April 17, 2025 • 31 mins
Acoustic inspection stands as a silent sentinel in semiconductor manufacturing, detecting microscopic defects that could lead to catastrophic failures in high-value applications. Bryan Schackmuth, Senior Product Line Manager at Nordson Test and Inspection, reveals how this technology has evolved from laboratory tools to production-line essentials.
When ultrasound encounters even the tiniest air gap—we're talking hundreds of angstroms—it reflects completely, making acoustic imaging uniquely powerful for evaluating bonds between materials. While optical inspection shows surface defects and X-ray reveals density variations, acoustic inspection peers between layers, identifying delamination and other hidden flaws that might otherwise escape detection until field failure.
The challenges of advanced packaging have driven significant innovation in acoustic inspection technology. As manufacturers stack more die, create complex interconnects, and push toward heterogeneous integration, the value of each wafer increases dramatically. Nordson's SpinSam system represents a breakthrough in this space, replacing traditional raster scanning with a rotational approach that achieves 41 wafers per hour—eight times faster than previous generation technology—while maintaining resolution down to 10 microns.
Beyond pure speed, the system's spinning scan technology offers unique advantages for edge inspection where defects are more common due to coefficient of thermal expansion effects. The modular design allows maintenance on individual scanners while others continue operating, maximizing uptime in production environments. Most exciting is the integration of AI and machine learning for defect detection, moving beyond simple thresholds to analyze complex multilayer images simultaneously.
Want to see how your inspection strategies might benefit from these advances? Check out Nordson's SpinSam technology at nordson.com and discover how acoustic inspection is helping manufacturers achieve higher yields and more reliable products in today's most demanding semiconductor applications.
Discover how technology is reshaping our lives and livelihoods.
Listen on: Apple Podcasts Spotify
Nordson Test and InspectionDelivering best-in-class test, inspection, and metrology solutions for semiconductor applications.
Disclaimer: This post contains affiliate links. If you make a purchase, I may receive a commission at no extra cost to you.
Become a sustaining member!
Like what you hear? Follow us on LinkedIn and Twitter
Interested in reaching a qualified audience of microelectronics industry decision-makers? Invest in host-read advertisements, and promote your company in upcoming episodes. Contact Françoise von Trapp to learn more.
Interested in becoming a sponsor of the 3D InCites Podcast? Check out our 2024 Media Kit. Learn more about the 3D InCites Community and how you can become more involved.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Françoise von Trapp (00:01):
This episode of the 3D Insights
Podcast is sponsored by NortonTest and Inspection leaders in
acoustic, optical and x-rayinspection and metrology systems
for the semiconductor and SMTmanufacturing markets.
Norton's next-generationSpinSam acoustic microimaging
system combines breakthroughscanning capability with
best-in-class defect capture andimage quality to enhance both
(00:24):
productivity and accuracy for100% semiconductor inspection.
The system sets an industrybenchmark for high throughput
and superior sensitivity,enabling precise defect
detection in wafer-basedassemblies.
Learn more at Nordsoncom.
Hi there, I'm Francoise vonTrapp, and this is the 3D
(00:55):
Insights Podcast.
Hi everyone, this week we aretalking about how acoustic
inspection is becoming a keytool in the semiconductor wafer
and package inspection toolbox.
Now, as semiconductor devicesbecome more complex, their value
is increasing at each step ofthe manufacturing process, and
(01:17):
that's why 100% inspection isbecoming critical at all stages.
There are options like optical,x-ray and acoustic imaging, and
these are all important methodsthat help ensure package
reliability.
So to learn more about this,I've invited Brian Schachmuth of
Nords Intestine Inspection tojoin me on the podcast.
Welcome, Brian.
Bryan Schackmuth (01:39):
Hey, thank you for having me Good to be here.
Françoise von Trapp (01:42):
So thanks for joining us from Korea.
I know it's early in themorning there.
Really appreciate your time.
So before we get started intoour topic, can you just tell me
a little bit something aboutyourself and your role at
Nordsen Test and Inspection?
Bryan Schackmuth (01:56):
Sure.
So currently I'm the SeniorProduct Line Manager for
Acoustic Products at NordsenTest and Inspection acoustic
products.
At Norton Test and Inspectionwe're part of the bigger segment
called ATS, which has threebusinesses the EPS, which is
(02:17):
like a Syntec dispensing, xrt,which is X-ray and test, and
then the division I'm in is OSMOptical Sensor Metrology.
Our main products are acousticand optical inspection as well
as wafer sense.
So I've been with Sonoscan,which was acquired by Norton,
for 28 years now, so justrecently over the last year,
taking the product line managerrole.
Françoise von Trapp (02:36):
Okay, so you were at Sonoscan like 10, 15
years ago when acousticmicroimaging really started
hitting the market.
Bryan Schackmuth (02:44):
Yeah, so I mean basically back then it was
mostly offline, manual typesystems, and then, I would say
probably 10 years ago, we reallystarted to get into more of the
production because of that needfor inspection that you were
talking about.
Françoise von Trapp (02:59):
Right, okay , so there is a history of using
all of these types ofinspection in R&D and
development, right?
And that's what you weretalking about with the manual
inspection.
Bryan Schackmuth (03:10):
Yeah, so you typically find a lab-type system
.
It's manual loading, manualunloading in the R&D lab, in
failure analysis and then alsoin new product development While
they're designing it, once it'sin the field and then any type
of field returns.
We would do the acousticinspection first because it's
non-destructive, so they don'thave to damage the sample and
(03:31):
they can look inside to see anytype of defects first before
they would go and cut the sample.
Françoise von Trapp (03:36):
So can you actually just give us a
high-level explanation of thosethree different types of
inspection?
Bryan Schackmuth (03:44):
Sure explanation of those three
different types of inspection.
Sure, so acoustic again, usingultrasound.
Typically we're anywhere frommaybe 15 megahertz up to about
300 megahertz At that highfrequency.
The key thing about ultrasoundis it won't travel through air.
So once we hit an air gap weget a hundred percent reflection
, basically a big signal back.
(04:04):
Even down to hundreds ofangstroms the thickness of an
air gap will be enough to stopthe ultrasound.
That makes it a very goodtechnique for bond evaluation.
So any type of materials thatare supposed to be stuck
together, if they're separatedwe can see it.
X-ray is very good at seeingvariation in density.
So like a very thin crack X-raywould not be able to see, but a
(04:29):
solder ball or a wire bond or avia they can see very easily
because of the variation in thedensity.
Acoustically we would haveproblems with that small of a
feature.
And then optical is really forsurface level defects what you
can visually see it doesn't lookinside the sample like x-ray or
(04:49):
acoustic.
Françoise von Trapp (04:49):
So all of these three are important to the
manufacturing process.
Why is that?
Bryan Schackmuth (04:55):
We don't compete with these different
things.
Most customers in an FA labwould have x-ray, acoustic and
optical, or have all three,depending on the application.
You might use one more than theother.
X-ray has some limitations asfar as, like memory devices,
x-rays will damage the memorychips.
You can still do analysis onthose as well.
Françoise von Trapp (05:15):
I always think about acoustic and
sonogram as being used inmedical situations.
Right For all sorts ofmeasuring density.
So there's usually some sort offluid involved.
How does that work in themicroelectronic space?
Bryan Schackmuth (05:29):
Sure.
