Many types of high-performance integrated circuits (ICs) cost only a few dollars—but price is not any indication of their quality. To the contrary, the reliability of die attachment, wire bonding, and plastic molding has reached the level where these processes are taken for granted.
Nevertheless, modern ICs are mechanically complex, and things can go wrong. X-ray inspection methods typically investigate IC-to-printed circuit board solder joints as well as some aspects of the ICs themselves. Acoustic microscopy is a complementary technique well-suited for examining internal IC packaging and bonding faults.
Acoustic microimaging systems use a scanning transducer that pulses very high-frequency ultrasound into a part and receives the return echoes. Ultrasound is very sensitive to internal interfaces and gaps. It is this sensitivity that has made acoustic microimaging valuable in analyzing internal defects in a wide variety of components and parts.
The standard acoustic image is the planar image. But in recent years, the usefulness of acoustic microimaging has been extended to imaging methods including sectioned acoustic solids, automated high-throughput imaging, and virtual cross sections. All methods nondestructively produce data and images about internal features.
The Planar Image
In making a standard planar (C-mode) image, the ultrasonic transducer scans over the area of the sample, pulsing ultrasound into the sample and receiving return echoes thousands of times per second. The frequency of the ultrasound varies with the nature of the sample and the purpose of the acoustic imaging.
Acoustic frequencies range from 5 MHz to 500 MHz. Low frequencies penetrate deeply, but have limited resolution. High frequencies provide less penetration but sharper resolution. In most electronic components, samples are typically relatively thin and made of materials such as polymers, silicon, and metals that are good transmitters of ultrasound.
When ultrasound is pulsed into the sample, it travels downward until it encounters an interface. If the interface represents the successful bonding of two dissimilar materials—molding compound to die face, for instance—a portion of the ultrasound is reflected back to the transducer.
The rest of the ultrasound travels deeper into the sample. In this way, the acoustic image can display features at various levels such as die face, die attach, or lead frame within the same device.
But in most situations, the operator limits the acoustic image to a defined level within the sample. The level of interest may be the die face or the lead frame. The return echoes are gated electronically so that only the echoes from the desired depth, which arrive at the transducer at different times than echoes from other depths, are used in making the acoustic image.
Ultrasound will not propagate across an air gap, even if the gap is as small as 1,000 A. Cracks, disbonds, delaminations, and voids are all air gaps and have the useful property of reflecting all of the ultrasound back to the transducer. Since these anomalies are internal, hard to find by other methods, and likely to grow until they cause field failures, acoustic microimaging systems are very useful in finding these defects.
The systems also detect and image other non-gap defects. Examples include the uneven distribution of filler particles in an epoxy and a tilted die.
Figure 1 is a planar acoustic image showing a plastic ball grid array (BGA). The yellow-black defect surrounding the die is a popcorn crack that extends upward and outward through the molding compound from the die level. The die attach (red) is completely disbonded as well.
The Acoustic Solid
One of the most recently developed types of acoustic microimaging is the acoustic solid. It provides all of the acoustic data from the entire volume of the sample in one on-screen display that can be electronically sectioned as desired.
Although it represents a high degree of sophistication and utility, the process of making an acoustic solid is heavily automated. Once the sample is in place on the C-mode scanning acoustic microscope, software routines find the sample, and the transducer measures the sample’s dimensions.
Then you specify the number of acoustic slices desired. The acoustic solid is built up from acoustic slices, and the number of slices can be anywhere from 2 to 200. Many samples have from 20 to 40 slices, depending on thickness, the complexity of the internal structure, and the size of internal features.
The transducer scans the sample and makes the required number of acoustic slices. Since the transducer is working nondestructively downward through the sample, some slices have more detail than others. A slice showing only defect-free molding compound is essentially featureless; one showing the lead frame is rich in features. You can view the slices immediately after scanning, either individually or as a slide show.
Advanced software then stacks the slices into the acoustic solid. The completed acoustic solid appears on screen as a 3-D rectangular figure. The walls are initially opaque, and the only internal features that are visible are those that reach an exterior wall. But the acoustic solid contains all of the data about both normal and anomalous internal features.
At this point, typically you rotate the acoustic solid to provide a better view and section it in various ways to show significant internal features. Sectioning may be as simple as removing several top slices or a smaller 3-D block from one corner. A die attach defect, for example, can be viewed in these ways.
You can remove parts of the acoustic solid according to their acoustic characteristics. This option makes it possible to remove irregularly shaped features such as die attach. You also can leave the exterior walls in place to give an impression of overall shape, but make them translucent.
Figure 2 is the acoustic solid of a PLCC that contains a popcorn crack. First, this acoustic solid has been sectioned vertically (a front portion has been removed). The top few slices, representing molding compound, have been left in place. Part of a surface mold mark is shown at the left side of the figure.
Next, several additional slices of molding compound were removed, but the crack running through the molding compound was retained. The final image gives a comprehensive view of the nature and extent of the popcorn crack.
Automated High-Throughput Imaging
Automated acoustic microimaging systems provide sample handling and high-speed image-recognition algorithms for analysis. They handle up to 10,000 parts per hour, give the same data and images as a laboratory system, and run unattended. These systems are most useful on the production floor, where they routinely screen large numbers of parts.
One innovation: The system’s transducer does not scan individual parts because it would be too slow, but scans an entire JEDEC tray as though it were a single part. Resolution can be the same as when scanning parts individually. The transducer also can travel laterally much faster than the transducer in laboratory instruments.
The parts stay in JEDEC trays throughout the process. The trays are unstacked, fed through, and restacked automatically. Computer analysis searches through the tray image and checks individual parts for defects while the next tray is being scanned. The output consists of stored images, identical to those which would be made by a laboratory instrument using the same parameters, and a variety of charts and spreadsheets.
Figure 3 is the image of a tray of 24 plastic quad flat packs (PQFPs). One of the devices has a hidden internal defect identified in the image by a red marker. Images of trays such as this one are stored, but the identification of defective parts also can be made from spreadsheets. When analysis of the defect is needed in addition to identification, the image can be viewed to study the defect.
Conclusion
These are three of the techniques used by acoustic microimaging to image and analyze hidden internal defects. Acoustic microimaging is a useful tool in finding and eliminating the hidden internal defects that lead to field failures in ICs.
About the Author
Tom Adams is a writer and consultant based in Lawrenceville, NJ. He has written extensively on semiconductor topics.
Sonoscan, 2149 E. Pratt Blvd., Elk Grove Village, IL 60007, (630) 766-7088, e-mail: [email protected].
Copyright 1999 Nelson Publishing Inc.
February 1999