Top x-ray companies
For Food & Product Inspection
+ Industry and shopping guide
The x-ray industry for product inspection is difficult to navigate. Even the worst players call themselves “global leaders” and the inconsistent way specs are listed doesn't help. This guide ranks companies on the most important metrics, explains them along the way, and ends with steps to compare those that match your needs.
TLDR: For difficult apps (you're looking for plastic, rubber, or very small metal), only reach out to high flexibility companies. If it's simple (your product is of the same consistency throughout and you're looking for metals) you can just find someone close to you. For a little more clarity jump to the comparison section, which will give you different steps based on your application difficulty. You can also fill out the RFQ and we will quote your top options. But if doing your own research, digesting this guide will save you 10s - 100s of thousands of dollars.
Our Offerings regardless of The brand
Note: This study is comprised of feedback from and research by humans, which are not known to be infallible—so for legal purposes everything herein should be considered opinion.
Company Rankings
This first section averages the scores given by distributors and end users across the industry. If they did not have experience with the company, they were asked not to contribute. Manufacturer’s scores were not combined for fear of sabotage, but some of their inputs regarding competitors are included in descriptions if they had positive things to say as well.
Contributors were asked to spread scores evenly so at least one company would get a 0 in each of the categories and at least one would get a 10, but many had trouble giving any company either score. So just remember that if someone got a 0 (which they didn't) it wouldn't mean they can't detect anything at all—the scores are just relative to one another.
Greyed scores are those with lower certainty/had 2 or less inputs.
Metric Details + EXPLANATIONS
Flexibility (Customizability)
This score combines 1, how many different applications a company will participate in and 2, how adaptable the machines are once installed (for when you modify products or swap them altogether), because the two typically go hand-in-hand. The next section will explain how flexibility impacts performance.
X-ray manufacturers can go one of two ways—they can make very flexible machines adaptable to any application or they can make inflexible machines targeting the simple, common applications. The benefit of flexibility is that your machines are better overall, and the down-sides are that they require more expert staff, both for manufacturing and for support. For this reason, the most common approach for manufacturers is to establish presets for simpler applications then market to those applications only. This is good because it makes things easy for sales and for techs at startup and lowers cost. It’s bad because the food world is big and even the producers of common products will at some point want flexibility that is not available to them.
Performance (Detection Accuracy)
This score averages how these companies are reported to compare to others in the applications they do.
Not all of these companies were always competing in their best application when observed, but it’s equally probable that your product is also not their ideal application. So if you’re considering a company with a low flexibility score and you make a product that is either easy or one of their specialties, then your experienced detection performance will be higher.
For example, Marel does well in their specialty, so they have high performance and low flexibility. Then there are other companies that do fine in simple applications, but because they take on complex ones and underperform, this drags down their performance score.
Why performance varies so much
How well a company performs is determined by their technology (single beam, dual, photon-counting), how much they develop that technology (hardware, algorithms, other software), and how it is applied to your product at install (company and technician skill, experience, and personal interest of the people involved). Each of these determining factors vary wildly between companies because they aren't something you can throw money at. X-ray development is complicated, tedious, and time consuming, and depends heavily on the people involved.
Technology
Eagle's introduction of single energy to food back in the 90's (they were a different company then) produced images like this:
Contaminants were visible, but large density differences were needed, and small contaminants hard to find. As you will see in the map of everyone's configuration options, Some companies still have single energy only.
Dual energy was later developed by Eagle and introduced to the food industry as MDX; it produced images like this:
But though developed in 2004, most companies still rely on single energy today and offer dual energy as a premium option.
Photon-counting was also introducted to the industry by Eagle, and produces images like this:
Technology makes a difference.
How much the technology is developed
Having a tool doesn't mean you can use it well, and different companies' performance will vary even within the same method of detection. Here are two different uses of the same technology (dual energy):
As you can see, the image on the right is more clear than the image on the left. This may not seem like a big difference, but over the course of thousands of products inspected, every little difference of accuracy means smaller contaminants found, and less good product being reworked (or thrown away). Every company should be continuously improving every technology they have.
