Originally Published MEM Fall 2005
Miniature PCB-mounted components that passed tests for high-reliability portable applications provide an example of component qualification for medical electronics OEMs.
New medical electronic products destined for wearable or portable applications have to be small and lightweight. The challenge for the product developer competing in this very specialized field is to introduce a device that will meet all performance and functionality expectations in a form factor that is significantly smaller and lighter than the product it is replacing. The engineer designing the medical device will likely look at the newer generation of integrated circuit (IC) packages offered by commercial suppliers, many of whom can furnish a packaged device that is only slightly larger than the silicon die element.
Two major considerations medical electronic equipment manufacturers must address when adapting these miniature chip-scale IC packages (CSPs) are reliability and performance. To successfully qualify an IC component package technology for medical electronic applications, the OEM must establish confidence in end-product reliability and, for some products, receive assurance of uninterrupted service.
|Figure 1. The µBGA package developed by Tessera Inc. (San Jose) looks like any miniature BGA.
(click to enlarge)
One package methodology that has been developed to miniaturize IC packages and improve their reliability in medical electronics is the micro ball-grid array (µBGA). This methodology provides a finished-component outline barely larger than the silicon die itself. The devices are furnished with either eutectic tin-lead or lead-free tin-silver-copper solder-ball contacts with a center-to-center pitch of 0.75 mm. The package is unique in that the die is mounted with its active surface facing down, providing the short electrical interface often needed to enhance performance to meet requirements. Although a µBGA package outwardly looks like any miniature BGA (see Figure 1), it is constructed as a physically compliant system of materials and die-to-package interfacing that compensates for the wide differences in coefficient of thermal expansion (CTE) between the silicon die and the printed circuit board (PCB) structure (see Figure 2).
|Figure 2. The system of materials and the die-to-package interface in the µBGA package compensates for the difference in coefficient of thermal expansion (CTE) between the die and board.
(click to enlarge)
The CSP selected for the reliability-testing regimen described in this article is a µBGA package optimized for a flash memory die, with 46 inputs and outputs (I/O), and a die size of 5.76 X 7.87 mm.
When designing a medical electronic product that integrates CSP technology, a number of testing procedures must be followed. Because reliability is a critical requirement, the following discussion outlines the rigorous reliability testing process necessary when qualifying CSP technology for integration into medical equipment. The CSP discussed here is a µBBA package with a flash memory die, 46 inputs and outputs, and a die size of 5.76 X 7.87 mm. For designers, this is a multistep process that begins with identifying a testing methodology. It includes selecting and undergoing an independent qualification program to support a company's own testing results.
The µBGA package technology is already in wide use throughout the semiconductor industry. Early qualification testing of components for medical applications included a number of considerations.
First of these was establishment of a predictive model, which related to the number of thermal cycles necessary to produce device failure. Most failures in electronic components mounted onto a circuit board occur at the solder-joint interface. These failures typically result directly from the strains developed at high and low temperature extremes. Different rates of expansion of different materials are exhibited during cyclic temperature variations typical of operational use. Temperature variations also result from different coefficients of thermal expansion, or cyclic differential expansion. This phenomenon typically is influenced by the range between maximum and minimum temperatures experienced during operational use or temperature-cycling tests. The failures that eventually result from temperature cycling during qualification testing are recorded and show the statistical distribution of failures due to wear.
To confirm that the solder-joint attachment on surface-mount circuit assemblies meets reliability expectations in the intended use environments, it is necessary to produce a reliability database. Accelerated fatigue tests may confirm reliability for some specified applications, but slower cycle testing typically provides more-useful data.
Before developing a product using any IC device, engineers must consider such factors as the physical features and construction of the device, suitable PCB substrate materials, surface finish, and the attachment methodology. Many companies rely on industry standards to define test methodology for qualifying the components for a particular application.
Physical tests typically are designed to replicate the actual use environments in which the electronic assemblies will operate. In addition, the test method establishes the expected levels of performance and reliability of both the component part and the solder attachments.
The benefits of reliability testing include:
Industry standards are in place to guide the engineer in identifying the product's use category and understanding the failure mechanisms that may affect its reliability over time. The IPC-SM-785 specification, for example, defines reliability as "the ability of a product [surface-mount solder attachment] to function under given conditions and for a specified period of time without exceeding acceptable failure levels."1
IPC-9701, a relatively new industry specification, establishes the specific test methods for evaluating both product performance and reliability.2 The test methods have been developed to assist the supplier and user in qualifying components, or the finished product, for specific use categories or operating environments.
The load condition that was the focus of the testing described below was cyclic differential thermal expansion, but vibration (usually experienced during transport) and mechanical shock (high acceleration force) are other loads whose effect needs to be investigated. These load conditions may exist individually, sequentially, or simultaneously.
