IN THIS ISSUE - WordPress.com · 2011-02-03 · kW), and 19x Model 7039(p655) populate this data center which comprises a mix of switching, communication and storage equipment [2]

SEPTEMBER 2010 | VoluME IV | ISSuE IX

pediaCFD Analysis of a Data Center

Feasibility Study of an LED-Based Lighting System Using Analytical Modeling

Multi-Chip Module Thermal Management

Spreading Resistance of Single and Multiple Heat Sources

Cooling News

IN THIS ISSUE

Advanced Thermal Solutions is a leading engineering and manufacturing company supplying com-plete 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.

EDITORKAVEH AZAR, Ph.D.President & CEO, Advanced Thermal Solutions, Inc.

MANAGING EDITORBAHMAN TAVASSolI, Ph.D.Chief Technology Officer, Advanced Thermal Solutions, Inc.

NORTH AMERICAADVANCED THERMAl SoluTIoNS, INC.89-27 Access Road Norwood, MA 02062 USA T: 781.769.2800 | F: 781.769.9979 | www.qats.com

EUROPEADVANCED THERMAl SoluTIoNS, B.V.De Nieuwe Vaart 50 | 1401 GS Bussum The Netherlands T: +31 (0) 3569 84715 | F: +31 (0) 3569 21294 www.qats-europe.com

All rights reserved. No part of this publication may be reproduced or

transmitted in any form or by any means, electronic, mechanical,

photocopying, recording or otherwise, or stored in a retrieval system

of any nature, without the prior written permission of the publishers

(except in accordance with the Copyright Designs and Patents Act 1988).

The opinions expressed in the articles, letters and other contributions

included in this publication are those of the authors; and the publication

of such articles, letters or contributions does not necessarily imply that

such opinions are those of the publisher. In addition, the publishers

cannot accept any responsibility for any legal or other consequences

which January arise directly or indirectly as a result of the use or adaptation

of any of the material or information in this publication.

ADVERTISINGTo PlACE AN AD IN QPEDIA: Contact John O’ Day at 781.949.2521Qpedia 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 eMagazine 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, contact John o’ Day at 781.949.2521.

SUBSCRIBE ON LINE AT:www.qats.com/qpedia

2

2216106

SEPTEMBER 2010 | VoluME IV | ISSuE IX

Features

6 CFD Analysis of a Data Center Data centers can be considered mission critical facilities, and it is not easy to obtain experimental data for these locations. Thus, much of their planning and optimization is carried out using simulation packages. As with all simulation or numerical results, it is beneficial to have at least one independent result to verify numerical values.

12 Feasibility Study of an LED-Based Lighting System using Analytical Modeling

In general, LED lighting systems are calculated with no solar load. This is a valid assumption if the lighting systems only operate at night. This article discusses the analytical modeling of an LED-based lighting sys-tem for an outdoor application when subjected to various environmental conditions.

20 Multichip Module Thermal ManagementThis article discusses three thermal management techniques used to cool an MCM: a thermal conduction module with Direct Solder Attach Cooling (DiSAC), a dual layer thermal interface (TIM) thermal design, and an MCM design with Small Gap Technology (SGT) and a hermetic seal.

24 Spreading Resistance of Single and Multiple Heat SourcesThe prediction of spreading resistance is not as straightforward as using simple equations. Several variables such as heat source and heat sink geometries, number of heat sources, thermal conductivity of heat sink, and convection on heat sink side must be considered to calculate the thermal resistance associated with the geometry change.

28 Cooling NewsNew products, services and events from around the industry.

3 |QpediaSeptember 2010

4

Two Thermal managemenT Powerhouses Join Forces

Aavid Thermalloy, LLC, the world’s oldest and one of the largest suppliers of thermal management products to the worldwide electronics marketplace, has signed a distribution agreement with Advanced Thermal Solutions, Inc. (ATS), a leader in innovative component cooling solutions. Aavid Thermalloy, with its global customer base and broad network of sales offices, representatives and distributors, will now have access to ATS’ high performance heat sink and attachment technologies to address the cooling of today’s and tomorrow’s electronic devices.

As part of the agreement, ATS will supply Aavid Thermalloy, with its complete line of maxiFLOW™ heat sinks whose patented design and thermal performance is unmatched in the electronics cooling market. In addition, Aavid Thermalloy will also distribute ATS’ patented maxiGRIP™ heat sink attachment systems. Its compact design securely attaches heat sinks to hot components on densely populated PCBs, and detaches quickly and cleanly when needed.

Please contact Aavid or ATS for more information regarding product availability.

aavid Thermalloy To globally disTribuTe aTs heaT sink and aTTachmenT ProducTs

“Around the world, the first name in heat sinks is Aavid Thermalloy. They’re the experts in thermal management technology and manufacturing and provide a worldwide distribution capability that marries uniquely to our innovative products...”

Dr. Kaveh Azar, President and CEOAdvanced Thermal Solutions, Inc.

“We’ve long recognized that ATS develops many of the industry’s most innovative heat

sinks and attachment systems, and that their products offer cooling solutions that can

benefit many of our worldwide customers...”Alan Wong, CEO

Aavid Thermalloy, LLC

Advanced Thermal Solutions, Inc. 89-27 Access Road | Norwood, MA | USAT: 781.769.2800 | F: 769.769.2800 www.qats.com

Aavid Thermalloy70 Commercial St. | Concord, NH | 03301 | USAT: 603.224.9988 | F: 603.223.1790www.aavidthermalloy.com

5 |Qpediaoctober 2009

Heat issues have you tangled up? let tHe experts at ats sHow you How to ensure product reliability, speed time-to-market and reduce costs witH expert tHermal management analysis and design services.

