A client provided several PCBAs for failure analysis of out-of-tolerance MELF resistors.

This is an optical image of a typical failed resistor showing localized delamination of the protective coating from the Tantalum thick film resistor material and Sn-Pb particles trapped at the interface.

This is a high magnification optical image of a Sn-Pb particle as-viewed through the outer coating that bridged the laser cut causing a short.

This elemental spectrum shows that the short is a Sn-Pb particle (sample particle as in previous image).

Sn-Pb particles were found in microsections sitting on top of the Tantalum thick film resistor material under ~ 40 microns of outer coating.

The root cause of the failures was tramp metal (Sn-Pb) contamination under the coating suggesting inattention to cleanliness in the manufacturing environment. I am left wondering why the manufacturing operation didn’t use AOI techniques (or human inspection of samples) to prevent this from getting out of the factory. The loss of reputation seems to be a cost greater than the investment in AOI.

This is an optical image (bottom-view) of a surface mount transformer.

This is a BSE SEM image of a microsection through the transformer.

 

If the image contrast is appropriate, image analysis software can be used to count and measure feature location and geometry.

In this case, the count and diameters of the conductors is summarized.

 

This is an optical image (top view) of an LED that failed open circuited.

The next image is a side view of the microsectioned LED. This fracture shows striations (red arrow) propagating away from the bond wire. There is also a discontinuity in the bond wire (white arrow) that appears to be the break that explains the open circuit condition.

How can this damage be explained? The CTE of cast epoxy (the body/lens) is ~55 ppm/C and the CTE of gold (the bond wire) is ~14 ppm/C. So the epoxy thermally expands ~4X more than the gold for a given temperature excursion. It seems likely that the break in the gold bond wire occurred due to an increase in temperature, which would put the gold wire in tension. On the other hand, it seems likely that the fracture in the lens material occurred due to cold temperatures (or alternatively due to pulse heating of the bond wire) where the epoxy just of the gold/epoxy interface would be in tension. In any case, our experience shows that LEDs tend to fail if forward current and ambient temperatures approach the maximum operating limits in the LED specification. Since lighting applications are often pushing for maximum lumens (maximum forward current and power dissipation), many failures could be avoided by applying derating principles at the design stage.

Chip components such as MLCCs, high value (~ 1Mohm) chip resistors, and transorbs that fail due to excessive electrical leakage are often not component failures at all, but rather are SMT design/process induced failures.  For example, the chip component illustrated below exceeded the component level specification of < 1 uA leakage.

In this case, the excessive leakage condition in-circuit was attributed to tin residue on the board surface under the chip component.  The suspect component was mechanically removed and the board surface was examined in a SEM.  The BSE SEM image (top) showed a residue on the board surface between the two terminals of higher average atomic number than the solder mask.  An elemental map for tin (bottom) showed that the tin residue was essentially continuous across the isolation space.

Three possible factors that can contribute to these residues are (1) design – the capillary gap, terminal spacing, and applied voltage are factors, (2) process – the cleaning process and/or solder flux selection may be inadequate, or (3) electromigration – the bias voltage, ionic residues, and humidity cause electromigration of tin across the isolation gap with time. The root cause is often a combination of these factors. 

A microcontroller IC was failing at apparently random I/O pins.

This is an optical image of the chemically decapsulated device.

 

This is a BSE SEM image of the decapsulated device.

 

This image shows a displaced ball bond with an elemental spectrum of the bottom of the crater.

 

This is a higher magnification image of the bond crater.

Bond cratering can occur at apparently random bond pads likely due to variation & non-perfect planarity of the die attachment.

A PCBA was provided to SEM Lab, Inc. for evaluation of some QFN solder joints.  This is a BSE SEM image of a microsection of the QFN.

The solder joint geometry appeared to be acceptable in this section. There was a uniform layer of Cu-Sn intermetallic at both the solder/pad and lead-frame/solder interfaces suggesting that the reflow profile was acceptable.  The interfacial intermetallic layers also suggest that there was no solderability problem associated with the QFN or PWB pads.

This is a 2nd section plane that included the thermal pad under the QFN.  The level of solder fill in the PTH-vias under the thermal pad was variable.

The “toe” of the solder joints shows an excess-solder condition possibly because of the balance of forces caused by solder surface tension between the signal solder joints and the thermal pad, which determines the solder joint height.

One possible solution for mitigating these issues is to fill the PTH-vias as illustrated in the image below.

SEM/EDS analysis is a useful technique for characterizing contamination on contact surfaces.  

The figures below show EDS spectra of contamination on a gold-plated connector contact.  The contamination was causing intermittent high contact resistance.

The EDS spectrum in this case shows C, O, Na, Mg, Al, Si, Au, S, K, Ca, Ni, & Cu.  Au is the only element that should appear in the spectrum.  The contamination appears to be a combination of “dirt”, ionic compounds, and corrosion product (i.e. of Ni & Cu).

The EDS spectrum in this case shows C, O, Zn, Mg, Al, Si, Au, S, K, Ca, & Fe.  Between these two spectra we see 10%+ of the elements on the periodic table of elements, where we should only detect gold.

