Revolutionizing Semiconductor Quality: The Dynamic Planar CT Advantage

WEBVTT
00:00:00.381 --> 00:00:04.791
This episode of the 3D Insights podcast is sponsored by Nordsen Test and Inspection.
00:00:04.791 --> 00:00:12.028
Leaders in acoustic, optical and x-ray inspection and metrology systems for the semiconductor and SMT manufacturing markets.
00:00:12.028 --> 00:00:22.329
Nordsen's next-generation 3D x-ray inspection software, dynamic Planar CT, reveals incredible details for Nordsen's zero-defect strategy.
00:00:22.329 --> 00:00:37.570
Nordsen's advanced software is two times faster than planar CT, with enhanced detail quality, superior reconstruction and a larger field of view for increased coverage and shorter cycle times all to improve yields, processes and productivity.
00:00:37.570 --> 00:00:40.067
Learn more at Nordsoncom.
00:00:40.067 --> 00:00:58.433
Hi there, I'm Francoise von Trapp, and this is the 3D Insights Podcast.
00:00:58.433 --> 00:01:15.894
Hi everyone, in this week's episode of the 3D Insights Podcast we're continuing a deep dive into the importance of 100% inspection in the semiconductor manufacturing environment, and this time we're going to focus on advancements in x-ray inspection and specifically for 3D imaging.
00:01:15.894 --> 00:01:25.950
So I am speaking with Andrew Mathers of Nord's Intestine Inspection to learn about what's going on in this area and pick his brain on his knowledge.
00:01:25.950 --> 00:01:27.766
So welcome to the podcast, andrew.
00:01:28.500 --> 00:01:29.786
Thank you so much for having me on.
00:01:30.700 --> 00:01:31.743
Before we dive in.
00:01:31.743 --> 00:01:33.990
I understand you're new to the company at Nordson.
00:01:33.990 --> 00:01:37.551
Can you tell me a little bit about yourself and the role that you've taken there?
00:01:38.441 --> 00:01:50.073
Sure, I joined Nordson about six months ago as a principal product line manager for their automated x-ray inspection product line and my background is really in x-ray.
00:01:50.073 --> 00:02:19.451
I've been in the x-ray world for about 10 years now, as starting out as a postdoctoral, postgraduate at university using x-ray to look at plants, to look at them in terms of their root development underground and their leaf development above ground, and then moved heavily into the industry side and have been in the industry for about eight years working for different x-ray system manufacturers and recently joined Nordson to look at electronics.
00:02:20.539 --> 00:02:24.985
So you're bringing this really varied background in.
00:02:24.985 --> 00:02:32.316
Are you finding the needs of electronic inspection a lot different than the other areas that you've studied?
00:02:33.539 --> 00:02:34.241
No, I don't think so.
00:02:34.241 --> 00:02:41.150
I think there's a commonality between what a lot of customers want from x-ray inspection.
00:02:41.150 --> 00:02:46.575
They want high resolution, so they need to be able to see very, very small things down to the micron level.
00:02:46.575 --> 00:03:12.771
They obviously want it to be done quickly, they want it to be cost effective and, most of all, they want it to help drive better products or better understanding of what they're imaging their imaging, and that we know is becoming very important in this industry, as the devices that we're inspecting are getting more and more complex and therefore more and more.
00:03:13.933 --> 00:03:26.752
I don't want to say valuable, but when you capture the defects can mean the difference down the road of having to start from scratch or being able to repair something and the loss there right.
00:03:28.099 --> 00:03:42.907
Absolutely, and I think this brings us purely around to 3D imaging, because in the past many of our customers, certainly many people operating in this space, have historically relied on 2D radiographic imaging for inspection.
00:03:42.907 --> 00:04:06.854
As manufacturing requirements become ever more stringent and the quality and the reliability of devices has to continue to increase to meet the demands from the suppliers, from the end customers and the market in general, 3d imaging, driving that higher performance, inspection, that better resolution, better quality, is becoming a much, much more common requirement to support these efforts.
