I I ELECTRONIC PACKAGING TECHNIQUES IN THE MARS OBSERVER CAMERA I Timothy E. Cushing Roger W. Applewhite* Altadena Instruments Corporation I Pasadena, California Abstract 1. Introduction I Volume constraints of small spacecraft place diffi­ The NASA Mars Observer mission is composed of a cult packaging requirements on electronic designs. Mars polar orbiting satellite carrying a number of High component counts per unit area and stiff, scientific instruments including a surface imager, the lightweight boards with large surface to volume Mars Observer Camera. The MOC was designed and I ratios will define the state of the art for PWB as­ constructed by the California Institute of Technol­ semblies. The Mars Observer Camera, a high reso­ ogy, with the assistance of the Optics Corporation of lutionimageron the Mars Observer mission, required America, Composite Optics, and Altadena instru­ I innovative developments in these design areas to ments. satisfy the difficult volume constraints placed on the instrument. These included qualifying a high-reli­ Spaceflight instruments generally are designed to ability "I-lead" surface mount technique used satisfy two sets of requirements. To gain the speci­ I throughout the electronic system. This enabled a fied performance, certain technical requirements are density with 4500 parts (including large DIPs and created. To assure reliability and reduce develop­ flatpacks), boards, and cabling in 900 cubic inches ment and operation risk, policy requirements are I and a 14 inch diameter PWBlhoneycomb board de­ promulgated. In the case of the MOC, both were sign. These PWBs were mounted through low fre­ especially difficult, requiring a carefully applied quency vibration isolators which provided dynamic system engineering approach. and thermal strain isolation. This paper outlines the I specific technical achievements gained with the Mars Accommodating a large number of discrete, leaded Observer Camera packaging design and the many components in a small volume is a difficult task. The I design rules developed from it. use of these components in a spaceflight mission,

v [ I r-- ""'"-r.- i?F~ I

.,..- .iF===1 I [g I :U( f-- J I "/ I ,-- L I" k 0 'r- ~- JL~

I NA Telescope Electronics Enclosure and Thermal Radiator Figure 1. Cross-section View of the Mars Observer Camera I *Member, AIAA 1 I I however, may be dictated by reliability, availability, edge-mounted boards in order to gain additional real I or policy constraints. The following paper discusses estate and eliminate layout conflicts within the mass how the I-lead packaging method was used to suc­ memory arrays. As a result, the packaging was re­ cessfullyassemble the Mars Observer Camera (MOC) quired to accommodate the substantial vibration­ I electronics. It will describe the requirements for the induced curvature anticipated during launch. No design, the packaging technology selected as the packaging solutions in the JPL Design Manual di­ base ofthe design, and the engineering accomplished rectly supported significant board curvatures, so to extend the selected base to meet the requirements. design extensions to highly compliant leads were I As examples. case studies of how the three most required. Common parts in the MOC are assembled will be given. 2.2 Policy ReQuirements I 2. ReQ.uirements Three unusual circumstances placed heavy political pressure on the design of the MOC to be conserva­ 2.1 Technical ReQllirements tive. First, the MOC is the first non-facility camera to I be flown in over 25 years. The principal investigator The MOC is a narrow-angle Ritchey-Chretien tele­ organized a team at the California Institute of scope with a 3.5-meter focal length as shown in Technology, using staff and a small group of sub­ figure 1. It uses the spacecraft's motion to scan a line­ contractors to design and build the flight hardware. array CCO across the Martian surface to create JPL acted mainly as program managers. Second, photographic swaths 3 km wide at l.4-meter per cameras are the most publicly visible instruments on pixel resolution from an altitude of 380 km. The planetary missions and return of images is very average data acquisition rate of the narrow angle important to public acceptance. Additionally, the system is approximately 5 Mbytes/sec, a factor of MOC acts as a data buffer for the Mars Balloon 1000 greater than the bandwidth of the spacecraft experiment, a French instrument on a Russian mis­ I data system. This requires the MOC to have a large sion. This increases the importance of reliability in image buffer. This in concert with a completely theMOC. block redundant system leads to a component count of over 4500. As shown in Figure 2, the electronics Therefore, as a contractual requirement, the MOC I were constrained to fit directly beneath the main electronics packaging needed to conform to the JPL telescope, while the height of the telescope base was design standards. A project waiver was required limited by the surrounding instruments. wherever the design rules were not covered by the I JPL Design Manual for flight hardware. This docu­ To help achieve the density, the Printed Wiring ment outlined the allowable packaging techniques Boards (PWBs) were designed to be large circular for JPL sponsored assemblies. Solder assembly was I

