Jansson

7/29/2019 Jansson

1/7

QEX July/August 2010 3

Dick Jansson, KD1K, ex WD4FAB and WA1QLI, ARRL Technical Advisor, AMSAT Thermal Designer

1620 Mayfower Ct, B-125, Winter Park, FL 32792; [email protected]

Heat Pipes

1Notes appear on page 9.

Your computer may be using this technology to cool the CPU.Learn how you can use heat pipes in your Amateur Radio station.

Many hams will think of heat pipes as

new devices, but in reality they have beenaround for more than 30 years. It is just that itis only relatively recently that these deviceshave been perfected and placed into thecommercial consumer marketplace. Morethan 20 years ago I studied the applicationof heat pipes to solve a difficult thermalproblem in the AMSAT Phase 4 satellite(geostationary ham radio communicationsplatform).1 At that time I found a (then) 10+

year old reference and description of heat

pipes.2Study Figure 1 and the following descrip-

tion and you will get the idea: The heat pipeis a thermal linkage of very high thermalconductivity. It is a closed, evacuated cham-ber lined with a wick. Heat is transportedby evaporation of a volatile uid, which iscondensed at the cold end of the pipe andreturned by capillary action to the hot end.The vapor passes through the cavity. Heatpipes consist of three zones or sections: theevaporator, the condenser and an adiabaticsection connecting the two. In some designs

the adiabatic section may be very short.

This device offers a number of importantproperties useful in electronic equipmentcooling systems. It has many times the heattransfer capacity of the best heat-conductingmaterials, while maintaining an essentiallyuniform temperature and transporting heatover distances of several feet. It requires nopower and operates satisfactorily in a zerogravity environment.

[For readers unfamiliar with the physicsof heat transfer, it may be helpful to turn to adenition of an adiabatic process, such as canbe found on Wikipedia (http://en.wikipedia.

Heat In

Vapor

Liquid

Evaporator

End

Condenser

End

Adiabatic

Section

Direction of Heat Flow

Shell

Wick

Heat Out

QX1007-Jansson01

Figure 1 Heat pipe basics. KD1K drawing

7/29/2019 Jansson

2/7

4 QEX July/August 2010

org/wiki/Adiabatic_process ). The shortdenition, taken from the Wikipedia source,is that an adiabatic process is a thermody-namic process in which no heat is transferredto or from the working uid. For example,adiabatic heating occurs when the pressureof a gas is increased from work done on it byits surroundings, such as by a piston. Dieselengines rely on adiabatic heating during theircompression stroke to elevate the temperaturesufciently to ignite the fuel. Adiabatic cool-ing occurs when the pressure of a substanceis decreased as it does work on its surround-ings. Ed.]

Putting this description into other terms,

Cold

Deep

Space

Shadow

Solar

Irradiation

on

Spacecraft

QX1007-Jansson02

Figure 2 Solar heating problem for stabilized, pointing satellites. KD1K drawing

Figure 3 Exploded view of AO-40 spacecraft, showing major components and heat pipes. Wilfred Gladisch, AMSAT-DL, drawing

7/29/2019 Jansson

3/7


a heat pipe is a linear thermal conductor thatis hundreds to thousands of times better thanthe purest of copper or silver bars they justdont get any better than that.

The applications and materials of heatpipes have been perfected over the years intosome truly meaningful devices. Spacecraftheat pipes are considerably more demandingthan those used in the commercial applica-tions, such as the cooling of modern-daycomputer CPUs. I shall try to illustrate thesedifferences and also to show how you can useheat pipes.

