The term ‘immersion cooling’ implies that a device is plunged entirely into a fluid. Following this course, traditional air cooling could be considered immer-sion cooling because an entire system is immersed in air, a gaseous fluid. However, immersion typically refers to a high power system that is immersed in some type of inert fluid, such as mineral oil or a 3MTM FluorientTM Electronic Fluid, e.g. FC-77. Although water is the best fluid for heat transfer, there are many others on the market for use in electron-ics cooling by immersion or in cold plate applications [1]. To gain an appreciation for the thermal transport capabilities of different fluorinert fluids, it is instructive to compare their properties with water. Table 1 shows such a comparison.
The use of liquids is an attractive propo-sition for heat transfer. It provides three broad spectra and distinct benefits:
• High heat transport capability
• No need for a secondary medium between the heat source and the sink
• Effective heat spreading on a larger surface area
The residual benefits of having the heat source immersed in a coolant are also significant. For example, by immersing the electronics in a liquid bath, the ther-mal resistance between the heat source and the sink is eliminated, as shown in Figure 1.
In this issue:
Future Cooling
Thermal Minutes
1
7
Thermal Analysis
Who We Are
14
Thermal Fundamentals19
Cooling News25
26
Table 1. Thermal Properties of Fluorocarbon Fluids and Water [2].
Figure 1. Immersion Cooling of Electronics in a Passive Mode, Providing a Direct Path Through the Liquid from the Source to the Sink.
Figure 1 shows a system where natural convection within an enclosure is used for cooling the devices on a double-sided PCB. If forced convection or boiling had been used, the thermal transport capability would have been significantly larger. Subsequently, higher heat dissipation from the electronics could have been achieved. Figure 2 demonstrates such a capability by comparing different heat transfer coefficients.
Figure 2. Heat Transfer Coefficients for Different Coolants and Cooling Configurations.
From Figure 2, it is quite evident that when fluorocarbons
are used in the boiling mode or single phase, the heat transfer coefficient is substantially larger than from other cooling modes. The cooling capability is clearly demon-strated by Figures 3 and 4 [3]. Figure 3 shows the cooling capacity of 50 W/cm2 in natural convection with moder-ate fluid temperature rise. If forced convection is used, the cooling capacity is increased appreciably at a lower temperature rise.
Figure 3. Comparison of Natural and Forced Convection in Fluorinert FC-72 [3].
In another example, Mudawar shows the thermal trans-port capacity of the same coolant, FC-72, in a boiling phase in different modes [3]. Jet impingement, mini-chan-nel and low-speed cavity flow are compared not only with each other, but against conventional conduction cooled chassis.
Figure 4. Comparison of Boiling FC-72 in Three Different Transport Modes vs Conventional Cooling [3].
Figure 4 describes thermal transport as a function of dif-ferent heat transfer modes in an immersion configuration.
Figure 4 clearly shows that a fluid’s flow delivery method can have a significant impact on its thermal transport capability, e.g., jet impingement, while allowing it to boil in a particular regime. When compared to conventional cooling, there is an order of magnitude change which corroborates the heat transfer coefficient data shown in Figure 2.
This fact suggests that the highest heat transfer can be achieved when the fluid boils. The boiling mode also provides another advantage: a constant component tem-perature that positively affects device reliability. However, there are no gains without penalties. Along with the pack-aging challenges to gain such high levels of cooling, ther-mal overshoots may occur when boiling begins. This is because, in these applications, the fluids tend to be low in surface tension and viscosity. This could create rapid dry out as a result of the liquid-to-vapor transition. Over the past decade, researchers have contributed many articles to the physics of boiling heat transfer and temperature overshoot. Such detail is not the scope of this article, but the reader should be aware of the potential challenges encountered by cooling with boiling heat transfer [4].
It is helplful to also note that both natural and forced convection can be rather effective for electronics cooling. Thermal transport capability may be best described by Figure 5, where the heat flux (W/cm2) is presented as a function of “wall superheat” or the surface-to-liquid tem-perature difference for a typical fluorocarbon coolant [5].
Figure 5. Heat Flux (W/Cm2) as a Function of “Wall Superheat” or the Surface-To-Liquid Temperature Difference for a Typical Fluorocarbon Coolant [5].
Such a high heat transport capability is attractive for electronics cooling when dealing with high power devices or an aggregate of lower power devices concentrated in a small area. Perhaps the most famous immersion cooling application on the market is the Cray-1 computer. Intro-duced during the 1970s, its unique electronics packag-ing, dictated by the need for the highest communication speeds between different devices, required novel cooling to make the system operational.
Figure 6. Cray Super Computer with Cooling Tower in the Background.