Establishing Thermal-Cycle Test Conditions
Temperature cycling established for the µBGA qualification test program subjects the sample assemblies to a thermal environment ranging between extremes of –40° and 125°C. Each full cycle requires a 20-minute ramp and 10-minute dwell for each extreme, making a total cycle period of 1 hour.
Two environmental-test-monitoring methods are available: continuous electrical monitoring and periodic sampling. Continuous monitoring typically requires specialized equipment, but this technique can detect random or intermittent failure. Periodic sampling detects failure outside of the test environment but may not detect random or intermittent failure.
Soft failures of the solder process or device can be detected only by means of continuous monitoring. By contrast, hard failures, those failures that interrupt device function, can be identified through periodic or posttest sampling. Mechanisms that cause device defects or failure are opens or cracks in the solder joint, a change in circuit resistance, and damage within the package structure.
IPC-9701 was developed to replace the somewhat dated MIL-STD-883. This industry specification establishes specific thermal-stress test conditions to create in evaluating board assembly–level product performance and reliability (see Table I). The test conditions and number of thermal cycles required for qualification are product dependent and typically defined by the product developer. The requirement for the test platform (the PCB) is not defined in the IPC standard, however. Users are advised to develop a test vehicle that closely emulates the physical features of the actual product in terms of size, thickness, and the number of circuit layers.
Numbe of Thermal Cycles
|TC 1: 0° to 100°C||NTC-A: 200|
|TC 2: -25° to 100°C||NTC-B: 500|
|TC 3: -40° to 125°C||NTC-C: 1000|
|TC 4: -55° to 125°C||NTC-D: 3000|
|TC 5: -55° to 100°C||NTC-E: 6000|
|Table I. Thermal-stress test conditions for evaluating PCB surface-mount solder attachments, from IPC-9701.|
Although early generations of the µBGA package met the basic requirements for most commercial applications, the package substrate design and assembly process have been refined since then in order to attain the even higher levels of reliability required for industrial, defense, and portable medical electronic applications. To establish that the reliability improvement was adequate, the µBGA was subjected to two environmental test programs.
One program was conducted by the independent Gintic Institute of Manufacturing Technology (IMT) in Singapore. In addition, Tessera performed a series of process evaluation tests at its test laboratory in San Jose, CA. The thermal-cycle test conditions selected for both test programs were typical of those defined in IPC-9701 as test condition 3 (TC 3).
Independent Qualification Test Program
A test platform was developed with three different soldering-land-pattern designs: blind-via-in-land, through-hole-via-in-land, and modified-dog-bone-style land connecting to a via-in-pad near the land.
The land pattern dimensions were engineered to facilitate observation of the results of stress and strain on solder joints. The stencil Gintic used for solder-paste printing in the assembly process was made of 0.127-mm-thick stainless steel with laser-cut openings. In order to maximize the amount of printed solder paste, a slightly larger square opening of 0.325 X 0.325 mm was used for each aperture in the array. The assembly of the µBGA test boards involved standard surface-mount processes. No assembly defects were apparent for either the blind-via or modified-dog-bone pad designs, and the solder joints formed on these pads appeared to have a regular and uniform shape.
The study found that a land pattern 0.275–0.325 mm in diameter provided the optimal condition for solder attachment. Surface finishes of two types were used for the test vehicles: organic surface protection, a chemical film coating, and electroless nickel-immersion gold plating.
Subsequent x-ray inspection revealed that some of the ball contacts had formed voids within the sphere. These voids typically occurred in the solder joints just above the blind-via hole. It was observed, however, that the via-in-pad design furnished the highest standoff, owing to the somewhat smaller soldering area. In addition, Gintic technicians measured the standoff height of each device soldered to assembled test boards and summarized the results in order to evaluate the effect the slightly varying land size and standoff height might have on reliability. No correlation with solder joint failure was reported.
The institute applied finite-element methods to study the package stress-strain distribution in the temperature-cycling test. A quarter model of the µBGA assembly with modified-dog-bone pad was used. The solder joint configuration followed the cross-sectioning picture of an actual joint. A total of 2690 eight-noded 3-D elements were used in the model. To simulate the deformation, elastic-plastic-creep analysis was performed following the temperature profile in the test.
|Figure 3. Weibull distribution of the time-to-failure data gathered during µBGA assembly solder-joint testing performed by Ginitic IMT (Singapore).
(click to enlarge)
The crack-initiation time for the solder joints was predicted to be around 1400 cycles. Indeed, scanning electron microscopy analysis of cross-sectioned samples discovered no cracks at the solder joints after 1000 cycles. The first failure occurred after 1800 cycles, a lifetime significantly higher than predicted. The Weibull distribution of the time-to-failure data is shown in Figure 3. The Weibull plot is used in the electronics industry to describe the reliability of a product and to predict its life cycle. It also allows the first failure to be extrapolated based on empirical results.