So, like I mentioned,ultrasound won't travel through
air, right.
So we need something to couplethe ultrasound to the sample.
In medical they use the gel andit's really more of a surface
contact type transducer For theacoustic.
We can't touch the sample, sowe're using deionized water.
It's readily available.
(05:49):
It's very safe.
You can either put the samplein a water bath like a water
tank, or one of the things thatNorton has for our systems is a
waterfall.
So instead of immersing thesample in water, we kind of
cascade or jet water just wherewe're inspecting, so that
(06:09):
minimizes the exposure to water.
So if you have some type ofdefects that are exposed to the
outside edge, water couldpotentially get inside the
sample.
So waterfall helps to minimizethat chance.
Françoise von Trapp (06:24):
Okay, so you know what are the trends
that you're seeing in thewafer-level packaging that is
impacting the needs of thesedifferent types of inspection.
Bryan Schackmuth (06:33):
We've seen significant growth in that
advanced packaging.
So whether it's system-inpackage, fan-out, wafer-level
package or heterogeneouspackaging, the market is really
driving to put as much aspossible into the smallest form
factor as possible.
It's getting much more complex.
With that complexity comesalong the chance for defects to
(06:54):
be introduced at all thedifferent stages, and so if we
can get in on those earlierstages, we can help to screen
out the defects, which improvesthe customer's yield and they
don't have to further process aknown defect earlier in the
process.
Right, so we're weeding out thepotential failures Exactly.
Françoise von Trapp (07:14):
Are we doing rework on those, or is
that pretty much just a sortingprocess?
Bryan Schackmuth (07:20):
Yeah, typically when we're looking at
a wafer, we'll map the wafer,give the location of the defects
so that further down theprocess we'll communicate with
their factory host.
We'll tell them OK, this chip A, b and C are defects.
Those chips will not get anyfurther processing downstream.
Françoise von Trapp (07:37):
So you don't lose the whole wafer.
You're able to select, andthat's what we're talking about.
Known good die yeah, correct,ok.
Able to select, and that's whatwe when we're talking about.
Known good dye yeah, correct,okay.
Beyond, like r&d.
Now we're starting to see theuse of these tools.
Can you give me some example ofhow these are used actually in
situ in the manufacturingenvironment?
Bryan Schackmuth (07:57):
sure.
So you know, for the waferinspection we're looking at
various applications, butessentially it's a bonded wafer.
You know.
So historically, many years agoa wafer was essentially two
dimensions.
You had the XY and then you hadone layer of metallization for
the chip.
You know, with this advancedpackaging they're now going in
(08:19):
the third dimension.
You know they're stackingdevices.
So each layer when you stackyou have a bond interface and so
that's what we're stackingdevices.
So each layer when you stackyou have a bond interface, and
so that's what we're looking at.
So it could be as simple assilicon on insulator or SOI.
That's just two bonded wafers.
We'll look at that bondinterface.
It could be the more complexMEMS devices where you have
(08:40):
multiple layers and then you getinto the stack die and TSV.
They're doing eight stack, 16stack.
So we'll be looking at thosedifferent applications,
determining if there's defectsin there and then mapping that
out and providing that to thecustomer and you're using the
acoustic imaging for that.
Françoise von Trapp (09:00):
Correct, yeah, okay, so what would be an
example of when you would useoptical or x-ray in the
manufacturing environment?
Bryan Schackmuth (09:25):
M8000, which is for wafer application X-ray.
X-ray is much higher resolution.
They can get down sub-micronAcoustic.
We're probably down to aroundthe 10-micron resolution.
So with X-ray at that very highresolution they'll go in and
look at the TSV because they'rewell-suited to look for that
variation in the density.
So they'll look for muchsmaller defects.
But typically they're doingkind of a spot check.
(09:46):
They wouldn't be doing a fullscan of the whole wafer because
of that very high resolution.
Françoise von Trapp (09:52):
So we're using all three of these in
manufacturing now.
Bryan Schackmuth (09:56):
Correct.
Françoise von Trapp (09:57):
So what determines when one would be
chosen over another, or is itnot like that?
Bryan Schackmuth (10:02):
A little bit like that.
Like I mentioned, if it's amemory application in production
, you can't use x-ray.
Françoise von Trapp (10:08):
Right, okay .
Bryan Schackmuth (10:09):
There are some things you can do to mitigate
the effects of the x-ray onmemory.
But typically they wouldn't beusing X-ray for memory but for
acoustics.
It's anytime they want to seethat level-to-level inspection.
So the benefit of acoustic iswe can isolate layer by layer so
we can look at the top layer,the middle layer, the bottom
layer and provide different setsof images for each of those
(10:33):
layers.
That helps them to kind ofinvestigate where the defect is.
Is it at their higher levels?
You know, maybe they have aprocess where they need to go
back and refine the process tohelp that top layer adhesion.
Françoise von Trapp (10:47):
So there are different types of defects,
and I know one of them is latentdefects, which are not as
obvious, something where adevice is going to fail down the
road in use.
Bryan Schackmuth (10:57):
Typically we might see a defect and it might
be a small, let's saydelamination.
It was two layers that aresupposed to be bonded together
and if you have a small defectas this chip gets processed, you
know there's the chemical,mechanical polishing.
Those defects can sometimescause almost like a blistering
(11:19):
effect or a slight bubble, andas you grind it that defect can
kind of expand it.
Can you know, as you polish thesurface that'll expand that
defect.
So there's certain cases wherewe would see defects that you
know would grow over time.
If it gets into a package andit goes into automotive you're
talking about extremetemperature variation and so
(11:42):
that expansion and contractioncan make that defect grow.
Françoise von Trapp (11:47):
So, basically, the acoustic
micro-imaging is making itpossible for you to find these
defects way sooner and pull themout of the line, so that you
don't even get to the point ofwhere they're going to get
bigger.
Bryan Schackmuth (12:00):
Exactly.
You know it depends on themarket.
If the chip is going into a toy, you know that's not high
reliability.
If it's going into a medical orsurgery, or if it's going in on
a satellite in space, it's veryexpensive to repair.
So the 100% inspection isgeared more towards that
high-end, high-expensive devices.
Françoise von Trapp (12:21):
So if you're weaving out the smaller
defects for the 100% inspection,can those get shipped off to A
use this?
This one's not a complete loss.
You can use this in a toy, Imean, do they do?
Bryan Schackmuth (12:33):
that they wouldn't even start with that
application from the start.
Françoise von Trapp (12:38):
Wouldn't that be great though.
So how has traditional acoustictechnologies, then, been
improved over the years?
Bryan Schackmuth (12:47):
Yeah, so traditionally the type of
scanning that's been done foracoustic microscopes and this is
, you know, since I've been withthe company 20 years is an XYZ
raster scan.
Imagine kind of an inkjetprinter scanning back and forth
over the paper, printing.
We're doing the same thing, butan XY over a sample and
generating an image.
(13:07):
It's that back and forth motion.
We use a transducer is what'scalled.
That generates the ultrasound,moves to the left, it stops, it
accelerates to the right, itstops, it accelerates to the
right, it stops and it goes backand forth.
So at the edges of the scanyou're constantly stopping, so
you're not acquiring data.
The spin Sam, what that does isa rotational or spinning scan.
(13:29):
So we start in the middle, wespin the wafer and we just move
to the edge.
I would say like a recordplayer but that would be dating
myself or like a CD player.