In this system all PCBs were installed horizontally and the entire system was immersed in coolant. Stacks of electronic module assemblies were cooled by a parallel forced flow of FC-77. Each module assembly consisted of 8 printed circuit boards on which were mounted arrays
Jet inpingement Sub-cooledBoiling FC-72
0 2 4 1086 12 14 16 18 20
AirF luorochemicalL iquids
AirF luorochemicalL iquidsWater
F luorochemicalL iquidsWater
Water
0.0001 0.001 0.01 0.1 10 100 1000h (W/cm2K)
NaturalConvection
S ingle Phase-ForcedConvection
B oiling
1
Mini-channel flow Sub-cooledBoiling FC-72
Low-speed Cacity Flow Sub-cooledBoiling FC-72
Air Force Liquid Cooled Frame ModuleChannel Flow, Single Phase Indirect Liquid CoolingPolyalphaolefin (PAO)
of single chip carriers. A total flow rate of 4.5 l/s was used to cool 14 stacks, each containing 24 module assem-blies. The power dissipated by a module assembly was reported to be 600 to 700 watts. Coolant was supplied to the electronics frame by two separate frames containing the required pumps and water-cooled heat exchangers to reject the total system heat load to customer-supplied chilled water [5, 6].
To get a scale of the cooling needed to make the Cray-1 operational, if air had been the coolant, the required volumetric flow rate would be on order of that created by a 747 jet engine. Obviously this would not only be physi-cally impossible, it still may not have provided the neces-sary device temperatures to enable the desired commu-nication frequency.
Another industry attempt in immersion cooling was made by IBM. The Liquid Encapsulated Module (LEM) devel-oped at IBM in the 1970s was designed for package-level cooling with pool boiling [5]. Figure 7 shows two sche-matic drawings of a substrate with integrated circuit chips (100) mounted within a sealed package-cooling assembly containing a fluorocarbon coolant (FC-72). One design used an air-cooled heat sink to reduce package complex-ity, while the other integrated a water-cooled cold plate. In either case, internal boiling at the chip surfaces created high heat transfer coefficients (1700 - 5700 W/m2K) to meet chip cooling requirements. Fins were placed internally to condense the vapor created as the result of the boiling and eliminate possible dry out. Either the air-cooled or water-cooled cold plate could be used to cool
the package. As stated by Simons
using this approach, it was possible to cool 4 W chips
(4.6 mm x 4.6 mm) and module powers up to 300 W.
Direct liquid immersion cooling has been used within
IBM for over 20 years as a means to cool high pow-
ered chips on multi-chip substrates during electrical
testing prior to final module assembly [5].
Figure 7. Air or Water-Cooled Liquid Encapsulated Module (LEM) Packages [5].
As mentioned at the start of this article, immersion cool-ing implies that the electronic device is immersed in some sort of inert fluid, whether mineral oil or fluorocarbon. As shown in Figures 6 and 7, the packaging required to make such cooling possible is a major departure from what is most commonly seen in the marketplace. Sys-tems requiring such a level of cooling are typically far more complex and tend to have dedicated facilities for their maintenance and operation. Such a level of care with respect to the cooling system is rather obvious. Foremost, fluorocarbon fluids are rather expensive and should be used in a closed system. In summary, there are many issues about immersion cooling to be consid-ered by the design engineer. These include:
• Pump cavitation prevention
• Vapor condensation
• Fouling – because fluorocarbons tend to pick up packaging materials from components and the boards, especially if boiling is involved
• Coolant compatibility with the seals and the plumbing system
• Surface cleanliness – surfaces that are in contact with the fluid should be properly cleaned
• Burnout management – some fluorocarbons at elevated temperature may produce hazardous gases
It is clear from the above that cooling capacity is not the only point of consideration. Besides, the packaging of a system, its operation and maintenance also play signifi-cant roles in its successful deployment. If these issues can be addressed, as shown by some systems that are currently deployed, immersion cooling can be a very ef-fective method for thermal management of high heat flux electronics.
References:
1. Advanced Thermal Solutions, Inc., Thermal Characteristics of Liquid Coolants for Liquid Loop Cooling, Qpedia, Novem-ber 2007.
2. Danielson, R. Tousignant, L., and Bar-Cohen, A., Saturated Pool Boiling Characteristics of Commercially Available
Perfluorinated Liquids, Proc. of ASME/JSME Thermal Engineering Joint Conference, 1987.
3. Mudawar, I., Assessment of High-Heat Flux Thermal Management Schemes, Itherm, Las Vegas, N.V., 2000.
4. 6th International Conference on Boiling Heat Transfer, Spoleto, Italy, May 2006. http://www.engconfintl.org/6ak.html
5. Simons, R., Electronics Cooling Application, Electronics Cooling Magazine, 1996.
6. Danielson, R., Krajewski, N., and Brost, J., Cooling a Superfast Computer, Electronic Packaging and Production, July 1986
Introduction The primary concern in electronics cooling is to design solutions that keep a device’s junction temperature below some prescribed limit. Maintaining the junction tempera-ture at or below this limit has an enormous impact on the device’s performance and reliability. Every 10°C reduction in junction temperature results in a doubling in lifecycle. To help with calculating junction temperature, a series of standards were developed by JEDEC (the Joint Electron-ic Device Engineering Council). These standards define the terminology and test methods used to thermally characterize electronic devices. They provide a common language that can be understood and applied by thermal engineers around the world.