How technology is applied
Each company develops algorithms for products as a category, and wants that algorithm to accomodate the extremes of the product category (the density ranges, density variance, volume, common contaminants, etc.). But your product will not fill that whole range (your cans of soup will have a narrower range of variance than all soup). So if you're applying the preset algorithm for all soup in the world to your soup, the detection will not be as good as it could be.
How well the algorithm is tuned to your product will depend heavily on the experience of the technician setting your machine up. This is why both machine flexibility and integrator skill impact performance.
Generator Lifespan
This is the average length people have reported seeing these companies’ generators last. This can be skewed if the interviewees encountered times when the generators were being run hotter than they should, or visa-versa. This is why we list generator power as one of the 4 things you should check when choosing between your top picks in the comparison section. Generator lifespan is important because it’s the most expensive component of an x-ray machine and the one that breaks down the fastest. It can be hard to get a straight answer on this from suppliers, so warranty length is the best indicator of quality itself. But generally, the more powerful a generator the longer it will last (if you have one metric to go by, wattage is the best indicator, amperage the second-best, and voltage the last). Users of small generators will accuse larger generators of “blasting through product”, but this is why they are de-tuned, and this is what makes them last longer. However, this metric exists because low-quality, high-power generators do as well, and this way, though flawed, is the only way to find them.
"But I heard smaller generators last longer!"
Assuming the generators are being run at full capacity, a lower-powered generator will last longer than a higher-powered one simply because the smaller one is generating less x-rays. But if a larger generator is turned down to generate as few x-rays as a smaller one, it will last much longer because its components are more robust. All this means is that a generator will last longer inspecting candy bars than large crates of meat. But you don't get to choose your product based on what will make a generator last longer, you have a fixed product. So from the end-user perspective, bigger is better.
A word about quality
You've heard the saying "you get what you pay for". Consumer products make us feel like this principle isn't true when cost does not equal quality (we're looking at you, Mercedes). But quality is not what you're buying in those scenarios, you're buying the appearance of success—or whatever that brand has paid to associate themselves with. This is usually not so in the processing industry; no one is "flexing" on their neighbors with the equipment they have at work. Here, the thing you're paying for is quality and overhead.
In the processing world where everything is wear parts, you quickly realize that better is cheaper—not just in parts cost, but maintenance and admin. And if it's true of anything, it's true of x-ray. X-ray machines have a lot of wear parts (in order of cost, the most common replacement parts are generators, rollers, belts, and curtains (unless you have a curtain-less option)). We chose generator lifespan as a metric for quality because it's the most expensive to replace and the most tempting to skimp on.
Tip - Bearings are what fail in rollers, and they cost ~$30 and are pretty simple to pack. Getting a whole new roller assembly will run you over $1000, so it will save you a lot of money down the road and will be simple for maintenance if they're even a little handy.
If you’re in a particularly tight spot now but know you won’t be later and financing isn’t an option, kicking trouble down the road with a cheap machine is one way you can "finance". But if you’re not in a tight spot and just trying to spend less overall, it pays to get the better equipment, which is more expensive most of the time.
Service
This score measures user experience with and general perception of: technician availability, technician competence, and parts ordering. Every end user knows the importance of having their equipment supported, so no further explanation is necessary. We can provide varying degrees of service on all x-ray equipment if you're already locked in and struggling.
Ease-of-Use
Ease of use is mentioned in most descriptions but removed from the table because it became meaningless for three reasons. First, flexibility and ease-of-use will be inversely related because with more flexibility comes more options—which someone has to learn. Second, even the most flexible machine can be the easiest to use if you don’t use all the features. And third, the company that doesn’t want to use all the features is not going to care about flexibility in the first place.
False Rejects
False rejects are not included in the table because they are synonymous with performance. The better the machine can see a contaminant, the better it can determine if it is one. (See the screenshots in the difficulty section for an example.) Anyone can get 0 false rejects by turning the machine off, and everyone can get sky-high false rejects by pushing the machine past where it’s supposed to go.
Top 5 Companies
Eagle PI.