Space ApplicationOne space application that has been expe-

rienced is that of the AMSAT AO-40 (for-merly the Phase 3 D (P3D) Program beforebeing launched and placed into service).Recalling the AO-40 effort will illustratethe evolution of its thermal design. Becauseof the spacecraft size, the thermal activi-ties happening on one side of the spacecraft

would not signicantly affect another side.In a stabilized, pointing mode orientation,one side would continuously face the Sun,while the opposite side would never see anysunshine. See Figure 2. As a result, the sunside would become quite warm while theback side would get dreadfully cold, all ofthis even though the structure is made of alu-minum, which is thought of as a pretty goodheat conductor. In these dimensions, sheetaluminum structures of the size of AO-40 areincapable of moving enough thermal energyto make a difference in this matter.

In addition to solar heating, the RF

transmitters installed in AO-40 had outputsup to 300 W PEP, with average power dis-sipation levels in the 25 to 75 W categories.This meant that there were power transistors(MOSFET devices, actually) with quite smallmounting surfaces trying to get rid of tens-of-watts of power. Numerically, heat uxes of30 W/cm2 could be expected at the mountingface of any single transistor. To give you afeeling for this type of heat ux, the sunshinethat you get at the beach on a clear summerday is in the order of 60 mW/cm2 or 1/500thof the ux at the mounting surface of eachof these transistors! To further heighten this

problem of cooling of high-powered solid-state electronics, the semiconductor junctioninside these devices should not be allowedto become any warmer than about 125C forreliability reasons. These devices cannot beallowed to glow red like the handy radiantheater in your house!

The solution to all of these basic thermalproblems on AO-40 was to employ the use ofheat pipes inside the spacecraft. This arrange-ment is shown in Figure 3, which illustratesthe basic six-sided prismatic structure ofthis spacecraft. The six Equipment Panels

were mounted about 220 mm (8.7 inches)below the exterior side panels. The key tothis design is that the equipment panels werein contact, actually bonded, with the fourheat pipes and that the heat pipes were notconnected directly to any external panel.The basic tenet here is that the waste heatremoved from one part of the spacecraft wasused to keep the cooler parts suitably warm,a thermal redistribution system, if you will.This thermal design concept provided forheat rejection from the spacecraft by radiantexchange between the equipment panels andthe side panels, which can get very cold, asintended. As you can also see in Figure 3,all of the basic electronic modules weremounted to those same equipment panels.I do not know of other satellite designs thathave used this principle.

The heat pipes used in AO-40 were a verysophisticated aluminum extrusion. Thesewere available in a maximum length of16 feet, just long enough to be formed into

a nearly complete hexagon on the inside ofthe equipment panels. Working with this par-ticular heat pipe extrusion was a continuationof efforts started with the earlier Phase 4 pro-gram. That experience was usefully assimi-lated into the AO-40 spacecraft. An enlargedview of this extrusion is shown in Figure 4.As you can see, the liquid wick is really agroup of 27 closely spaced channels in thewall of the extrusion. The geometry of thesechannels allowed a very substantial ow ofliquid returning from the condenser sectionto the evaporator section, thus providingfor an equally substantial thermal transfer

rating for the heat pipe. In AO-40 use, therewas no clear geometric denition for theseevaporator and condenser sections, as thepipes were exchanging thermal energy with

the spacecraft in many locations, but this waswhat was intended for this design.

While the bonding of the heat pipes to theequipment panels was sufcient for the heatexchange with the panels, such methods wereinsufcient for the high-powered transmittermodules. Coupling high power transistors tothis heat pipe system therefore required someadditional measures to be taken. Figure 5shows a cross-sectional view of this arrange-ment. At the high heat uxes presented bythese transmitters, coupling to just one sideof the heat pipe was not adequate. Thereforea specially machined aluminum Heat PipeClamp was made of high purity aluminum(for maximum thermal properties) and care-fully bonded to each side of the heat pipeusing a special xture. Such a design stepwas intended to bring at least two other sidesof a heat pipe into thermal contact with themodule heat sink. These heat pipe clampswere, in turn, bonded to the equipment pan-els at the same time as the heat pipes. On the