Of particular interest to engineers are the thermal re-sistance θjx and thermal characterization parameter Ψjx values for a device. With this information one can calcu-late the junction temperature of a device and design an appropriate thermal solution.
Theory and application To make the best use of the JEDEC values for θjx and Ψjx a thermal engineer must understand the JEDEC test conditions for obtaining these values. First, the general JEDEC thermal resistance equation θjx can be written as shown in Equation 1:
(1)
Where,
θjx = Thermal resistance from the device junction to the specific environment [°C/W]
Tj = Junction temperature of the device under steady- state conditions [°C]
Tx = Reference temperature for the specific environment at steady-state [°C]
P = Power dissipated by the device [W]
This general thermal resistance value assumes that all of the device’s heat flows from the junction to the specific environment X. In most cases, X is either the top of the device (Ttop, Tc, Tcase), the circuit board to which the device is attached (Tboard, Tb), or the ambient air (Tamb, Ta). The test procedures for determining each of these thermal resistances are discussed at length in the JEDEC stan-dards. Here, we will focus on three of the most commonly used thermal resistance values [1]. The first of these is θjma which is commonly used in forced convection prob-lems. It is defined in Equation 2:
(2)
Where,
θjma = Thermal resistance from the device junction to moving air [°C/W]
Tma = Average temperature of the moving air under steady-state conditions [°C]
A schematic diagram of the θjma case is shown in Figure 1.
Figure 1. Diagram of a Typical θjma Thermal Resistance Network.
JEDEC-Driven ResistancesHow to use θjx and ψjx
q−
= j xjx
T TP
q−
= j majma
T TP
Tma
Tj
θ jma
P
Airflow
Device
BGAPCB
Tcase
Tj
θ jc
PDevice
BGAPCB
Tboard
Tj
θ jb
PDevice
BGAPCB
Ttop
Tj
Ψjt
PDevice
BGAPCB
Tboard
Tj
Ψjb
PDevice
BGAPCB
The diagram shows that the device power flows from the junction, through all surfaces of the device to the moving ambient air. The ambient air temperature is measured upstream of the device, and is sometimes referred to as the approach air temperature. The air flow velocity is averaged across the cross-sectional area of the test wind tunnel. All velocity and temperature values are taken at steady-state conditions [1].
The value of θjma is often used by engineers to obtain a quick assessment as to whether or not a device will require a heat sink. If the junction temperature calculated using θjma is below the specified limit, a heat sink is usu-ally not required. But when a heat sink is needed, the thermal resistance values of θjc and θjb become important. These can be obtained using Equations 3 and 4:
(3)
(4)
Where,
θjc = Thermal resistance from the device junction to the device case [°C/W]
θjb = Thermal resistance from the device junction to the device base [°C/W]
Tboard = Temperature of the PCB taken at the base of the device [°C]
Tcase= Temperature of the device case taken on the sur-face of the case [°C]
These properties are defined according to the following diagrams:
Figure 2. Diagram of a Typical θjc Thermal Resistance Net-work.
Figure 3. Diagram of a Typical θjb Thermal Resistance Net-work.
These diagrams show that the values of θjc and θjb are “true” thermal resistances (i.e., all of the device’s power goes either to the case or the board). To obtain the θjc value, the JEDEC standards require that all of the sur-faces of the device, except the top surface or case of the device, be insulated. This includes the board to which the device is attached. By insulating in this manner, the heat from the device is effectively channeled to the case sur-face. A cold plate is usually attached to the case surface to keep the surface isothermal (at constant temperature) as required, and to increase the heat transfer to the sur-face, thus reducing measurement error. Similarly, for the θjb value, all of the device surfaces are insulated and cold plates are attached to the perimeter of the JEDEC test board. The following diagrams summarize the JEDEC test setup for these values.