00:04:08.080 --> 00:04:12.372
So how long has 3D imaging been an option in the electronic space then?
00:04:13.241 --> 00:04:16.870
3D imaging has been around for a while in electronics imaging.
00:04:16.870 --> 00:04:22.031
However, the quality of it has always been something that's quite difficult to achieve.
00:04:22.031 --> 00:04:53.240
One of the problems with this style of imaging is you're looking at a planar object, you're looking at something that's typically very flat, like a semiconductor wafer or a printed circuit board, and to image these objects you use planar scanning, or laminography as it's also known, and this can suffer from lots of imaging artifacts, which are features or distortions that are present in your images that aren't truly present in reality in the sample, for example, in electronic samples.
00:04:53.240 --> 00:04:59.024
These make it very, very difficult to clearly visualize between the different layers and the devices.
00:04:59.024 --> 00:05:13.310
However, the 3D imaging side of things allows us to very, very clearly discriminate between these layers and see things in clear detail, which is incredibly valuable.
00:05:14.430 --> 00:05:14.670
Okay.
00:05:14.670 --> 00:05:26.115
So I want to step back just one minute because I want to make a clear distinction when you're talking about planar, you're talking about using planar images to create a 3D image versus planar for a 2D image.
00:05:27.375 --> 00:05:33.437
So when I say planar, I mean using planar CT, which is designed for flat samples.
00:05:33.437 --> 00:05:51.673
Ok, so you are building a 3D volume of your images at the end, but because of the nature of the sample being so flat and the way that you have to rotate around it to get that computed tomography approach, it tends to be quite difficult to avoid having quite nasty artifacts in the image.
00:05:52.399 --> 00:05:56.737
Okay, and the artifacts aren't actually on the sample that you're using.
00:05:56.737 --> 00:06:00.064
These are actually introduced because of the imaging process.
00:06:00.824 --> 00:06:01.365
Absolutely.
00:06:01.365 --> 00:06:08.744
It's generated features or distortions that are present in the image that aren't there in reality in the sample.
00:06:08.744 --> 00:06:16.911
But you're quite right, they come from the imaging approach that you're using, but it is the best imaging approach for these planar samples.
00:06:18.000 --> 00:06:25.627
So with laminography they're taking different slices of images and then putting them together using software.
00:06:25.959 --> 00:06:42.326
So laminography, you tend to have a static sample and you rotate your source and detector, or just your detector, in a motion around the point that you're inspecting your region of interest within the sample and you still reconstruct and build up a 3D volume.
00:06:42.326 --> 00:06:47.435
However, you tend to inspect this in slices in the Z-axis.
00:06:47.435 --> 00:06:48.923
So you're looking top down.
00:06:48.923 --> 00:06:59.422
You have very good clarity in those slices, but once you look at a vertical slice and you look at it from side on, you tend to see quite bad artifacts that disrupt the data.
00:06:59.882 --> 00:07:03.209
So for a 3D x-ray, what are we getting as the output?
00:07:04.091 --> 00:07:11.538
You're getting data which is non-destructively looking at the structure of something, so you won't just get surface detail.
00:07:11.538 --> 00:07:15.779
Yes, indeed, you're getting the surface externally and internal surfaces and anything in between.
00:07:16.661 --> 00:07:20.740
Okay, and the image that's output is a 3D image.
00:07:22.011 --> 00:07:29.161
The reconstruction of the scan is a 3D volume, but you can intersect that volume at any angle.
00:07:29.161 --> 00:07:32.934
You like to create a 2D image as a cross-section.
00:07:33.956 --> 00:07:39.932
How has Nordson addressed these challenges that you've been describing about planar CT?
00:07:40.713 --> 00:07:52.214
So in the past we've had a solution to acquire images called planar CT, which takes images by moving our detector around and keeping the sample and the x-ray source static.
00:07:52.214 --> 00:07:58.036
But, as I mentioned, there is a lot of artifacts that these types of scans can suffer from.
00:07:58.036 --> 00:08:08.382
So one thing that we've released more recently and we're really excited about is dynamic planar CT, and this is our next generation of 3D inspection.
