Long-term product reliability is a concern for any production manager responsible for the plastic-encapsulated microcircuits (PEMs) going on printed circuit boards used in military, aerospace or medical applications. Similar concern for reliability is exercised in manufacturing some commercial products, especially where competition is intense and a word-of-mouth reputation for field failures can be harmful.
It is widely understood that images of internal anomalies made by acoustic micro imaging can be used to evaluate component reliability before assembly, and thus help to predict product reliability in the field. Typically the required quantity of a given component is imaged, and evaluation of each image is performed according to one of the limited number of existing standards. One of the functions of Sonoscan’s laboratory division, known as SonoLab, is the screening of large and small lots of components. The components involved may be lead-frame-based or substrate-based PEMs, ceramic IC packages, various types of flip chips, ceramic chip capacitors, or other types. This article discusses PEMs.
The ultrasonic pulse that an acoustic micro imaging system’s transducer sends into a component is reflected to the transducer by the material interfaces it encounters and converted into a pixel whose brightness depends on the amplitude of the return echo. Internal defects – the conditions that may destroy field reliability – are generally material gaps such as delaminations, cracks or voids. More information about the gap comes from the polarity constituent of the echo. Even if a gap is 0.01µm thick, the interface between the gap and the solid material above it reflects effectively all the ultrasonic pulse. The transducer scanning a tray of components sends thousands of pulses per second into the sample and produces acoustic images that show well-bonded material interfaces in medium tones and gaps as bright white or brightly colored features.
Selecting and modifying a standard. A great deal of the predictive value that can be gained from acoustic micro imaging is determined in the early stages of planning. The basic problem is that none of the various standards (Table 1) for acoustic inspection is likely to be perfect for achieving reliability for a given component in a given application. Those involved need to define precisely what is to be acheived. Acoustic micro imaging will tell the size and location of internal defects. Typically, the goal is to have the highest possible field reliability by removing defective or questionable components before assembly. With careful planning, that goal will probably be reached. If an inspection standard is chosen at random, it may fall short.
[Ed.: To enlarge the figure, right-click on it, then click View Image, then left-click on the figure.]
There are significant differences among the standards listed in Table 1. MIL-STD-883 Method 2030 is limited to die attach evaluation; PEM-INST-001 mentions die attach, but gives no standard. A defect may be called a crack in one standard and a delamination in another. The trick is to find the standard that can best fit a particular component. Often, a standard gives the best results if it is modified to match the component being imaged.
The lots of parts arriving at one of the three SonoLabs are typically destined for use in military, aerospace automotive or medical applications. Some will wind up in high-end commercial products such as servers, in competitive retail products such as cellphones, and in other commercial or retail products where competition or the application demands high reliability. Parts are scanned in trays. Depending on the volume of parts and customer needs, the parts may be scanned on a manual or partly automated C-mode scanning acoustic microscope (CSAM), or on a fully automated tray scanner.
Those who want to classify a part for a certain moisture sensitivity level using J-STD-020 in order to predict reliability may be in for a disappointment. J-STD-020 is really a manufacturing specification. It gives plenty of information about the component, but tests between moisture uptake, baking and reflow may not be what are needed to determine whether components will survive in the field. It’s not safe to assume that acceptable moisture sensitivity level (MSL) information will lead straight to field reliability.
If a key concern is the possibility of field failures caused by defects in the die attach material, the better standard may be MIL-STD-883 Method 2030. This standard wasn’t designed for PEMs; its original purpose was the inspection of power devices in ceramic packages for military applications. More recently, Method 2030 was rewritten to be made more generic and more easily applied to PEMs. The method defines as rejectable any die attach that has voids covering more than 50% of the intended bond area, a single void covering more than 15% of the intended bond area, or a single corner void covering more than 10% of the intended bond area.
J-STD-020 has a similar definition (“No delamination/cracking >50% of the die attach area in thermally enhanced packages or devices that require electrical contact to the backside of the die”) for metal leadframe-based PEMs, but has a different overall purpose. J-STD-020 also requires “no delamination/cracking change >10% through the die attach region” for substrate-based PEMs from before soak until after reflow.”