Best all-around, mandatory for most difficult
Eagle is the company that pioneered x-ray in the food industry, they developed most auxiliary (item count, weight, package integrity, etc.) benefits available today, and they continue to get the best results with them in the field. They were the first to functionally apply photon-counting to food, and their performance in this field is also the best available. They have been the industry’s premium brand from the start, making generally the highest performing, toughest, and most customizable machines. The customizability is great for performance in each application, but needs distributors or experienced staff to set machines up or optimize them for significant product changes. The down-side of all the features and build is up-front cost, which is in the industry's top 20%. However, they claim the lowest long-term cost because (as covered in the generator section) the better x-ray components are to begin with, the cheaper they are long-term. It’s possible they were also factoring in performance (lower false rejects) but we need to confirm this. Down-sides include their smallest machine coming stock at 52.8” (which can be requested down to 36”) and smaller aperture-size to belt-width ratios on some models.
Notable Data - Does not retarget web visitors, sell your data, or deanonymize it.
Mettler Toledo/Safeline
#1 for standard/simple applications
Mettler Toledo is the high-end of the mid-range machines. They’re in the mid-range category because they use established, pre-set algorithms to go after standard applications with standard machines. They’re at the high-end of this spectrum because they’re [at least one of] the most experienced in this category, they make good units, and have sanitary designs on-par with Eagle. The down-sides of Mettler are the machine and service costs for their category. Mettler locks the equipment down so only they can service it, which means it’s expensive, they’re incentivized not to reduce the service needed, and support can become hard to get—all techs come from the same location so one bout of Covid can make machines unserviceable. The restricted access also means you are limited in the settings you can adjust. Because of this, they go through reps instead of distributors, but large accounts are generally happy because they don’t have to fiddle with support and get taken care of. On the plus-sides, they have a more formalized version of our rollout package that makes installation easy—on engineering, QA, maintenance, and operators. They have great software and data, good performance, and modular reject options—there is one universal plug that most any type of reject can be plugged into. They are also standardizing with quick-ship programs and bolt-on options.
Aicon
#2 for nonstandard/difficult applications
The guys who founded Aicon were Eagle’s Poland distributors until Mettler got more involved with Eagle. So when they were left with the short end of the stick, they made what they knew best—Eagle machines. You saw their similarity to Eagle in the table, and that’s no coincidence. This is the goal of Aicon and their performance reflects it so far. Aicon entered the market very recently and they don’t have a huge install base, but we expect them to become a major player because of their knack for sophisticated systems and programs.
Notable Data - Does not retarget web visitors, sell your data, or deanonymize it.
Anritsu
#3 for nonstandard/difficult applications
Anritsu is a good competitor out of Japan that manufactures metal detectors, x-ray, and check-weighers. They are experienced with dual energy and have good detectors, generators, and software, so they do quite well in custom applications. The down-side here is that changes for custom applications need to be made back in the land of the rising sun. They were also the first company to respond to the introduction of photon-counting, and we expect them to be the next best at it. Their cost was scored the same as Eagle on a 1-10 scale, but actual quotes are needed for confirmation.
Ishida
#2 for standard/simple applications
Ishida (also Japanese) is essentially the cheaper alternative to Mettler as they also target simpler applications to get better economies of scale. They specialize in maximizing the use of single beams, and for that reason they are great for low difficulty applications and—like Anritsu and Mettler—can even compete directly with Eagle on simpler applications. Like Anritsu and Mettler they are not exclusive to x-ray, but their portfolio is even more spread out than the former to include labelers, desiccant inserters, leak detectors, etc. They also promote dual beam and even photon-counting, but no one interviewed had seen either of them in the field. Complaints included difficulty in sourcing parts and more than one instance of being sent the wrong machine for install (not maliciously, of course). Ishida goes through Heat and Control as their sole US distributor.
Notable Data - Does not retarget web visitors, sell your data, or deanonymize it.
Specialty Companies
These companies focus on specific applications at the exclusion of others
Tomra
Best for dry bulk over 30 tons/hr
Tomra was known for their [good] vision systems in the agriculture world before they made x-ray machines, and their x-ray machines are made for that same kind of throughput and operate from the sorting perspective (where unwanted materials are a significant portion of overall product). Bulk players like Foss and Eagle go up to about 30 tons an hour—Tomra does up to 150. As you might expect, this means their machines are dramatically larger than everything else and harder to place in processing lines. A user added that glass detection was "spotty" (which is common) but in keeping with the rest of the industry, their detection of metal is usually still better than what you can get with a metal detector. Tomra bought Odenberg in 2011 and Best in 2012.