Françoise von Trapp (13:39):
You can say record player.
Well, I was just thinking aboutthe difference between scans.
When my kids I had twins andwhen I was pregnant, you
couldn't tell what baby was new,and now you see these amazing
images where they can really seewhat the baby looks like.
So you know, take that and putthat into the wafer application.
(14:01):
I would imagine like theresolution is a lot better.
Bryan Schackmuth (14:04):
Yeah.
So we've improved kind of thefrequency you know, going to
higher frequency.
Generally speaking you wouldsay lower frequency is lower
resolution but it can go throughmore material.
Higher frequency is higherresolution but it can go through
more material.
Higher frequency is higherresolution but it can't go
through as much material.
So we've improved thetransducer design, optimized the
frequency and the lens shape tooptimize the focal spot so we
(14:29):
can get that much higherresolution, and then, along the
way, you know, makingimprovements to the RF chain
that generates the ultrasound,maximizing, you know, the output
of the signal, as well asminimizing, you know, signal to
noise ratio.
Françoise von Trapp (14:47):
So you mentioned SpinSam, which is a
fairly new product for NORD'sintestine inspection.
Bryan Schackmuth (14:53):
Yeah, it was just released last year towards
the end of FY24.
That's our 100% 300-millimeterwafer inspection tool, really
geared for high throughput inthe wafer inspection area.
Françoise von Trapp (15:08):
For the purposes of our audience?
How can it be used to addressthe needs of wafer-level
packaging and advanced packaging?
Bryan Schackmuth (15:16):
Due to the complexity of these samples and
the wafers themselves, the costis extremely high, so they
really need high reliability.
You can put the SpinSam systemin your facility and do 100%
inspection on these samples.
A lot of the stuff we'relooking at today is stack dye,
whether 8-stack or 16-stackthese samples A lot of the stuff
(15:38):
we're looking at today isstacked die, whether you know
eight stack or 16 stack.
Françoise von Trapp (15:44):
And again, it's all about finding those
defects kind of early on in theprocess.
Bryan Schackmuth (15:46):
So that's the memory application.
Yeah, correct, Then juststandard bonded wafers.
One thing we've seen is atemporary bond.
So they'll use kind of acarrier wafer and they'll
temporarily bond a processedwafer on there.
That carrier wafer willbasically act like a substrate
for the actual wafer.
So we'll inspect that temporarybond because if the temporary
(16:08):
bond is not good you couldpotentially induce defects into
the sample later.
Françoise von Trapp (16:13):
So what about like fan out wafer level
packaging?
Is that an application spacewhere the spin sam would be
helpful for like doinginspection on RDL layers?
Bryan Schackmuth (16:23):
In the wafer area they build up the fan out
wafer level package so we canlook at those different layers,
the dielectric layers, the low Kdielectric or the
redistribution layer, and theneventually that wafer would get
cut and diced and then thatwould be put onto a substrate
that goes into kind of the finalpackage.
(16:44):
At that stage we would go fromour wafer inspection to more of
our tray inspection system, moreof a backend process to look at
that final assembly.
Françoise von Trapp (16:54):
Okay, so the spin sam is definitely for
the wafer itself.
Bryan Schackmuth (16:58):
We wouldn't say it's very front end, we
would say more like middle end.
So front end is more of likeyour UV processing, metal
deposition and that we're kindof after that process where they
start stacking those wafers.
Françoise von Trapp (17:13):
Okay, so it's the finding.
The known good die before iswhere it's used.
It's not for final packageinspection.
Bryan Schackmuth (17:23):
Correct.
Yeah, so definitely before.
If there's a thousand die onthe wafer, using the acoustic
microscope they can pick outthose defect parts and then, you
know, put them into a rejectbin.
There's no use in packagingthat into the final assembly,
saving your time.
Françoise von Trapp (17:40):
Okay, that's good to know, cause when
you were talking before aboutthe spin process, I was thinking
about how things are shiftingto panel level packaging and how
would that be adapted for usein panel level packaging.
But you're actually using itbefore we even get to that die
placement on the panel.
Bryan Schackmuth (18:00):
Correct.
Yeah, we would have a largearea scanner because you can use
the acoustic microscope throughthe different stages.
At the wafer level it getsdiced, gets put into a package,
but then we would go to adifferent form factor for the
acoustic inspection.
You know a large area scanner.
Typically we would be doinglower resolution at that stage,
lower frequencies, maybe 30megahertz, 50 megahertz, 100
(18:24):
megahertz, because the defectsize is not as critical in the,
you know, back end.
Françoise von Trapp (18:31):
Okay, so that would be a different tool
than the SpenSam.
Bryan Schackmuth (18:34):
Correct.
Françoise von Trapp (18:35):
Okay, so let's talk about the SpenSam a
little more.
Can you give me some of thebenefits?
Things like UPH, speedresolution, edge scanning
capabilities, things like that.
Bryan Schackmuth (18:46):
Sure.
So the driving reason that wewent to develop the SpenSam is
so we had a previous product, awafer inspection system AW300,
using that traditional rasterscan.
But really the push was tooptimize throughput.
But we kind of pushed it evenfarther.
What we want to look at iswafers per hour, per footprint.
(19:08):
So clean room space is veryexpensive, so we want to
maximize the throughput in thesmallest footprint.
So we want to maximize thethroughput in the smallest
footprint.
So with that spin scan we canget at 100 micron resolution
about 41 wafers per hour, whichis, I think, eight times faster
than our previous machine.
And then the resolution.
(19:34):
So currently the spin scan cango down to about 10 micron
resolution.
So you can do that full 300millimeter wafer at a 10 micron
pixel.
Again, the way we're scanningit, because of that spinning
scan we can do, you know, anedge scan, or you can think of
it kind of like a donut.
Or if I go back to the recordplayer playing the last track,
on the record.
You could never do that type ofscan with a traditional raster
scan, like you could never dothat type of scan with a
(19:54):
traditional raster scan.
You know we're just doing aring around the outside of the
wafer and the reason that'simportant is the coefficient of
thermal expansion when you'redoing the bonding process is
more extreme at the edges Right,and so customers are seeing
that the edge of the wafer iswhere they're finding more
defects.
In the center of the wafer wemight not see as many defects
(20:17):
because that's easier to bond.
And so with the SpinSam it'svery well suited to do a lower
resolution scan in the center,where you don't expect defects,
and then maximize the resolutiontowards the edge to see those
edge defects.
And that all goes to improvingyield.
If they can isolate thosedefects at the edges, remove
(20:39):
those dye from the process.
Françoise von Trapp (20:41):
Okay.
So while they're doing that inthe volume setting and basically
identifying the defects andremoving them, are they also
getting any knowledge that theycan feed back to the process
guys to say, hey, we keep seeingthis happen over and over.
Can you tweak the process?
Bryan Schackmuth (21:02):
Definitely so.
The acoustic wave again, like Imentioned, you can isolate
layers so as the ultrasoundtravels it travels in time and
then we can basically pickregions of time to look at.
If they see a trend thatdefects are occurring at the
base layer or the deepest layer,a trend that defects are
occurring at the base layer orthe deepest layer, you know they
might go back into theirprocess and say what is it about
the process that's causingthese defects?
(21:22):
Just at this layer?
We're not seeing it at theupper layers, it's always at
this lower layer.