00:08:08.382 --> 00:08:16.072
It's still a planar approach, it's still laminography, but it has some major advancements and improvements over planar CT solution.
00:08:17.033 --> 00:08:18.596
So how does it work?
00:08:18.596 --> 00:08:19.218
What does it do?
00:08:19.800 --> 00:08:35.100
It's a software-based product and it offers high-speed, high-resolution 3D imaging of electronic devices to identify critical defects during the production, particularly in the semiconductor market and the SMT market.
00:08:35.100 --> 00:08:50.416
And it's a lot, lot faster than plain SET, so it acquires images at twice the speed, which obviously has huge benefits to the customer because it has much higher throughput and much higher units per hour.
00:08:50.416 --> 00:09:00.403
You certainly don't want the bottleneck in your manufacturing process to be waiting for your x-ray system to check the quality of what you've produced.
00:09:00.403 --> 00:09:07.514
You want to keep your line or your fab moving at high speed to get your products out the door, and that's what this enables.
00:09:15.730 --> 00:09:17.352
So with planar CT that you were describing, you were talking about laminography.
00:09:17.352 --> 00:09:22.043
Does the dynamic planar CT still use laminography or is it a different image capture process?
00:09:23.951 --> 00:09:28.379
It does still use laminography, but it is a slightly different image capture process.
00:09:28.379 --> 00:09:37.263
It's far smoother and it uses a much wider field of view, so you can acquire images of a much wider area, which means it can be done a lot more quickly.
00:09:37.263 --> 00:09:53.495
And it also means that, critically, you are dosing your electronic samples up to a much lower level, and X-ray dose is really quite critical for a lot of x-ray inspection components in the electronics world, like high bandwidth memory.
00:09:53.495 --> 00:10:06.037
So we take an orbit with our detector which sweeps around the point that we're inspecting and we can take many, many more projections, many, many more 2D radiographs.
00:10:06.037 --> 00:10:15.839
When we do this, and with a brand new reconstruction algorithm to build these all together, we get a much higher quality 3D model and data out of the other end.
00:10:17.910 --> 00:10:20.679
So how does this impact the image quality?
00:10:23.211 --> 00:10:24.517
So it's greatly improved.
00:10:24.517 --> 00:10:31.450
As I mentioned, one of the biggest challenges is these artifacts that you can't see that makes it very difficult to see between different layers.
00:10:31.450 --> 00:10:41.825
It also means if you've got quite a dense component on top of something else in your circuit board within your semiconductor sample, it can be very, very hard to see underneath it.
00:10:41.825 --> 00:10:56.624
And in 3D we can separate out the different layers and we can inspect each one individually, so we can see details that were previously obscured or masked by those dense or really complex components.
00:10:57.690 --> 00:11:04.484
If it's software based, does it remove artifacts or does it prevent them from happening altogether?
00:11:07.051 --> 00:11:13.831
The artifacts to a certain degree with this type of imaging are always present, but we take the images in a slightly different way.
00:11:13.831 --> 00:11:15.535
We take far more images.
00:11:15.535 --> 00:11:36.914
We take far more individual shots from different angles, far more projections and we have in the past used a more iterative approach to reconstruction, Whereas now, because we've taken far more images, we can leverage different reconstruction techniques and therefore it builds a much higher quality, detailed model in 3D of the sample.
00:11:38.721 --> 00:11:40.368
Are you implementing AI for this?
00:11:42.041 --> 00:11:43.988
No, not currently, not this product.
00:11:43.988 --> 00:11:49.892
We are looking into AI in all sorts of areas, but I'm currently not able to speak about those.
00:11:50.533 --> 00:11:52.601
Okay this tool.
00:11:52.601 --> 00:11:58.193
Is it something that's used more in development, or is it actually used in the manufacturing side of things?
00:11:59.320 --> 00:12:04.873
It's used more in the manufacturing side with my product line with automated X-ray inspection.
00:12:04.873 --> 00:12:06.659
So it needs to happen extremely quickly.