Advanced Thermal Solutions, Inc89-27 Access Road | Norwood, MA | 02062 | USAT: 781.769.2800 | F: 769.769.9979 |www.qats.com

Advanced Thermal Solutions, Inc.89-27 Access Road Norwood, MA USAT: 781.769.2800 F: 781.769.9979www.qats.com

66

CFD Analysis of a Data CenterData centers can be considered mission critical facilities, and it is not easy to obtain experimental data for these locations. Thus, much of their planning and optimization is carried out using simulation packages. As with all simulation or numeri-cal results, it is beneficial to have at least one independent result to verify numerical values. This article describes the physical data center, CFD modelling and assumptions. The results of the simulations will be compared to experimental data previously obtained.

ExperimentalDataThe layout of the data center (232 m x 322 m) shown inFigure 1 indicates the locations of the heat generating racks, the twelve computer room air conditioning (CRAC) units and the perforated tiles that supply chilled air to the room [1].The raised floor height is 711 mm, while the distance from the raised floor to the ceiling is 2.74 m. The servers are lo-cated in a cold aisle/hot aisle arrangement with the aisle width approximately 1.2 m. The cold aisles are populated with 40% open tiles. Detailed measurements were taken of the electronic equipment power usage, perforated floor tile airflow, cable cutout airflow, CARC airflow and electronic equipment inlet air temperatures.

Two primary server racks: 101x IBM Model 7040(p690), (6.7 kW), and 19x Model 7039(p655) populate this data center which comprises a mix of switching, communication and storage equipment [2]. The total heat load from the electron-ics was 1.187 MW and, after adding the CRAC and lighting components, the total data center heat load was 1.268 MW [1]. The total airflow rate in the facility was 77.56 m³/s. This included the flow from the perforated tiles, the cable opening cutouts and leakage from the perimeter of all the floor tiles. The average flow rate of the CRAC units was 6.46 m³/s. The average data center heat flux was measured at 1.9 kW/m². However, within a portion of the data center (59 m x 85 m) that encompasses high-powered servers, the heat flux was 5.5 kW/m².

CFDModelingandAssumptionsThe data center was modeled using a commercial CFD package [3]. The three-dimensional CFD model was solved using the k-epsilon turbulence model. The CFD contains many simplifying assumptions about the room architecture and infrastructure. In particular, the floor plenum is not ex-plicitly modeled in this study. Instead, measured airflow andtemperature data are prescribed as boundary conditions along the raised-floor in the CFD model [1].

For convenience and to be consistent with the available ex-perimental data, the racks were modeled in a compact man-ner without explicit representations of internal components. The airflow rate for p655 racks dissipating approximately 19 kW is assumed to be 1.13 m3/s. The other rack airflow rate is calculated as 0.0755 m³/s·kW, which is consistent with an 11°C temperature rise across the rack. The p690 rack type populates most of the right section of the data center in Fig-

Figure 1. Layout of the Data Center [1].


ure 1. The numerical model for this type of rack is divided into three vertical sections. Each section has been assigned a percentage of the total measured rack power and flow rate, simulating the simplified model for these racks. The densely populated middle section is modeled to be 813 mm high, with 85% of total heat load and 56% of total airflow rate. Sim-ilarly, the bottom section is modeled as 831 mm high with 6% of the total measured heat load and 14% of total airflow rate. The topmost section is modeled to be 356 mm high, with 9% of the total measured heat load and 30% of the total airflow rate. For instance, the racks (p690) dissipating a total of 6.7 kW of power are modeled with the middle section dissipating 5.7 kW and with an airflow rate of 0.283 m3/s (600 CFM), thebottom section dissipating 0.4 kW and with an airflow rate of 0.071 m3/s (150 CFM), and the top section dissipating 0.6 kW and with an airflow rate of 0.151 m3/s (322 CFM). Theother types of electronic equipment (p655 racks, switches, PDUs, storage units etc.) populating the data center are modeled to dissipate power that is uniformly distributed through the entire height of the cabinet. Perforated tiles are modeled with the square dimensions of 0.6 m (2 ft) and with an airflow rate and temperature data from the measure-ments. The cable openings are modeled behind the racks with the dimensions, airflow rate and temperature of the cool air collected by the measurements.

The CRAC units are modeled as black boxes, with the mea-sured airflow rate prescribed uniformly over the CRAC return area. The average return air temperatures to the CRACs arenumerically calculated and compared with the measured values.

The numerical model also includes estimates for leakage flow (defined as the difference between total CRAC and total perforated-tile plus cable-cutout airflow) and the light-ing load. Leakage airflow is modeled as a source of airflow spread uniformly over the entire raisedfloor area. From the experimental measurements, the total leakage as defined above is 3.2 m³/s or 4.2% of the total airflow supplied by the CRACs. Note that if the cable-cutout airflow is included in the leakage figure, the leakage is 38.4 m3/s or 49.5% of the total airflow supplied by the CRACs. The total lighting load is divided equally between the ceiling and the floor. The light-ing load prescribed on the floor simulates the reflected part of the lighting load.

ResultsTable 1 shows a comparison of the numerical and measured values of the average return air temperature to the CRACs. The agreement is generally good to excellent; though with CRAC 2 and 4, the temperature is over and under predicted by 4°C, respectively.

Another method of comparison is to look at the rack inlet tem-peratures. This is what the outcome of the numerical simula-tion is about for a data center. The major cooling challenge in high-density data centers is to maintain the temperature of the cold air entering the racks at a specified level. At higher inlet temperatures, equipment becomes more vulnerable tofailure and reliability problems. Accordingly, rack inlet tem-perature is used as the metric for comparing numerical and experimental values. At the middle of the racks, the rack inlettemperature is reported at 1.75 m above the raised floor and two inches in front. Figure 2 shows the percentage frequen-cy histogram for ΔT for all racks. More than 83% of thecalculated rack inlet temperature values from the numerical analysis are within ±7ºC of the measured values. However, the mean absolute difference and the standard deviation ofabsolute difference are observed to be 4ºC and 3.3ºC, re-spectively. The overall agreement is reasonable, consider-ing the uncertainties in measurements and simplifications in the numerical model.

Table 1. Average CRAC Return Air Temperature [°C].

