SMALL SCALE REFRIGERATION SYSTEM FOR ELECTRONICS COOLING WITHIN A NOTEBOOK COMPUTER Rajiv Mongia, Kuroda Masahiro, and Eric DiStefano Intel Corporation 2200 Mission College Blvd. Santa Clara, CA , USA 95052 Phone: (408) 765-9955 Fax: (408) 765-6553 Email:
[email protected] Jim Barry, Weibo Chen, and Mike Izenson Creare Incorporated Fabricio Possamai and Augusto Zimmermann Embraco Masataka Mochizuki Fujikura Limited ABSTRACT The cooling of high power components in notebook computers is uniquely challenging due to space constraints that limit the size of the thermal solution. As a result, for some applications, a method of inserting a “negative” thermal resistance into the heat flow path may be required in order to achieve higher component powers. In this paper we describe a small-scale refrigeration system for the cooling of high power components in notebook form factors. The small-scale refrigeration system includes a compressor, cold plate, condenser, and throttling device. These components are designed for a vapor-compression cycle with iso-butane as the working fluid. All of these components are designed such that the entire system can be incorporated within a notebook form factor. In order to achieve the targeted performance, the cold plate and condenser contain microchannels to efficiently transfer heat to and from the refrigerant. Prototypes of each of the components were built and tested in order to assess their individual performance. A complete form factor loop was also built and tested in order to determine overall system feasibility and performance. The test results show that the targeted performance of the system (COP > 2.25) is achievable in this form factor at the moderate temperature rise expected in this application. KEY WORDS: electronics cooling, vapor-compression refrigeration, evaporators, condensers, compressors COP DF P T W h
NOMENCLATURE coefficient of performance, dimensionless density factor, dimensionless pressure, bar temperature, oC electrical power, W enthalpy, J/kg
0-7803-9524-7/06/$20.00/©2006 IEEE
m&
q
mass flow, g/s heat power, W
Greek symbols Θ thermal resistance (oC/W) η isentropic efficiency, dimensionless Subscripts amb ambient comp compressor cond condenser evap evaporator f fluid j junction sat saturation conditions sup superheat sys system effect INTRODUCTION The use of a remote heat exchanger is an established method of cooling high power components in a notebook computer. The advantage of this configuration is that it allows for a larger amount of heat exchanger surface area than would be possible if a heat sink were attached directly to the CPU. As a result, the remote heat exchanger configuration will have high heat dissipation while still allowing for a light and compact notebook computer. In order for this technique to be useful, the heat from the processor must be transported to the remote heat exchanger, where the heat is dissipated to the air passing through the remote exchanger. Currently, heat pipes are used in order to transport the heat from the microprocessor to the remote heat exchanger. The resulting power dissipation capability of such a configuration can be approximated as:
751
Pressure
T j − Tamb − Tsys
(1) Θ j − amb Where Tj is the maximum junction temperature, Tamb is the ambient temperature, Tsys accounts for system level effects (proximity and preheating), and Θj-amb is the thermal resistance between the junction and ambient, which can in turn be broken down into a series of resistances along the heat flow path. As shown in Equation 1, the power can be increased by reducing the overall thermal resistance. This can be achieved by improving heat pipe and thermal attach technologies or by increasing the size of the heat exchanger. These improvements, however, will eventually reach an asymptotic limit for a given size restriction. Further increases in power will require insertion of a “negative” thermal resistance into the heat flow path; this can be achieved by the use of refrigeration [1, 2, 3, 4]. Power ≈
2
Psat(Tcond)
Psat(Tevap)
1s 1
4
3
Enthalpy
∆hevap
4'
∆hsup ∆hcomp
Figure 2. Thermodynamic state point diagram for a vapor compression cycle. Table 1. State Point Identification
One type of refrigeration system that has been proposed for electronics cooling is a vapor compression configuration. A schematic of such a device is illustrated in Figure 1. As shown, a vapor-compression refrigeration system has four major components: an evaporator, a compressor, a condenser (heat exchanger), and a throttling device. Figure 2 shows a thermodynamic state diagram of the system in Figure 1. Table 1 defines the statepoints.