And so they might find that,you know, they have some
unintended particle generationthat's putting particles at the
edge of the wafer during thatprocess and those particles then
cause the defects.
So they can go back and kind ofmaximize their process earlier
(21:44):
on to minimize defects down theroad.
Françoise von Trapp (21:46):
Yeah.
So that's great because they'renot just using it to remove the
defects and that immediateyield, they're actually taking
that information for improvingyields in the future.
Bryan Schackmuth (22:00):
Yeah, and improving their process
altogether.
Françoise von Trapp (22:01):
Right, okay , so I was reading through some
of the information about theSpinSam and there's something
called global tool matching.
What is that and why is itimportant?
Bryan Schackmuth (22:13):
So the SpinSam has four scanners, so there's
four scan stages.
There's a eFEM that loads allthe wafers, so we're using four
different transducers.
There can be slight variationsin the transducers, so what
global tool matching does isit's basically a matching
network.
So it'll look at the signal ofall four transducers and to do
(22:35):
this we have kind of a referencewafer that we'll use.
So it's a you could call itkind of a calibration wafer
that'll go on each of the fourscanners.
We take that referencemeasurement and then we optimize
the signal so that everythingis matched.
So kind of the end goal is, ifyou have a spin SAM in Singapore
running a wafer and that samespin SAM in Taiwan running a
(22:57):
same wafer, what we want toachieve is the same recipe, same
system, same wafer, same result, so that multinational
companies can use the same dataset or the same parameters and
not have to change or modifydepending on the system.
Françoise von Trapp (23:17):
Okay, and what about maintenance?
Bryan Schackmuth (23:20):
Yeah, so another key thing that we were
trying to design into theSpinSAM system is its modular
design.
So each of those four scannersI mentioned is a scan module,
and then there's also another RFmodule which generates the
ultrasound.
And so with that modular designyou can take one scanner
offline to do preventativemaintenance or servicing, but
(23:43):
the other three scanners willcontinue to run Traditionally.
If you had to do preventativemaintenance, the whole machine
goes down and you lose 100% ofyour throughput.
With modular design on SpinSan,we can take each scanner offline
, one at a time, but the otherthree scanners can continue to
scan, and so it makes it mucheasier to service, as well as
(24:06):
minimizing the downtime on thesystem.
Françoise von Trapp (24:09):
Are there any other advantages to being
modular design besides that?
Bryan Schackmuth (24:14):
In the future, if we were to develop, you know
, higher frequency, let's say,or develop a new RF feature that
gives a better image, we canjust remove that RF module in
the spin sand and upgrade it tothe latest RF module, and so you
don't have to replace the wholesystem to get the latest image
capability.
(24:34):
We're looking at ease ofrepairability as well as ease of
upgradability in the future.
Françoise von Trapp (24:40):
One of the things we've been talking a lot
about with different membercompanies is integrating machine
learning, ai and digital twintechnology into their tools.
Is there any of that involvedwith the spin Sam?
Bryan Schackmuth (24:57):
Yeah, that's a great question.
So traditionally we've had asoftware package called DIA,
digital image analysis.
It's using a threshold.
So you have a, imagine, agrayscale image.
You set a threshold, sotypically the defects are bright
, or, you know, we get a largereflection so we have a bright
defect.
So we would say any pixelthat's brighter than a certain
(25:18):
value is a defect.
That was good when the deviceswere simple.
But with today's complexlayering, the contrast in the
images is no longer easy to usea simple threshold technology.
So there's a team at Norton, youknow they work with that whole
(25:40):
segment, all three divisions,and they're really developing
the AI ML for us and so for theacoustic system we've already
introduced kind of the firstrendition of the what's called
MGIA or multi-gate imageanalysis, and so instead of
setting that threshold, there'sapplications today where if you
set a threshold you'll geteither overkill or underkill,
(26:01):
meaning you're going to rejecttoo much or you're not going to
catch the defect.
But with the AIML we can gothrough and teach the system
what a defect is and then youknow with the machine learning
over time you can build up arobust model that be able to
isolate the defects but notoverkill or underkill, and so
(26:28):
that's been introduced.
It can do.
The MGI is multi gate, meaningmultiple images, so you can take
, you know, three images, stackthem all together and then
analyze that as a whole package.
So it's really been useful for,you know, for the more complex
structures that we're seeingtoday.
Françoise von Trapp (26:40):
Are you getting more of a 3D image with
that, then?
Bryan Schackmuth (26:43):
We can do 3D Typically.
That takes a little bit longerso it's not really done in
production.
But what they'll do is if anoperator looks at the three
images, they're looking kind ofone by one.
If an operator looks at thethree images, they're looking
kind of one by one, and so theymight see a defect in the first
one, nothing in the second one,maybe something in the third
image.
What the AIML can do is look atall of them at the same time,
(27:10):
and so it might see a defect atthe first layer and the second
layer, but then, looking at bothof those at the same time, it
can interpolate.
Maybe that defect actually goesthrough the middle layer, and
so it can really see more thanan operator would be able to see
.
Françoise von Trapp (27:22):
So what does this do as an advantage
over a tool that wouldn't havethat?
Bryan Schackmuth (27:27):
We've seen applications, you know, in the
past where an operator is doingthat analysis, because the
contrast is not good enough todo the traditional threshold.
It's an operator making thatdecision, and so you introduce a
certain element of human error.
Right, okay, no if it's closeto lunchtime, maybe they're
thinking about lunch and notthinking about the defects in
(27:49):
the wafer.
Françoise von Trapp (27:50):
That would be me.
So does it speed up the process, makes it more accurate?
Yeah, definitely so, speedingup the process makes it more
accurate?
Bryan Schackmuth (27:56):
Yeah, definitely so.
Speeding up the process,automating it you know a lot of
our customers are drivingtowards that lights out factory,
where they don't want operatorsmaking any decisions and then
also for the traceability.
So you know, all the wafers aretraced, we're communicating
with the host and themanufacturing site and so all
that data goes along with thatwafer.
Françoise von Trapp (28:17):
Okay, so is the SpinSam a one-of-a-kind in
the industry, or are therecompetitive offerings?
Bryan Schackmuth (28:24):
I would say it's one-of-a-kind because of
that spinning technology.
So any of the competitiveofferings and even our older
generation system are using thatraster scan.
So there's really no way theycan catch up to the throughput
and footprint.
When you're doing that rasterscanning you have a mechanism, a
gantry over the wafer and soyou have a lot of moving parts
(28:46):
on top of the wafer, which isnot ideal for particle
generation because you'll havebearings or belts or motors over
the surface of the wafer.
Particles would tend to drop onthe wafer.
With SpinSam we just have onetransducer that just moves from
the center to the outer edge, sowe don't even have any moving
(29:06):
parts essentially over the topof the wafer.
So it really minimizes particlegeneration for the clean room
environment.
Françoise von Trapp (29:14):
So because you don't want to introduce
something that would cause adefect further down the line
during the inspection process.
That would be counterproductive, correct, okay?
Last question for you what canthe customers gain by installing
the SpinSam in their facility?
Bryan Schackmuth (29:33):
Yeah.
So I would say Spin SAM hasbest-in-class wafer per hour at
100 microns, 41 wafers per hour,and then in our roadmap we're
going to be targeting about twotimes that throughput and
that'll be coming kind oftowards the end of this year.