Peco Inspx
Great for talls (bottles etc.), good for low-mid. bulk, only for packet/pouch webs
Peco Inspx is a Californian company that specializes in tall, liquid product inspection including fill level monitors and dud detectors. They are an overall good company with talls (side-shooter x-rays for bottles etc.) that stand out. Although they only use up to 3 beams and Eagle uses up to 5 (but usually 4) we expect Peco's 3 to be comparable though we haven’t had a head-to-head with them yet. We have seen their machines last a long time in field, but their generators are reported to last as little as a year. They also do well with bulk, which would explain this problem if the gist we're getting of smaller generators turns out to be correct. Notably, they also appear to be the only company with a solution for continuous webs of packets or pouches that feed straight into the machine.
Marel
Good for bulk meat (non-crated)
Marel focuses on the meat industry with weighing, processing, and inspection equipment. X-ray is not their focus in the broader sense, but they’ve got meat figured out. Their site currently says “best in the industry”, “unrivaled”, and “more accurate than any other system available”, but their claims to support this are lackluster. They say they can go “down to 2 mm” bone on single chicken breasts, and that was legitimately impressive 6 years ago, but TDI Packsys claims better and an Eagle machine will do that out of the box, and better when tuned to the cuts. That said, the PXT integration isn’t quite done for Eagle’s largest bulk machine (3.3 ft. wide belt), so Marel will likely be the best for applications needing anything above a 2.4 ft. wide belt through 2026, though they are reportedly the most expensive in this category by a wide margin. They also do FA and are third in performance in this area, behind Eagle and Foss respectively.
Data Warning - The only company that retargets web visitors, sell your data, AND deanonymizes it
Foss
Good for FA (Fat Analysis) (crated)
Foss is a meat analysis company making samplers and analyzers for labs and the processing floor alike. Unlike the rest of the rest of the companies on this list they use x-ray primarily for Fat Analysis instead of contaminant detection (though they do it). They are listed in the specialty section because they are the next closest to Eagle's performance in FA, though they are reported to show different CL values for ground vs. loose trim on the same meat, which has resulted in some recalls. However, they appear to be the only company besides Eagle that can do frozen blocks and cartons, and can take blocks up to 7.87" high instead of 7". Foss's list pricing is higher than Eagle with starkly higher service and parts costs, but they will mark down upfront costs. Their machines appear to be made in Denmark, China, and Hungary.
Standard Ranking [Resumed]
Sesotec
Good for small, dry products
Sesotec can be thought of as a cheaper alternative to Mettler/Safeline and Ishida. The machines have traditionally been lower quality and lagging in performance, but their distributor (who is a trustworthy person) tells us they’ve gotten better recently and continue to improve. We include this because one of their competitors told us they have been investing a lot in support and sales in the last few years, so it’s easy to believe those aren’t the only things improving. Their competitor also noted they do free startups and the machines are easy to use. They are a German company that split off of S&S and appear to do a lot of manufacturing in the Asian countries.
Wipotec
Wipotec provides good sensitivity at good cost, their machines are easy to use, and they have strong software packages and data collection. Interviewees did not think they had dual energy and we couldn't find it on their site, but there is a press release with them talking about the benefits of it, so it might be forthcoming. They have a 200W variable energy generator and their units are modular in that you can attach their check-weighers or vision systems to them.
Notable Data - Mentions retargeting web visitors and defers to Salesforce's data policy which allows for deanonymization, but this doesn't mean Wipotec participates.
TDI Packsys
TDI Packsys comes from the security x-ray world and very recently threw (a lot of) resources at the food market. They make entry-level machines with average line speeds and detectors, but they do have dual energy that they use well (although it's at a disproportionately high price). The machines are reportedly difficult to use, have abnormally high maintenance costs, and we got mixed reviews on service. However, they have a large catalog of machines, keep many in stock, and the machines are reportedly "plug-and-play".
Notable Data - No privacy disclosure we could find
Loma
Checks the “X-ray box”
Loma is a mid-range manufacturer with a lot of machines in the field. No interviewees thought they had dual energy yet, but they put out a press release of what appeared to be their first go in June of 2025. The machines are mid-builds (neither chintzy nor robust) and the cost of service is high unless you’re close to a technician (you are billed from the home address of the visiting tech). As far as praise, they do well with entry-level customers, they have a good bit of machines in stock, and the machines are simple to use.