#L.-I _____ ------'Il;\-----\ ___ I Analog Board I System #1 Digital Board Viton Elastomeric __-..IIc-'JI~;;~~~~~~.~F~Wgi~~3i~~~n;;$ I Vibration Isolators ~~~~~~= I 14.8" Diameter I Figure 2. MOC Electronics Enclosure

2 I I I I acceptable, but with the additional constraint that the 2.3.2 Quality Control and Inspection design be suitable for hand-soldering, 100% visual inspection, and rework/repair at any stage. Although originally constrained to comply with the I supplied JPL QA personnel, we ultimately retrained To accomplish its program management function, them for the unique reliability issues of the final ~I JPL formed design review and quality assurance design. Extensive engineering discussions were re­ teams to monitor the packaging design. Lacking the quired to overcome the traditionally accepted views, budget or schedule to develop truly new packaging but the end result was in fact much improved by the technologies, we were constrained to remain within retraining process. I technologies at least somewhat familiar to most of the reviewers. The review boards mixed experts 3. The Selected Packaging Technology from development engineering and current programs with 'grey beard' specialists, creating a group sus­ With the above constraints in mind, we reviewed I picious of anything that was not already flight proven. possible packaging methods. An effort was made to compare costs and capabilities between candidate In practice, however, there are no truly 'qualified' technologies. This led to the final selection. packaging technologies. Each instrument subassem­ I bly, and preferably the entire instrument, should be Certain design candidates were eliminated early on. subjected to a variety of environmental and fabrica­ For instance, JPL's flight design rules prohibit thru­ tion stresses to assure proper application of the se­ hole assembly. All components must be either sur­ I lected packaging technology. As such, the task was face mounted or wired to terminal posts. Epoxy or actually one of convincing the review board that the polyimide printed wiring boards are specified, re­ assembly could be expected to pass the qualification quiring that components be mounted with compliant tests, and then agreeing on what tests constituted leads to withstand extended thermal cycle and launch I acceptance of the finished unit vibration stresses. JPL requires special multi-layer PWB processing to fill all vias with epoxy. Blind and 2.3 Additional Requirements buried vias are not permitted. Leadless surface mount I designs were not generally acceptable. Although 2.3.1 Flight Component Selection Constraints hybrids have flown on virtually every spacecraft to date, hybrid packaging was considered too special­ Flight components were generally constrained to be ized for the Design Manual, and was not an accept­ I off the shelf due to the extremely low production able solution for general electronics packaging. volumes. Parts from JPL' s Preferred Parts List, espe­ cially from existing flight stores, were chosen wher­ The packaging technology ultimately selected, as I ever possible. If not available, industrial parts were described below, was initially spurred by recent purchased outright. This required accommodating development work at JPL. Previous applications had both JPL and industry specifications for similar parts. failed to address various key issues, however, and the Parts from flight stores were often quite old, requiring resulting failures left a pre-judgment against our use I aggressive lead preparations and tinning verification.

Special fabrication lots for a preferred package type I were generally not possible, leading to a mix of package styles for similar parts. Similarly, not all NO SOLDER parts could be purchased with extra long leads; lead IN LEAD BENDS I forming designs had to accommodate the shortest leads within that package style. Parts were either n purchased to Class S specifications, or screened to a subset ofClass S, as applicable. A few key parts were I not available even to military specifications. These were subjected to rigorous physical analysis, and special screening tests were devised to assure reli­ I able performance. I-LEAD SOLDER JOINT (DIP) SCALE 10:1 I Figure 3. 3 I I of this technique. To further complicate the matter, mining where JPL's DM could be applied to the 1- I JPL retracted the relevant sections from their Design lead method without modification throughout the Manual following our initial design reviews, forcing design. us to obtain waivers for virtually the entire instru­ I ment. Fortunately, manufacturing engineering was The basics of instrument assembly and handling able to provide solutions to the historical shortcom­ were applicable, as were general materials specifica­ ings, most notably the ability to inspect the lead tip­ tions and the specialized details for PWB fabrication. to-pad gap with confidence. Many circuit board design issues were not covered in I detail, but reliance on IPC and MIL specifications The method selected for use throughout the assembly was generally accepted. Likewise, the most basic was the I-lead surface mount solder joint. In this level of component mounting and soldering re­ I joint, the lead is terminated normal to the pad (see quirements were applicable to the design of reliable Figure 3), minimizing joint area. Traditional termi­ joints and assemblies. nal-mounting of parts clearly required too much real estate and overhead, while the drill holes for the Many fabrication-specific details required separate I terminal shanks would interfere with good circuit documentation to clarify and condense the appli­ layouts. Gull-wing leads also required too much real cable specifications. Assembly subcontracting re­ estate, and most IC leads were too short to meet the quired extraction of relevant paragraphs into a pro­ I highly compliant design requirement with enough cess specification, with direct references back to the lead left over for a gull foot. originating JPL specification.