00:12:06.659 --> 00:12:17.892
We're talking about capturing a scan of a specific area in under three seconds typically closer to two seconds and it needs to reconstruct very quickly and perform the analysis very quickly as well.
00:12:17.892 --> 00:12:32.255
So this is for inline manufacturing very, very high speed, short tack time and really high throughput to keep the samples moving down the line or keep samples being analyzed at high speed.
00:12:33.400 --> 00:12:39.274
So how do you distinguish between when something is automated versus, I guess, manual?
00:12:44.220 --> 00:12:44.600
I guess manual Sure.
00:12:44.600 --> 00:12:56.240
So in the manual x-ray world you'll typically have an operator that loads samples on a tray or individually into an inspection area, into a system and they will kick the scan off themselves with the software, wait till the scan's completed and then exchange the sample.
00:12:56.240 --> 00:13:19.027
The difference here with automated x-ray inspection is samples are typically fed into the systems on a conveyor belt or via a loader, like a magazine loader which you might have in a fab environment for semiconductor, and the samples are analyzed and then taken out the other side or they come out the same side back into the loader and then the next sample is put in.
00:13:19.027 --> 00:13:20.530
So it's completely automated.
00:13:20.530 --> 00:13:22.221
There's no human interaction needed.
00:13:22.221 --> 00:13:23.924
It's all done autonomously.
00:13:24.645 --> 00:13:35.890
Okay, so this is specifically designed for use in volume manufacturing semiconductor fabs and also in SMT and surface mount technology.
00:13:35.890 --> 00:13:36.532
Okay.
00:13:36.873 --> 00:13:37.072
And what?
00:13:37.092 --> 00:13:38.743
about in the advanced packaging facilities.
00:13:38.743 --> 00:13:40.365
Is it also suitable for there?
00:13:41.288 --> 00:13:42.991
No, so that's a different product line.
00:13:42.991 --> 00:13:46.504
That's our metrology x-ray product line.
00:13:46.504 --> 00:13:53.475
Anything that's a wafer that's populated, we would do with AXI.
00:13:53.475 --> 00:14:08.690
But AXM, which is very much the wafer fab very, very beginning level of building a semiconductor wafer, that would be what would be used for that and they don't use dynamic planar CT for that.
00:14:09.539 --> 00:14:17.100
What are some of the applications that this tool is ideally suited for, and where in the manufacturing process would you find the tool?
00:14:18.542 --> 00:14:26.955
So this is really good for the semiconductor back end or SMT line in a factory.
00:14:26.955 --> 00:14:32.019
It's really useful for any 3D application with electronics X-ray inspection.
00:14:32.019 --> 00:14:34.288
But focus on those two markets.
00:14:34.288 --> 00:14:47.346
It's really really useful for looking at voiding present in ball grid arrays, imprinted circuit boards in SMT or looking at voiding in micro bumps in flip chip devices in the semiconductor backend.
00:14:47.346 --> 00:15:05.688
Being able to visualize each layer of the sample and clearly discriminate these voids in high resolution without artifacts from the neighboring or dense, complex components around is really really valuable for our customers and is something they desperately need to ensure high quality of their products.
00:15:06.801 --> 00:15:25.711
So, basically, your customers are able to identify these defects and pull off the line the wafers that are impacted, so that they don't go through continued processing before they end up being packaged into some device that later fails.
00:15:26.919 --> 00:15:29.967
Absolutely, that's the entire.
00:15:29.967 --> 00:15:46.570
Purpose is to ensure that products that aren't of the necessary quality to perform at a high level will be identified, removed from the manufacturing process, to ensure that the quality of what is going out to the end customer is of the highest quality and will perform to the top level.
00:15:47.820 --> 00:15:53.379
Does the tool provide feedback to help in process development further down the road?
00:15:53.379 --> 00:16:05.181
If you keep finding the same void or defect at a certain stage of the process, can they use that information later to, for instance, improve processes and yields?
00:16:07.105 --> 00:16:07.586
Absolutely.
00:16:07.586 --> 00:16:15.390
And one of the big movements in the manufacturing world at the moment, and in particular electronics, is something called Quality 4.0 or Industry 4.0.
