The True Cost of Hermetic Seals for Aluminum Housings

The True Cost of Hermetic Seals for Aluminum Housings

Jack Pollock

Special Hermetic Products Inc.

Wilton, NH

Hermetic seal reliability has been a major concern in the microwave industry for the past 35 years or more. While sealing problems have occurred with various materials and processes, none have approached the difficulties encountered with sealing aluminum modules. An informal survey conducted over the last seven years of companies soldering contemporary feedthroughs into aluminum modules has revealed a wide range of results when tested to meet military specifications. A few companies indicated no difficulties whatsoever, while the majority indicated a 15 to 40 percent failure rate of seals during in-house testing prior to shipment. Some companies reported higher failure rates, including one with as high as 98 percent in-process failures on a major program. While this study was not conducted formally, it provided considerable insight into the magnitude of hermetic sealing problems in aluminum housings. Although these findings appear staggering on the surface, they are perfectly understandable when the typical contemporary feedthrough design is analyzed for strain in required cyclic thermal environments. In fact, what is more difficult to understand is how any significant yield could be obtained at all using contemporary designs. The root cause of the problem lies in the fact that seal designs being marketed to the microwave industry generally predate the advent of aluminum as a primary hermetic packaging material. Over the years, the focus on this problem has been one of process improvement, plating changes and solder improvements. While these activities produced some increase in yields, the basic problem remained and companies became accustomed to poor yields as a way of life. As frequently happens, contact between companies provided comfort mainly in the fact that most were encountering similar problems and, apparently, no one had the answer. Some changes that were introduced and used by many companies with marginal results included replacement of gold plating with nickel, tin or silver plating to avoid gold embrittlement; multiple pretinning and wicking of gold parts to remove gold prior to final soldering; and use of indium solders to introduce more flexibility in the solder joint. Analyses and extensive experience have shown that contemporary feedthrough designs have a significant reliability problem even with the elimination of gold embrittlement and the use of more flexible solders. These problems are attributed to the fact that contemporary designs do not provide for strain control of the solder joint within acceptable limits. Without this specific control, the solder will be subjected to excessive strain during thermal cycling and will break down after relatively few cycles. The change introductions reflected previously present additional problems and/or limitations. None of the alternate platings solder as reliably and consistently as gold and, therefore, lead to more rework. Nickel poses a particularly difficult obstacle to achieving consistent solder results even with stringent impurity and shelf life controls. The tin plate limits solder temperatures, thereby limiting choices of solder temperature for subsequent solder operations. While pretinning and wicking of gold has reduced or eliminated the embrittlement problem, the process also precludes the achievement of consistent hermeticity results and is costly. In many cases, pretinning alters the precision dimensioning of parts sufficiently to preclude their subsequent assembly. Finally, use of indium solders can be expensive and limits the temperature of subsequent processes.

Hermetic Seal Cost

The true cost of hermetic seals must include the cost of dealing with in-process failures in addition to the initial cost. A more appropriate title for this article might be "The Cost of Hermetically Sealing Aluminum Housings." It is a difficult and frequently endless task to assess the overall cost associated with a design flaw. The intent of this article is to provide some guidelines to the relative cost of various methods of providing hermetic feedthrough seals in aluminum packages. More specific costs can be obtained by substituting actual costs from company histories in any of the areas given typical or average treatment herein.

A Cost Model

Table 1 lists a cost model that accommodates each of the sealing methods considered. When dealing with the model it is important to reduce costs to averages. Any attempt to include all costs in terms of actuals might become cumbersome. Sensitivity testing will allow determination of those factors that most affect the final outcome or decision process. These driving factors then may be evaluated or reconsidered more closely to provide a more refined analysis without wasting effort on nondrivers. Cost-incurring elements include thermal cycling, leak testing, rework, re-inspection, material review board (MRB) action, electrical retests resulting from failures and scrap. Thermal cycling is not an insignificant cost. Extensive testing driven by high failure rates may cause demand to dictate further capital investment. Eliminating testing can reduce capacity requirements and possibly allow company growth without further capital investment. It is important to note that high yields through optimum seal design can potentially eliminate some current standard testing as well as added failure-related testing. Typically, current stress testing programs are based on historically high failure rates and are used to accelerate failures to the early process stages. Leak testing cost can be reduced in the same way as thermal cycling. Proper design can eliminate extra testing as well as current basic testing. Rework obviously can represent a sizable cost, particularly if it occurs in the latter stages of the process. Rework increases significantly after lidding, but also can be extensive on modules already populated with components. Re-inspections are required after any rework to assure that cleanliness and reliability have not been compromised. The further along the assembly is in the process, the more costly this inspection may become. Electrical retest may be required after rework to ensure that functionality of the module has not been affected before standard testing is resumed. MRB action may be necessary if failures occur at final acceptance. The added administrative cost of conducting this procedure may be significant depending on lot size in which the failure occurs and frequency of lot failures. Scrap can be costly should failures be nonrepairable. Looking at some failure cost relationships based on assumed or typical cost factors will shed some light on the true cost of hermetic sealing and allow knowledgeable choices to be made regarding the design approach.