J-Iead and very minimal gull-wing solder joints could 4.1 Creating Design Rules I possibly have been made to fit the available real estate. However, I-lead pads were smaller still, pro­ Due to the lack of schedule or budget for an empirical viding some margin in the layout estimates. Further, development program, creating the design rules for I development testing at JPL showed no reliability the packaging engineering relied heavily on model­ gains for J-Ieads, and there was no inheritance from ing and analysis. A high degree of conservatism was previous flight instruments. Tooling for J-Ieads and built into the design goals to cover the uncertainties gull-wing leads was also substantially more complex. of models that were not to be physically verified. I

I-leads permit simple production tooling: a good lead The basic design rules center around lead bending, bender to locate the leads precisely over the pads, and part location, and bonding. Lead bending relieves I a flush-cut or shearing lead trimmer for proper co­ board curvature stress while selection of the location planarity of the leads with the board. To reduce of the part on the board is dictated by vibration production complexity further, component packages loading. Thermal cyclic fatigue is mainly handled by were generally grouped into one of several standard ensuring solder fillets have the proper volume and I lead forming specifications. The slight reduction in shape, and that the lead is the proper height above the layout optimization was more than balanced by mini­ pad in the solder fillet. Where lead design and solder mizing likely tooling and inspection errors. fillets were not sufficient to handle the mechanical I loads, the part was bonded directly to the PWB. All I-leads also provided the smallest pad dimensions of these rules are centered around the lead and fillet and shortest lead length requirements. The joints engineering detailed in the following sections. could be hand soldered to meet the reliability and I longevity requirements, and the production tooling Manufacturing concerns make up the rest of the to ensure repeatable assembly results was inherently design rules. Lead preparation for I-lead joints rep­ simple. resents a substantial reliability concern. With gull­ I wing leads, a rather long portion of the lead is held 4. Applying the Selected Technology to the MOC within the joint, where a poorly tinned patch would Environment represent a small portion ofthe overall joint strength. For an I -lead, however, the entire joint is concentrated I As described above, the MOC contract required within the last 0.020" of the lead; a poorly tinned either compliance with the JPL Design Manual, or an patch would represent a substantial reduction in the approved waiver covering aspects outside the Design overall joint strength. For added assurance, we in- I Manual. A notable engineering task involved deter-

4 I I I

corporated a wetting balance solderability test for all 4.2.1 Failure Criteria and Modes I-leads in addition to 100% visual inspection of the lead tinning. Two modes of mechanical susceptibility were iden­ tified for discrete components. The first was the I-leads required several joint characteristics that were quasi-static stress limitations in the leads and solder in direct conflict with traditional flight soldering fillets, and the second was cyclic fatigue and creep in rules. The most difficult I-lead requirement to enforce the solder. Each environmental input affected one or turned out to be the need for maximizing the solder both ofthese areas. volume. Traditional soldering rules call for concave I joints where one can see the outline of the lead. In the In both cases, failure was defined as a discontinuity case of an I-lead, this leaves virtually no joint at all. in electrical conductivity of the joint or the lead Both assembly technicians and inspectors required before the completion of the mission life plus mar­ focused training and visual gauges to assist in keeping gin. This allowed for fillets to experience partial I the joints nearly conical and at least 0.020" high. cracking in life tests while still not being considered failed. This is significant, as fillets which incur A second critical process area was ensuring that the vibration or thermal stress in test will often have I lead tips were within the required 0.005" of the surfaces that appear "frosty", a condition usually copper pad. Icicleing, heavily tinned pads, etc. re­ denoted as a failure. However, if the joint still quired touch-ups to ensure proper registration of the maintains electrical continuity, it passes the test. lead tips. Every joint required two stages: adhesive or I solder-tack mounting with 100% visual inspection of 4.2.1.1 Quasi-static Stress the lead tip-to-pad gap, followed by final soldering with 100% inspection of the final joint appearance. The quasi-static stress is the instantaneous stress in a I part under dynamic load. Dynamic and static loading 4.2 Packaging Engineering Process can be assessed in terms of the engineering values of yield and ultimate strength of materials. Margins of The success of adapting a packaging technology to a safety are developed which account for process and I very large variety of discrete components depends on materials variations as well as unknowns in the carefully breaking down the environmental conditions mission environment. Generally ,leads were designed into manageable pieces. After this is done, the different such that the yield stress of the lead material was at I ways parts react to each of these environments can least 1.5 times that of the stress developed under also be broken down into generic mechanisms. Table loading. While it can be argued that the energy 1 outlines the different environmental inputs to the present in dynamic loading may not be large enough MOC electronics and the generic mechanisms by to effect damage even if the quasi-static limit is I which reliability of the mounting scheme can be exceeded, especially for soft materials such as sol­ compromised: der, this method represented a simple and conservative way to select between candidate lead configurations. I Table 1. Environmental Inputs and Reaction Mechanisms 4.2.1.2 Cyclic Fatigue Environment Quasi-static stress cyclic fatigue Inertial loading Cyclic fatigue and creep, on the other hand, represent I the summation of stress or strain activity in a com­ Static accelemtion ..j ponent. Low yield, low melting point materials such Dynamic acceleration ..j as solder absorb small amounts of energy at every application of stress. This has the cumulative effect I PWB curvature Dynamic of concentrating grain boundaries in the material which eventually lead to cracks. Therefore, the en­ vironmental history of a part plays a large role in its I Thermal Stress Deep cycle ..j (creep) ability to resist further stress . Small ..j I 4.2.2 Stress/Strain Investigation To adequately describe the stress/strain state of any I of a wide variety of components when subjected to a 5 I I