other side of this thermal path, the transmittermodules were equipped with two high-purityaluminum heat sink plates for the mount-ing of the power transistors. These moduleswere also closely mounted to the equipmentpanels as well as being bolted through to theirrespective heat pipe clamp. The nal step,shown in Figure 5, was to mount the powertransistor to the heat sink, using space-ratedsilicone grease. All of these bonding and bolt-ing steps are necessary to insure a very highquality thermal path from the transistor to theheat pipe uid (anhydrous ammonia, NH3).Figures 6 and 7 are photo illustrations of these

heat pipe clamps as a separate assembly andas installed in the spacecraft.The thermal performance of this complete

high-power coupling of transmitter transis-

Figure 4 AO-40 heat pipe extrusion cross section. KD1K drawing

0.550

QX1007-Jansson04

7/29/2019 Jansson

4/7


tors to the heat pipe system was the subjectof a fairly extensive thermal analysis as wellas a specially constructed physical thermalmodel, including a short section of activeheat pipe. Building such a thermal modelwas one of the extra-special fun things that Idid for the AO-40 program (thermal design-ers are strange people, after all). The needfor such elaborate evaluation is that in thevacuum of space, the satellite does not havethermally conductive air between parts, and

nearly none of the parts will be in perfectcontact. These interstitial spaces needed tobe lled with a compliant material to permitthe thermally conductive joining of the parts.The analyses and testing checked out the useof these various materials and I made trade-off evaluations as a result. The bottom lineof this effort was a transistor thermal mount-ing with an effective thermal resistance(between the transistor case and the heat pipeuid) ofsf = 0.27C/W; that is quite low,even lower than that of the transistor junc-tion to its case. The results of this effort were

Figure 5 AO-40 module mounting to heat pipe cross section view. KD1K drawing

Interface Silicone

Grease

Interface Bonding RTV

Silicone Compound

Heat Pipe

Heat Pipe Clamp

Equipment Panel

Module Heat

Sink Plate

Printed Circuit BoardRF Power

Transistor

TUBE

10-32 Socket HeadScrew

QX1007-Jansson05

Figure 6 AO-40 heat pipe and clamp. KD1K photo

7/29/2019 Jansson

5/7


very gratifying, as the initial concerns that itwouldnt work well had been great.

I have emphasized the use of heat pipesin space, so far and for good reason.Heat pipes are sensitive very sensitive to environmental accelerations, a fancydescription of the Earths gravity. On Earth,if the evaporator section of an ammonia heatpipe is higher than the condenser section bymuch more than 0.1 inch in a foot of length,then the capillary forces of the liquid wickmay not be strong enough to draw the liquidup hill and overcome these gravitationaleffects. If, however, the evaporator sectionis lower than the condenser section, thengravity will help the ow of liquid back tothe evaporator section which is how heatpipes work for computer users here on Earth.In the gravity freedom of space, this is notvery much of a problem, as spaceight isthought of as being in zero gravity. This isnot an absolute condition, however, as space-craft motions can cause local accelerations

that can pose problems. Let me describe justsuch a case.AO-40 was designed as a cylindrical

spaceframe, albeit a hexagonal cylinder. Itwas also designed to initially spin about itsprincipal axis until the mission was readyto stop the spin, deploy the solar panels andoperate as a three-axis stabilized satellite. Forreasons not germane to this discussion, thestable mode of operation was never achievedfor AO-40. Spin-stabilized operation is use-ful for times when the orbit-adjusting propel-lant motor was operated, because such a spinwill average out any small errors in propel-

lant motor alignment. As a result, for motoroperation, AO-40 was spun up to more than10 rpm. Since the heat pipes were placed onthe walls of a cylindrical satellite, they werenominally equidistant from the spin axis andthus saw a uniformly distributed centrifugalaccelerationexcept. That exception wasthat the heat pipes were part of a hexagonalcylindrical form, not a circular cylinder. Thatmeant that the corners of the hexagon werefarther from the spin axis than the ats ofthe hexagon. As a result these corners hada higher centrifugal acceleration (caused bythe spin motion) than for the at sides. At

the low spin rate of AO-40 in normal opera-tion, these acceleration differences were notso large as to overcome the capillary forcesinside of the heat pipes, and all operated justne.