The most effective use for JEDEC thermal resistances is in developing thermal network models for predictive analyses. Because JEDEC-defined thermal resistances are determined through testing procedures that are diffi-cult, if not impossible, to replicate in real world situations, they are less useful for predicting junction temperature with experimental test data. For this purpose, the JEDEC thermal characterization parameter Ψjx is a more useful option. The general term is defined as:
(5)
Where,
Ψjx = Thermal characterization parameter from the device junction to the specific environment [°C/W]
Tj = Junction temperature of the device under steady-state conditions [°C]
Tx = Reference temperature for the specific environment [°C]
P = Power dissipated by the device [W]
The two most useful thermal characterization parameters are shown in Equations 6 and 7:
(6)
(7)
Where,
Ψjt = Thermal characterization parameter from the device junction to the top-center of the device case [°C/W]
Ψjb = Thermal characterization parameter from the device junction to the PCB [°C/W]
Tboard = Temperature of the PCB taken at the base of the device [°C]
Ttop = Surface case temperature of the device (top center of device)
Figures 6 and 7 show the situation schematically:
Figure 6. Diagram of a Typical Ψjt Thermal Resistance Network.
Figure 7. Diagram of a Typical Ψjb Thermal Resistance Network.
The thermal characterization parameter is defined in a similar manner to that of the thermal resistance, but is different in one key aspect: the heat dissipated by the device is not defined to pass through a particular refer-ence point, but can dissipate as it would under normal real-world conditions (i.e., heat will conduct to the board, case of the device, environment, and all exposed device surfaces). Also, the temperature used to calculate the parameter need not be defined on an isothermal surface as required for the thermal resistance values.
While Ψjt and Ψjb are useful to the thermal engineer for an approximate estimate of a device’s junction tempera-ture, it is important to be aware of the limitations. There is significant thermal coupling between the elements that comprise real systems (components, boards, etc.). If this coupling is not properly considered, grossly inac-curate junction temperature values may be obtained if the JEDEC values are used alone. To illustrate this point, a table with predicted junction temperatures from an actual thermal characterization experiment is shown in Table 1.
The approach ambient air temperature and device case temperature obtained by measurement were used with the device manufacturers’ reported JEDEC values for θjma and Ψjt to calculate the junction temperatures listed in the table.
The numbers marked in blue are values within 10°C of the maximum allowable junction temperature, and indicate the need for cooling. Using θjma there are nine (9) devices identified as needing additional cooling. Using Ψjt only two, (2) devices were flagged as being in trouble thermally. Finally, the last column shows the difference in predicted junction temperature, using the two JEDEC standards. All the numbers marked in red show a differ-ence of at least 10°C or more. In some cases the differ-ence is between 20°C and 30°C! This table helps illus-trate why it is important to always have two independent solutions to any thermal problem.
conclusion The JEDEC thermal resistance and thermal characteriza-tion parameters can be used by as a “measure of good-ness” by thermal engineers to model the performance of electronic devices. A solid understanding of the meaning of θjx and Ψjx, as defined through JEDEC standards, will allow for their use in a way that is consistent with the standard, minimizing any errors. As always, care must be taken when using these or any other device standards to theoretically predict junction temperature. These parame-ters should not be used as a singular reference for calcu-lating junction temperature. Electronic systems by nature are complex and unique, and there is significant thermal coupling between system components that must be fully understood for accurate thermal modeling. Finally, a sec-ond independent solution is recommended to verify and validate all theoretical and experimental results.
The use of liquid cooling in electronics is growing fast. Designers and end users are overcoming hydrophobia and embracing liquid cooling as a viable solution for high heat flux applications. Among other factors, an effective liquid cooling system depends on its heat exchanger’s ability to remove heat. These heat exchangers come in different types, sizes and configurations, depending on the application.
Basic equationsFigure 1 shows a typical, water-to-air, liquid cooled, closed loop system. Coolant liquid is pumped through a cold plate that is in contact with the IC component. The heat absorbed from the component is removed in the wa-ter-to-air heat exchanger and the cooled liquid continues its path through the loop.
The heat that is transferred from the water to the air can be calculated from Equation 1
(1)
whereTw2 and Ta are liquid and air temperatures entering the heat exchanger, Cmin is the smaller of the air (Ca) or water (Cw) heat capacity rates that is the product of the mass flow rate and the specific heat at constant pressure.
The effectiveness of the heat exchanger, ε, is defined as the ratio of the actual to the thermodynamically maximum possible heat transfer in a heat exchanger. The surface temperature of the IC component where it contacts the cold plate is calculated from Equation 2.
(2)
Where Tw1 is the cold plate inlet water temperature and Rcp is the thermal resistance of the cold plate from the attachment of the IC component to the inlet of the cold plate, including the interfacial resistance between the component and cold plate.
The temperature rise of water resulting from component heat absorption can be estimated using Equation 3. (3)
In this equation, Cw is the heat capacity rate of the liquid that is the product of the mass flow rate and the specific heat of the liquid. Substituting and manipulating Equa-tions 1-3 provides Equation 4, which relates the surface temperature of the IC component to the performance of the cold plate and heat exchanger. (4)
In the above equations, the subscripts w and a should be changed to the generic subscripts, c and h respectively, to indicate cold and hot fluid properties if the heat ex-changer uses different fluid types.
heat exchanger frontal areaThe surface areas of heat exchangers must be very large, especially on the side exposed to air. This is be-cause the heat transfer coefficient in air cooling is much smaller than that of liquid cooling. Increasing the surface
Compact Heat Exchangers for Electronics Cooling
q C T TL w a= −ε. ( )min 2
T q R Tcom L cp w= +. 1
2 1L
w ww
qT TC
- =
T q RC C
Tcom L cpw
a= + - +.( (.