The next thing would be thedefect detection, so we can see
down to that 10 micron defectsize and we can do it in a very
(29:55):
small form factor.
So it's really about wafers perhour per footprint, so
maximizing the cost of ownershipof the system, and then the
quality standards, so theability to do that global tool
matching, to have each systemmatch, no matter where it is,
with any facility across theglobe.
Françoise von Trapp (30:17):
Okay, so where can people go to learn
more?
Bryan Schackmuth (30:19):
Yeah, so you can go to nortoncom and look up
SpinSam.
You'll see probably my face.
Françoise von Trapp (30:27):
We'll put a link in the show notes so
people can find it.
Thanks so much.
It's been a while since I'vetalked about acoustic
microimaging, so it was kind offun to catch up on that.
So I really appreciate yourtime.
Bryan Schackmuth (30:38):
Thanks for having me.
I really appreciate the timetogether.
Françoise von Trapp (30:44):
If you want to learn more about
semiconductor inspection, besure to mark your calendars for
an interview with Norton Testand Inspections' Andrew Mathers
to learn all about dynamicplanar CT with automated x-ray
systems dropping May 16th.
There's lots more to come, sotune in next time to the 3D
Insights podcast.
The 3D Insights podcast is aproduction of 3D Insights LLC.
This episode of the 3D Insights
Podcast is sponsored by NortonTest and Inspection leaders in
acoustic, optical and x-rayinspection and metrology systems
for the semiconductor and SMTmanufacturing markets.
Norton's next-generationSpinSam acoustic microimaging
system combines breakthroughscanning capability with
best-in-class defect capture andimage quality to enhance both
(00:24):
productivity and accuracy for100% semiconductor inspection.
The system sets an industrybenchmark for high throughput
and superior sensitivity,enabling precise defect
detection in wafer-basedassemblies.
Learn more at Nordsoncom.
Hi there, I'm Francoise vonTrapp, and this is the 3D
(00:55):
Insights Podcast.
Hi everyone, this week we aretalking about how acoustic
inspection is becoming a keytool in the semiconductor wafer
and package inspection toolbox.
Now, as semiconductor devicesbecome more complex, their value
is increasing at each step ofthe manufacturing process, and
(01:17):
that's why 100% inspection isbecoming critical at all stages.
There are options like optical,x-ray and acoustic imaging, and
these are all important methodsthat help ensure package
reliability.
So to learn more about this,I've invited Brian Schachmuth of
Nords Intestine Inspection tojoin me on the podcast.
Welcome, Brian.
Bryan Schackmuth (01:39):
Hey, thank you for having me Good to be here.
Françoise von Trapp (01:42):
So thanks for joining us from Korea.
I know it's early in themorning there.
Really appreciate your time.
So before we get started intoour topic, can you just tell me
a little bit something aboutyourself and your role at
Nordsen Test and Inspection?
Bryan Schackmuth (01:56):
Sure.
So currently I'm the SeniorProduct Line Manager for
Acoustic Products at NordsenTest and Inspection acoustic
products.
At Norton Test and Inspectionwe're part of the bigger segment
called ATS, which has threebusinesses the EPS, which is
(02:17):
like a Syntec dispensing, xrt,which is X-ray and test, and
then the division I'm in is OSMOptical Sensor Metrology.
Our main products are acousticand optical inspection as well
as wafer sense.
So I've been with Sonoscan,which was acquired by Norton,
for 28 years now, so justrecently over the last year,
taking the product line managerrole.
Françoise von Trapp (02:36):
Okay, so you were at Sonoscan like 10, 15
years ago when acousticmicroimaging really started
hitting the market.
Bryan Schackmuth (02:44):
Yeah, so I mean basically back then it was
mostly offline, manual typesystems, and then, I would say
probably 10 years ago, we reallystarted to get into more of the
production because of that needfor inspection that you were
talking about.
Françoise von Trapp (02:59):
Right, okay , so there is a history of using
all of these types ofinspection in R&D and
development, right?
And that's what you weretalking about with the manual
inspection.
Bryan Schackmuth (03:10):
Yeah, so you typically find a lab-type system
.
It's manual loading, manualunloading in the R&D lab, in
failure analysis and then alsoin new product development While
they're designing it, once it'sin the field and then any type
of field returns.
We would do the acousticinspection first because it's
non-destructive, so they don'thave to damage the sample and
(03:31):
they can look inside to see anytype of defects first before
they would go and cut the sample.
Françoise von Trapp (03:36):
So can you actually just give us a
high-level explanation of thosethree different types of
inspection?
Bryan Schackmuth (03:44):
Sure explanation of those three
different types of inspection.
Sure, so acoustic again, usingultrasound.
Typically we're anywhere frommaybe 15 megahertz up to about
300 megahertz At that highfrequency.
The key thing about ultrasoundis it won't travel through air.
So once we hit an air gap weget a hundred percent reflection
, basically a big signal back.
(04:04):
Even down to hundreds ofangstroms the thickness of an
air gap will be enough to stopthe ultrasound.
That makes it a very goodtechnique for bond evaluation.
So any type of materials thatare supposed to be stuck
together, if they're separatedwe can see it.
X-ray is very good at seeingvariation in density.
So like a very thin crack X-raywould not be able to see, but a
(04:29):
solder ball or a wire bond or avia they can see very easily
because of the variation in thedensity.
Acoustically we would haveproblems with that small of a
feature.
And then optical is really forsurface level defects what you
can visually see it doesn't lookinside the sample like x-ray or
(04:49):
acoustic.
Françoise von Trapp (04:49):
So all of these three are important to the
manufacturing process.
Why is that?
Bryan Schackmuth (04:55):
We don't compete with these different
things.
Most customers in an FA labwould have x-ray, acoustic and
optical, or have all three,depending on the application.
You might use one more than theother.
X-ray has some limitations asfar as, like memory devices,
x-rays will damage the memorychips.
You can still do analysis onthose as well.
Françoise von Trapp (05:15):
I always think about acoustic and
sonogram as being used inmedical situations.
Right For all sorts ofmeasuring density.
So there's usually some sort offluid involved.
How does that work in themicroelectronic space?
Bryan Schackmuth (05:29):
Sure.
So, like I mentioned,ultrasound won't travel through
air, right.
So we need something to couplethe ultrasound to the sample.
In medical they use the gel andit's really more of a surface
contact type transducer For theacoustic.
We can't touch the sample, sowe're using deionized water.
It's readily available.
(05:49):
It's very safe.
You can either put the samplein a water bath like a water
tank, or one of the things thatNorton has for our systems is a
waterfall.
So instead of immersing thesample in water, we kind of
cascade or jet water just wherewe're inspecting, so that
(06:09):
minimizes the exposure to water.
So if you have some type ofdefects that are exposed to the
outside edge, water couldpotentially get inside the
sample.
So waterfall helps to minimizethat chance.
Françoise von Trapp (06:24):
Okay, so you know what are the trends
that you're seeing in thewafer-level packaging that is
impacting the needs of thesedifferent types of inspection.
Bryan Schackmuth (06:33):
We've seen significant growth in that
advanced packaging.
So whether it's system-inpackage, fan-out, wafer-level
package or heterogeneouspackaging, the market is really
driving to put as much aspossible into the smallest form
factor as possible.
It's getting much more complex.