Table 2. Lead Stresses Due To Vibratory Loads and Thennal Distortion I

HEIGHT E CTEOF DlAOF LeD.L1 Leu.U BENDING DlFF. JNTILD AD (t.DCIL) LEAD LEAD B.c. BODY STRESS EXP. CRY I lJG I RNSO 2 0.167 1.600+07 9.2OE.{)6 0.016 0.290 0.170 18 RWR81 2 0.165 1.600+07 9.2OE.{)6 0.020 0.440 0.270 28 LVR3 2 0.234 1.600+07 9.2OE.{)6 0.032 0.822 0.570 63 CCR7S 2 0.165 2.80E+07 9.20E.{)6 0.020 0.320 0.170 21 LI641 2 0.238 I.60E+07 9.2OE.{)6 0.025 0.615 0.420 85 10 9.20E.{)6 I 0035 2 0.165 2.80E+07 0.020 0.320 0.180 16 10 007 2 0.165 2.8OE+07 9.2OE.{)6 0.020 0.440 0.300 35 21 1018 3 0.330 2.80E+07 9.2OE.{)6 0.020 0.400 0.2S0 123 4 1039 3 0.330 2.80E+07 9.20E.{)6 0.020 0.520 0.370 361 7 8 O. . 70 125 • I

PART HEIGHT E CTEOF LEAD LEAD LEAD Le... LII Leu.Ul Leu.L1Z t.Dc.G. LEAD DIM 1 DlMZ HEIGHT B.c. BODY BODY I DIPS S.88E.{)6 1.9SE+07 DlPI4 1.2lE4.5 1.9SE+07 MFPI4 2.11E.{)6 t.9SE+07 DIPI6 t.26E"(),5 1.9,5E+07 MFPI6 2. 11 E.{)6 1.9SE+07 I DlPI8 1.55E4.5 1.95E+07 DlP24 2.43E"()S 1.9,5E+07 4. S +07 I PART AD

\3 I 69 15 88 107 109 I

NOTE: Leads "'" ..oumed to be DIM 2 (or ball Ibe vertic:a1 rail aod all o( Ibe boriztmlal NO. Ibea 3" DIM 2 (or tile remainder of the vertical run. NOTE: L11 is distance betweeojoiDtcenteni, _Ibe IC. LI2 is Ibe width of tile IC body. L21 islbe overall lead distance along Ibe IC. NOTE: Direction 1 (-I) is across Ibe IC; direc:tio1l2 (-2) is along the IC. NOTE: D-RES FREQ is DtOn! applicable fordoublco-bend leads. S-RES FREQ is mOle applif:able for smgle-bend leads. I

SUMMARY l.25mC .2SmC MX: 19.84 I CURVE; LEAD ~ot STRESS CRV DESIGN (KSO LiMIT 0.21 0.21 1% I.5S 0.07 2% 0.40 0.40 2% 2.48 0.14 4% I 1.37 • 0.84 0.84 4% 0.84 0.28 7% 3.64 0.51 0.51 3% 2.07 0.13 3% 1.07 • 0.49 0.49 2% 0.49 0.15 4% 4.16 0. .51 0.51 3% 1.73 0.15 4% 2.84 0.70 0.70 4% 3.30 0.31 8% 0.78· I 0.32 0.32 2% 0.32 0.06 1% 0.45· 0.41 0.41 2% 0.41 0.10 3% 04 0.41 2% .41 4%

SUMMARY I

PART CURVE ~of STRESS DESIGN DIPS I D!P14 9.01 MFP14 0.47 DlPl6 6.88 8.43 MFPI6 5.95 1.01· 0.52 DlPI8 7.53 1.80 9.41 I DlP24 8.85 1.77 • 2.54 .73 2.55