Figure 1 shows a nomogram, which reflects the added unit cost of a seal as a function of three variables: feedthrough failure rate, quantity of feedthroughs per module and estimated module repair cost. Entering the nomogram (arrow), a feedthrough failure rate should be chosen based on assumption or actual history. Proceed vertically to a feedthrough quantity per housing and, hence, laterally to the vertical axis where a maximum module failure rate is indicated. This failure rate is a maximum since in a large housing lot the feedthrough failures may not necessarily be distributed evenly among the modules and, therefore, may present a slightly lower module failure rate. Now, proceed laterally to an estimated unit module repair cost and, hence, downward to the horizontal axis where an amortized cost per module is indicated. Continue down the line, again noting the quantity of feedthroughs per module (same as that used at entry) and, hence, laterally back to the vertical axis where an added cost per feedthrough is indicated. This cost must be added to the basic procurement cost to obtain the total cost of a hermetic feedthrough seal. The sample case shown (dotted line) indicates a four percent failure rate of feedthroughs and reflects a module with five feedthroughs (that is, nominal complexity). The module repair cost is $150, resulting in an amortized added module cost of $30 or an added cost per feedthrough of $6. The four percent failure rate example is not uncharacteristically low for contemporary seals. As discussed previously, the average survey failure rates reported were considerably higher. It is clear from the nomogram that the only factor that can reduce hermetic feedthrough sealing costs significantly is the failure rate given the fact that a fixed number of feedthroughs are required and improvements cannot be made in reducing failure-related module rework costs. It should be emphasized that the cost of repair presented in the nomogram must reflect the total added failure-related cost as presented by the model.

Table 2 lists the most common hermetic seal design options available currently for aluminum housings. Designs A, B and C are contemporary designs of straight-barrel or headed feedthroughs. The total cost range of these designs is significantly higher than that of the other three options. This discrepancy is due to the vast rework requirements resulting from solder joint failures. The cost of repair used here was a reasonably conservative $6 to $8 per feedthrough, which is based on a relatively low solder failure rate of four percent. As indicated previously, typical reported rates have been much higher. The laser-weldable feedthrough (option D) reflects a slightly lower total cost, which results entirely from its basic procurement cost. It has been assumed for the purpose of this article that a typical failure rate for this design is zero when, in fact, some failure rate could be expected due to the high cyclic stress levels encountered with this design. Option E reflects a variant of option A whereby gold-based solders are used in lieu of soft solders. This design allows the use of contemporary feedthrough seals that are slightly lower in cost, but the lower price is offset by the unusually high cost of gold-based solder preforms, which produce an installed material cost approximately the same as option F (the aluminum-compatible feedthrough). Two potentially costly problems exist with the use of option E. The first possible problem is the added thermal stress on the aluminum plating. Plating on aluminum, when applied properly, stands up well at temperatures of 250° to 275°C and allows for timely flow of product from the plater. The gold-based solders (Au/Sn, Au/Ge and Au/Si) require a minimum process temperature of 310°C and up to 400°C. Processing plated aluminum at these temperatures carries a significant risk of blistering. The cost impact of such failures is extremely high in that rework likely incurs the loss of the feedthrough seals as well as the cost of stripping, replating and reassembling the parts. If the housings have been post machined for lid laser weldability or plated with a masking process, the cost of plating rework is increased greatly. The second risk factor is the design risk associated with new geometries when using gold-based solders. These solders are quite strong and actually approach the strength of a braze. The fact that the seal glass is placed in high compression is good unless a geometric situation is encountered. In this case, the aluminum yields locally adjacent to the seal and causes the seal to crack. A number of commonly desired configurations preclude the reliable use of gold solders. When the aluminum yields locally adjacent to the seal, a total loss of the module results in the form of cracked seals with severe cost effects. The only recovery from this problem is a fallback to the use of soft solder aluminum-compatible feedthroughs, which may require housing changes and/or special feedthroughs to rectify, with the associated cost and schedule impacts. Many cases have been observed where a fallback design was not considered and modules could not be delivered to the required specification, resulting in untold cost and program delays.

Option F reflects the aluminum-compatible design, which utilizes a particular geometry and soft solders to provide a conformal solder joint that reliably absorbs all of the thermally created strain, allowing significant numbers of thermal cycles without loss of hermeticity. Figure 2 shows a curve reflecting the solder strain/radial clearance relationship for a typical glass-to-metal seal soldered into a 6061 aluminum housing. This case represents a Kovar® seal of 0.1" (dia) over a temperature range of -65° to +125°C. While this scenario is typical, the degree of solder strain (elongation percentage) will vary approximately directly with the total temperature range, the differential expansion of the seal and housing materials, and the seal diameter. The type of soft solder will not affect the degree of strain appreciably, but only the level of stress (solder yield) developed during thermal cycling. Two ranges of radial clearance are shown, representing contemporary seal designs and the aluminum-compatible design. Comparing the radial gap (solder thickness) of the two designs reflects the significant difference in the degree of strain (elongation) to which the solder joint is subjected. While this strain does not occur instantaneously with temperature change, it does occur essentially, in total, after hours of dwell at temperature extremes. The aluminum-compatible design reduces radial solder strain by approximately an order of magnitude over contemporary soft-soldered feedthrough seals. Zero failure rates are achievable, which results in a unit cost of a seal that is equal to that of its initial acquisition and assembly cost. The elimination of seal failures then can lead to reduction and/or elimination of current in-process testing when product confidence is gained.

Jack Pollock received his BSME degree from the University of New Hampshire. He was with Sanders Associates for 26 years as a development engineer and program manager, leading such programs as the AN/ALQ-137 and AN/ALQ-137-4 countermeasures systems for the FB-111 and EF-111 aircraft, respectively, as well as the AN/USM-464 computerized flight line test set. Pollock founded Product Engineering Associates Inc., a consulting firm, and Special Hermetic Products Inc. in 1986 and 1990, respectively. He is a registered professional engineer in New Hampshire.