With that complexity comesalong the chance for defects to
(06:54):
be introduced at all thedifferent stages, and so if we
can get in on those earlierstages, we can help to screen
out the defects, which improvesthe customer's yield and they
don't have to further process aknown defect earlier in the
process.
Right, so we're weeding out thepotential failures Exactly.
Françoise von Trapp (07:14):
Are we doing rework on those, or is
that pretty much just a sortingprocess?
Bryan Schackmuth (07:20):
Yeah, typically when we're looking at
a wafer, we'll map the wafer,give the location of the defects
so that further down theprocess we'll communicate with
their factory host.
We'll tell them OK, this chip A, b and C are defects.
Those chips will not get anyfurther processing downstream.
Françoise von Trapp (07:37):
So you don't lose the whole wafer.
You're able to select, andthat's what we're talking about.
Known good die yeah, correct,ok.
Able to select, and that's whatwe when we're talking about.
Known good dye yeah, correct,okay.
Beyond, like r&d.
Now we're starting to see theuse of these tools.
Can you give me some example ofhow these are used actually in
situ in the manufacturingenvironment?
Bryan Schackmuth (07:57):
sure.
So you know, for the waferinspection we're looking at
various applications, butessentially it's a bonded wafer.
You know.
So historically, many years agoa wafer was essentially two
dimensions.
You had the XY and then you hadone layer of metallization for
the chip.
You know, with this advancedpackaging they're now going in
(08:19):
the third dimension.
You know they're stackingdevices.
So each layer when you stackyou have a bond interface and so
that's what we're stackingdevices.
So each layer when you stackyou have a bond interface, and
so that's what we're looking at.
So it could be as simple assilicon on insulator or SOI.
That's just two bonded wafers.
We'll look at that bondinterface.
It could be the more complexMEMS devices where you have
(08:40):
multiple layers and then you getinto the stack die and TSV.
They're doing eight stack, 16stack.
So we'll be looking at thosedifferent applications,
determining if there's defectsin there and then mapping that
out and providing that to thecustomer and you're using the
acoustic imaging for that.
Françoise von Trapp (09:00):
Correct, yeah, okay, so what would be an
example of when you would useoptical or x-ray in the
manufacturing environment?
Bryan Schackmuth (09:25):
M8000, which is for wafer application X-ray.
X-ray is much higher resolution.
They can get down sub-micronAcoustic.
We're probably down to aroundthe 10-micron resolution.
So with X-ray at that very highresolution they'll go in and
look at the TSV because they'rewell-suited to look for that
variation in the density.
So they'll look for muchsmaller defects.
But typically they're doingkind of a spot check.
(09:46):
They wouldn't be doing a fullscan of the whole wafer because
of that very high resolution.
Françoise von Trapp (09:52):
So we're using all three of these in
manufacturing now.
Bryan Schackmuth (09:56):
Correct.
Françoise von Trapp (09:57):
So what determines when one would be
chosen over another, or is itnot like that?
Bryan Schackmuth (10:02):
A little bit like that.
Like I mentioned, if it's amemory application in production
, you can't use x-ray.
Françoise von Trapp (10:08):
Right, okay .
Bryan Schackmuth (10:09):
There are some things you can do to mitigate
the effects of the x-ray onmemory.
But typically they wouldn't beusing X-ray for memory but for
acoustics.
It's anytime they want to seethat level-to-level inspection.
So the benefit of acoustic iswe can isolate layer by layer so
we can look at the top layer,the middle layer, the bottom
layer and provide different setsof images for each of those
(10:33):
layers.
That helps them to kind ofinvestigate where the defect is.
Is it at their higher levels?
You know, maybe they have aprocess where they need to go
back and refine the process tohelp that top layer adhesion.
Françoise von Trapp (10:47):
So there are different types of defects,
and I know one of them is latentdefects, which are not as
obvious, something where adevice is going to fail down the
road in use.
Bryan Schackmuth (10:57):
Typically we might see a defect and it might
be a small, let's saydelamination.
It was two layers that aresupposed to be bonded together
and if you have a small defectas this chip gets processed, you
know there's the chemical,mechanical polishing.
Those defects can sometimescause almost like a blistering
(11:19):
effect or a slight bubble, andas you grind it that defect can
kind of expand it.
Can you know, as you polish thesurface that'll expand that
defect.
So there's certain cases wherewe would see defects that you
know would grow over time.
If it gets into a package andit goes into automotive you're
talking about extremetemperature variation and so
(11:42):
that expansion and contractioncan make that defect grow.
Françoise von Trapp (11:47):
So, basically, the acoustic
micro-imaging is making itpossible for you to find these
defects way sooner and pull themout of the line, so that you
don't even get to the point ofwhere they're going to get
bigger.
Bryan Schackmuth (12:00):
Exactly.
You know it depends on themarket.
If the chip is going into a toy, you know that's not high
reliability.
If it's going into a medical orsurgery, or if it's going in on
a satellite in space, it's veryexpensive to repair.
So the 100% inspection isgeared more towards that
high-end, high-expensive devices.
Françoise von Trapp (12:21):
So if you're weaving out the smaller
defects for the 100% inspection,can those get shipped off to A
use this?
This one's not a complete loss.
You can use this in a toy, Imean, do they do?
Bryan Schackmuth (12:33):
that they wouldn't even start with that
application from the start.
Françoise von Trapp (12:38):
Wouldn't that be great though.
So how has traditional acoustictechnologies, then, been
improved over the years?
Bryan Schackmuth (12:47):
Yeah, so traditionally the type of
scanning that's been done foracoustic microscopes and this is
, you know, since I've been withthe company 20 years is an XYZ
raster scan.
Imagine kind of an inkjetprinter scanning back and forth
over the paper, printing.
We're doing the same thing, butan XY over a sample and
generating an image.
(13:07):
It's that back and forth motion.
We use a transducer is what'scalled.
That generates the ultrasound,moves to the left, it stops, it
accelerates to the right, itstops, it accelerates to the
right, it stops and it goes backand forth.
So at the edges of the scanyou're constantly stopping, so
you're not acquiring data.
The spin Sam, what that does isa rotational or spinning scan.
(13:29):
So we start in the middle, wespin the wafer and we just move
to the edge.
I would say like a recordplayer but that would be dating
myself or like a CD player.
Françoise von Trapp (13:39):
You can say record player.
Well, I was just thinking aboutthe difference between scans.
When my kids I had twins andwhen I was pregnant, you
couldn't tell what baby was new,and now you see these amazing
images where they can really seewhat the baby looks like.
So you know, take that and putthat into the wafer application.
(14:01):
I would imagine like theresolution is a lot better.
Bryan Schackmuth (14:04):
Yeah.
So we've improved kind of thefrequency you know, going to
higher frequency.
Generally speaking you wouldsay lower frequency is lower
resolution but it can go throughmore material.
Higher frequency is higherresolution but it can go through
more material.
Higher frequency is higherresolution but it can't go
through as much material.
So we've improved thetransducer design, optimized the
frequency and the lens shape tooptimize the focal spot so we
(14:29):
can get that much higherresolution, and then, along the
way, you know, makingimprovements to the RF chain
that generates the ultrasound,maximizing, you know, the output
of the signal, as well asminimizing, you know, signal to
noise ratio.
Françoise von Trapp (14:47):
So you mentioned SpinSam, which is a
fairly new product for NORD'sintestine inspection.
Bryan Schackmuth (14:53):
Yeah, it was just released last year towards
the end of FY24.