6 I I I' I particular environment, it was necessary to create be applied and the resulting stresses and fatigues generic models which could be applied to any part. analyzed. I Simple beam models were created for the part leads which could be easily modified to fit any lead con- 4.2.3.1 Inertial Loading figuration. Finite Element Model (FEM) investiga- tions of the solder fillet led to generalized elastic When the PWB on which a part is mounted acceler­ I boundary conditions for the lead models forming a ates, the reaction ofthe inertia of the part mass creates complete elastic description of the part. Stress and a stress in the lead and fillet The acceleration can strain approximations could then be computed for either be static, such as the 6.5 g launch acceleration any lead and body configuration. This was the method of the Titan III launch vehicle, or it can be dynamic, I used to create spreadsheet models, as shown in Table as in the 16.5 grrns random vibration of the spacecraft 2, of each part for quasi-static stress analysis. during launch.

I Analyzing fatigue began with a non-linear/plastic Determining the reaction loading in the static accel­ FEM of candidate solder ftllets. The purpose was to eration case is simply a process of applying the load determine the strain distribution in the fillet as a factor to the simplified models. The vibration load­ function of strain input The largest values of strain ing, however, requires an understanding of the har­ I were then used as the input into a solder fatigue law. monic nature of the part. The simple beam models were not capable of determining the fundamental H. D. Solomon and the Sandia Corporation have mode directly, as several rigid body modes exist for I published data for the fatigue behavior of eutectic multi-leaded (and hence high degree of freedom) solder for a variety ofloading conditions and strains. components. FEM analysis was required to deter­ In each case, the amount of strain accumulated in a mine likely lowest mode shapes for several multi­ solder sample could be correlated with the amount of lead configurations (from axial components to DIPs I "life" which had been consumed. The well known and flatpacks). With the mode shape determined for Coffin-Manson power law regressed fairly well to any class of part, the simple beam method could then the Solomon data and was used as the function for be used to find the frequency of the lowest mode. I analyzing the effect of each environment which contributed to fatigue, including thermal strain and The vibration any part might experience depends board curvature. largely on its location on the PWB. The construction of the circuit board assemblies attempted to mitigate I The error of fitting the power law to the solder data, the need for high stiffness and hence low curvature as well as the large variability in the data (not un- against the compliance needed for low vibration common for fatigue testing) led us to adopt a cyclic amplification and hence low reaction loads in the I fatigue life margin of 10 in the design. Unlike quasi- board mounting brackets. static stress, the exact mode of fatigue failure is relatively unknown, accounting for the larger margin. Multilayer GlassIPoIyimide PWB

I Using the power law, qualification tests can also be designed which transform the differing cyclic fa-

I tigue environments into a small number of large I thermal cycles which can be reasonably reproduced in a thermal test chamber. Fatigue life can be con- sumed at an accelerated rate using a smaller number of deeper cycles which are transforms, in terms of I total fatigue, of the very large number of small cycles encountered during the mission. Tests which simu- I late the mission can be performed in the matter of a Figure 4, Circuit Board Assembly I week, and tests which simulate multiple lifetimes can also be performed in a similar amount of time. The circuit board design uses two 14" diameter 4.2.3 Environmental Conditions PWBs bonded to a 112" thick aluminum honeycomb I core resulting in a very stiff assembly. The honey­ I With the models completed, the environments could comb sandwich construction constrained each cir- 7 I I

Table 3. Mars Observer Camera Thermal Profile I IMISSION PHASE Tmin Tmax AT # of cycles

I ASSEMBLY (POST REFLOW) I Fabrication 20°C 700 e sooe 4 Ambient ON/OFF 20De 3Soe lOoe 100 Rework Omaior 3 minor) 20De 70De sODe 10 I

IHANDLINGrrESTING INSTRUMENT TESTING Bakeout 200e 6S De 4Soe 4 I ThermaVVac I -30°C sSDe 8SDe 2 ThermaVV ac IIIlI1 -30De ooe 30°C 10 Non-operarlng -3SDe 20De ssoe 1 I Overatin~ 200e 7Soe 9Soe 1

SYSTEM TESTING Bakeout 20°C 6Soe 4Soe 4 I Thermal Balance -20oe 200 e 40 0 e 3 System TN -30oe 20De sooe 1 I IGROUND HANDLING 200e 25°e soe 1000

