Scanning acoustic microscopy (SAM) can map hidden separation at a copper-to-ceramic interface without cutting the substrate. It does this by listening for reflected ultrasound and isolating the echoes from the interface of interest. A map only answers the inspection question if the lab can show which interface was scanned, which regions were readable, and how the suspected separation was distinguished from normal board geometry.
Consider a direct bonded copper (DBC) substrate being qualified for a power module. Its normal metal pattern already creates changes in the acoustic image, while a small separation at the copper edge can matter more than a larger indication in a noncritical location. The aim is not to obtain a dramatic image. It is to determine what should be rescanned, confirmed by another method, or evaluated against the drawing.
Scanning Acoustic Microscopy (SAM): How It Works
The instrument sends a short ultrasonic pulse and records when echoes return and how they differ from the reference signal. Echoes form where the acoustic properties of two materials differ. An intact copper-to-ceramic boundary can therefore reflect sound even when nothing is wrong. A gap filled with air creates a very different boundary and may produce a strong echo and a change in polarity relative to a bonded reference. Display settings determine which colors show those changes.
A shallow air gap can also block sound from reaching a deeper bond. The darker area underneath is not proof of a second defect; it may simply be an acoustic shadow. The first step is to locate the reflecting interface in the time trace rather than interpret color alone. The Fraunhofer ENAS method overview explains how acoustic reflections reveal internal material discontinuities.
Acoustic Microscopy of Semiconductor Packages: Find the Right Interface
Start with the assembly cross-section and name the joint to be evaluated. A bare DBC coupon might ask one question about the copper-to-ceramic bond. A finished module may add die attach above it and a substrate-to-baseplate joint below it. Those interfaces are not interchangeable: the same X-Y location may have several echoes from different depths.
Draw the layers in the order the pulse meets them from the scan side. Record whether the top copper is patterned, which areas have no metal, and where thicker features may block the beam. The Aalborg power-module study used a DBC stack with 0.30 mm copper, 0.63 mm aluminum nitride and 0.30 mm copper, plus a die and baseplate. Those dimensions describe its sample, not a standard DBC stack. If the ceramic or copper thickness changes, do not assume the previous timing window still isolates the joint.
C Scan Ultrasonic Testing: Set the Time Gate
A C-scan is a plan view assembled from echoes in a selected time window, called a gate. Start with an A-scan, the echo trace at one point on a comparable intact area. Match echo arrivals to the stack, then inspect a B-scan, a view along a line that shows how the echo positions change across the part. Only then set the C-scan gate around the target echo. A single wide gate can mix two neighboring interfaces into one plausible-looking image.
For the example coupon, compare an intact area beneath the copper island with a line that crosses the island edge and the suspect region. Check whether the same echo stays inside the gate on both sides of the suspect region. If the surface height varies, a gate referenced to the front-surface echo may help; validate it against the known layer sequence. If two reflections cannot be separated, record the interface as unresolved rather than assigning the strongest one to the nearest layer by guesswork.
Scanning Acoustic Microscopy Resolution: What Controls Detection?
There is no single resolution that applies to every ceramic PCB. A higher-frequency focused probe can improve lateral detail but generally leaves less useful signal at depth. A smaller scan step samples the image more densely; it does not shrink the acoustic spot or fix a weak echo. A fine-looking pixel grid can therefore overstate what the setup can actually resolve.
| Setting | What to record | What the lab must demonstrate |
| Probe and focus | Frequency in MHz, probe identity and target focus | Usable reference echo and repeatable response at the joint depth. |
| Scan sampling | X/Y step in µm and scan area in mm² | Smallest relevant reference feature can be recognized on repeat scans. |
| Time gate | Start and width in ns or µs, reference echo and scan side | The target echo is isolated across representative copper geometry. |
| Display and gain | Gain, polarity or amplitude map, and palette | Comparable scans use the same settings or document why they changed. |
Set the qualification condition on a representative stack or calibrated reference feature, not on the nominal pixel size. If a deeper echo disappears when switching to a higher-frequency probe, return to a configuration with useful penetration and check the smallest feature it can reliably detect.
Acoustic Microscopy vs X Ray Inspection PCB Methods: What Can Each See?
Use acoustic microscopy for suspected flat interface separation; use X-ray when the main question is density or geometry in the metal or solder. Conventional radiography records attenuation through the object, so a very thin flat gap can contribute little to the projection even if it reflects ultrasound strongly. The methods are complementary; a clean result from one does not invalidate a relevant finding from the other. A Nordson engineering comparison shows how the two modalities respond to different defects in power devices.
| Method | Choose it for | What can still be missed |
| Pulse-echo SAM | Mapping a thin separation at a known, accessible bond plane. | Regions below strong reflectors or layers whose echoes overlap. |
| 2D X-ray | Comparing projected solder voiding or metal geometry without wet coupling. | Thin interface gaps with little projected contrast; stacked structures can overlap. |
| Targeted microsection | Confirming the actual layer and local morphology at a mapped location. | Features outside the cut plane; the inspected sample is consumed. |
When a scan flags the DBC bond but X-ray looks normal, first ask whether radiography had a realistic chance of seeing the suspected thin gap. If not, repeat the acoustic scan with documented settings and consider a section through the registered coordinates. If instead the primary question is a void in the baseplate solder, start with the method demonstrated to show that joint on the actual assembly.
