{"id":3459,"date":"2026-09-29T04:01:50","date_gmt":"2026-09-28T20:01:50","guid":{"rendered":"http:\/\/www.christianfort.com\/blog\/?p=3459"},"modified":"2026-09-29T04:01:50","modified_gmt":"2026-09-28T20:01:50","slug":"what-is-the-minimum-detectable-defect-size-of-an-offline-x-ray-inspection-system-4b28-e5f7eb","status":"publish","type":"post","link":"http:\/\/www.christianfort.com\/blog\/2026\/09\/29\/what-is-the-minimum-detectable-defect-size-of-an-offline-x-ray-inspection-system-4b28-e5f7eb\/","title":{"rendered":"What is the minimum detectable defect size of an Offline X-ray Inspection System?"},"content":{"rendered":"

If you\u2019ve ever walked through our quality control lab here, you\u2019ve probably seen the team huddled around an Offline X-ray Inspection System, staring at a monitor where tiny shadows of metal, plastic, and solder paste are crisp enough to count individual micro-vias in a printed circuit board (PCB). More times than I can count, I\u2019ve had customers stop me mid-presentation and ask, \u201cWhat\u2019s the smallest defect this thing can actually catch?\u201d It\u2019s not a trivial question\u2014buyers don\u2019t just want a number from a spec sheet; they want something they can trust will keep their products from failing in the field, from a medical device that powers a patient\u2019s insulin pump to a automotive ECU that controls braking systems. As a supplier of these systems, I\u2019ve spent years testing, refining, and arguing about this exact number, and today I want to break down exactly what goes into defining that minimum detectable defect size (MDDS) for an offline X-ray inspection system, no corporate jargon overloaded, just real-world context from the teams that use our equipment every single day. Offline X-ray Inspection System<\/a><\/p>\n

<\/p>\n

First, let\u2019s get one thing straight: there is no universal MDDS. It\u2019s not like a ruler where every system has a single, agreed-upon smallest inch it can measure. The number depends on three non-negotiable factors: the type of defect, the sample you\u2019re scanning, and the inspection settings you use. Let\u2019s start with defects because that\u2019s where most people get confused. When we talk about \u201cdefects\u201d in X-ray inspection, we\u2019re not just talking about a hole in a metal trace. There are three core defect types that matter for MDDS: voids, misalignments, and foreign material (FM) inclusions. Each interacts with X-rays differently, so their detectable sizes are wildly different.<\/p>\n

Take voids, for example. If you\u2019re inspecting solder joints\u2014one of the most common use cases for offline X-ray\u2014voids are gaps in the molten solder that don\u2019t bond properly, and too large a void can weaken a joint enough to break under stress. Voids absorb less X-ray energy than solid solder, so they show up as darker (wait, no\u2014actually, let\u2019s clarify that: X-ray systems work by measuring how much radiation passes through a part. Denser materials block more, so they appear brighter on the detector. A void, which is air, has almost no density, so it lets more X-rays through, appearing darker. That means void contrast is based on the difference in density between solder (about 8.4 g\/cm\u00b3) and air (0.0012 g\/cm\u00b3)\u2014a huge difference, right? So for a typical tin-lead solder joint on a PCB, a well-calibrated offline X-ray system can pick up a void as small as 10 micrometers (\u00b5m) in diameter. But wait\u2014if the solder is lead-free? Solder densities shift a little (tin-copper is about 7.2 g\/cm\u00b3), so the void contrast drops just enough to push the MDDS to 15 \u00b5m. That\u2019s a real, tangible difference we\u2019ve measured in our in-house test lab, and it\u2019s not something you\u2019ll see advertised in a spec header.<\/p>\n

Now, misalignments. These are defects where two components don\u2019t line up correctly when assembled\u2014for example, a surface-mount resistor that\u2019s shifted half its width off a pad, or a die attach that\u2019s offset from the lead frame. Misalignments rely on edge detection, not contrast, so their MDDS is tied to spatial resolution, which is a different metric than contrast resolution. Spatial resolution is how sharp a system can distinguish two separate points from each other. For most mid-tier offline X-ray systems we supply, spatial resolution is about 5 \u00b5m per pixel. That means the smallest misalignment you can reliably detect is roughly two pixels wide, or 10 \u00b5m. But again, this changes with the part size. If you\u2019re scanning a large component like a power module lead frame, you might use a lower magnification to cover the whole part, which cranks your spatial resolution down to 20 \u00b5m. Suddenly, that same 10 \u00b5m misalignment is invisible. We\u2019ve had a automotive client come to us once saying their old system missed a misalignment that caused a brake failure, and it turned out they were running at too low a magnification to pick up the 12 \u00b5m shift\u2014something we caught when we re-ran their tests at optimal magnification.<\/p>\n

