outer edge of the substrate

or the chassis.

IPC-279-D1

Figure D−1 Thermal Vias and Planes

IPC-279-B2

Figure D−2 Other Methods of Conductive Heat Transfer

IPC-D-279 July 1996