State point
Heat exchanger (condenser)
Location
1
Compressor outlet/condenser inlet
1s
Compressor outlet, isentropic
2
Condenser outlet/capillary tube inlet
3
Capillary tube exit/evaporator inlet
4
Evaporator exit/superheater inlet
4'
Superheater exit/compressor inlet
Fan Throttling device
Compressor Cold plate (evaporator) Refrigerant
High power component (Tj)
Component power
Figure 2. Schematic representation of a refrigeration system applied to electronics cooling The use of refrigeration breaks the thermal circuit into two legs interrelated by the characteristics of the refrigerator. The power dissipation capability of the refrigeration system shown in Figures 1 and 2 can be expressed as two coupled thermal circuits:
component _ power ≈
HX _ power ≈
T j − T fluid ,evap Θ j − fluid ,evap
T fluid ,cond − Tamb − Tsys Θ fluid ,cond −amb
(2) (3)
Where Tfluid,evap is the temperature of the refrigerant in the evaporator, Tfluid,cond is the temperature of the refrigerant in the condenser, Θj-fluid,evap is the thermal resistance between the junction and the refrigerant in the evaporator, and Θfluid,cond-amb is the thermal resistance between the refrigerant in the condenser and the ambient.
Equations 2 and 3 are coupled by the performance of the refrigeration system through the coefficient of performance (COP) and the temperature rise (Trise) of the refrigeration system: 1 (4) HX _ power = component _ power ⋅ 1 + COP Trise = T fluid ,cond − T fluid ,evap (5) Each of the components of a vapor-compression refrigeration cycle and their requirements for use in a notebook computer will be discussed in subsequent sections. Cold plate/evaporator The cold plate is the component through which the working refrigerant flows and where the heat is transferred from the high power component to the refrigerant. We designed a cold plate where the liquid passes through an array of microchannels that are oriented parallel to each other. Compressor A critical component for the vapor-compression cycle shown in Figures 1 and 2 is the compressor. The compressor will generate a pressure rise (and consequently temperature rise) in the refrigerant. It is the performance of the compressor that will have the highest impact on the overall COP of a vapor compression cycle. For a notebook application, it is desirable that the compressor have the following characteristics:
752
• A small form factor in order to fit within a thin and compact notebook • Lightweight • Highly reliable • High isentropic efficiency • Quiet Since OEM requirements vary widely, it is not possible to determine precise specifications for all applications, rather it is expected that the above characteristics will be evaluated for each proposed compressor design. Condenser Once passing through the compressor, the high temperature refrigerant is condensed to a saturated (or subcooled) liquid. This is accomplished in a microchannel-based condenser. Throttling device The refrigerant liquid is then passed through a throttling device which decreases the pressure (and consequently the temperature) of the refrigerant prior to entering the evaporator. In this work, we use a coiled capillary tube as our throttling device. Working fluid selection We evaluated over 40 candidate refrigerants for this application and concluded that isobutane offered the best efficiency at low pressure ratios and was readily available. Although isobutane is flammable, the miniature refrigerator requires only a very small quantity (only a few milliliters) which can be very well contained.
Figure 3. Microchannel evaporator used to cool the simulated hotspot Condenser Figure 4 shows the microchannel condenser used to reject heat from the refrigeration loop. The condenser was designed to achieve a thermal resistance of approximately 0.5 oC/W, controlled mainly by heat transfer on the air side. The condenser was manufactured by Fujikura.
COMPONENT DESIGN Evaporator As shown in Equation 2, a refrigeration system for hotspot cooling requires a low thermal resistance between the junction and the refrigerant. In order to achieve this low thermal resistance, the evaporator shown in Figure 3 was designed and constructed. The copper evaporator cooling the hotspot has internal flow passages comprising an array of many parallel microchannels (80 µm wide) that enable efficient heat transfer. The copper wall between the microchannels and the hotspot is very thin, enabling a very low overall thermal resistance. The copper evaporator was fabricated by Fujikura. In order to ensure that the mixture entering the compressor is superheated, a second copper evaporator with the same external form factor as the evaporator but a simplified internal structure was used in the loop. The second evaporator contained 21 heat transfer channels that were 350 µm wide on a 700 µm pitch, providing cooling to an area 15 mm square. This evaporator was built by Embraco.