Thermo Fisher
Good if you don't want x-ray
Thermo Fisher, who you probably know for their electronics and lab equipment, also makes x-ray. There aren’t a lot of options, you can’t really customize them, and the lcd displays are reportedly finicky. They use 0.8 mm detectors with an option for 0.4, and only have single energy. However, they have a 900 W generator for their 'heavy duty' machines with their smaller units having 500 W generators. This suggests their machines are good for people who need x-ray but want to forget about it quickly; the machines will have a long life of not finding much. Of course the lifespan assumption is only true if they de-tune them, but this needs confirmation as we got no feedback on lifespan. We also got no entries regarding their support, but know from general experience that they have a good reputation.
Fortress
Fortress is one of the main metal detector manufacturers in the industry, and that’s probably what you know them for. They are new to x-ray (too new to assess well) but so far they have stuck to basic machines and applications, which makes sense.
Minebea
Minebea is a Japanese company with very high-end weighing products. They very recently made some x-ray equipment but haven’t been able to do well with their applications yet. Our daughter company, A-1 Scales, carried their checkweighers for a while but encountered some transactional and support/discontinuation issues, so they’re a company we would bookmark for some years down the road. Unlike the rest of their equipment, their x-ray machines are inexpensive and reportedly easy to operate.
Mekitec
Mekitec is a Finnish company with a hub in Texas that has their machines made in China—they have not historically performed well and have always been hard to get service and parts on. They use Ai to inspect images which sounds interesting, but this creates a low cap on line speeds because the Ai needs lots of extra time to process the image. Case-in-point, a customer of our ingredient handling division is using a Mekitec machine and has staff inspecting images visually because the Ai is too slow for the line, although you could put the reject further down the line if you're confident on timing. Also worth noting, your method of analyzation can only be as good as your image—if you're not getting a clear image, analyzing that image all day won't do any good.
Notable Data - Does not retarget web visitors, sell your data, or deanonymize it.
Cassel
Extremely wide, flat products (carboard, fabric)
Cassel is a german company and the only one (we've seen) that makes huge machines for flat products. Their larger machines aren't included in the machine mapping table below because it would become even harder to read, but their largest has a detection area width of 126". Their machines have decent specs but are very rarely encountered in the US so getting a service guarantee in writing is advisable.
Antares
Gravity-drop
Antares has traditionally been a vision company but they've recently started making x-ray machines. They seem to be targeting cosmetics, though surely they do food as well. No respondents or interviewees had experience with them, but interestingly they seem to make the only gravity-drop x-ray available. Other companies don't do this because you get better results dropping into a chute on a belt, but it could be worth a shot if that's not an option for you.
Available configurations
This table shows the types of machines every manufacturer provides, minimum and maximum specifications, and a few other details about the company so you can quickly identify or eliminate candidates.
Numerical values are the minimums and maximums we were able to find on their sites.
Yes's signify we found a claim.
No's signify we did not find it and the specs availability leads us to believe they don't have it.
Blanks mean me couldn't find it but have low confidence it means they don't have it.
Are you a supplier listed and we missed something? Please let us know. We reached out to all companies for inclusion.
Table Key
Simple Items
Washdown - Maximum found washdown rating
≥Length - Smallest machine length (for top-down (standard) machines)
≤Speed - Max belt speed - see next section for a warning and deep-dive on this topic.
≥Temp - Minimum environmental operating temperatures (everyone can inspect frozen/IQF products).
Pipeline - Have solution for pumped product
Tall - Have side-shooter machines for tall products and those with foil tops
3A - Have 3A-certified machines. Nearly all manufacturers mention dairy, but we were not able to find statements of 3A certs. for most. If they have a washdown rating of 69, though, and have hygienic designs, they could probably pass the certification.
≤Beams - Maximum number of beams (dual energy, triple energy, etc.). This includes tall/bottle machines which almost always involves more beams than standard inspection. But if they have 2 or more for talls, they almost certainly have dual energy for standard products.
FA - Do fat analysis
Photon - Have photon-counting (latest technology that bypasses scintillator crystals to improve resolution).