Foreign material inclusions are the third big category, and these are the trickiest because FM can be anything: a speck of copper dust from manufacturing, a plastic shaving from a reel of components, even a fragment of silicon from a broken die. FM contrast depends entirely on what the speck is made of. A copper particle has a density almost as high as solder, so it shows up very brightly, with excellent contrast. We\u2019ve detected copper specks as small as 5 \u00b5m in our lab. But if the FM is a low-density material like plastic or paper, contrast drops way down. A plastic speck might only be detectable at 50 \u00b5m, and even that requires the system\u2019s software to adjust threshold settings to pick up faint contrast differences that are easy to miss if the machine is not calibrated daily.<\/p>\n

That leads us to the second big factor in MDDS: the sample itself. No two parts are identical, and even tiny variations in a part can change what you can detect. For example, if you\u2019re inspecting a thin PCB for solder joints, the underlying substrate (fiberglass and copper) is relatively transparent to X-rays, so more radiation reaches the detector, giving you a stronger signal and better contrast. If you\u2019re inspecting a thick automotive power module, which is made of stacked copper, ceramic, and molded plastic, the total amount of material in the path of the X-ray is much higher. That means more radiation is absorbed before it reaches the detector, increasing noise in the signal. Noise is the enemy of MDDS\u2014if the signal from a tiny defect is the same brightness as random background noise, you can\u2019t tell them apart. For thick power modules, that same 10 \u00b5m solder void we found on a thin PCB jumps to 25 \u00b5m because of background noise. We worked with a aerospace parts manufacturer last year who was struggling to inspect thick turbine blade mounts, and they were stuck with a system that could only detect 100 \u00b5m voids\u2014until we upgraded their detector to a higher-sensitivity model, which cut that MDDS by 60%. It was a small adjustment that made a huge difference for their compliance requirements.<\/p>\n

The third factor is the inspection setup, and this is where customers often cut corners, leading to inaccurate MDDS numbers. Two settings matter most here: magnification and exposure time. Magnification is how much you zoom in on the part. Higher magnification means you\u2019re using a smaller area of the detector to scan the same part, which increases the number of X-ray photons that hit each pixel\u2014reducing noise and improving contrast. But higher magnification also means you can only scan a small area at a time, so you have to move the stage more, slowing down inspection. Exposure time is how long the X-ray source fires at the part for each frame. Longer exposure means more photons hit the detector, again reducing noise, but it also makes the inspection cycle slower, which is a problem for high-volume manufacturing.<\/p>\n

Here\u2019s the trade-off we walk customers through all the time: if you need the smallest possible MDDS, you run at high magnification and long exposure. But if you\u2019re inspecting 10,000 parts an hour, that\u2019s not feasible. For a mid-volume PCB assembly line, our most common recommendation balances at a magnification of 20x and an exposure time of 100 milliseconds, which gives an MDDS of 10\u201315 \u00b5m for solder voids and 10 \u00b5m for copper FM. For a low-volume, high-precision medical device line, we might suggest going to 50x magnification and 500 milliseconds exposure, which brings MDDS down to 5 \u00b5m for voids and 3 \u00b5m for copper FM\u2014but that cuts your throughput by 70%. It\u2019s not a one-size-fits-all number, and anyone who tells you their system has a fixed MDDS is either lying or selling something that doesn\u2019t fit your needs.<\/p>\n

Now, let\u2019s talk about how we actually test MDDS in our lab, because anyone can pick a number out of thin air. We don\u2019t trust spec sheet numbers, and we make our customers watch us run these tests so they know we\u2019re not fudging results. The standard method I use with every customer is to scan a calibration sample from a reputable metrology company\u2014specifically a standard test block with drilled holes of known sizes, varying contrast materials, and intentional misalignments. For void testing, we use a test block with 5 to 50 \u00b5m holes drilled into a tin-copper alloy; for FM testing, we have small specks of copper, aluminum, and plastic placed at varying sizes; for misalignments, a test die with pads that are intentionally shifted by known amounts.<\/p>\n

We run this test block through our systems with all the typical inspection settings a customer would use, and then we look at what size feature the system\u2019s software reliably flags as a defect\u2014three times in a row, to rule out random noise. We also do blind tests: we hide a set of test features from the operator and have them flag defects, then cross-reference with the known sizes to see how many they catch. In our lab, we recently ran a test for a semiconductor packaging client and found that our mid-tier system caught 92% of 5 \u00b5m copper FM, but only 78% of 50 \u00b5m plastic FM\u2014exactly matching the contrast expectations we\u2019d discussed earlier. That\u2019s data, not marketing, and that\u2019s what we share with every customer.<\/p>\n

I\u2019d be remiss if I didn\u2019t mention the software, too. A good offline X-ray inspection system\u2019s software isn\u2019t just a viewer\u2014it has built-in algorithms that enhance contrast, reduce noise, and flag defects based on your tolerance limits. A lot of lower-tier systems use generic algorithms that can\u2019t adapt to different parts, so their MDDS is much higher. Our software, for example, has adaptive thresholding that adjusts to the density of the part you\u2019re scanning, so if you switch from a thin PCB to a thick power module, it automatically adjusts the signal processing to keep the MDDS as low as possible without increasing inspection time. We also let customers customize defect limits\u2014so if you only care about voids larger than 20 \u00b5m, you can set the software to ignore smaller ones, which speeds up inspection but doesn\u2019t compromise the defects that matter.<\/p>\n