  • Oxygen (O) – mostly as corrosion product, i.e. oxidation 
  • Iron (Fe) – unknown source, possible corrosion product 
  • Magnesium (Mg) – likely as oxide mineral or glass constituent 
  • Aluminum (Al) – likely as oxide mineral or glass constituent 
  • Silicone (Si) – likely as oxide mineral or glass constituent 
  • Chlorine (Cl) – corrosive, likely a chemical contaminant 
  • Sulfur (S) – corrosive, likely a chemical contaminant 
  • Potassium (K) – likely as oxide mineral or glass constituent 
  • Calcium (Ca) – likely as oxide mineral or glass constituent 
  • Copper (Cu) – possible corrosion product 
  • Nickel (Ni) – possible corrosion product 
  • Zinc (Zn) – possible corrosion product

Dirt, ionic compounds, and corrosion product are likely to cause high contact resistance for a connector like this one.   The EDS spectra provide a good basis for understanding the nature of the contact resistance problem.

 

It can be challenging for failure analysts to know if device damage caused electrical breakdown or did electrical breakdown cause the damage.  Here’s an example where a failed Power MOSFET device shorted source-to-drain.  The image below shows the device geometry and pin descriptions.  The device package is a small-outline-no-leads (SON) with 26-mil pad pitch.

The solder joints were suspected to be a potential failure cause by our client, so microsection analysis was chosen as the appropriate analysis technique (versus chemical decapsulation of the device and SEM analysis of the device die).  The figure below is a BSE SEM image of the microsection of the device as-mounted on the PCBA.

This is one of the source solder joints, which has a void and shows an excess solder condition at the toe.  Neither of these issues likely contributed to the failure.

The thermal pad also shows solder voids, but they are not so severe as to significantly increase the thermal impedance of the device.

The source-to-drain short was found at the front edge of the source lead frame.

The die attachment had failed on the left end of the die as shown below.

The molding compound separated from the die surface on the right end of the die (see below).  This damage raises a question about whether thermal stresses might have caused the separations (e.g. popcorn damage) causing the die to to overheat and break down electrically?  Or, did the device break down causing the thermal damage due to the heat generated by the short?

The failure current at breakdown can be estimated from the radius of the alloyed region (e.g. 1-mil radius =~ 1 ampere) [1] .  The short site (melt pipe) diameter was ~ 265 microns (or 10.3 mils), suggesting a failure current of ~ 5.2 amps, which is within the acceptable operating range at 25C (i.e. 13 amps max continuous operating current) per the device specification.  

The safe operating conditions diagram for this device is shown below.  The brown line at 5.2 amps is the current estimated from the radius of the melt pipe.  The diagram suggests that the device likely failed due either to a voltage spike above 30 volts or an intermediate voltage exceeding the pulse width indicated by the colored-dashed-lines.

It seems likely that thermal damage was caused by the failure event rather than being the cause of the failure.  The rapid thermal expansion in the region of the melt pipe likely failed the interfaces that were found to have separations.

In summary, this device most likely failed due to an over-voltage condition on the source signal.

REFERENCES

[1] Microelectronic Failure Analysis – Desk Reference 3rd Edition, ASM Press, ISBN 0 -87170-479-X, p.327.

 A shorted PWB was submitted for microsection analysis.  The client reported that there was a known issue involving break out of the board from the panel (i.e. misalignment of the V-score).  The client attempted to mitigate the break out damage through application of a bead of epoxy on the edge of the board (see optical image below).

 

 The BSE SEM image below shows the location of the short (dashed oval) and a laminate fracture that propagated from the edge of the board (arrows).

 The higher magnification image below of the short site showed carbonized epoxy resin and melted glass bundles.

 A region just outside of the short site showed conductive anodic filaments (CAF) in epoxy fractures and de-bonded glass/resin interfaces.

 EDS analysis confirmed that the filaments were copper.

So, why didn’t the epoxy bead on the edge of the PWB prevent the CAF short?  The likely explanation is either one or both of these reasons, (1) epoxy is non-hermetic so moisture can ingress in spite of an epoxy bead or (2) moisture and perhaps process chemistry were trapped in the capillary spaces created by the break out damage before the epoxy bead was applied.

Some edge card connector solder joints were reportedly failing open.  This is an optical image of the connector that showed fractures (red arrows) in the epoxy staking material apparently used to attempt to mitigate the problem.

This is an optical image at intermediate stage of micro-sectioning into the suspect solder joints. The displacements (yellow arrows) are very large relative to what the solder joints could be expected to survive.  Note also that the back face of the connector body is not normal to the PWB as it should be, which is the likely cause of the mechanical loading on the solder joints.  When the subject connector is mated with its mating connector, the loads on the solder joints are exactly as indicated by the displacement vectors shown in this image.

These are SEM images of solder fractures as viewed in a parallel and traverse section of the solder joints.  The evidence that these are creep rupture failures includes (1) the large upward displacement of the connector leads and (2) the fact that the fracture propagated through the bulk solder joint rather than through an intermetallic compound layer or at an interface.

So, why didn’t the epoxy staking material work to prevent these failures?  Because epoxy has a very low mechanical stiffness relative to the connector leads, so in spite of the addition of staking material the majority of the applied load is still carried through the leads to the solder joints.