Wet bulk - Do wet bulk (usually meat; does not include talls or pipeline)
Dry bulk - Do dry bulk (for dry and wet (meat) bulk alike, most companies who do it will claim up to 50 tons per hour, but the practical limit is realistically around 30. Tomra for dry bulk is the only exception we've found.)
Aux - List at least two auxiliary functions (item count, weight, package integrity, etc.) Effectiveness with these features will swing as wildly as performance and depends on the hardware, software, and configuration at install.
Since - When first x-ray machine was released. There are many low-certainty values because few people listed this; low-certainty dates are the dates of the first press releases we could find from them about x-ray.
Manufacturing - Manufacture location — note that some companies will make different models in different places.
West - Manufactured in the West (including Japan)
Domestic - Manufactured in the States
Generator - Maximum generator power of largest machine (if you have a normal product, this will not be their best spec. available to you, which is why we will recommend you gather it when comparing picks.)
Detector - Minimum diode spacing available (not everyone's best spec. will be available on their machine for your app., which is why you should gather this spec when comparing picks.)
[Line] Speed
≤Speed - Fastest belt speed mentioned
All brands with 0.4 mm detectors (except Mekitec) should produce standard results up to ~150 fpm, and will find the contaminants they claim they can up to ~75 fpm. And this is fine for most people. But if you want great detection over ~75 fpm or standard detection over ~150 fpm, you will need to go with a brand with high customizability (spend more) or lower your standards. Here's an explanation if this is a hot topic for your product.
The tradeoff between line speed and performance
How fast your line can go will depend on how long it takes for an x-ray machine to take and process an image. How much time you need for each image is determined by:
1. How much data you need to process - In short, higher-resolution detectors will have lower speed limits to give the computer time to capture the data.
X-ray detectors are made of diodes spaced out in a grid, which are scanned one line at a time. A 0.4 detector (which has diodes spaced every 0.4mm) is going to generate twice as much data as 0.8 detector per line, and then there are going to be twice as many lines to scan. If the product is moving too fast, the lines of the image will lag behind, resulting in warped images, which will lower accuracy. (Software can “un-warp” the images, but it doesn’t restore the original resolution.) Since a lower resolution detector will generate less data than a high resolution one, your speed limits will be higher with lower resolution detectors. Diode spacing can be anywhere from 0.1 to 2mm, with 0.4 being the standard.
At full performance, assuming they are upgrading other hardware to handle it:
- 0.4 detectors will max out around 160 fpm.
- 0.2 detectors will max out around 105 fpm.
- 0.1 detectors will max out around 50 fpm.
- PXT (photon-counting) maxes out around 180 fpm.
2. How long it takes for your detector to register the x-ray beams - In traditional (non-photon-counting) models, the detectors do not receive x-rays directly. Instead, they rely on the light cast by scintillators, which generate light when hit by x-rays. This means inherent lag. Photon-counting (PXT) models generate much more data but don’t have this lag, which is why their line speed is similar to legacy models.
3. Processing time - The complexity of the algorithms being run on each image, the software running them, and the speed at which the hardware can execute them all impact processing time.
An often-overlooked and hard-to-assess element of a manufacturer’s quality is their software. Everything else can be great but if the software’s inefficient, the whole line will suffer. Also worth noting, a benefit of customizable machines is that we can add hardware to speed things up. For example, we can include one lith card (the analog-to-digital converter) per layer of detector, which will dramatically reduce processing time. Then we’ll just add cpu, ram, and hard drive space to accommodate it if needed. So this is why brands with high customizability should retain detection specs. at higher speeds.
Regarding High line speeds above ~160 fpm
Most suppliers will simply list the fastest speed their belt motor will go as their max line speed, but you’ll pay a price for going that fast with most suppliers because of the factors listed above.
So if you’re considering two suppliers for line speeds above ~150 fpm, have tests run with your product at lower and higher speeds, then ask for the images. This will allow you to understand the tradeoff you’re paying and determine whether the images are getting warped. Around 90% of applications we see are less than 120 fpm so you probably don’t need to worry about this. Applications with small items like bars (breakfast, energy, candy, etc.) are the most common exception, with speeds up to 450 fpm.
Data / Privacy
The Surveillance State doesn't usually innovate their own methods; the majority of them come from the private sector turning over its own data and programs. This metric determines how much these companies feed this beast by purchasing/fueling those services.