8

Figure 3 shows the temperature contours at planes 0.5 m, 1 m, 1.5 m, and 1.75 m above the raised floor. Temperature stratification in the data center is obvious, as the average temperature increases with the vertical distance from the raised floor. The temperature of the air entering the lower section of the racks is close to the supplied air temperature.However, the infiltration of the exhaust hot air from the racks and mixing with the cold air in the cold aisle increases the temperature of the air available to the inlets of the upper sec-tion of the racks.

Figure 4 shows the temperature contours at various z planes (z = 0.8 m, 6.5 m, 12 m, 16 m, 20 m) and demonstrates the level of detail included in the numerical modeling. Hot spotsthat consist of racks dissipating ~19 kW can easily be locat-ed. Because of the insufficient amount of cool air supplied in this region, the racks draw warm air from the hot aisle, resulting in local hot spots. Also, as shown in Table 1, CRAC 1 and CRAC 2 have the highest average return air tempera-tures of the 12 CRAC units.

Figure 2. Percentage Frequency Histogram for the Difference of Numerical Values of Rack Inlet Temperature and the

Corresponding Measured Values [1].

Figure 3. Temperature Contour at 0.5 m (A), 1.0 m (B), 1.5 m (C) and 1.75 m (D) [1].

9 April 2010 |Qpedia

PTM-1000TM

PRESSURE TRANSDUCER MODULE

•Provides highly accurate data to within 1% of the full scale•Low pressure measurement from 0-0.15 PSI or 0-0.30 PSI•Additional pressure ranges available as custom options from 0-100 PSI•User friendly stagePRESSURETM software application for easy data viewing •USB connection provides power and signal communication•0-5V output analog voltage for integration into control circuits that require analog voltage as input signals

89-27 ACCESS ROAD, NORWOOD, MA 02062 USA | t: 781.769. 2800 | WWW.QATS.COM

Relieve Some Stress With ATSUnder Pressure?

10

To evaluate the parameters of the data center, the ceiling height was changed from 2.7 m to 3.6, 5.1 and 6.6 m. Table 2 compares the values of the rack inlet temperatures for a few randomly selected racks. Increasing the ceiling height from 2.7 m to 3.6 m improves the thermal performance of the data center, especially in the region of high-density rackpowers; the greater ceiling height provides more space for the hot exhaust air to rise and mix with cooler surrounding air before being recirculated into the rack inlets. However, increasing the ceiling height above 3.6 m does not have any significant influence on the thermal performance. Similar conclusions for the impact of ceiling height on the thermalperformance of data centers were reported in [4].

Figure 4. Temperature Contours Shown for Different Z-Planes [1].

SummaryThis article has discussed the numerical comparison of a large data center to experimental data. The numerical predi-cations of rack inlet temperatures are within ±7°C of the mea-sured values. The mean absolute difference is 4.0ºC and the standard deviation of absolute difference is 3.3ºC. An ex-cellent agreement was observed in the regions of moderate rack powers and for the racks located near the CRACs and for racks located along the aisle. The greatest differences occurred in the region of high density, suggesting a need for more careful and detailed collection of data for such regions and perhaps a more detailed numerical model.Increasing the ceiling height shows no appreciable improve-ment in thermal performance for ceiling heights above 3.6 m. The article shows a good agreement may be achieved between a numerical model and measurement with a prac-tical level of modeling detail and effort – based on readily available facility and rack data, appropriate CFD software, and common computing resources.

References:1. Shrivastava, S., Iyengar, M., Sammakia, B., Schmidt, R. and VanGilder, J., Experimental-Numerical Comparison for High-Density Data Center: Hot Spot Heat FluxesExcess of 500 W/ft², ITherm 2006..

2. Schmidt, R., Iyengar, M., Beaty, D., and Shrivastava, S., Thermal Profile of a High Density Data Center: Hot Spot Heat Fluxes of 512 W/ft², ASHRAE 2005.

3. FLOVENT v5.1 Software now 6Sigma by Future Facilities, London, http://www.futurefacilities.com.

4. Shrivastava, S., Sammakia, B., Schmidt, R. and Iyengar, M., Comparative Analysis of Different Data Center Air flow Management Configurations, IPACK 2005.

Table 2. Rack Inlet Temperature at 1.75 m above Raised Floor for Data Centers with Different Ceiling Heights [1].

11January 2009 |Qpedia

Outdoor LED lighting systems must not only withstand rain and dust, but also solar radiation. In general, LED lighting systems are calculated with no solar load. This is a valid assumption if the lighting systems only operate at night. This article discusses the analytical modeling of an LED-based lighting system for an outdoor application when subjected tovarious environmental conditions.

Lighting SystemThe system consists of a rotatable LED-based lamp fixture which is mounted in an enclosure as shown in figure 1. The enclosure consists of an aluminum compartment on which a dome is placed.The LED lamp fixture is an actively cooled heat sink with 25 cold white LEDs. The junction temperatures of the LEDs need to be determined for different forward currents. Inside the enclosure, components such as motors, drivers and an LED power supply unit (PSU) dissipate heat.

Analytical ModelA thermal resistance diagram is derived to analyze the lighting system as shown in Figure 2.The lighting system is mounted vertically with the dome up or down. It is assumed that the lighting system is not in a good contact with the ground.

DimensionsThe LEDs are mounted on a 200 mm x 200 mm metal core board (MCB). The heat sink base dimensions have the same area as the MCB. The heat sink is 100 mm high. The heat sink is cooled by two fans dissipating 3 W together. The heat sink/fan combination has a thermal resistance of 0.6 K/W.The enclosure is round with a diameter of 400 mm and is 400 mm high.

Feasibility Study of an lED-Based lighting System using Analytical Modeling

12

Figure 1. Sketch of the Lighting System.

Figure 2. Thermal Resistance Diagram of the Lighting System Given in its Electrical Equivalent.