During one of the motor-burn phases ofthe early AO-40 operation, the heat pipesstopped operating. There was some consid-erable concern by those controlling AO-40,and I was consulted on the matter. As theheat pipe operation had been very well char-acterized, the problem was conrmed and aprojection was made of just what spin rate

Figure 7 Heat pipe assembly with clamp mounted inside of AO-40. KD1K photo

Figure 8 SilverStone CPU cooling heat pipe. SilverStone Tek photo

would again achieve operation and that com-putation was conrmed by the controllers asthe spin rate was slowed, much to their relief so much for zero gravity in space. Thereare always caveats to be addressed.

Terrestrial ApplicationsAll successful heat pipe operations

depend upon some fairly tight technical

details, regardless of where a heat pipe isused. In addition to the challenging design ofthe uid wick on the walls of the heat pipe,

the purity of the working uid is paramount.The uid in the AO-40 heat pipes was anhy-drous ammonia, NH3, owing to the need forthe heat pipe to operate at temperatures below0C and up to 40C. The steps used to makesure that the inside of the heat pipe was trulyclean was enough to make one wonder justwhat does clean really mean. Any impuritycan seriously reduce the operating capabilityof a heat pipe, because it will result in a non-condensable gas that impairs the workinguid from getting to where it is needed.

7/29/2019 Jansson

6/7


Using a powerful and hazardous reagentlike ammonia for a consumer-market devicewould be out of the question but there isno need for such a heat pipe to operate atlow temperatures. So the ground rules arechanged. Computer heat pipes need only tooperate from room temperature, 20C up toperhaps 70C. This opens the opportunity touse one of the very best heat transfer uidsthat is available for us water, just plain (butvery pure) water, H2O. The pressure insideof water heat pipes is less than 1 atmosphere(14.7 psia). For example, at 20C the pres-sure would be 2.34 kPa (0.34 psia) and at50C a pressure of 12.3 kPa (1.8 psia) wouldbe observed. Operating at lowered pressures,such as noted, is not a great concern becausea heat pipe must be truly sealed, and waterdoes not care until the temperatures wouldget nearly to freezing, 0C.

Heat pipe wick operation will be quitegood with water because the surface tensionof water is relatively high, meaning that a

water heat pipe wicking operation will besomewhat better in the anti-gravity situationthan many other uids can support, includingammonia. Nevertheless, the water heat pipesthat I have found all are positioned in thegravity-assist position the condenser sec-tion is located above the evaporator section.

Consumer application heat pipes comein two basic varieties, based on the materialsused for the heat pipe and the CPU heat sink.You will nd these heat pipes in aluminumand copper and these differences will driveother factors such as in the details of the inter-nal uid wick design and fabrication.

Figures 8 and 9 illustrate heat pipes thatare manufactured for computer CPU cool-ing. This is a relatively recent applicationfor heat pipes, driven by the ever-increasingcomplexity and power density of todaysCPU devices. Traditionally, as electroniccircuits have been miniaturized, made morecomplex, and made to operate with fasterclock speeds, their power dissipation hasalso dramatically increased. The days of trulylow-power CMOS devices are long gone.Consequently the computer designers havehad to become much more inventive in theirschemes to cool these devices the opening

for useful heat pipe applications.Using copper in the design of a CPU heat

pipe is promoted as an advantage over usingaluminum, because copper has a thermalconductivity that is more than twice as highas aluminum. A disadvantage of copper heatpipes is the weight of copper assemblies,which all bear more heavily on the printedcircuit board. This issue becomes a concernof PCB designers, as it is a hazard for theelectrical trace integrity of the PCB itself.