- Red "X" - Retargets you from web visit (fingerprints your device (using IP, device ID, keystroke patters, bluetooth, GPS, or other identifiers) to "anonymously" identify you and deliver ads on other websites you access)
- Black "X" - Uses a third party to "de-anonymize" your above data to get your name, phone number, email, etc. A grey "X" indicates their policy doesn't mention it directly, but includes policies of services that do.
- Blue "X" - Sells your data
- Blank - Their privacy policy does not discuss it
- Border - The information needed for this study was hidden behind forms (material because it usually means they want you to agree to their practices)
These indicators rely on statements from their privacy policy so it is possible that some do these things without disclosing, as it's possible they don't while their statement allows for it. It is also possible that they retarget based on supplied emails instead of web visits.
Machine Mapping
This section maps out all machines in the industry by inspection area width (at the belt) in inches. It includes all bulk and packaged product units, but naturally excludes pipelines and talls. Greyed values are not low-certainty, but awkward values that were approximated.
The X's
Each x indicates a model from the corresponding supplier in that size that we found.
If you are a supplier and we missed one, let us know!
Multiple x's will indicate they have different "models" in that size, but different suppliers divide their models differently. Some segment their models by build, and not capabilities: for example, they will consider single energy, dual energy or photon-counting as options to be placed in the same model, and then will have different "models" for dairy, meat, etc. based on washdown and other features. Some segment their models by capabilities instead of build: for example, they will consider different levels of washdown (standard, dairy, meat) options to be placed on the same model, then they will call different levels of performance (single, dual, photon) their different models.
So if you have low detection expectations and don't need washdown, a single x means that supplier has a model for you. If you have simple application with washdown or high detection expectations, you need a supplier with either a high flexibility score or multiple x's in the size you want. (Companies with a flexibility score over 7 will probably deliver any model in any configuration you need.)
How To Tell What Width You Need
The machines are listed by inspection area at the belt, not belt width, so you don't need to make that conversion.
X-ray generators cast x-ray (light) through your product in the below shape, and this shape varies little if at all between manufacturers. Ideally you want your entire product scanned, so the minimum width necessary will depend more on the height of your product at the top than the bottom.
For example, the white box in the image below represents a product that is 22" wide and 10" tall (length typically doesn't matter). So by putting it in the diagram you can see the minimum belt width you'll need is 32".
The above image is for larger products, so for smaller ones ones you can use the images below.
Mid-sized products
Small products
Bulk products
Don't worry if this isn't clicking; any RFQ will ask for your product dimensions and your quote will be for the smallest machine suitable for your product.
How To Choose An X-Ray Supplier
Hopefully a few companies have stood out to you, and now you need to compare them. You can reach out to the companies individually or submit the form here and we’ll forward it on.
If you have an easy application:
Anyone experienced should do fine with it given they have the right hardware. To assess this, compare:
- Detector Resolution - this will tell you the spacing between diodes, which determines the resolution of the image that your machine will be inspecting. 0.4 mm detectors are still considered standard, and they can almost universally go up to ~150 fpm. 0.2 mm is pretty widely available, with a few going as far as 0.1 mm. Line speeds for a 0.1 mm detector will be capped around 50 fpm, thus why PXT (now adopted as photon-counting by others) was introduced to get the same performance without sacrificing speed. See the notes on line speeds for more information.
- Generator - Even if you don’t have dense product, higher power generators will last longer, and they are the highest-cost wear part by a massive margin. Wattage is the best determiner of longevity (unless a high-wattage generator is cheaply made, so COMPARE WARRANTIES). Beryllium generators are slightly better than glass but have slightly narrower scan widths.
- Service - Talk to a fellow processor about service in your area. Everyone says they’re the best, but can you get in touch with them, and do they know what they're doing?
- Beam count - Even on simple products, having more beams will make contaminants easier to see as they get more similar to your product's density. If you aren’t looking for anything like that (rubber, plastic, wood, etc.), and you have a simple product, single energy is fine and you can skip this.
- Testing on your product - Hopefully results won’t vary much, but everyone will do it for free and sending product is easy.
If you have a difficult application...