It is assumed that the wall thickness of the enclosure and dome has a negligible effect on the analysis. Therefore, the conduction resistance of the walls is ignored.It is assumed that only the enclosure transfers heat to the outside. The dome consists of a 200 mm high straight section with the same diameter as the enclosure.

Ambient ConditionsThe solar load (direct, indirect and diffuse) is assumed to apply to half the outside area of the enclosure and dome.

For the lifetime condition, the solar load is 700 W/m² at an ambient of 20°C.For the maximum condition, the solar load is 1020 W/m² at an ambient of 40°C.The solar absorptivity, αsolar-dome, of the dome is assumed to be 0.4.The solar absorptivity, αsolar-EN, of the enclosure is assumed to be 0.4.The emissivity of the enclosure is assumed to be 0.85.The calculations will be done at no wind and with a wind speed of 2 m/s using the heattransfer coefficients of 5 and 15 W/m²·K, respectively.

LEDsA total of 25 cold white LEDs will be evaluated at 350 to 1000 mA. The evaluation uses the Luxeon Rebel reliability and lifetime data for the B10, L70 condition with required lifetime of 40,000 hours. The lifetime condition B10, L70 implies that for a specific lifetime, 10% of the LEDs are expected to fail at the specified junction temperature and forward current. Thefailure criterion is when the light output of the LED has been reduced to 70% of its original light out.

The required junction temperatures are calculated using lifetime and maximum temperature conditions. The lifetime conditions are taken from Figure 3, while the maximum condition is from [2]. The lifetime and maximum temperatures are calculated using Equation 1 and are shown in Table 1. Although Table 1 shows that the lifetime conditions are more severe than the maximum conditions for certain forward current ratings, both lifetime and maximum conditions are evaluated in this article.

The LEDs can be mounted on an FR4 PCB or a metal core board. The LEDs have a junction to-board thermal resistance of 10 K/W. The heat dissipation of the LEDs is calculated using Equation 2, where If is the forward current in amps, Vf is the voltage in volts and ηL is the light efficiency. The values used in the study are given in Table 2.

(1)

(2)

Figure 3. Expected (B10, L70) Lifetimes for Ingan LUXEON Rebel LEDs [1].

Table 1. The Maximum and Lifetime Junction Temperatures Required [1, 2].

Table 2. LED Heat Dissipation Assuming a 20% Light Efficiency.

14

Motor and DriversThe motors and drivers are used in the enclosure to rotate the lighting system. The total power dissipation from these devices is assumed to be 10 W.

LED Power Supply Unit (PSU)The LED power supply unit will also dissipate heat and is given by Equation 3, where N is the number of LEDs and ηPSU is the PSU efficiency, which is assumed to be 85%.

Calculation ProcedureThe first calculation step for this analysis is to determine the outside wall temperature. This can be done by applying a control volume around the enclosure. Applying a steady-state energy balance, Equation 4, to the control volume yields Equation 5, which is solved iteratively for the enclosure wall temperature, TEN. Please note that in the calculations, the temperature is in Kelvin and not degree Celsius. This is because the radiation equations use temperature in Kelvin to calculate heat flux. The factor FEN in Equation 11 is the increase in the enclosure

surface area. Any increase in this factor over 1 means that the enclosure surface has become wavy or corrugated. Andtherefore, it has a higher surface area than before.The second step is to calculate the heat sink temperature. This is done by applying a control volume to the heat sink, as shown in Figure 5. Applying an energy balance equation to the control volume yields Equation 13 which is then solved for the heat sink temperature, THS.The inside air is assumed to be well mixed. Therefore, the heat transfer coefficient in Equation 14 is assumed to be 8 W/m²·K.

Figure 4. Control Volume around the Enclosure Wall.

Figure 5. Control Volume around Heat Sink.

(3)

(4)

(5)

(6)Where

(7)

(8)

(9)

(10)

(11)

(12)

First Volume of Qpedia Thermal eMagazine Articles Now Available in Hardcover Book!

ATS has published Qpedia Thermal eMagazine, Volume 1, Issues 1-12, a hardbound, full-color book that compiles 46 technically comprehensive articles offering expert-level coverage on a wide range of heat management issues that impact virtually all of today’s electronic devices.

Qpedia Thermal eMagazine, Volume 1, Issues 1-12, ISBN-978-0-615-23660-5, the hardbound, full color book can be ordered for $94.95 (USD) by calling 781-769-2800 or by email [email protected]. Contact ATS or visit www.qats.com/qpedia.asp for details.

The junction temperature can be calculated applying conduction resistance from the LED junction to the heat sink, as shown in Equation 15. This can then be rearranged to Equation 16 to calculate the LED junction temperature.

15

Advanced Thermal Solutions, Inc. 89-27 Access Road | Norwood, MA | 02062 | USAT: 781.769.2800 | F: 769.769.9979 |www.qats.com

Need Something Special?

LET THE ExPERTS AT ATS DESIgN A CUSToM HEAT SINk SoLUTIoN THAT wILL ENABLE yoU To ENSURE PRoDUCT RELIABILITy, SPEED TIME-To-MARkET AND REDUCE CoSTS. LEARN MoRE ABoUT ATS’ ExPERT THERMAL MANAgEMENT ANALySIS AND DESIgN SERVICES By VISITINg www.QATS.CoM.


Figure 6. Resistance Diagram from Heat Sink to LED Junction.

(13)

(15)

(14)

(16)

16

Several studies have been performed to analyze the various parameters of the lighting system. A summary of these studies is listed in Table 3.

Table 3. Summary of the Studies to be Performed.

ResultsCalculation results are given in Table 4 for different case studies. Figure 7 (a) shows the temperature difference between components. This gives a visual indication of the largest temperature differences and where improvements can be made. Figure 7 (c) shows the heat energy applied to the system from various sources as an absolute value. Values in Figure 7 (d) are similar to (c), but are given as a percentage of the total.For the baseline model, the studies show that LED junction temperatures are above specification when the LEDs are run at 700 mA (S3 and S4) and 1000 mA (S5 and S6).When the improved model was analyzed, it was found that the junction temperature was still above specification for 1000 mA. However, for 700 mA, it was found that the junction temperatures were within specification. When the other temperatures were examined, it was found that the air temperature was 99°C. This will be far too high for the fans, PSU, motor and drivers. Considering the maximum LED storage temperature of 135ºC [2], it can be concluded that the components of the LED, e.g. encapsulate, are at too high a temperature.