That's our 100% 300-millimeterwafer inspection tool, really
geared for high throughput inthe wafer inspection area.
Françoise von Trapp (15:08):
For the purposes of our audience?
How can it be used to addressthe needs of wafer-level
packaging and advanced packaging?
Bryan Schackmuth (15:16):
Due to the complexity of these samples and
the wafers themselves, the costis extremely high, so they
really need high reliability.
You can put the SpinSam systemin your facility and do 100%
inspection on these samples.
A lot of the stuff we'relooking at today is stack dye,
whether 8-stack or 16-stackthese samples A lot of the stuff
(15:38):
we're looking at today isstacked die, whether you know
eight stack or 16 stack.
Françoise von Trapp (15:44):
And again, it's all about finding those
defects kind of early on in theprocess.
Bryan Schackmuth (15:46):
So that's the memory application.
Yeah, correct, Then juststandard bonded wafers.
One thing we've seen is atemporary bond.
So they'll use kind of acarrier wafer and they'll
temporarily bond a processedwafer on there.
That carrier wafer willbasically act like a substrate
for the actual wafer.
So we'll inspect that temporarybond because if the temporary
(16:08):
bond is not good you couldpotentially induce defects into
the sample later.
Françoise von Trapp (16:13):
So what about like fan out wafer level
packaging?
Is that an application spacewhere the spin sam would be
helpful for like doinginspection on RDL layers?
Bryan Schackmuth (16:23):
In the wafer area they build up the fan out
wafer level package so we canlook at those different layers,
the dielectric layers, the low Kdielectric or the
redistribution layer, and theneventually that wafer would get
cut and diced and then thatwould be put onto a substrate
that goes into kind of the finalpackage.
(16:44):
At that stage we would go fromour wafer inspection to more of
our tray inspection system, moreof a backend process to look at
that final assembly.
Françoise von Trapp (16:54):
Okay, so the spin sam is definitely for
the wafer itself.
Bryan Schackmuth (16:58):
We wouldn't say it's very front end, we
would say more like middle end.
So front end is more of likeyour UV processing, metal
deposition and that we're kindof after that process where they
start stacking those wafers.
Françoise von Trapp (17:13):
Okay, so it's the finding.
The known good die before iswhere it's used.
It's not for final packageinspection.
Bryan Schackmuth (17:23):
Correct.
Yeah, so definitely before.
If there's a thousand die onthe wafer, using the acoustic
microscope they can pick outthose defect parts and then, you
know, put them into a rejectbin.
There's no use in packagingthat into the final assembly,
saving your time.
Françoise von Trapp (17:40):
Okay, that's good to know, cause when
you were talking before aboutthe spin process, I was thinking
about how things are shiftingto panel level packaging and how
would that be adapted for usein panel level packaging.
But you're actually using itbefore we even get to that die
placement on the panel.
Bryan Schackmuth (18:00):
Correct.
Yeah, we would have a largearea scanner because you can use
the acoustic microscope throughthe different stages.
At the wafer level it getsdiced, gets put into a package,
but then we would go to adifferent form factor for the
acoustic inspection.
You know a large area scanner.
Typically we would be doinglower resolution at that stage,
lower frequencies, maybe 30megahertz, 50 megahertz, 100
(18:24):
megahertz, because the defectsize is not as critical in the,
you know, back end.
Françoise von Trapp (18:31):
Okay, so that would be a different tool
than the SpenSam.
Bryan Schackmuth (18:34):
Correct.
Françoise von Trapp (18:35):
Okay, so let's talk about the SpenSam a
little more.
Can you give me some of thebenefits?
Things like UPH, speedresolution, edge scanning
capabilities, things like that.
Bryan Schackmuth (18:46):
Sure.
So the driving reason that wewent to develop the SpenSam is
so we had a previous product, awafer inspection system AW300,
using that traditional rasterscan.
But really the push was tooptimize throughput.
But we kind of pushed it evenfarther.
What we want to look at iswafers per hour, per footprint.
(19:08):
So clean room space is veryexpensive, so we want to
maximize the throughput in thesmallest footprint.
So we want to maximize thethroughput in the smallest
footprint.
So with that spin scan we canget at 100 micron resolution
about 41 wafers per hour, whichis, I think, eight times faster
than our previous machine.
And then the resolution.
(19:34):
So currently the spin scan cango down to about 10 micron
resolution.
So you can do that full 300millimeter wafer at a 10 micron
pixel.
Again, the way we're scanningit, because of that spinning
scan we can do, you know, anedge scan, or you can think of
it kind of like a donut.
Or if I go back to the recordplayer playing the last track,
on the record.
You could never do that type ofscan with a traditional raster
scan, like you could never dothat type of scan with a
(19:54):
traditional raster scan.
You know we're just doing aring around the outside of the
wafer and the reason that'simportant is the coefficient of
thermal expansion when you'redoing the bonding process is
more extreme at the edges Right,and so customers are seeing
that the edge of the wafer iswhere they're finding more
defects.
In the center of the wafer wemight not see as many defects
(20:17):
because that's easier to bond.
And so with the SpinSam it'svery well suited to do a lower
resolution scan in the center,where you don't expect defects,
and then maximize the resolutiontowards the edge to see those
edge defects.
And that all goes to improvingyield.
If they can isolate thosedefects at the edges, remove
(20:39):
those dye from the process.
Françoise von Trapp (20:41):
Okay.
So while they're doing that inthe volume setting and basically
identifying the defects andremoving them, are they also
getting any knowledge that theycan feed back to the process
guys to say, hey, we keep seeingthis happen over and over.
Can you tweak the process?
Bryan Schackmuth (21:02):
Definitely so.
The acoustic wave again, like Imentioned, you can isolate
layers so as the ultrasoundtravels it travels in time and
then we can basically pickregions of time to look at.
If they see a trend thatdefects are occurring at the
base layer or the deepest layer,a trend that defects are
occurring at the base layer orthe deepest layer, you know they
might go back into theirprocess and say what is it about
the process that's causingthese defects?
(21:22):
Just at this layer?
We're not seeing it at theupper layers, it's always at
this lower layer.
And so they might find that,you know, they have some
unintended particle generationthat's putting particles at the
edge of the wafer during thatprocess and those particles then
cause the defects.
So they can go back and kind ofmaximize their process earlier
(21:44):
on to minimize defects down theroad.
Françoise von Trapp (21:46):
Yeah.
So that's great because they'renot just using it to remove the
defects and that immediateyield, they're actually taking
that information for improvingyields in the future.
Bryan Schackmuth (22:00):
Yeah, and improving their process
altogether.
Françoise von Trapp (22:01):
Right, okay , so I was reading through some
of the information about theSpinSam and there's something
called global tool matching.
What is that and why is itimportant?
Bryan Schackmuth (22:13):
So the SpinSam has four scanners, so there's
four scan stages.
There's a eFEM that loads allthe wafers, so we're using four
different transducers.
There can be slight variationsin the transducers, so what
global tool matching does isit's basically a matching
network.
So it'll look at the signal ofall four transducers and to do
(22:35):
this we have kind of a referencewafer that we'll use.
So it's a you could call itkind of a calibration wafer
that'll go on each of the fourscanners.
We take that referencemeasurement and then we optimize
the signal so that everythingis matched.