Scanning Acoustic Microscopy Delamination: Rule Out False Indications
A repeatable indication centered at the target echo is stronger evidence than a spot on an isolated image. Mark the suspect coordinates, overlay the artwork, and compare the same geometry on a reference coupon. An indication following every patterned copper edge may reflect the geometry or gate choice. One that moves or disappears after remounting can point to coupling or a trapped bubble. Neither observation by itself proves the part is good.
Capture A- and B-scan views both through the indication and through an adjacent intact region. If the part and fixture allow it, rescan from the other side using a gate established for that side; do not compare raw colors from different setups. If the finding remains important and ambiguous, section a representative sample at a registered feature.
Scanning Acoustic Microscopy Images: Read the Report, Not the Palette
The scan report should distinguish the size of the indication from the size of the area that was actually evaluated. Suppose a specified joint is 10 × 8 mm, a 2 × 8 mm strip is unreadable, and a 4 mm² indication lies in the readable area. That gives 64 mm² of readable bond and 16 mm² not evaluated. The observed indication covers 4 ÷ 64 = 6.25% of the readable area; 4 ÷ 80 = 5% of the drawing area is merely the known indicated fraction of the whole. Neither proves that the unreadable strip is free of defects.
This is an arithmetic example, not a measured defect or pass limit. Provide both areas, the exclusion mask and the location of the indication. A small feature at a defined edge or under a die may trigger review even when the aggregate area is low. Two reports using different denominators, thresholds or zone definitions are not directly comparable.
C-SAM Testing: What Makes a Result Ready for Acceptance?
Put the target interface, required coverage and disposition criteria in the inspection request before the parts are scanned. Specify the sample condition, scan side, reference setup, regions excluded by design, unacceptable locations and what to do when coverage is incomplete. Require sample ID, stack revision, scale, orientation, gate settings, reference/repeat records and annotated images. Agree on escalation for an ambiguous feature rather than asking the lab to improvise a pass percentage.
IPC/JEDEC J-STD-035A addresses acoustic microscopy for non-hermetic encapsulated electronic devices. It provides relevant method terminology but does not automatically establish acceptance limits for a bare ceramic substrate. Use the applicable product drawing and qualification plan for disposition. A passing scan is evidence about the tested interface under the documented conditions, not a measurement of bond strength or a prediction of field life.
Frequently Asked Questions
Can a normal-looking ceramic substrate have hidden delamination?
Yes. The copper can look intact from the surface while a portion of the buried bond is separated. Use visual inspection for chips and exposed damage, and a qualified acoustic scan for the buried plane. Neither observation alone proves that all interfaces are sound.
Does a red spot always mean delamination?
No. Color represents the instrument’s selected mapping and threshold. Ask which gate and display mode produced the image, then check the raw echo or a cross-sectional view along the suspect location. Compare it with an intact area of similar geometry before assigning a defect.
Can a powered assembly go in a water bath?
Do not assume it can. Acoustic labs normally inspect unpowered samples and agree the coupling method, immersion compatibility and post-test handling before inspection. For a finished module that cannot be immersed, ask whether a local water-coupling fixture or scanning an earlier assembly stage would answer the same question.
Will a scan distinguish a void from a crack?
Not from color alone. Shape, orientation, echo timing and whether the indication persists across planes can help narrow the possibilities. If the distinction changes the corrective action or acceptance decision, confirm the morphology at the registered location using a method suited to that structure.
Should samples be scanned before or after thermal cycling?
Use both when the goal is to detect growth or new indications. A baseline tracks what was already there. Register the before-and-after scans to the same part orientation, gate, region and scale, and document any settings change before interpreting an apparent difference.
What if the test lab cannot see the entire copper bond?
Ask for a map of the unreadable area and a proposed alternative: a different probe, an opposite-side scan, inspection before assembly, or a targeted destructive check. If a mandatory region remains unreadable, report that explicitly and use the agreed escalation path.
Do repeated indications on the same edge prove poor bonding?
No. A recurring location can be consistent with a real process problem, but also with a repeating change in stack geometry or acoustic focus. Register the scan to the copper pattern, compare with a reference part and verify the suspect joint on a representative sample before blaming the bonding process.
What should I give the lab with a failed field sample?
Provide the layer cross-section, sample revision and orientation, known operating or stress history, relevant failure locations and an intact comparison sample if available. Tell the lab whether it may permanently alter the part. Those details help distinguish evidence-preserving screening from destructive confirmation.
Prepare the Ceramic Substrate Review
Start with the stack drawing and mark the joint and any edge or die-footprint zones that require separate disposition. For fabrication context, see ceramic PCB options and the guide to DBC copper edge relief. To discuss a Best Technology ceramic PCB drawing, send the layer dimensions, material, copper artwork and requested inspection evidence to sales@bestpcbs.com. Confirm the test arrangement and acceptance criteria separately.