00:16:15.390 --> 00:16:21.032
And this isn't just something where you're looking at the same defect over time and reporting.
00:16:21.032 --> 00:16:32.904
This is reporting in a very connected smart factory back to the manufacturing defect over time and this is a reporting in a very connected smart factory back to the manufacturing process to try and remove that defect or critical problem in the samples that are being created.
00:16:32.904 --> 00:16:47.109
So as a customer, not spending a lot of time making that adjustment, you can make it on the fly and you can quickly correct it and ensure the product coming through is of the quality needed to make the final cut and be sent to the end customer.
00:16:48.572 --> 00:16:53.486
Okay, so to just wrap things up, tell me about Nordson's automated x-ray solution.
00:16:53.586 --> 00:17:01.052
So Nordson's current portfolio of automated x-ray solutions can be split into three categories.
00:17:01.052 --> 00:17:04.421
We have three different system series that serve different markets.
00:17:04.421 --> 00:17:19.726
We have our XS series, which focuses on the semiconductor market, the X series for SMT and the X hash series, and that supports two markets the power hybrid market and the final assembly, test and pack market.
00:17:19.726 --> 00:17:36.874
But the thing that all these systems, these different markets, have in common is that they all provide high speed, high resolution and, most critically, high quality imaging with critical defect inspection analysis driven by our advanced in-house algorithm library.
00:17:36.874 --> 00:17:50.573
We have a direct presence in over 35 countries worldwide, working with the biggest names in the wide world of electronics, and we have an install base in excess of 2000 AXI systems alone.
00:17:52.039 --> 00:17:57.645
What would you say are the three key takeaways you'd like listeners to come away with after listening to this episode?
00:17:58.386 --> 00:18:13.613
I think the first one is that there is a real movement going on at the moment away from 2D radiographic imaging alone to judge whether samples are of the highest quality, and there is a move towards 3D imaging to give more detail and higher performance.
00:18:13.613 --> 00:18:15.382
So, I would say, is the first one.
00:18:15.382 --> 00:18:29.509
Secondly, what this 3D imaging provides is greater speed of inspection, is greater speed of inspection, so it's faster cycle time, higher units per hour and throughput, whilst it's also reducing the dose given to the samples while they're under x-ray.
00:18:29.509 --> 00:18:39.869
So this gives you more confident detection of defects due to increased imaging quality and greatly reduced presence of artifacts.
00:18:39.869 --> 00:18:55.295
And finally, it's a lower cost of ownership to customers, because we don't now need specialized clamping apparatus for these samples, which had to be held in place previously, because of the new, smoother way that we acquire the images.
00:18:55.295 --> 00:19:04.186
And, fundamentally, dynamic Planar CT helps drive better quality from our customers' products, supporting their development and their success.
00:19:05.471 --> 00:19:08.327
Okay, well, where can people go to learn more then?
00:19:09.039 --> 00:19:12.445
They can go to our website, they can go to nortoncom.
00:19:12.445 --> 00:19:20.088
We also have a YouTube channel and some very nice new videos just created on the Dynamic Planar CT we discussed today.
00:19:21.141 --> 00:19:31.782
So on any of our socials, our websites and just Google around on the internet, Great, and you know what we will do that Googling for them and put some links in the show notes.
00:19:31.782 --> 00:19:35.451
So thank you so much for joining me today, Andrew.
00:19:35.451 --> 00:19:36.073
It's been a pleasure.
00:19:37.140 --> 00:19:37.803
You're very welcome.
00:19:37.803 --> 00:19:48.799
Thank you for having me on them
00:19:49.080 --> 00:19:54.292
Next time on the 3D InCites podcast, join me and my guest Pradyush Kamal from Siemens, as we discuss the evolution of EDA since the dawn of 3D integration.
00:19:54.292 --> 00:19:59.551
There's lots more to come, so tune in next time to the 3D InCites podcast.
00:19:59.551 --> 00:20:04.969
The 3D InCites podcast is a production of 3D InCites LLC.