Method 2030, by itself, might give adequate field reliability for the aforementioned component. The actual danger levels, however, are likely to differ from the percentages in 2030. How long a given PEM will function when it has a particular percentage of delamination in the die attach depends on many factors: the physical characteristics of the die attach material and the substrate material, the rate at which heat is dissipated from this particular package design, and (especially for power devices) the precise locations of hot spots on the die itself. This component might, in the environment in which it will be used, survive for the desired lifespan with 50% of the die attach area voided, or even with 70% of the area voided. But it might also fail in service if only 10% of the die attach area is voided. An acoustic overlay is sometimes made for power devices. When placed over a map of the die, the overlay shows whether any voids lie directly beneath hot spots. Some find a better fit by using MIL-STD-883 Method 2012, which was written for x-ray inspection, but the failure criteria are relevant to acoustic micro imaging. Among its failure criteria: 50% or more total die attach delamination, or a single largest delamination >10% of the die attach area.
To achieve the best results, standards may be modified in other ways. Method 2030 and other standards become more useful when their criteria are modified according to the user’s experience with a particular component. Users may be encouraged to make such modifications in order to prevent the pass/fail criteria from resting solely on definitions that may have been written to pass judgment on a very different component. The three components in Figure 1 have not only varying areas of voids in the die attach area; they have varying distributions of those voids, especially in part C, where most of the 18.43% void area is in a single large void. For one component these percentages and distributions might be acceptable; for another, especially an application where no single void >10% of the die attach area is permitted, part C might be a reject. The user’s judgment, based on experience with the parts being screened, is critical.
Some of the user’s knowledge about a given component may come from previous field failure data. It is often worthwhile for the customer to conduct life tests in order to examine in detail how various defects change as the component ages and goes through temperature cycles. Customers bringing PEMs to a SonoLab often wind up using Method 2030 as the standard, but they often acquire additional information to refine the criteria.
Figure 2 shows acoustic images of a PEM before life testing (left) and after 1000 cycles (right). Before testing, there are several delaminations (red) on or near wire bonds on the lead fingers. After 1000 cycles, some of these delaminations have grown, and a few new ones have appeared. But the die paddle surrounding the die is now largely delaminated from the mold compound. In service, this delamination might grow to extend under the die.
Other standards are also available for screening components to improve long-term reliability. One is J-STD-035. Unlike J-STD-020, this standard does not address moisture sensitivity. Instead, it simply creates very generic definitions for the anomalies that may appear when transducer pulses ultrasound into the circuit side or the non-circuit side of the package. The seven anomalies are shown in Figure 3.
The evaluation of some defects is essentially cast in stone. Die face delaminations are practically always a cause for rejection, and are thus at the head of the list. Delaminations near lead finger wire bonds are surely risky, but the wire bond is typically a wedge bond, and may, in a particular package, be somewhat better at enduring a delamination.
There is no quantitative definition of the area of an anomaly in J-STD-035. Type 5, for example, seen in Part #7 in Figure 4, simply defines any delamination of the encapsulant from either surface of a lead finger. But the two sides are not equally risky. A delamination on the top side could allow corrosion of the circuit or may expand until it reaches a wire bond, but a delamination on the bottom side is relatively innocuous, unless it somehow propagates to the top side.
J-STD-020 is more comprehensive and rejects any component having an “internal crack extending more than 2/3 the distance from any internal feature to the outside of the package.” This definition encompasses lead finger delaminations, as well as delaminations along tie bars and heat sinks. The relative length of a lead finger delamination may be modified, for example, if the component being imaged has a small die, long wires, and very short lead fingers. The usual modification in this case is to consider a lead finger delamination of any length as cause for rejection.
Some customers use PEM-INST-001, a standard written by NASA before IPC/JEDEC put together J-STD-020. The two standards are similar but not identical. For example, PEM-INST-001 lists as a cause for rejection “Delaminations extending more than 2/3 the length of the internal part of the leads.” J-STD-020 calls such an anomaly an “internal crack extending more than 2/3 the distance from any internal feature to the outside of the package.”
Acoustic micro imaging is most successful at predicting reliability and identifying components likely to fail when engineers 1) select an inspection standard that is well-suited to the part; and 2) modify the standard according to the data they have from life testing, field failures and other sources.
Tom Adams is a consultant to Sonoscan (sonoscan.com); firstname.lastname@example.org.