17September 2010 |Qpedia

Table 4. Summary of Results.

Figure 7. Temperature Difference between System Components (a), Junction Temperatures as aFunction of Solar Absorptivity (b) and Heat Energy Applied to System Shown in Absolute Values (c) and

Percentage of Total (d).

18

Though the previous studies have shown that 700 mA LEDs are not feasible for the baseline model, the improved model was also studied for the effect of change in the solar absorptivity of the enclosure wall. Although the heat load to the system has been reduced by 272 W (S11) to 188 W (S15), the air temperature remains too warm. Also, they areconsidered to be too shiny for a surface to have an absorptivity of 0.2, Over time, the shininess will be reduced by a dust layer, wind born dust that sand papers the surface, etc.

Looking at the temperature difference plot in Figure 7 (a) for S11, the largest temperature difference is between the enclosure and ambient. It is shown in Figure 7 (d) that the solar load to the enclosure is 50 % of the total heat load to the system. Because the enclosure solar load is 50% of the total load, it is recommended that a solar shield be used todecrease the temperature difference between the enclosure and ambient. The drawback of a solar shield is that wind will have a smaller effect on the system when no solar shield ispresent. In S16, it was shown that the junction temperature decreases by 17°C when the system is in a wind of 2 m/s.The second largest temperature difference is between the enclosure and internal air. To decrease this temperature difference, the air inside the enclosure must be able to movealong all surfaces of the enclosure and dome. This will increase the heat transfer coefficients and can be done by baffling the inside of the enclosure. However, this must be analyzed by computational fluid dynamics.

For the baseline study, it was shown that at 350 mA (S1), the junction temperature is within specification for the lifetime condition. However, the air temperature is quite high for the maximum condition (S2). This was improved by using the superior model (S7).

SummaryThis article has shown how to set up an analytical model of an outside system and perform severity studies. It was found that the solar load for the enclosure was on average 50% ofthe entire heat load for the system. This shows that for an outside system, the solar radiation and therefore the solar absorptivity are important parameters to consider when designing the system. The system is not feasible for a forward current of 1000 mA, while further improvements would have to be evaluated using computational fluid dynamics to determine whether 700 mA is feasible. When the LEDs were run at 350 mA with the improved model, they were within specification. The air temperature was also found to be reasonable for the fans, power supply unit, motor and drivers.

Safety is also an important factor in a system design. One of the parameters needed for CE certification is that the surface temperature not be above 75°C. None of the maximum condition studies with still air were found to satisfy this condition. However, when the effect of wind was included, it was found the enclosure temperature is reduced.

References:1. Luxeon, LUXEON® Rebel Reliability Datasheet RD07, 2007.2. Luxeon, LUXEON® Rebel Technical Datasheet DS56, 2009.

19January 2009 |Qpedia

Features:

» Quick Access Quickly change the test specimen through the front access test section

» Sensor Ports Measure pressure, velocity and temperature through the sensor ports

» Data Center View data and monitor events (with optional controller)

» Flow Characteristics High quality flow with very low turbulence intensity

applications:

» High Temperature Testing Evaluate the effects of elevated temperatures on component and PCB response and reliability

» Heat Sink Characterization Characterize a variety of heat sink sizes for natural and forced convection cooling

» Sensor Calibration Precision temperature and velocity controls allow accururate calibration of sensors

» Component Testing Utilize for individual or mul-tiple component testing

» Multiple PCB Testing Test actual or simulated PCBs for thermal and flow distribution

The CLWT-067 is a research-quality closed loop wind tunnel that provides a convenient, accurate system for thermally characterizing PCBs and individual components at controlled temperatures from ambient to 85°C.

The CLWT-067 wind tunnel produces air flows up to 7.5 m/s (1500 ft/min). With customization, it can generate flows up to 50 m/s (10,000 ft/min) using orifice plates (available optionally). The clear Lexan test section lets the user to view the test specimen and allows for flow visualization.

Unlike open loop wind tunnels, the CLWT-067 re-circulates internal air. This allows the system heater to quickly warm the air to a specific temperature. The testing of boards and components in hot air is a requirement in some NEBS and other standards. The precise controls and temperature range of the CLWT-067 wind tunnel allows its use for testing heat sink performance and for calibrating air and temperature sensors.

The complete wind tunnel fits on most lab benches and is powered from standard AC outlets. It has a smaller footprint than traditional, closed loop wind tunnels or environmental test chambers.

The wind tunnel’s test section can be accessed from the top door for mounting and repositioning of boards, components and sensors. Internal rail guides provide an easy mechanism to install test specimens of different sizes (e.g., PCB, heat sink).

Instrument ports (6) are provided in the front and side walls of the test section for placing temperature and velocity sensors such as thermocouples, Pitot tubes and hot-wire anemometers.

Custom options are also available. Contact ATS for details.