Figure 4. Microchannel condenser with attached cross-flow air heat exchanger Throttling device A capillary tube was used as the throttling device in the refrigerator. The tube has an inner diameter of 0.4 mm (0.016 in.), and total pressure drop was varied by using different lengths of tubing. Compressor A discussed previously, the design of the compressor is critical in obtaining the necessary level of refrigerator performance. The miniature compressor shown in Figure 5 was designed and built by Embraco. The unit is a positive displacement compressor that runs off an inverter using 12V
753
DC power. Typical operation provides 12.5 cm3/s inlet volumetric flow and a pressure ratio of 2.3:1. The control electronics allow adjustment of the operating frequency. During testing the casing temperature was monitored and air cooling was provided to prevent the case temperature from exceeding 70 oC. In a notebook computer, this cooling flow would probably be part of the normal air flow through the case.
closely as possible, instrumentation was limited to externallymounted thermocouples, two absolute pressure transducers, and the temperature sensors internal to the thermal test vehicles (which were used to simulate the microprocessor and chipset). No instruments were used that would lead to unrealistic fluid volumes in the working portion of the refrigerator. Electric power inputs were measured for the heaters in the thermal test vehicles and the compressor. The flow rate of air was measured using a rotameter upstream from the wind tunnel. Figure 8 shows the locations of the thermocouples and also the two pressure transducers, which were connected into the system (via tees) directly upstream and downstream from the capillary tube. Data from thermocouples attached to the refrigerant tubes were used to approximate the fluid temperature at those locations. The beads of these miniature, surface mount, Type T thermocouples were attached to the refrigeration tubing using Omega thermal epoxy, and the first 0.25 in. of lead was insulated when possible. In addition, one thermocouple was attached to the underside of the upstream end of the microchannel evaporator at the thermal interface with the first test vehicle. This temperature was used as an estimate for the evaporator base temperature.
Figure 6. Embraco miniature compressor FORM FACTOR TEST LOOP The components described above were integrated into a formfactor test loop in order to demonstrate the overall system performance as well as observe any system level interactions. This form factor test loop is described in detail in the following sections. Description of refrigeration loop Figure 6 is a schematic showing the main components in the form factor refrigerator. The refrigerator comprises an evaporator, a superheater, a compressor, a condenser, and an expansion device. The microchannel evaporator cools a thermal test vehicle (DTTV) to simulate a high power component. The superheater is a conventional finned evaporator that cools a second thermal test vehicle (YTTV). The miniature compressor accepts low-pressure, superheated vapor from the superheater and compresses it to a higher pressure/temperature for heat rejection to the environment. The compressed vapor flows through the microchannel condenser, emerging as subcooled liquid. The high-pressure, subcooled liquid then flows through the expansion device and emerges as a low-pressure, two-phase flow. In early tests a variable-area needle valve was used for expansion, while in later tests a fixed capillary tube was used. Cooling for the condenser is provided by a small “wind tunnel” that enables good control and measurement of the cooling air flow rate and temperature. Table 2 lists overall design parameters for the refrigeration system. Figure 7 is a photograph of the complete refrigeration system.
Component test data showed the pressure losses in the evaporators and condenser were negligible compared to the pressure rise produced by the compressor, so the two absolute pressure transducers provide all the important data needed to assess system performance. As shown in Figure 6, the pressure transducers are coupled with the loop via short lengths of refrigerant tubing. During operation, the transducers and coupling tubes will cool and fill with stagnant, liquid refrigerant. Although this leads to a slightly higher overall fluid inventory than would be found in an uninstrumented loop, the presence of liquid in these lines will not affect loop operation in any significant way. The metering volume (Figure 6) allows measurement of the quantity of isobutane used to charge the system. In order to avoid unrealistic fluid volumes in the loop, there is no direct measurement of the refrigerant flow rate. Instead, we used calorimetry to deduce the mass flow rate of refrigerant based on heat input from the thermal test vehicles and the change in state of refrigerant between the inlet of the evaporator and exit of the superheater. Data acquisition system All key test data were recorded using a computerized data acquisition system programmed with LabVIEW™ software. Real-time data displays were used to help control the loop during testing. The data were recorded once every second and saved for later analysis.