So kind of the end goal is, ifyou have a spin SAM in Singapore
running a wafer and that samespin SAM in Taiwan running a
(22:57):
same wafer, what we want toachieve is the same recipe, same
system, same wafer, same result, so that multinational
companies can use the same dataset or the same parameters and
not have to change or modifydepending on the system.
Françoise von Trapp (23:17):
Okay, and what about maintenance?
Bryan Schackmuth (23:20):
Yeah, so another key thing that we were
trying to design into theSpinSAM system is its modular
design.
So each of those four scannersI mentioned is a scan module,
and then there's also another RFmodule which generates the
ultrasound.
And so with that modular designyou can take one scanner
offline to do preventativemaintenance or servicing, but
(23:43):
the other three scanners willcontinue to run Traditionally.
If you had to do preventativemaintenance, the whole machine
goes down and you lose 100% ofyour throughput.
With modular design on SpinSan,we can take each scanner offline
, one at a time, but the otherthree scanners can continue to
scan, and so it makes it mucheasier to service, as well as
(24:06):
minimizing the downtime on thesystem.
Françoise von Trapp (24:09):
Are there any other advantages to being
modular design besides that?
Bryan Schackmuth (24:14):
In the future, if we were to develop, you know
, higher frequency, let's say,or develop a new RF feature that
gives a better image, we canjust remove that RF module in
the spin sand and upgrade it tothe latest RF module, and so you
don't have to replace the wholesystem to get the latest image
capability.
(24:34):
We're looking at ease ofrepairability as well as ease of
upgradability in the future.
Françoise von Trapp (24:40):
One of the things we've been talking a lot
about with different membercompanies is integrating machine
learning, ai and digital twintechnology into their tools.
Is there any of that involvedwith the spin Sam?
Bryan Schackmuth (24:57):
Yeah, that's a great question.
So traditionally we've had asoftware package called DIA,
digital image analysis.
It's using a threshold.
So you have a, imagine, agrayscale image.
You set a threshold, sotypically the defects are bright
, or, you know, we get a largereflection so we have a bright
defect.
So we would say any pixelthat's brighter than a certain
(25:18):
value is a defect.
That was good when the deviceswere simple.
But with today's complexlayering, the contrast in the
images is no longer easy to usea simple threshold technology.
So there's a team at Norton, youknow they work with that whole
(25:40):
segment, all three divisions,and they're really developing
the AI ML for us and so for theacoustic system we've already
introduced kind of the firstrendition of the what's called
MGIA or multi-gate imageanalysis, and so instead of
setting that threshold, there'sapplications today where if you
set a threshold you'll geteither overkill or underkill,
(26:01):
meaning you're going to rejecttoo much or you're not going to
catch the defect.
But with the AIML we can gothrough and teach the system
what a defect is and then youknow with the machine learning
over time you can build up arobust model that be able to
isolate the defects but notoverkill or underkill, and so
(26:28):
that's been introduced.
It can do.
The MGI is multi gate, meaningmultiple images, so you can take
, you know, three images, stackthem all together and then
analyze that as a whole package.
So it's really been useful for,you know, for the more complex
structures that we're seeingtoday.
Françoise von Trapp (26:40):
Are you getting more of a 3D image with
that, then?
Bryan Schackmuth (26:43):
We can do 3D Typically.
That takes a little bit longerso it's not really done in
production.
But what they'll do is if anoperator looks at the three
images, they're looking kind ofone by one.
If an operator looks at thethree images, they're looking
kind of one by one, and so theymight see a defect in the first
one, nothing in the second one,maybe something in the third
image.
What the AIML can do is look atall of them at the same time,
(27:10):
and so it might see a defect atthe first layer and the second
layer, but then, looking at bothof those at the same time, it
can interpolate.
Maybe that defect actually goesthrough the middle layer, and
so it can really see more thanan operator would be able to see
.
Françoise von Trapp (27:22):
So what does this do as an advantage
over a tool that wouldn't havethat?
Bryan Schackmuth (27:27):
We've seen applications, you know, in the
past where an operator is doingthat analysis, because the
contrast is not good enough todo the traditional threshold.
It's an operator making thatdecision, and so you introduce a
certain element of human error.
Right, okay, no if it's closeto lunchtime, maybe they're
thinking about lunch and notthinking about the defects in
(27:49):
the wafer.
Françoise von Trapp (27:50):
That would be me.
So does it speed up the process, makes it more accurate?
Yeah, definitely so, speedingup the process makes it more
accurate?
Bryan Schackmuth (27:56):
Yeah, definitely so.
Speeding up the process,automating it you know a lot of
our customers are drivingtowards that lights out factory,
where they don't want operatorsmaking any decisions and then
also for the traceability.
So you know, all the wafers aretraced, we're communicating
with the host and themanufacturing site and so all
that data goes along with thatwafer.
Françoise von Trapp (28:17):
Okay, so is the SpinSam a one-of-a-kind in
the industry, or are therecompetitive offerings?
Bryan Schackmuth (28:24):
I would say it's one-of-a-kind because of
that spinning technology.
So any of the competitiveofferings and even our older
generation system are using thatraster scan.
So there's really no way theycan catch up to the throughput
and footprint.
When you're doing that rasterscanning you have a mechanism, a
gantry over the wafer and soyou have a lot of moving parts
(28:46):
on top of the wafer, which isnot ideal for particle
generation because you'll havebearings or belts or motors over
the surface of the wafer.
Particles would tend to drop onthe wafer.
With SpinSam we just have onetransducer that just moves from
the center to the outer edge, sowe don't even have any moving
(29:06):
parts essentially over the topof the wafer.
So it really minimizes particlegeneration for the clean room
environment.
Françoise von Trapp (29:14):
So because you don't want to introduce
something that would cause adefect further down the line
during the inspection process.
That would be counterproductive, correct, okay?
Last question for you what canthe customers gain by installing
the SpinSam in their facility?
Bryan Schackmuth (29:33):
Yeah.
So I would say Spin SAM hasbest-in-class wafer per hour at
100 microns, 41 wafers per hour,and then in our roadmap we're
going to be targeting about twotimes that throughput and
that'll be coming kind oftowards the end of this year.
The next thing would be thedefect detection, so we can see
down to that 10 micron defectsize and we can do it in a very
(29:55):
small form factor.
So it's really about wafers perhour per footprint, so
maximizing the cost of ownershipof the system, and then the
quality standards, so theability to do that global tool
matching, to have each systemmatch, no matter where it is,
with any facility across theglobe.
Françoise von Trapp (30:17):
Okay, so where can people go to learn
more?
Bryan Schackmuth (30:19):
Yeah, so you can go to nortoncom and look up
SpinSam.
You'll see probably my face.
Françoise von Trapp (30:27):
We'll put a link in the show notes so
people can find it.
Thanks so much.
It's been a while since I'vetalked about acoustic
microimaging, so it was kind offun to catch up on that.
So I really appreciate yourtime.
Bryan Schackmuth (30:38):
Thanks for having me.
I really appreciate the timetogether.
Françoise von Trapp (30:44):
If you want to learn more about
semiconductor inspection, besure to mark your calendars for
an interview with Norton Testand Inspections' Andrew Mathers
to learn all about dynamicplanar CT with automated x-ray
systems dropping May 16th.
There's lots more to come, sotune in next time to the 3D
Insights podcast.
The 3D Insights podcast is aproduction of 3D Insights LLC.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!