IFLIGHT Bakeout -20°C -lSoe SoC 1 I Cruise, inner -> outer -30oe 200 e sooe 1 Cruise, rotation -30oe 200 e

cuit board to a single surface for component assem­ g's at the center. Parts could be moved to different bly, but eliminated all mounting hardware from the areas on the PWB according to their susceptibility to layout area. The lack of obstructions in the layout vibration. I area was a key element in achieving the required component density. See Figure 4. By taking the random vibration spectrum, deter­ mined by FEM of the boards, the quasi-static loads I In order to reduce the harmonic amplification char­ could then be determined by assuming the parts were acteristics of the board assembly from an unreason­ rigid bodies below their fundamental frequencies. To ably high value of 90 discovered in test, it was insure no resonant amplification of vibratory inputs necessary to isolate the boards from the aluminum would occur at the parts, a design rule was created I chassis of the MOC with Viton elastomeric vibration which insured each part had a fundamental fre­ isolators. Without adding any unacceptable low quency of at least 4 times that of the circuit board frequency rigid body modes, the isolators effectively assembly or that its body was bonded directly to the I reduced the amplification of the board system to less PWB, relieving the leads of any structural responsi­ than 20. bility.

The result of this design is a circuit board system 4.2.3.2 Board Curvature I which floats as a rigid body on shock absorbers with a fundamental frequency of around 400 Hz. The rms Quasi-static as well as cyclic fatigue is important acceleration components experience during launch here. Parts which have large spanning dimensions I range from 109's at the board edges to as high as 60 across the boards such as long DIPs and flatpacks are

8 I I I I especially susceptible to board curvature. Although before being integrated into the prototype unit. Pro­ the board assembly is very stiff as mentioned, the tot test levels were used. intending to show suit­ I PWBs do exhibit some curvature while vibrating. abil of the packaging but not its ultimate life. The extent of the curvature was estimated from a FEM model of the circuit board. Even with vibration Based on the results of the prototype testing, process isolators, substantial board curvature was expected revisions were made to address observed weaknesses, I due to launch vibrations. Component leads were thus mostly in improved soldering techniques and more required to include compliance in the mounting plane, quantitative inspection criteria. Some flight hardware effectively requiring every lead to have both a verti- was already in process, requiring rework or re-in­ I cal and a horizontal run between the case and the spection. solder joint. Unfortunately, more compliant leads lower the fundamental frequency ofthe parts. Curva- Final hardware was subjected to "workmanship" and ture loading, therefore, can lead to a large number of acceptance tests. Specifically, vibration and thermal I parts being bonded directly to the PWBs. cycle stresses were applied, followed by 100% re­ inspection of the assemblies. In response to continu­ 4.2.3.3 Thermal Stress ing skepticism within NASA, further testing was I applied to the prototype units, including reworking The PWBs are constructed of a poly imide/glass fiber hundreds of joints after application of the Parylene composite bonded in layers. The coefficient of ther- conformal coating followed by repeating the envi­ mal expansion (CTE) of the boards is generally ronmental tests. Final judgment of the JPL special­ I larger than that of many of the ceramic or metal ists was that the packaging was acceptable as pre­ component packages. Changes in temperature over sented. the mission create loads in the leads and solder. I 5, Case Studies The temperature profile of the MOC is composed mainly of two periodic components. One correlates RNC50 Case Study to the orbital period of the spacecraft about Mars, and I the other relates to the orbit of Mars about the sun. Axial leaded parts, such as the RNC50 resistor, ar:e The first causes temperature swings on order 5° C generally characterized by a small, light body sup­ while the other can be as high as 50° C. See table 3. ported by two cylindrical leads. Exceptions include I the occasional large inductor or capacitor, which Interestingly, even though the temperature excur- require an adhesive mount to augment the leads, or sions are large, quasi-static stress is not an issue. rectangular cross-section leads which only affect the Solder exhibits a large creep behavior even at very solder pad dimensions. Having only two leads, axial I low stress levels, tending to relieve any load. Unfor- parts are inherently less stressed by board curvature tunately, ifthermal cyclic fatigue is a problem, making than components such as a long DIP part with many the component leads more compliant will not help. leads in a row. I Any application of strain of a period longer than a few minutes will usually cause fatigue damage in solder. I • 0..Q60.':J· O.170."n ."LM~n _ The FEM model of the solder fillets was incorporated ~ IG ~"C[O.0.16" t~ into a software package which took the fatigue envi- o ".{ ...... iE '".". TYP d ronments listed above and systematically applied MAX them to the part, accumulating a damage total. The O.OlT I INSIDE solder fillet volume was sized in this way to ensure each part had adequate margin. RADIUS-=rfV=l:::::====Ril1--. SOLDER ~ I 4.3 Test MASK\ - ~~~~~~~~~+ Although dedicated test hardware could not be built, I prototype hardware for engineering was assembled 0..360." GLASS I POL YIMIDE per the intended flight specifications. These boards RNC50 I-LEAD MOUNTING EXAMPLE were subjected to vibration and thermal cycle tests I SCALE 4:1 9 I I