Instrumentation Since one of the main objectives of the form-factor refrigeration loop was to model a real loop geometry as 754
YTTV Superheater
DTTV Evaporator
Compressor Expansion Valve
Fill Port Condenser Pressure Transducers “Wind Tunnel”
Figure 6. Schematic of Main Components in the Form Factor Refrigeration Loop. Components to the left of Valve B constitute the subsystem for charging the loop.
Figure 7. Form Factor Loop for Refrigeration System Testing. Needle valve shown for expansion device, as used in early tests. Later tests used capillary tube. Loop shown without insulation. Insulation added to reduce heat leaks during testing.
Table 2. Overall loop parameters Isobutane (R600a or 2methylpropane)
Working fluid Max. pressure (bar)
17
3
Fluid volumes (cm ): Compressor
4.5
Evaporator
0.57
Superheater
3.9
Condenser
1.0
Capillary tube
0.16
Tubing
1.2
Pressure transducers
∼0.2
Isolation valve
∼0.1
TOTAL
12 3
Nominal liquid volume (cm )
5.2
Active liquid volume (cm3)
1.2
Tube material
Copper
Tube outer diameter (mm)
3.175
Tube inner diameter (mm)
1.65
Temperature measurements
Type T thermocouples
Pressure measurements
Omega model PX105200G5V (0 to 200 psi) Omega model PX105500G5V (0 to 500 psi)
Figure 8. Temperature and pressure instrumentation Data reduction During testing, periods of steady operation were identified and data were averaged over these periods. These average data were then used to calculate parameters to characterize the steady-state performance of the system and individual components. Thermodynamic properties of isobutane were computed using NIST standard reference database 23 (REFPROP 7.0). Subscripts that identify state points in the cycle use the convention shown in Figure 3 and Table 1. Overall COP The overall coefficient of performance for the refrigeration system was computed based on the power inputs measured for the compressor, DTTV, and YTTV: 755
COP =
q DTTV + qYTTV Wcomp
LOOP AND COMPONENT PERFORMANCE
(6)
where qDTTV is the heat input from the first test vehicle (W), qYTTV is the heat input from the second test vehicle (W), and Wcomp is the electric power input to the compressor (W).
This section presents data from a series of steady-state operating points, in which key parameters were varied to test a range of conditions that simulate operation in a notebook computer. Data from these tests demonstrate the overall performance of the refrigeration system as well as the performance of individual components in the loop.
Estimation of mass flow rate
& , g/s) is computed by The mass flow rate in the loop ( m calorimetry as follows:
m& =
h4 ' − h3 q DTTV + qYTTV
(7)
Baseline operating conditions and test matrix Table 3 lists the baseline test conditions. This operating point is intended to correspond to a typical operation of the refrigeration system. A matrix of test points was selected by varying key parameters around this baseline condition. Table 4 lists the complete set of steady-state tests.
In this equation, h3 is the enthalpy (J/kg) of the fluid at the inlet to the evaporator, which is identical to the enthalpy of the fluid at the entrance to the capillary tube since the expansion process is isenthalpic. Therefore h3 is calculated using REFPROP as h3 = h2 = h(T2,Pcond), where T2 is the fluid temperature (oC) at the exit of the condenser and Pcond is the condenser pressure (Pa). The enthalpy at the exit of the condenser is easily computed because the fluid there is singlephase, subcooled liquid. h4' is also easily computed since the fluid leaving the superheater is single-phase, superheated vapor: h4' = h(T4',Pevap). Temperature T4' was measured by thermocouple TC2 in the refrigerator (Figure 9). Compressor efficiency Typically compressor efficiencies are stated as isentropic efficiencies. The isentropic efficiency is the efficiency relative to a “perfect” compressor in which: (1) there is no heat transfer to or from the fluid during compression, and (2) all the work input to the compressor is transferred without loss to the fluid. Defining the efficiency this way makes it simple to use in computing overall cycle performance. The isentropic efficiency (η) was calculated as follows:
η=
(h1s − h4' ) m&
(8)
Wcomp
where h1s is the enthalpy computed for an isentropic compression from point 4' at the compressor inlet: h1s = h(s4',Pcond). Evaporator thermal resistance A key metric for cooling performance involving the evaporator is the junction-to-fluid resistance:
Θ j− f =
q DTTV T j − Tsat ( p evap )
(10)
where Θj-f is the junction-to-fluid resistance (°C/W), Tj is the junction temperature (°C) of the junction measured by the DTTV, and Tsat(Pevap) is the saturation temperature (°C) corresponding to the evaporator pressure.
