Structural Component Defects Screened Out By Acoustic Micro Imaging
New AMI developments have made component screening more informative and more effective, a boon to the task of scrutinizing avionics parts before assembly and removing those likely to cause field failures.
When a plastic-packaged IC fails in service, the cause is just as likely to be a pre-existing internal structural defect in the package, as it is to be a defect in the circuitry on the chip. The vast majority of internal structural defects are gap-type defects—voids, non-bonds, delaminations, cracks and the like. In the field, they may expand in size, collect moisture and contaminants that percolate through the plastic and initiate corrosion. In any case, the result is often a broken connection in the package.
X-ray has a hard time imaging gap-type defects, but they are easily imaged and analyzed by acoustic micro imaging (AMI) tools, which pulse high-frequency ultrasound into the package. The key to imaging a defect is the material interface between a solid (the mold compound, die, die paddle, etc.) and the air (or vacuum or another gas) inside the gap. The physical properties of the solid and the air are so different that more than 99.99% of an arriving ultrasonic pulse is reflected by the interface and received by the transducer of the AMI tool.
The extraordinarily high ultrasonic reflectivity of the gaps means that it is feasible to inspect parts before assembly to remove those parts whose internal defects are likely to cause field failures in avionics and other applications. The more expensive a part, or the more critical its application, the more likely it is to be imaged by an AMI tool. Ceramic chip capacitors, which are relatively inexpensive, are frequently imaged if they are intended for critical systems such as avionics control systems. Components going into low-cost systems where failure is not critical are generally not pre-scanned. (But during the 2008-2009 business recession some component suppliers screened components acoustically before shipping them as a way of gaining a competitive advantage on reliability.)
Today, though, components are typically screened by the assembler, or by outside laboratories specializing in such screening. In addition to plastic-packaged ICs, the components may include ceramic chip capacitors, insulated gate bipolar transistors (IGBTs) and other parts. Generally in-line automated systems that handle trays of parts perform high-volume screening (thousands or millions of components), image acoustically and report the results to factory information systems.
Medium-volume screening is carried out by a semi-automated system that requires a technician but provides non-stop throughput. The components in low-volume screening are often high-reliability military or aerospace items, and are usually imaged on laboratory-type systems. For one of the companies in this field, Sonoscan, it is not unusual to receive small lots of components that require very advanced inspection.
Amplitude Mode Imaging
There are about a dozen AMI imaging modes, each making different use of the ultrasound echoes from layers inside the component. One common mode of screening is the amplitude mode. A JEDEC-style tray of parts rests below the transducer, which scans the area of the tray. It pulses VHF or UHF ultrasound into the parts and receives the return echoes, often from a specific depth of interest (e.g., die attach) known to be a location for gap-type anomalies. Pulse-echoes occur thousands of times per second as the transducer scans.
Figure 1: Two trays on components on the stage of an AMI tool. The ultrasonic transducer is about to begin scanning.
Echoes are returned from any material interface in the part, but only those echoes whose arrival time coincides with the depth of interest are used to make the acoustic image. Thus a PQFP may be gated on the depth from the top of the die to the die paddle, since most critical gap-type anomalies will occur within this gate.
Figure 1 shows the arrangement for scanning trays of parts. The transducer here is stationary above the two trays; when scanning begins, it will move at speeds >1 m/s and collect echoes from thousands or millions of x-y coordinates on each part.
Figure 2 is the acoustic image of the gated depth (top of die to die paddle) of a tray of plastic-encapsulated ICs. In each IC package the die, die paddle and lead fingers are visible because the interfaces (e.g., mold compound to lead finger) reflect a portion of the ultrasound at medium amplitude. The red feature in the second IC in the second row from the top is a non-bond between the mold compound and the die paddle—red here indicating echoes having the highest amplitude. This IC would be rejected because a non-bond in this location is likely to expand underneath the die. If it separates the die from the die paddle, the resulting gap would interfere with heat removal from the die and could cause electrical failure.
Figure 2: Acoustic image of a portion of a tray of parts. The red area is a significant internal structural defect.
Thru-Scan is probably the second most commonly used acoustic micro imaging mode. The amplitude mode, described above, uses one transducer to pulse ultrasound into the top of the component and to receive the return echoes from material interfaces. Thru-Scan uses two transducers: one on top to pulse ultrasound, and one beneath the part to receive the ultrasound that is transmitted through the entire thickness of the part. A pulse sent into the top of the part will be entirely blocked from reaching the transducer below by gap-type defects. It will be partly attenuated by well bonded solid-to-solid interfaces.