In spite of the low body mass, keeping a high I (mechanical) resonant frequency is a concern for axial parts. To accommodate the wetting balance solderability testing, the leads were required to be I relatively long. The basic analysis results, shown in Table 2, include the frequency of the lowest reso­ nance. Those failing to meet the given design limit Q5;; <0Q.. ~ require an adhesive mount, and are so designated. o o I TYP CASE Table 4, below, lists key lead fonning and trimming -----'- design rules, relevant to axial leaded parts, resulting ...... I from stress and manufacturing analyses. In addition, 0.020" ::.-:.: all the standard lead and lead tinning criteria apply, such as no nicks, tinning voids, icicles, etc. Table 5, INSIDERADIUS ~-=19:q~~:O-:::i~P~\---.- below, lists key assembly and inspection criteria for I installing axial I-leaded parts to maximize the assembly's reliability over vibration and thennal cycle. Again, conventional solder appearance crite­ I ria, such as no granulation, pits, etc., also applies.

Table 4: Key Leadfonnjn& Desi&n Rules At least 2 lead diameters from body undisturbed. I At least 1 lead diameter inside radius for bends.

Bends should be fully 90°, after springback. 14------10.340" \ At least 2 lead diameters between bend and joint 0.500" I At least 0.12" below body for solderability test. GLASS I POLYIMIDE Flush cut tips: «tOOl" step, <0.001" ridge. TO-IS I-LEAD MOUNTING EXAMPLE SCALE 4:1 Table 5: Key Axial I-lead Installation Criteria I Leads rise vertically from pads (within 10°). TO-18 Case Study Lead tips must be within 0.005" of copper pad. Lead tips must be fully within pad (>0.005"). Devices packaged in TO-style cans, such as the TO- 360° solder fillet, extending to edge of pad. 18 case fortransistors, are generally characterized by I Fillet Height: >0.020" average, >0.005" local. a large, heavy body supported by two to eight leads. Fillet Shape: nearly conical, slightly concave Exceptions include various high-power packages, Lead bends must be free of solder. which require unique mounting designs to increase I heat dissipation. TO-can packages have all leads Finally, Table 6, below, gives a cursory description exiting from one end, arrayed in a circle. The leads of the parts preparation and solder assembly steps are splayed out radially, then bend back towards the developed for axial I-lead parts. Custom tooling was case. This mounting sty Ie is commonly referred to as I created to pennit rapid and accurate lead fonning and "dead-bug" mounting. Alternate mounting styles trimming. Solderability was tested using the wetting were considered, but failed to meet all of the design balance on a Multicore Universal Solderability Tester. constraints. I Table 6j Production Process SUItlIIlatY The device case requires mechanical support due to Clean and inspect parts; pre-tin gold plated leads. the long lead lengths. In addition, the long and Form and trim leads using designated tooling. springy leads were difficult to keep aligned with the I Test lead tip solderability within 8 hours of trim. solder pads without actually holding them during Inspect workmanship; touchup tinning as needed. soldering (causing built-in stresses). The final solu­ Bond mounting pedestal to PWB, as required. tion involved adding a collar for the leads to the TO I Solder tack lead tips to pads, stress-free. pedestal. The basic analysis results are shown in Bond body to mounting pedestal, as required. Table 2. Inspect configuration and lead tip to pad distance. Final solder and clean, keeping leads stress-free. Tables 7 and 8 list the leadfonning rules and instal­ I Inspect workmanship; 1 rework/joint permitted. lation criteria.

10 I I I I I I I

0.040---," -1-1,,,,,,1'- I 0.010" I INSIDE TYP H H r-----i I--i RADIUS 0.025". LEAD 0.020" 0.040" 0.100" REF 0.060" (BOlli 0.045" I BENDS) SOLDER MASK GLASS I POLYIMIDE