Another concern in these CPU applica-tions is that of the thermal coupling between

Figure 9 AeroCool CPU cooling heat pipe devices. Note that the unit at the bottom isshown upside down to show the CPU mounting surface and the heat pipes. This unit

includes a fan for extra cooling. AeroCool photos.

7/29/2019 Jansson

7/7


Figure 10 W1TKZ 1976 Field Day 6m station; L-R unknown youth; Ed Jansson, WD4DTC (then WA1UTJ); (thesomewhat younger) KD1K (then WA1QLI); with the liquid cooled 6m amplier and the authors Yaesu FT620B

transceiver. Chris Wiles, WA1RGA, photo

the CPU device and the heat pipe. CPUdesigners are faced with a nearly impossiblenightmare task of moving the device heat

out to the surface of the package to a pointwhere a cooling device can be attached. Theheat pipe designer also has the challengingproblem of getting this CPU heat to the heatpipe uid, which is the area that involves athick copper plate (and its attendant weight)or an aluminum plate (with its lowered ther-mal properties). These coupling plates can beseen in Figures 8 and 9. There was a reviewof several of these CPU coolers in a specialissue ofComputer Power Usermagazine in2005. This review gives considerable insightinto the thermal issues involved with thesecoolers.3 (You can also do a Google searchon heat pipe to nd this article and muchmore information.)

An Amateur Radio Application?I am going to get inventive now and

hypothesize how these CPU heat pipescan be used in an Amateur RF application.Suppose you have a transmitter that employsceramic-cased high-power transistors in theoutput stage. Normally these devices arescrew-mounted through openings in a PCBto a heat sink. This device mounting couldbe done instead to the underside the CPU

side of a water heat pipe cooler, along withits PCB. The transistors would be hangingupside down, but they do not really care at all

about that position. Such transistor mountingwould have to be very carefully done so thatany screw hardware does not penetrate theseveral heat pipes. Alternatively, an inter-mediary aluminum or copper plate could beused between the transistors and the cooler.

This conjecture reminded me that about35 years ago I constructed a somewhatsimilar transmitter cooler for the home-brewcompact 100 W amplifier of my 6 meterstation at that time. While this cooler didnot use any wicking agent inside the cooler,it had all of the other elements: a copperevaporator section in a portable cabinet, with

an air-nned condenser section above thecabinet. This was operated at atmosphericpressure and employed one of the suitable(now banned) Freon liquid cleaning agentsavailable at the time. This transistor coolerworked very well, bringing quite amplecooling into the tight space typically seenin a high power solid-state amplier. Figure10 shows this amplier in action during theField Day 1976 operations by the WellesleyAmateur Radio Society, W1TKZ. (I alsovery well remember that right after the open-ing bell, my rst contact was YV5ZZ. That

opened up the eyes of the HF operators.) Asthis was a liquid cooler operated at atmo-spheric pressure, it depended upon having

liquid ooding the evaporator section and thenecessity of a sight glass to insure that suf-cient liquid was available. In a heat pipe, thesidewall wick makes that assurance.

So you can see that I have had off andon experiences with liquid cooling of powerelectronics, some of it considerably beforebecoming involved in the AO-40 experiencestarting in 1990. I do not currently employany liquid coolers in my current computerand station installation, as air-cooled heatexchangers have sufced for these applica-tions, but I have not abandoned the concept.

Let your imagination be your guide.

Notes

1Dick Jansson, WD4FAB, SpacerameDesign Considerations or the Phase IVSatellite, AMSAT-NA Technical Journal, Vol1, No. 2, Winter 1987-88.

2Military Handbook Reliability/Design ThermalApplications, MIL-HDBK-251, 19 January1978, Chap.12, p 405.

3This Computer Power Userreviewarticle about some CPU heat pipe cool-ers is available on the Internet at: www.computerpoweruser.com/Editorial/article.asp?article=articles/archive/u0904/52r04/52r04.asp&guid=

Jansson

Documents

Transcript of Jansson