Now, let\u2019s get real about common misconceptions. I can\u2019t tell you how many times a customer has asked, \u201cCan this system see a 1 \u00b5m defect?\u201d The short answer is: technically, the detector might be able to pick up a signal that small, but it will look exactly like noise, so you\u2019ll get hundreds of false positives an hour that your operators have to manually check, which defeats the purpose of an offline inspection system. Offline systems are for non-destructive testing (NDT) of parts that are too critical to ship, not for 100% inspection of every tiny particle. For that, you\u2019d need an inline system, which is a different category entirely. There\u2019s no reason to sacrifice throughput or sanity for a defect so small it\u2019s irrelevant to your product\u2019s reliability.<\/p>\n

Another common myth: the MDDS is the same across the entire detector. That\u2019s not true. Edges of the detector have slightly different pixel sensitivity than the center, so defects near the edge might have a 5\u201310 \u00b5m higher MDDS than ones in the middle. We account for that in our calibration, which we recommend doing every 100 inspection runs or whenever you change parts, because even small temperature shifts in the X-ray source can shift sensitivity over time. Calibration is a step most customers skip, and that\u2019s why their actual MDDS is worse than the spec sheet says.<\/p>\n

So, putting this all together, what\u2019s the real-world MDDS for an offline X-ray inspection system? It depends. For most standard PCB assembly applications, it\u2019s between 10 and 25 \u00b5m for voids, 5 and 20 \u00b5m for copper foreign material, and 10 and 20 \u00b5m for misalignments. For high-precision applications like medical devices or semiconductor packaging, that number drops to as low as 3 \u00b5m for high-density FM and 5 \u00b5m for voids, but that requires optimized settings and slows down inspection. For thick, dense parts like power modules or aerospace components, expect that number to be closer to 20\u201350 \u00b5m, because the material absorbs so much X-ray energy.<\/p>\n

<\/p>\n

At the end of the day, the MDDS isn\u2019t a technical detail to hide in a spec sheet\u2014it\u2019s the difference between a product that works as intended for 10 years in the field and one that fails prematurely, costing you thousands in recalls or lawsuits. As a supplier, our job isn\u2019t to give you the smallest number possible; it\u2019s to give you the number that\u2019s right for your products, your throughput, and your quality requirements. If you\u2019re tired of spec sheets that promise a 1 \u00b5m MDDS and deliver garbage when you run real parts, reach out to our team. We don\u2019t sell one-size-fits-all systems, and we don\u2019t hide our test data. We\u2019ll walk you through our lab, run the exact tests for your parts, and give you a realistic MDDS number you can trust. Stop guessing about whether your inspection system is catching the defects that matter. Contact our procurement and engineering team to learn more about custom configurations tailored to your quality control needs.<\/p>\n

Micro-CT<\/a> References<\/p>\n

    \n
  1. ASTM E1865-18, Standard Practice for Computed Tomography (CT) for Industrial Nondestructive Testing<\/li>\n
  2. ISO 16232:2018, Road vehicles \u2013 Cleanliness of components of fluid circuits \u2013 Method of analysis<\/li>\n
  3. International Electronics Manufacturing Initiative (iNEMI), 2023 Solder Joint Inspection Best Practices<\/li>\n<\/ol>\n
    \n

    Shanghai Focus Intelligent Technology Co., Ltd.<\/a>
    With abundant experience, we are one of the most professional offline x-ray inspection system manufacturers and suppliers in China. We warmly welcome you to buy customized offline x-ray inspection system made in China here from our factory. If you have any enquiry about quotation, please feel free to email us.
    Address: No. 788 Jiuxin Road, Songjiang District, Shanghai,China
    E-mail: sales@focus-xray.com
    WebSite:     https:\/\/www.focus-xray.com\/<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

    If you\u2019ve ever walked through our quality control lab here, you\u2019ve probably seen the team huddled … What is the minimum detectable defect size of an Offline X-ray Inspection System?<\/span>Read more<\/a><\/p>\n","protected":false},"author":841,"featured_media":3459,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[3422],"class_list":["post-3459","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-industry","tag-offline-x-ray-inspection-system-4e4b-e63021"],"_links":{"self":[{"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/posts\/3459","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/users\/841"}],"replies":[{"embeddable":true,"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/comments?post=3459"}],"version-history":[{"count":0,"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/posts\/3459\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/posts\/3459"}],"wp:attachment":[{"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/media?parent=3459"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/categories?post=3459"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.christianfort.com\/blog\/wp-json\/wp\/v2\/tags?post=3459"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}