))min
1 1eε
area on the air side will decrease thermal resistance and allow higher heat transfer between the air and liquid.
Even though large surface areas are needed for efficient heat removal, many cooling applications do not have the space for bulky heat exchanger units. This is particularly true with component and board level applications. It is important to use the available space in selecting the heat exchanger, and to enhance such parameters as the mass flow rate, the use of liquids with higher specific heat, or increasing the power of the cooling fans.
Heat Exchanger Fin TypesThe air side of a heat exchanger is usually intensely finned to increase the thermal exchange between the air and the liquid. The fin types must be suitable for the spe-cific applications. In their classic textbook, Compact Heat Exchangers, Kays and London investigated a number of fin configurations [2]. Among these configurations, straight fins, louvered fins, strip fins, wavy fins and pin fins are shown in Figure 2. Marthinuss and Hall summarized the Kays and London air cooled heat exchanger test data and provided guide-lines on how fin configurations are optimized by combining
heat transfer, pressure drop, size, weight and costs [3].
In this study, the Colburn j-factor (JH) and friction factor (f) correlations are used as heat transfer and pressure drop indices and are described in the following equations. (5) (6)
Where ρ and v are the density and kinematic viscosity of the fluid.
In Equation 5, St is Stanton Number, defined as: (7)
In Equations 5-7, Pr, t, v, NU, and Re are the Prandtl number, wall shear stress, fluid kinematic viscosity, Nusselt number and Reynolds number, respectively. Figure 3 shows the ratio of the Colburn j-factor and fric-tion factor (JH/f) versus the Reynolds number for the fin configurations shown in Figure 2.
Figure 2. Heat Exchanger Fin Configurations Studied by Kays and London [2].
2 / 3PrHJ St=
212
f t
r n=
RePrNUSt =Nu
Thermal analysIs
Figure 3. Heat Transfer/Pressure Drop for Different HEX Fin Configurations [2].
Figure 4. Heat Transfer per Unit Height for Different HEX Fin Configurations [2].
As is evident from Figure 3, when the critical design factor for a heat exchanger is the heat transfer per unit of pressure drop, the straight fin configuration is the most efficient, followed by louvered, wavy, offset and pin fin configurations.
Another important design factor in selecting heat exchangers is their size. For instance, many telecommunication chassis that are packed with electronics components do not have enough space for bulky cooling devices. In Figure 4, the ratio of heat transfer per unit height to the Reynolds number suggests that pin fins are most suitable configuration.
This is followed by louvered, offset, wavy and straight fin configurations.
Another important design factor, e.g. for optimizing heat exchangers used in avionics, is the weight. Here, Marthinuss and Hall ranked louvered fin configuration highest, followed by wavy, offset, pin and straight fin configurations [3].
compact heat exchanger liquid coolant Depending on the system and application, different types of fluids are used in compact heat exchangers. The advantage that a certain fluid brings to a specific application is owed mainly to its larger transport capacity of heat per fluid volume and to more effective heat spreading. To demonstrate this, consider thermal transport in an open system resulting from a change of enthalpy, as shown in Equation 8.
(8)
Where , (ρ is the fluid density, V is the velocity, and A is the cross sectional area), and Cp is the specific heat at constant pressure. If we consider the velocity and cross section as constants, the Cp and ρ will dictate the extent of heat transfer when different fluid is used. Table 1 shows the values of Cp and ρ, μ and k for ethylene glycol, water, and air at 300 °K.
Table 1. Thermodynamic Properties of Typical HEX Coolants.
Table 1 indicates that fluids with higher density and ther-mal capacity are capable of removing greater amounts of heat. In high heat flux applications, using such liquids dramatically increases the heat transfer. However, using liquids with superior thermal transport requires higher pumping power to push the liquid through the system. To
decrease the power required to circulate the coolant, boil-ing heat transfer is employed in heat exchangers. In such devices, a refrigerant coolant vaporizes after it absorbs the heat from a hot source. Based on the complexity of the system, the hot coolant is pumped through a finned condenser section where the coolant is cooled and returned to liquid phase. Because the coolant changes its phase from liquid to gas and becomes less dense, the required pumping power to circulate the coolant is substantially less.
The use of phase change heat exchangers in electronics applications involves other items, including compres-sors, that not only increase the cost, but also bring up the issue of liability. This type of system should obviously be considered as the last resort if nothing else works, for the aforementioned reasons. However, as the need for sophisticated, two-phase heat exchangers increases, manufacturers will develop improved products that are more compact, reliable and less expensive than those currently available.