Gintic concluded that the packaging process employed for the 46-I/O µBGA device was highly reliable and well suited for applications with stringent service conditions. The institute performed low-acceleration tests that closely mimicked expected field conditions. Companies lacking the time and resources required for low-acceleration testing may have to conduct a higher-acceleration test. It is important to note, however, that while highly accelerated testing may be acceptable in some commercial electronics applications, uncertainty and risk goes up with the acceleration level. As such, high-acceleration testing may not be suitable to meet the more-stringent reliability requirements of a broad range of medical electronic products.
Developer's Reliability Test Program
Using the same µBGA device, test samples were prepared that had both lead-free-alloy solder-ball contacts and a traditional eutectic-alloy composition. The general test method used is classified as a highly accelerated thermal cycle. The cyclic damage and the fatigue life exhibited in accelerated testing may not be equivalent to those in slower testing processes because of the time-dependent creep and stress-relaxation properties of solder. Slower tests typically take 1 hour to complete each high-to-low-to-high-temperature cycle. The accelerated test program is usually preferred for general package assembly monitoring because the tests are from 100 to 500 times faster than the cycle test procedure.
A recent test prepared devices with lead-bearing-solder alloy contacts and also with lead-free-alloy contacts. The lead-free test comprised nine PCB assemblies, each board having 10 46-I/O, die-size µBGA packages. The assembled boards were prepared using components with Sn96.3/Ag3.2/Cu0.5 solder balls, attached to the PCB by means of a solder paste having a Sn96.5/Ag3.5 alloy composition. Other samples were assembled using the same solder-paste alloy but with Sn96.5/Ag3.5-alloy balls. Additional assemblies—the control samples—used the Sn63/Pb37 eutectic-alloy ball and solder paste.
In each of the board sets, the solder paste was printed onto the array pattern using a laser-cut stainless-steel stencil. Devices were placed into the deposited paste and reflow-soldered using a commercial in-line convection-type furnace.
The melting points of the lead-free and eutectic solder paste alloy compositions were very different. The temperature profile for reflow soldering the lead-free alloy is roughly 30°C higher than that for the traditional tin-lead eutectic composite. That is because the temperature of both alloys must be elevated beyond the liquidus threshold by 20°–30°C to promote proper wetting or joining.
Following Joint Electron Device Engineering Council (JEDEC) standard component-preparation requirements, technicians included preconditioning with a 24-hour bake at 125°C, 168 hours at 85°C and 85% relative humidity (RH), three exposures to temperatures experienced in reflow soldering of each component's alloy composition, and 1000 hours at 85°C and 85% RH. The components were then held for 168 hours in an autoclave set at 121°C and 100% RH, after which they went into high-temperature storage at 150°C for an additional 1000 hours. Temperature cycling established for this program was according to IPC-9701, Test Condition 3-C: the sample assemblies were subjected to 1000 cycles ranging between a low of –40°C and a high of 125°C.
|Figure 4. Weibull distribution of the time-to-failure data for an assembly having three different solder-alloy materials, gathered during high-acceleration µBGA solder-joint testing performed by Tessera Inc. (San Jose). The graph shows results for a TV46 with onboard cycling from -40° to 125°C.
(click to enlarge)
Parallel testing of the assemblies with lead-free versus lead-bearing solder-ball and attachment alloy produced consistently positive results, actually exceeding the 1000-cycles-before-failure requirement. In fact, all 10 of the sample assembly configurations successfully completed more than 2500 cycles without electrical failure. The Weibull distribution of the time-to-failure data for the assembly having three different solder-alloy ball and attachment materials is shown in Figure 4.
Thermal-stress testing of the µBGA package yielded performance results that exceeded the established criteria, in part because of the package design and the base materials used in its construction (as shown in Figure 2). The package technology furnishes a compliant, stress-absorbing structure. Physical compliance is achieved through a unique package-to-die lead-bonding process and the use of a specially formulated polymer compound for in-package encapsulation. The polymer material enables a limited amount of elasticity within the component's structure to absorb any physical stresses experienced during thermal excursions while the product is in use.
Proven to offer a physically robust and reliable package, µBGA devices are being manufactured in high volume by several IC suppliers, providing memory for medical, commercial, and industrial device applications. The package technology, which is versatile and capable of adapting to a variety of die designs, lends itself to manufacturing via methods consistent with traditional IC packaging. It is supported by a well-established material and equipment supplier infrastructure.
Vern Solberg is senior application engineer for Tessera Inc. (San Jose). He can be reached at email@example.com.
Copyright ©2005 Medical Electronics Manufacturing