For further technical information, please contact Advanced Thermal Solutions, Inc. at 1-781-769-2800 or www.qats.com

*RoHS Compliant

89-27 ACCESS roAd, norWood, mA 02062 USA | t: 781.769. 2800 f: 781.769.9979 | WWW.QAtS.Com

ovErAll dimEnSionS (l x W x H) 143.6 cm x 49.3 cm x 67.7 cm (56.5” x 19.4” x 26.6”)

tESt SECtion dimEnSionS 41.8 cm x 22.5 cm x 9.5 cm (16.4” x 8.9” x 3.7”)

mAtEriAlS ALUMINUM, PLExIgLAS

floW rAngE 0 To 7.5 m/s (1500 ft/min)

tEmpErAtUrE rAngE Up to 85°C (185°F)

WEigHt 70.7 kg (156 lbs.)

r0_0409

CLWT-067BEnCHtop, CloSEd loop Wind tUnnEl for ElEvAtEd tEpErAtUrE tESting of BoArdS And ComponEntS

CLoSeD LooP WinD TunneL

20

Multi-Chip Module Thermal Management

A multi-chip module (MCM) is a specialized electronic package where multiple integrated circuits (ICs), semiconductor dies or other discrete components are packaged onto a single substrate, facilitating their use as a single component. A cutaway assembly view of an MCM is shown in Figure 1, with various associated components indicated. MCM packaging is frequently used for high-end server systems. In the consumer market, the Intel Core 2 and i7 processors can be considered MCMs, but their power dissipation values are very different from a high-end server. A typical i7 processor has a maximum power dissipation of 130 W or 49.5 W/cm². The recently announced IBM z9 server system MCM uses a total of 16 chips, in which eight chips dissipate 640 W and the total power dissipation is nearly 1 kW [1]. The Hitachi MP6000 has 20 chips, with some chips dissipating nearly 600 W for a total maximum power dissipation of 6.5 kW or 100 W/cm² [2].

It is understandable that packaging several high-power chips in an MCM presents considerable thermal and mechanical challenges [1]. These can include:

• Achieving and maintaining the thermal gaps due to the close proximity, noncoplanarity and tilts of the multiple chips,• Chip and capacitor re-work,• Sealing the MCM to prevent dry-out of the thermal paste,• Corrosion of the C4s (Controlled Collapse Chip Connections), and• Maintaining the package mechanical integrity during the assembly process and operating life

This article discusses three thermal management techniques used to cool an MCM: a thermal conduction module with Direct Solder Attach Cooling (DiSAC), a dual layer thermalinterface (TIM) thermal design, and an MCM design with Small Gap Technology (SGT) and a hermetic seal.

Thermal Conduction Module with Direct Solder Attach Cooling (DiSAC)The thermal conduction module is an MCM concept in which the chip is in physical contact with a water-cooled jacket. The evolution of the thermal conduction module can be seen inFigure 2. The current design of the Hitachi MP6000 is connected to the cooling jacket via solder, the Direct Solder Attach Cooling (DiSAC) method. Because all the components are connected to solid materials with different thermal expansion coefficients, there is an increase in strain due to the components. In previous generations, the chips were mechanically separated from the heat sink by thermal grease/micro fins in order to reduce the stress on the controlled collapse chip connections (C4).

Figure 1. Cutaway Assembly View of a Multichip Module (MCM) [3].


According to [2], the DiSAC design causes nearly all the load induced by module distortion to be supported by the C4.

With the help of finite element analysis and experiment work, Yamada, et al. [2] have been able to clarify the mechanism of C4 strain, improving the module structure to reduce strainand improve the assembly process to reduce defects. Some of these changes include:

• Reducing the contact area of the solder attachment which makes the temperature on the chip more uniform.• Additional C4 connects were used, although this increases the foot print of the module.• In addition to the increased number of C4 connections, the outer C4 connections were also reinforced.• The micro carrier (MCC, see Figure 3) was changed from a glass/copper to a tungsten structure. The height of the MCC was also increased.

Dual Layer Thermal Interface Material Thermal DesignThe dual layer thermal interface material thermal design is physically similar to the thermal conduction module, but instead of using solder between the top of the chip and the jacket, the dual layer design uses two different interface materials in combination with a heat spreader material. This is shown in Figure 4. The design discussed in this section uses four high-power chips, each with two processors and integrated L2 cache [3]. This results in a highly non-uniform power distribution, with regions exceeding 100 W/cm².To address the high power density, each chip uses an individual heat spreader bonded with adhesive [3]. The thermal resistance of the bond line is minimized by using a very thin layer of thermally conductive adhesive, with an effective conductivity of 1.23 W/m·K. A silicon carbide (SiC) heat spreader was selected as it has a coefficient of thermal expansion (CTE) similar to that of the chips and a thermal conductivity of 275 W/m·K. Having matched CTEs avoids thermal stress problems when the module heats up. The heat spreader is optimized by its own thickness as well as the chip spacing. Each heat spreader is individually attachedso that thin bond lines are maintained. The heat spreader is coupled to the copper hat via an adhesive thermal compound.

The benefit of using the ATI, SiC heat spreader and ATC (Advanced Thermal Compound) in comparison to a solution that only uses an ATC, is that the thermal resistance of the ATC layer is significantly reduced by having the heat distributed over the spreader. The heat spreaders have more than twice the area of the chips. The combined thermal resistance of the composite structure is less than the ATC alone.

Figure 2. Historical Progress from the M-880 to the MP-6000 MCM Thermal Conduction Module Concepts. Adapted from [2].

Figure 3. Build Up Sketch of the Thermal Conduction Module with DiSAC [2].

Figure 4. MCM Cross Section Showing Adhesive Thermal Interface (ATI) Cooling [3].

22

MCM with SGT Technique with Hermetic SealThe last of the multi-chip modules discussed in this article uses a thermal paste in combination with Small Gap Technology (SGT) with a hermetic seal as shown in Figure 5.The thermal paste is a non-silicone oil based [1]. This minimizes contamination concerns during chip rework.

The Small Gap Technology (SGT) design uses soldered pistons in the copper hat. The pistons are located over the higher power chips. The paste gap between the chip and piston can be individually customized to a required level by reflowing the pistons during the assembly. The high thermal conductivity of the piston and cap allows effective spreading of the heat before it is conducted to a modular refrigeration unit (MRU). After the pistons have been reflowed, the parts are removed and the effective ATC gap is measured. This is done in order to verify that the hat meets the required specifications. Thereafter, the hat is machined before the MRU is attached. The MRU uses a thin layer of oil as interface material.