Compact heat exchangers are becoming an important part of electronics cooling solutions due to the steady increase in power of microelectronics. To use these devices to their fullest potential, it’s important to under-stand their concept, benefits and limitations. To properly utilize a heat exchanger system, design factors such as pressure drop, heat transfer, size, weight and cost must first be prioritized. The simple rule is to choose a simple system that is proven to work before considering a more exotic system.
Light-emitting diodes (LEDs) have long been used in instruments and computers as visual indicators for signal integrity and operations status., LEDs are ideal choices due to their high reliability, low power use and little to no maintenance needs. More recent market interest in LEDs is in their use, not only as indicators, but also as lighting devices. However, as illumination becomes the focus, the power consumption of LEDs has risen dramatically. Device heat fluxes are rivalling those of CPUs and other semiconductor packages. As a result, thermal manage-ment of LEDs has taken center stage for successful implementation.
It is important to remember that an LED is not a high temperature, filament-type lighting device. While a single LED is a cold and efficient light source, high-power LED applications, including arrays of LEDs, need thermal management similar to other semiconductor devices. High temperatures not only degrade an LED’s lifetime, but also result in lower or non-uniform light output, which can significantly affect their application.
Most LEDs are designed in SMT (surface mount technol-ogy) or COB (chip-on-board) packages. In the new 1~8W range of surface mount power LED packages, the heat flux at the device’s thermal interface can range from 5 to 20 W/cm2. These AllnGaP and InGaN semiconductors have physical properties and limits similar to other tran-sistors or ASICs (application specific integrated circuit). While the heat of filament lights can be removed by infrared radiation, LEDs rely on conductive heat transfer for effective cooling.
As higher powers are dissipated from LED leads and central thermal slugs, boards have changed to move this heat appropriately. Standard FR-4 technology boards can still be used for LEDs with up to 0.5 W of dissipation, but
metallic substrates are required for higher levels. A metal core printed circuit board (MCPCB), also known as an insulated metal substrate (IMS) board, is often used un-derneath 1W and larger devices. These boards typically have a 1.6 mm (1/16 inch) base layer of aluminum with a dielectric layer attached. Copper traces and solder masks are added subsequently. The aluminum base allows the heat to move efficiently away from the LED to the system [1].
Thermally dissipating PCBs are not always adequate or suitable for LED applications. Other cooling design choices are available, and it can be challenging to select the most appropriate and cost effective solution for a given application. In this article, we show the required approach for the thermal management of LEDs. This method enables the designer to select the appropriate cooling solution based on the LED’s junction temperature and not on the total power dissipation.
Figure 1. Luxeon K2 Power LED (courtesy of Lumileds).
adVanced Thermal soluTIons89-27 Access RoadNorwood, MA 02026 USA P:[email protected] | www.qats.com
InTroducIng heaT sInK Technology for leds from adVanced Thermal soluTIons
create something amazing.
Advanced Thermal Solutions can solve the most difficult LED/Thermal chal-lenges. ATS provides innovative and cost-effective thermal and mechanical packaging solutions for challenging LED lighting applications.
Our facilities include the most up-to-date software tools for analytical and computational modeling as well as, full thermal and fluids laboratories for experimental characterization of components, boards and systems.
Two parameters play a pivotal role in the success of an LED. These are the cooling method and the optical lens. These factors affect the shape, size and construction of the luminaire that comprises the overall lighting unit. Because long life and fail safe operation are essential for any LED, the cooling process is uniquely critical. An LED’s plastic body is not thermally conductive and the device does not radiate heat. The only effective cooling method is to conduct the heat away through the bottom of the device. Therefore, highly thermally conductive materi-als are commonly used to take the heat from the LED’s back side (see Figure 1). Depending on power dissipa-tion and light emission uniformity, the method of cooling can be passive (heat sink in natural convection) to active (fan-sinks) or can use liquid cooling.
With their basic, robust construction, LEDs can be used in environments ranging from ornamental to such critical illumination needs as automotive headlamps. There-fore, their cooling systems must be designed with the ambient temperature and the specific end use in mind. For example, a car’s headlamp with an under-the-hood temperature of 85-100°C and power dissipation values of 42-90 W requires unique consideration for cooling and re-liability. In other applications, to get the same light output
as an incandescent lamp, the LED lamp will often run on comparable power dissipation values. However, the LED device’s maximum allowable junction temperature is limited to 120-135 °C (up to 185 °C in recent develop-ments). If we compare these limits to an incandescent lamp, which allows filament operating temperatures of 1500-3000 °C, the thermal challenge for LEDs, especially in harsh environments, is the major obstacle to their suc-cessful implementation.