I DIP I-LEAD MOUNTING EXAMPLE SCALE 4:1 Table 7: Key Leadfonning Design Rules Table 9: Production Process Summary I Transipad required to protect glass lead exits. Clean and inspect parts; pre-tin gold plated leads. Fonn 1ST bend over the transipad. Bond transipad in place before bending leads. At least 2 lead diameters from case to 2ND bend. Fonn and trim leads using designated tooling. At least 1 lead diameter inside radius for bends. Test lead tip solderability within 8 hours of trim. I Bends should be fully 90°, after springback. Inspect workmanship; touchup tinning as needed Flush cut tips: <0.001" step, <0.001" ridge. Bond lead collar to PWB, aligned with pads. Solder tack lead tips to pads, stress-free. I Table 8: Key TO-can I-lead Installation Criteria Bond case to lead collar as mechanical support. Insulate metal case from PWB. Inspect configuration and lead tip to pad distance. Use lead collars to align lead tips with pads. Final solder and clean, keeping leads stress-free. Leads rise vertically from pads (within 10°). Inspect workmanship; 1 rework/joint permitted. I Lead tips must be within 0.005" of copper pad. Lead tips must be fully within pad (> 0.005"). DIPl6 Case Study 360° solder fillet, extending to edge of pad. Fillet Height: > 0.020" average. > 0.005" local. Devices using DIP-style packages, such as the DIP I 6 I Fillet Shape: nearly conical. slightly concave. package for integrated circuits, are characterized by a large, heavy body supported by 8 to 28 leads. DIP The production process for TO-can parts is similar to packages have half their leads exiting from opposite I that for RNC50's, and is summarized in Table 9. sides, forming a row. Two lead styles are common: These devices require adhesive mounting for me­ side-brazing and frit-seals. Since the leads are ar­ chanical support. Small tabs were added to the TO­ ranged in rows, a horizontal run must be formed into can mounts to align the leads directly over the pads, the leads to accommodate board curvatures. Using I eliminating the built-in stress that would have re­ an I-lead joint design maximizes the lead available sulted from holding the leads during soldering. for compliance. I Where additional mechanical support is required due to especially massive bodies or the need for spot I radiation shielding, amounting pedestal is employed. 11 I I

The basic analysis results are shown in Table 2. techniques, training, and prejudices, creates a design I statement that may seem impossible to physically Tables 10 and 11 list the leadforming rules and realize. installation criteria. We have found that a systems engineering approach I Table 10: Key Leadfonnin~ Desi~n Rules to packaging design, where requirements and inter­ Leads must be clamped during fonning. faces with other portions of instrument design are Fonn 1ST bend at least 0.030" below body. broken down and closely managed, can yield a work­ I Fonn 2ND bend at least 0.030" from 1ST bend. able design where inheritance from past tradition is At least 1 lead thickness inside radius for bends. married to innovation. The Mars Observer Camera, 0 Bends should be fully 90 , after springback. as an entire system, was created in this way. Future Trim leads at least 0.025" below 2ND bend. small satellite missions will find increasing pressure I Flush cut tips: <0.001" step, <0.001" ridge. to meet the wisdom of the old and the possibilities of the new. Table 11: Key DIP I-lead Installation Criteria I Lead tips must be within 0.005" of copper pad. 7. Acknow1edgrnents Lead tips must be fully within pad (>0.005"). 3600 solder fillet, extending to edge of pad. The design of the MOC electronics packaging was Fillet Height: >0.020" average, >0.005" local. undertaken by several members of the Mars Ob­ I Fillet Shape: nearly conical, slightly concave. server Camera Design Team. They include: Rob Calvet of Ad Astra Engineering for FEM analysis of Table 12 summarizes the production process. Many leads and solder fillets and the creation of cyclic I ofthese devices required spot radiation shields, which fatigue estimation software, Mike Blakely of JPL for were installed after electrical test. electronics packaging technology consulting and Dave Carter of Caltech for component procurement. Table 12: Production Process Summmy The authors also wish to thank Ed Danielson, the I Clean and inspect parts; pre-tin gold plated leads. MOC instrument manager, and Dr. Mike Malin, Fonn and trim leads using designated tooling. MOC principal investigator for supporting this effort. Test lead tip solderability within 8 hours of trim. Elaine Lindelef and Ken Andrews of Altadena In­ I Inspect workmanship; touchup tinning as needed. struments provided invaluable assistance preparing Bond mounting pedestal to PWB, as required. the final document. This work was funded by the Solder tack comer lead tips to pads, stress-free. Mars Observer JPL Contract No. 957575 and Altadena Bond body to mounting pedestal, as required. Instruments. I Inspect configuration and lead tip to pad distance. Final solder and clean, keeping leads stress-free. 8. References Inspect workmanship; 1 rework/joint pennitted I M. C. Malin, G. E. Danielson, M. A. Ravine, and T. A. Soulanille, "Design and development of the Mars Observer Camera," Int. J. lmag. Syst. Tech., Vol. 3, 6. Summary pp. 76-91, 1991. I Engineering of a packaging design is always driven The Jet Propulsion Laboratory, "JPLDesignManual," by a difficult combination of requirements. Often JPL Des. Req. DM509306 D, Vol. I - III relegated to late in the design, packaging can be I forced to fit within a stringent set of volume and Harvey D. Solomon, "Fatigue of 60/40 Solder," environmental constraints. This, combined with a IEEE Trans. Components, Hybrids, Manu! Technol., dizzying array of available packages, manufacturing I I I

12 I I