A multi-chip module requires that chips can be reworked and replaced if found to be electrically defective. Therefore, the chips are not underfilled. The well-matched coefficients of expansion of the glass ceramic substrate and the silicon chip allow for the required fatigue life of the C4 connections even when not underfilled. Because the C4s are then exposed to the ambient, corrosion concerns arise.An additional concern is the drying and associated performance loss of the ATC when exposed to the ambient [1]. To mitigate both the C4 corrosion and paste drying concerns, a hermetic seal is achieved by a C cross-section ring inserted between the substrate and hat.

The C-ring force is supported by a thin polymer cushion which couples the carrier to the steel base plate..

SummaryThis article has shown that complex design, encapsulation and occasional measurement techniques are required for cooling MCMs with significant power dissipation. It has also been discussed that the reliability of an MCM is dependent not only on the effective junction temperature of the individual chips, but also on the mechanical and thermally induced mechanical strain. Even with sealing, thermal paste TIMs remain susceptible to degradation over the life of a product. The mechanism of thermal degradation is the apparent separation of the oil from the filler matrix.

References:1. Sikka, K., et al., Multi-Chip Package Thermal Management of IBM z-Server Systems, ITherm 2006.2. Yamada, O., Sawada, Y., Harada, M., Yokozuka, T., Yasukawa, A., Moriya, H., Saito, N., Kasai, K., Uda, T., Netsu, T., and Koyano, K., Improvement of the Reliability ofthe C4 for Ultra-High Thermal Conduction Module with the Direct Solder-Attached Cooling System (DiSAC), ECTC 2001.3. Knickerbocker, J., Leung, G., Miller, W., Young, S., Sands, S., and Indyk, R., IBM System/390 Air-Cooled Alumina Thermal Conduction Module, IBM J. Res. & Dev., Vol. 35, No. 3, 1991.

Figure 5. Cross Section of the MCM [1].

23 |QpediaAugust 2010


Heat sinks are widely used in electronic devices to improve the heat transfer from hot components to air and therefore reduce the components’ operating temperature.In most cases of enhanced cooling of electronic components, the heat sink size is much larger than the size of the component package. When heat flows through the electronic component to the heat sink, there is a thermal resistance associated with the changing of cross-sectional area. This thermal resistance is normally referred to as spreading resistance when heat flows from a small cross-sectional area to a large cross-sectional area. On the other hand, the thermal resistance is called constriction thermal resistance when heat flows from a large cross-sectional area to a smallcross-sectional area. Take Figure 1 as an example: a finned heat sink sits on top of a component with a thermal interface material (TIM) compressed between them. The overall thermal resistance from component junction to ambient Rja is,

Where Rjc is the thermal resistance from component junction to case; RTIM is the thermal resistance of thermal interface, including contact resistance; Rs is the spreading resistance due to the cross-sectional area change; Rcond is the conduction resistance inside the heat sink; and Rconv is the convection resistance between the heat sink fins and the air.

As is the trend in the semiconductor industry, electronic packages are becoming smaller as their power density gets larger. Therefore, the spreading resistance becomes relatively more important when determining the overall thermal resistance.The prediction of spreading resistance is not as straight forward as using simple equations. Several variables, such as heat source and heat sink geometries, number of heat sources, thermal conductivity of the heat sink, and convection

24

Spreading Resistance of Single and Multiple Heat Sources

Figure 1. Thermal Resistance of Electrical Package with Heat Sink.(1)


on the heat sink side, must be considered to calculate the thermal resistance associated with the geometry change. Many studies have been made on the effects of heat source geometry on spreading resistance. Yovanovich et al. [1] found that the dimensionless constriction resistances are a weak function of the shape of the contact configuration when the square root of the contact area is used as the characteristic length and the area ratio is kept constant. Naraghi and Antonetti [2] demonstrated this weak relationship with numerical studies of constriction resistances on a single heat source of various shapes located on a heat-flux tube. In Figure 2, is dimensionless constriction resistance, As is the heat source area, Ap is the cold plate area, and Rc

is contraction resistance.

There are few special cases of spreading resistances that have been solved analytically. Because of the idealized boundary condition assumptions of these calculations, they rarely happen in electronics cooling applications. Lee et al. [3] proposed an analytical model with that assumption of a heat source in contact with a large cold plate which is cooled convectively on the opposite side. This model, seen in Figure 3, is pretty close to real applications in electronics cooling.

Lee et al. [3] derived dimensionless equations, in the form of infinite series, to compute the average and maximum spreading/constriction resistances as a function of relative contact size, plate thickness and the Biot number. To simplify the calculation, they proposed approximate equations which are widely used. Their approximate equations for single heat source are discussed here.For Figure 3, the dimensionless heat source radius is:

The dimensionless plate thickness is:

The effective Biot Number is:

Where k is the thermal conductivity of the cold plate.The dimensionless average and maximum resistances are defined as:

And the thermal resistances are:

Figure 2. Dimensionless Contraction Resistance versus Dimensionless Area Ratio [2].

Figure 3. Model for Spreading Resistance Calculation [3].

(2)

(3)

(4)

(5)

(6)

Where

With

For non circular geometries:

Where As is the heat source area, and Ap is the cold plate area.In most electronics cooling applications, the cold plate in Figure 3 should be replaced with a heat sink with multiple fins. Therefore, the thermal resistance Rf should comprise the conduction resistance of the fins and the convection resistance between the fins and air as shown in Figure 1. Song et al [4] showed that Equation 6 has an accuracy of 10% in predicting spreading resistance using heat sink.

For spreading thermal resistance associated with multiple heat sources, analytic solutions are scarce and their applications are limited. Kim et al. [5] proposed a method of transformation of multiple heat sources to single heat source, see Figure 4.

26

(7)

(8)

(9)

Figure 4. Transformation of Multiple Heat Sources to a Single Equivalent Heat Source [5].

Figure 5. The Dimensionless Size of a Single Equivalent Heat Source for Various m/l [5].