Table 3. Baseline operating conditions for refrigerator Evaporator temperature (°C) Evaporator pressure (bar) Superheat at compressor inlet (°C) Condenser temperature (°C) Condenser pressure (bar) DTTV power input (W) YTTV power input (W) Density factor in DTTV Pressure ratio Mass flow rate of isobutane (g/s) Air inlet temperature (°C) Air flow rate (SCFM) Length of capillary tube (m)
50 6.85 10 90 16.4 40 10 1.0 2.4 0.26 50 ∼2.8 1.1
Refrigeration efficiency The refrigeration system demonstrated good overall efficiency. Figure 10 shows the coefficient of performance computed for the tests. Under normal conditions, the COP was about 2.25. The plot shows that the COP was not sensitive to the density factor (as expected). COP did increase when the overall thermodynamic conditions were more favorable; for example the maximum COP in this series of tests (3.70) was achieved in Test 5 when the DTTV power was reduced and the evaporator and condenser temperatures increased and decreased by 10°C from baseline conditions, respectively. Increased air flow rate in Tests 8 and 9 boosted the COP slightly by changing the saturation conditions in the condenser. While the COP alone is of interest, another useful performance metric for this microscale refrigeration system would be to compare its performance to the theoretical ideal COP for a Carnot refrigerator. This ratio is plotted in Figure 11 as a function of pressure ratio. As shown, the microscale refrigeration system shown here achieves approximately 2530% of the Carnot efficiency. This is comparable to the efficiencies (relative to Carnot) that are achievable in today’s household refrigerators.
756
DF = 2.7
qevap=45 W 45°C Tair
3.00
2.00
Baseline
3.8 SCFM (50°C / 80°C)
1.00 0.50 0.00 1
2
3
4
5
6
7
8
9
10
3.8 SCFM (50°C / 80°C)
0.40 Baseline
0.20
qevap = 35 W (60°C / 80°C)
0.10 0.00
0.95 m
3.2 SCFM
0.30
1
qevap = 45 W 45°C Tair
2
3
4
5
6
7
8
9
10
11
12
0.70
0.53
0.29
0.27
0.28
0.26
0.26
0.25
0.27
0.29
0.26
0.40 Compressor Isentropic Efficiency
FFL
0.35
Separate Effects Tests
0.30 0.25 0.20 0.15 0.10 0.05 0.00 0.0
0.5
1.0
1.5
2.0
2.5
3.0
Pressure Ratio
Figure 13. Compressor isentropic efficiency
In order to demonstrate that the components described in this work can fit within a Thin and Light mobile system, a mockup of a refrigeration system was integrated into an Intel Concept Platform. This mock-up was non-operational, but actual components were used to demonstrate that the pieces can fit within a mobile platform. This mock-up is shown in Figure 14. To the authors’ knowledge, this is the first time that a refrigeration system has been developed that can fit within the tight confines of a notebook platform and yet still operate with high refrigeration efficiencies.
0.95 m
3.2 SCFM
1.50
DF = 2.0
0.50
NOTEBOOK INTEGRATION
2.50
COP
0.95 m (60°C / 90°C)
0.60
Figure 12. Junction-to-fluid thermal resistance
0.95 m (60°C / 90°C)
qevap = 35 W (60°C / 80°C)
DF = 2.0
3.50
DF = 2.7
0.70
theta-jf 0.28
Compressor performance Figure 13 plots the compressor isentropic efficiency as a function of pressure ratio for the tests. The efficiency generally increases with increasing pressure ratio, and achieves a value in the range 0.33 to 0.35 at a pressure ratio of 2.4 corresponding to the baseline operating point. Higher efficiency at higher pressure ratio is not surprising, since the compressed gas temperature increases at higher pressure ratios. The hotter compressed gas will lose more heat to the environment, which will tend to decrease the compression work relative to isentropic (i.e., insulated) compression. Compressor performance in the form-factor loop was consistent with the data from prior component tests. 4.00
0.80
Junction to Fluid Thermal Resistance, ◊ jf (°C/W)
Junction-to-fluid thermal resistance The junction-to-fluid thermal resistance (Θj-f, °C/W) is based on the DTTV power input and the difference in temperature between the DTTV junction and the evaporator saturation temperature. As expected, Θj-f varies little with the cycle operating conditions but is very sensitive to the power density factor. Figure 12 shows the values of Θj-f measured in the tests. The thermal resistance value for the baseline case (about 0.28 °C/W) is consistent with what is expected for the microchannel evaporator. The resistance remains constant for changes in loop operating conditions, but reaches values as high as 0.70 °C/W for operation with high density factor. Consistent with expectations, Θj-f scales almost directly with the density factor.