These thermal constraints typically need to be consid-ered:
• Tjunction LED max < 120-185°C
• Tjunction LED lifetime <100~110°C
• PLED = 1-8 W
• Light output is strongly dependent on temperature
cooling options The cooling options for LEDs range from simple natural convection in air to liquid cooling, where a cold plate and liquid loop form the required cooling system. Because most market applications for LEDs shy away from liquid
cooling, the focus of this section will be on air cooling of LEDs.
Most LED lamps employ familiar heat sinking techniques. In some cases, the metal fixture of a luminaire can act as a heat sink, but the thermal requirements of its LEDs must be considered when designing the unit. Increas-ing power density, a higher demand for light output and space constraints are leading to more advanced cool-ing solutions. High-efficiency heat sinks, optimized for convection and radiation within a specific application, will become more and more important.
As with any semiconductor package, thermal resistance plays a significant role in the thermal management of LEDs. The highest thermal resistance in the heat transfer path is the junction-to-board thermal resistance (Rj-b) of the package [2]. Spreading resistance is also an impor-tant issue. Thermally enhanced spreader materials, such as metal core PCBs, cold plates and vapor chambers for very high heat flux applications are viable systems to reduce spreading resistance. [3]
Linear heat sinks are available specifically for LED strips, such as OSRAM SYLVANIA’s DRAGONstick® linear LED strips, which are widely used in architectural light-ing. For example,the maxiFLOW™ linear heat sink from Advanced Thermal Solutions has a patented spread fin array that maximizes surface area for more effective con-vection (air) cooling, particularly when air flow is limited, such as inside display cases.
Active thermal management systems can be used for high-flux power LED applications. These include water cooling, two-phase cooling and fans. Although active cooling methods may not be energy-justifiable for LEDs, reasons for using them include ensuring lumen output, or
maintenance-free operation, or to meet specific wave-length requirements.
Figure 2. An LED-Based Spotlight with a Round, Finned Heat Sink, Visible and IR Views [5].
The led Thermal design and cooling solution selection Process The thermal design of any electronic component, including an LED, consists of three steps [4].
1. Analytical (Integral) Analysis
2. Computational (Numerical) Analysis
3. Experimental Analysis/Verification
Thermal design - analytical analysis Analytical analysis is used to develop a first-order solu-tion. This approach identifies the problem areas (compo-nents and system layout) and ascertains the magnitude of the problem (device junction temperature and required level of cooling). Some analyses can be performed quickly to evalueate the scope of the problem – the so-called “what if” scenario.
computational analysis Computational analyses are used to develop second-or-der solutions to verify results from Step 1. The problem must be well understood in order to develop a model that accurately represents the problem.
CFD (computational fluid dynamics) gives a total 3D picture of the problem. Both heat transfer and flow will be calculated. CFD is typically used to characterize the effect of spreading resistance within the PCB, the flow around the LED lamp and the thermal performance and optimization of a heat sink. Figure 3 shows some results from a CFD study of the LED-based spotlight discussed earlier.
Experimental Analysis/Verification The final product must be tested experimentally, whether for compliance or effective operation. For an LED-based application, the junction temperature is measured by the forward voltage characteristic. The LED has to be cali-brated first with a 10 mA constant current source. During the operational test, the current measurement source is on all the time, then, after stabilization, the operational current is switched off. After turning off the current, the drop in the forward voltage is measured. The thermal mass of the junction is small, which results in a fast cool-down time. This temperature change occurs in less then 1 msec, so the forward voltage has to be measured in microseconds after the event. More information can be found in Farkas, et al. [6].
The forward voltage together with the calibration curve will give the junction temperature under operational con-ditions. This junction temperature must be within specifi-cations for both maximum and typical ambient conditions.
Figure 3. Results of a CFD Study on an LED-Based Spotlight
conclusion This article highlights the importance of thermal manage-ment to the successful use of LEDs. Selecting a cooling solution based on device (LED) junction temperature ensures that the most critical parameter, one that can ad-versely impact its reliability and performance, is identified and thermally managed. More importantly, developing a cooling solution based on an independent analysis ap-proach, i.e., analytical, experimental and computational, provides a high degree of confidence for identifying the most effective cooling solution for high power LEDs.
References:
1. Petroski, J., Thermal Challenges in LED Coolingi, Electron-ics Cooling Magazine, November 2006.
2. Siegal, B., Measurement of Junction Temperature to Confirm Package Thermal Design, Laser Focus World, November 2003.
3. Azar, K., Cooling Electronics Theory and Application, a Short Course in Thermal Management of Electronic Sys-tems, Advanced Thermal Solutions, Inc.
4. Azar, K., Chapter 1, Electronics Cooling and the Need for Experimentation, Thermal Measurements in Electronics Cooling, CRC Press, New York, NY, 1997.