...and cash is particularly tight now, the steps are the same except that testing is mandatory because performance will depend heavily on factors that will get too messy even for this guide (company and technician experience with your application, algorithm performance, software optimization, and small hardware and software differences that add up).
If you have the money to invest now, you can start saving time now by going with Eagle; test anyone you like against them.
Notes
Difficulty
The factors that determine an application’s difficulty are primarily how messy the image is, the density difference between your contaminant and the product, and any other unique needs you have.
- "Messy" (noisy) images - X-rays are usually sent one-way through the product to create a 2D image. So if you have a single block of cheese, that’s going to create a very clean image and contaminants will stick out like a thumb that's not going to be sore because you saw it way ahead of time. A bag of french-fries, on the other hand, will have tons of 3D overlapping objects flattened into a 2D image, which is going to create a very chaotic—or noisy—image.
As you can see in the example below, the image of cheese is very clean and easy. With the peanuts, however, dual energy is required at a minimum or you are going to miss some things—or have very high reject rates.
- Density difference - Even if you have a simple, clean image, contaminants will be harder to spot if they don’t differ enough in density. And this is why better machines are still valuable for simpler products. Single energy will be nearly as good at finding stainless steel in cheese as dual beam, triple beam, or even photon-counting. But it will not do nearly as good for rubber in cheese or golf balls in potatoes.
- Product mass - How much material is in each scanned product (your volume and density) can be a sole factor taking your product from easy to difficult because the more product, the harder it is to get x-rays through it. For example, a small cheese block is easy—anyone can get through it, and because it’s consistent, everyone will do well. But if it’s a wheel of cheese, fewer suppliers will be able to do it. We are setting mass as a difficulty metric because more product means a better generator, and the “better” the brand, the more likely they are to have better generators.
- Line speed - If you want to find standard contaminants that everyone can find, and your line speed is over ~150 fpm, you need a better machine. If you want to find smaller contaminants that some may have trouble with, and you want to go over ~100 fpm, you also need a better company. This is explained at length in the “line speed” section.
- Auxiliary function needs - Auxiliary functions aren’t necessarily hard, but few enough companies do them sufficiently that needing them will push you further up the list. These include checking for anything aside from contaminants like: item count, item weight, package integrity, fat analysis, fill level, missing items, voids in bread or potatoes, etc.
- Non-standard requirements - Not everyone makes x-ray machines you can wash down, run in cold temperatures, or inspect product in a pipeline. So any non-standard need you have will restrict your options. The down-side is that this will eliminate some entry-level options, but the up-side is that the more unique applications a company can accommodate, the better their machines are likely to be.
What this means for you is the easier the application, the more companies will do well with it—if you’re only looking for stainless in cheese you can pick about anyone. But the messier the image, the more similar the contaminant, the higher mass, the more unique needs you have, the more you will want to stick to the top of the list.
About ABM & Interest Conflict
ABM Equipment builds (engineers, fabricates, installs) systems (handling, processing, packaging, inspection) for many products (mostly food). As a supplier of systems, we've come to service equipment regardless of brand, including x-ray—sometimes because it's part of our system, sometimes because they're short on techs. Over time, we've come to distribute and/or rep. for companies whose equipment makes our customers' (and our) lives easier. Two of those companies are Eagle and Mettler Toledo.
There is a short video overview of our company here.
Our (Conflict of) Interest
We are Eagle’s primary distributor with a territory in the Northwest and a Metter Toledo rep. nationwide (for all their equipment). We argue this is an alignment instead of conflict of interest because our interest aligns with reality—Eagle isn't at the top of this list because we're with them, we're with them because they're at the top of this list. We have also been doing x-ray since its early days in the industry, and it's not like we don't have a choice.
A story to illustrate the point:
One of our global customers—for whom we’d done a system—needed x-ray. As one of the top 5 food producers with many hundreds of lines, they wanted to save money and ruled out Eagle from the start. We recommended some companies but they largely did their own sleuthing, had tests run with the top companies, and then invited us for a head-to-head against the best of those. Long story short, we became very familiar with the top companies, and that’s reflected in the ranking. Again, these are great companies in respect to the industry as a whole. But in the end, when they factored in the replacement part costs and false reject rates, there was only one real option. Of course they also got us with the deal (the best x-ray support group in the Americas) but that is probably (definitely) biased.