For the case of four heat sources, the calculated equivalent heat source area is shown in Figures 5 and 6. Figure 5 shows the dimensionless single equivalent heat source area according to the heat source size. In this Figure, A is the total area of the four heat sources, and Aeq is the area of the single equivalent heat source.Figure 6 shows the dimensionless single equivalent heat source area according to the distance between heat sources.Based on their numerical simulation, Kim et al. [5] proposed the following correlation for an equivalent single heat source:

For a single heat source, the equations proposed by Lee et al. [3] are very useful tools to evaluate spreading/constriction thermal resistance and are widely used in the thermal engineering community. For multiple heat sources, only a few cases, such as Figure 4, have corresponding correlations. For other cases with multiple heat sources, the calculation of spreading resistance is largely completed by using numerical simulation software, especially for irregular geometries.




(10)

Figure 6. The Dimensionless Size of a Single Equivalent Heat Source for Various d/l [5].

References:1. Yovanovich, M. and Schneider, G., Thermal Constriction

Resistance Due to a Circular Annular Contact, AIAA Progress in

Astronautics and Aeronautics, Thermophysics of Spacecraft and

Outer Planet Entry Probes, Vol. 56, New York 1977.

2. Naraghi, M. and Antonetti, V., Macro-Constriction Resistance of Distributed Contact Contour Areas in a Vacuum Environment, ASME 1993. 3. Lee, S., Song, S., Au, V., and Moran, K, Constriction/Spreading Resistance Model for Electronics Packaging, ASME/JSME 1995. 4. Song, S., Lee, S., and Au, V., Closed-Form Equations for Thermal Constriction/Spreading Resistances with Variable Resistance Boundary Condition, IEPS Conference 1994. 5. Kim, Y., Kim, S., and Rhee, G., Evaluation of Spreading Thermal Resistance for Heat Generating Multi-Electronic Components, Itherm 2006.

28

ThermalGapFillerProvidesHighToleranceStackUpThe Tflex™ XS400 Series from Laird Technologies is the latest thermal pad in the Tflex™ thermal gap filler line, of-fering a compliant elastomer gap filler specifically designed to provide moder-ate thermal performance with a ther-mal conductivity of 2.0 W/mK. This soft interface pad conforms with minimal pressure, resulting in minimal thermal resistance, even at low pressure with little or no stress on mating parts.Available in thicknesses from 0.020 inch (0.50 mm) through 0.200 inch (5.0 mm) in 0.010 inch increments, the Tflex™ XS400 thermal material is nat-urally tacky for easy assembly and no adhesive coating is required.

Low Resistance Thermal GapFillerFeaturesMeshLayerSarcon 100GR-FL from Fujipoly is a low resistance, durable thermal inter-face gap filler pad. The gap filler in-cludes an integral nylon mesh layer to prevent distortion and stretching during die-cut operations. Used between hot high-performance semiconductors and heat sinks, Sarcon 100GR-FL trans-fers heat with a thermal conductivity of 2.8W/m-K and a thermal resistance of .67°Cin2/W at 14.5 PSI. The flame re-tardant TIM is 1.0 mm thick and avail-able in sheets up to 300 × 200 mm.

TestStationforThermalAnalysisofElectronicsChassisThe iTHERM-200™ for Advanced Thermal Solutions, Inc. is an integral system of instruments for precisely measuring and recording airflow ve-locity and temperature data at multiple points inside electronics housings and on circuit cards. The new system in-cludes a freestanding wind tunnel, an automated wind tunnel controller, sen-sors and a temperature and air velocity scanner. The system’s large test cham-ber and eight candlestick-style sen-sors allows characterization testing on active or prototype boards and racks, including single ATCA, MicroTCA, cPCI and AMC.

Cooling News New Products, Services and Events from around the Industry

29 |QpediaSeptember 2010 29

Small Fan Suits Space-LimitedApplicationsThe new Superflow-Micro fan from Jaro Components is just 15 x 15 x 4 mm, provides a 0.57-cfm output, a rat-ed speed of 19,000 rpm, and has an IP 57 (water-resistant) rating.The fan protects against harsh environ-ments where dust, water, or humidity could be a problem. It has an oper-ating voltage range of 4.5 to 5.5 Vdc and a rated current of 0.035 A. The micro fan consumes 0.175 W and suits space-limited applications including PDA, MID, smart phone, CCTV, GPS, mask and goggle, pocket projector, mi-cro fuel cell, and air-sensing usages. It provides a minimum 30,000 hours of product-life.

Air Nozzles Provide BlowingForceWithoutMarringAir Nozzles Provide Blowing Force Without Marring Exair Corporation now provides PEEK Super Air Nozzles for deliver blowing force while provid-ing non-marring protection should the air nozzle come into contact with an-other surface. The thermoplastic con-struction resists chemicals, fatigue, and temperatures up to 320°F. The air nozzles provide 80 psig at a <76-dBA sound level. The Model 1110- PEEK Nano Super Air Nozzle with an M6 x 0.75 inlet measures 0.78 in. long, has adiameter of 0.25 in., an air consump-tion of 8.3 SCFM, and produces 8.1 oz of blowing force. The 1102-PEEK Super Air Nozzle has a 1/8 NPT inlet, measures 1.19 in. long, has an air con-sumption of 10 SCFM, and produces 9 oz of blowing force. The 1100- PEEK Super Air Nozzle has a 1/4 NPT inlet, is 1.75 in. long, has an air consumption of14 SCFM, and produces 13 oz of blow-ing force.

30 30 |Qpediaoctober 2009

Getting your company’s message out to over 17,000 engineers and 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 17,000 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 thermalmanagement of electronics.

www.qats.com | 781.769.2800 | 89-27 Access Road Norwood, MA 02062 USA

For more information on how you can generate awareness about your company or products, contact John O’Day at 781.949.2521.

Documents

IN THIS ISSUE - WordPress.com · 2011-02-03 · kW), and 19x Model 7039(p655) populate this data center which comprises a mix of switching, communication and storage equipment [2]