5. LED assembly provided by Universal Science Ltd.
6. Farkas, G., Haque, S., Wall, F., Martin, P., Poppe, A., van Voorst Vader, Q., and Bognar, G., Electric and Thermal Transient Effects in High Power Optical Devices, Twenti-eth IEEE SEMI-THERM Symposium, 2004.
advanced Thermal solutions, Inc. offers custom design solutions for every application. contact us at 781.769.2800 for details.
www.qats.com | 781.769.2800 | 89-27 Access Road Norwood, MA 02062 USA
● Produces flow velocities from 0 to 6 m/s (1,200 ft/min)
● Test Section Dimensions (L x W x H): 34 cm x 29 cm x 8.5 cm (13.25” x 11.5” x 3.25”)
● 12 Sensor ports
● Produces flow velocities from 0 to 5 m/s (1000 ft/min)
● Test Section Dimensions (L x W x H): 61.0 cm x 61.0 cm x 20.32 cm (24” x 24” x 8”)
● 18 sensor ports
● Measures velocities at temperatures from 20°C to 65°C (±1°C)● Capable of controlling velocities from up to 50 m/s (10,000 ft/min) depending on the fan tray ● Features a user friendly, labVIEW based, application software
Accssories
● Produces flow velocities from 0 to 10 m/s (2000 ft/min)
● Test Section Dimensions (L x W x H): 46 cm x 46 cm x 15 cm (18” x 18” x 6”)
● 10 Sensor ports
● Flexible, robust, base-and-stem design allows continuous repositioning and reading ● Measures temperature and velocity Narrow and low profile minimizes the disturbance flow ● Temperature range from -15°C to 120°C (±1°C)
Candlestick Sensor
CWT-100™
CWT-108™
CWT-PCB™
WTC-100
RESEARCH QUALITYLABORATORY TESTED
OPEN LOOPWIND TUNNELS
New Star Heat Sinks Cool High Heat Flux LEDs
Advanced Thermal Solutions, Inc. (ATS) has in-troduced the Star™ Series of heat sinks designed specifically for cooling higher power, surface mount power LED packages. High temperatures not only degrade an LED’s lifetime, but also result in lower light or non-uniform output – effects that can significantly affect their application. The 32 new Star heat sinks provide enhanced conductive heat transfer for effective LED cooling without fans or blowers. Cooling performance in the series (∆Ths-ambient) reaches more than 60 K at up to 16 Watts of power dissipation.
All Star heat sinks are made from light weight alu-minum in a cylindrical shape that fits common LED lamp applications. Their cooling fins are arrayed in a round, star-like cross section that optimizes thermal performance through radiative cooling to the local air flow. A flat surface at one end of the heat sink provides a base for direct mounting of LEDs. Integral threads on the base perimeter allow attachment of brackets and other hardware. All standard sizes are available with an inner thread for convenient attachment of LED lens mounts.
ATS’ Star heat sinks are available in 25, 45, 50 and 75 mm (0.98, 1.77, 1.96 and 2.95 inches) lengths with a standard diameter of 45 mm (1.77 inches). Each heat sink features a black anodized finish that provides corrosion resistance, electrical insula-tion, and improved thermal performance. All heat sinks within the series are RoHS compliant.
Unit pricing for Star Series LED cooling solutions start at less than $8.00 in production quantities. For more information, visit the Advanced Thermal Solutions website, www.qats.com, or call 1-781-769-2800.
Advanced Thermal Solutions is a leading engineering and manufacturing company supplying complete thermal and mechanical packaging solutions from analysis and testing to final production. ATS provides a wide range of air and liquid cooling solutions, laboratory-quality thermal instrumentation, along with thermal design consulting services and training. Each article within Qpedia is meticulously researched and written by ATS’ engineering staff and contributing partners. For more information about Advanced Thermal Solutions, Inc., please visit www.qats.com or call 781-769-2800.
adVanced Thermal soluTIons, Inc. United States89-27 Access Road, Norwood, MA 02062T: 781.769.2800 | F: 781.769.9979 | www.qats.com
getting your company’s message out to over 21,000 engineers & Industry Professionals has never Been easier.advertise in qpedia Today! Qpedia was launched in 2007 as a technology eMagazine focused on the thermal management of electronics. It is designed as a resource to help the engineering community solve the most challenging thermal problems.
The newsletter is published monthly and distributed at no charge to over 21,000 engineers worldwide. Qpedia is also available online or for download at www.qats.com/qpedia.
Qpedia’s editorial team includes ATS’ President & CEO Kaveh Azar Ph.D., and Bahman Tavassoli Ph.D., the company’s chief technologist. Both Azar and Tavassoli are internationally recognized experts in the thermal management of electronics.
for more information on how you can generate awareness about your company or products, contace ray santos at 782.769.2800.