11
12
Test Number
Figure 10. Overall coefficient of performance
COPact/COPideal
0.40
0.30
0.20
0.10
0.00 1.0
1.5
2.0
2.5
3.0
Pressure Ratio
Figure 11. COP ratio vs. pressure ratio
Figure 14. Mock-up mobile platform with miniature refrigerator components 757
REFERENCES CONCLUSIONS In this paper, we have described a miniature refrigeration system that is capable of obtaining high performance in a form factor compatible with today’s thin and light notebooks. This refrigeration system is enabled by a high-efficiency miniature compressor and microchannel-based condenser and evaporator. The miniature refrigeration system described here is capable of obtaining high thermal efficiencies (COP > 2.25) at temperature rises of interest in the mobile space. This is the first time to our knowledge that a miniature refrigerator consistent with a notebook form factor has been designed and demonstrated at these levels of efficiency.
[1] Phelan, P.E., 2001, “Current and Future Miniature Refrigeration Cooling Technologies for High Power Microelectronics”, Semiconductor Thermal Measurement and Management Symposium, San Jose, CA, March 20-22, 2001, pp. 158-167 [2] Maveety, J.G., et. al., 2002, “Thermal Management for Electronics Cooling Using a Miniature Compressor”, International Microelectronics and Packaging Society, Palo Alto, CA, October 24-26, 2002 [3] Peeples, J.W., 2001, “Vapor Compression Cooling for High Performance Applications”, Electronics Cooling, Vol. 7, pp. 16-24. [4] Trutassanawin, S. and Groll, E.A., Refrigeration Technologies for High Electronics Cooling”, Proceedings of Refrigeration and Air Conditioning Lafayette, IN, July 12-15, 2004
2004, “Review of Heat Dissipation the International Conference, West
Table 4. Summary of test conditions Evap T.sat (ºC)
Air Inlet Temp. (ºC)
Air Flow Rate (SCFM)
Capillary Length (m)
Test
Variable
Cond T.sat (ºC)
1
Baseline
87.7
51.2
49.6
2.87
89.6
52.2
49.5
89.1
51.7
89.0
2 3 4 5 6
Density Factor Density Factor Baseline Evap Power Evap Power
Thermal Power Input
Density Factor
DTTV (W)
YTTV (W)
Total (W)
1.10
40.3
10.1
50.4
1.0
2.87
1.10
40.8
10.5
51.3
2.7
49.7
2.87
1.00
40.3
10.6
50.9
2.0
51.9
49.9
2.87
1.10
40.0
10.6
50.6
1.0
80.5
59.6
50.2
2.87
1.10
34.0
10.0
44.0
1.0
91.0
47.7
45.1
2.87
1.10
45.1
10.4
55.5
1.0
7
Baseline
88.3
51.7
50.4
2.75
1.10
40.1
10.1
50.2
1.0
8
Air Flow
84.6
49.8
50.2
3.21
1.10
40.2
10.2
50.4
1.0
9
Air Flow
80.8
47.7
49.8
3.83
1.10
40.5
10.3
50.8
1.0
10
Baseline
89.2
51.9
50.5
2.75
1.10
40.0
10.1
50.1
1.0
87.8
49.0
50.3
2.75
0.95
40.0
9.9
49.9
1.0
86.5
61.0
50.9
2.75
0.95
39.6
10.8
50.4
1.0
11 12
Capillary Tube Capillary Tube
758