Thru-Scan images of two TQFPs are shown in Figure 3. Ultrasound pulsed into the top surface traveled with some attenuation through most areas of both components. Bright regions have the least attenuation and the highest amplitude in echo signals exiting the bottom surface, while regions where the ultrasound is fully reflected are black because no echo signal reached the bottom surface.
In the TQFP at right (Figure 3) numerous small black features are visible. These are small voids in the mold compound. Because the Thru-Scan pulse travels through the entire thickness of the component, there is no way to tell the depth of each of these voids, but unless they are in direct contact with wires (for example) they pose little risk.
Figure 3: Thru-Scan images of part with (at left) serious internal defects, and (right) numerous small voids in the mold compound.
The TQFP at left shown in Figure 3 has numerous voids, as well as a large black ultrasound-blocking feature. This could be a delamination or a popcorn crack, but its position at the edge of the die makes it dangerous, and this component would probably be rejected.
Each assembler establishes his own rules for accept/reject. Typically the rules specify that an acoustic anomaly above a specified size and in a specified location will be cause for rejection. High-reliability products naturally have more rigorous standards. The TQFP at right, for example, might be rejected if the performance demands on the product it is going into are especially high.
Some imaging modes are just beginning to be used in screening of numbers of components. These AMI modes take advantage of the ways in which the transducer receives and handles return echoes from the interior of a sample. When the transducer sends a pulse of ultrasound into the sample, echoes are returned at various times from multiple interfaces located at various depths. Each echo reports time data (which tells the depth of the interface), amplitude data (which identifies well-bonded and non-bonded interfaces), and polarity data (whether the acoustic impedance—density times velocity—increases or decreases at the interface).
One AMI mode takes advantage of the time-since-launch data to permit the user to set a number of gates before starting the scan. Each gate receives echoes only from a specified vertical distance within the part. If there are ten gates, for example, then ten depths of interest are being imaged separately, and each will produce its own acoustic image. The gates can be of the same or different vertical width (thicknesses), and can be adjacent to each other, overlapping or separated. Imaging the TQFPs in Figure 3 in this way would have resulted in multiple images that would reveal the depth of each of the multiple voids. Setting multiple gates does not increase the scan time.
The plastic-packaged IC shown in Figure 4 was “sliced” horizontally into a total of 50 gates as a demonstration; most component screening operations would use a handful of gates. But using 50 gates shows how powerful this method can be.
Figure 4: Acoustic images of gates 22 through 24 (out of 50 gates) in a plastic-encapsulated IC package.
The topmost 15 or so gates, as you might expect, were featureless (solid black) because there were no ultrasonic reflections from the mold compound. At gate 22 in Figure 4, the die and lead fingers are just coming into view. In gate 23, there are numerous non-bonds (red) between the lead fingers and the mold compound, and also along the tape. Gate 24 lies slightly deeper; only scattered defects on the tape and on the lead fingers are visible.
Time Difference Mode
A recently developed AMI mode lets the user measure and map the flatness of an internal feature. It is handy for determining whether the die in an IC package is tilted, as well as mapping features that should be flat in other parts. Called the Time Difference Module, it records the time, in nanoseconds, between the launch of a pulse and the arrival of the returned echo. It does this for each of the millions of pulses launched by the transducer during the scan. A color is assigned to each range of arrival times, which determines the distance to the interface from which the echo was reflected. The result is a color map showing the tilt of the die, or, with other features, the details of warping.
This method was used to map the flatness of one ceramic raft in an IGBT module, as shown in Figure 5. The rainbow-like bands indicate the distance of each point on the raft’s surface from the transducer. The magenta area at bottom is farther from the transducer and thus the lowest portion of this warped raft. Moving upward, the red-orange area is highest, except for the narrow black band above it. To make this image, the transducer pulsed ultrasound into the surface of the heat sink at the bottom of the IGBT module, and echoes were gated on the solder bonding the heat sink, and on the raft. The black band represents a region where warping has pushed the solder aside and made contact with the heat sink.
Figure 5: Acoustically imaged contours of a warped ceramic raft inside an IGBT module.
The scattered irregular features are voids (air bubbles) in the solder. Their colors indicate their depth in the solder. Two large red voids, for example, are close to the heat sink.
Most large-scale component screening today uses Amplitude mode or Thru-Scan mode. There are also innovations that do not involve new imaging modes. One example: programming the scanning of a tray holding a few large parts so that only the parts, and not the unoccupied areas of the tray, are scanned. Overall, new AMI developments have made component screening more informative and more effective.
Tom Adams is a consultant to Sonoscan, Inc.