Invited to the facility along with just a few outlets, Canon offered us a deep look into their manufacturing processes, a look that no journalists have had before. Even the Canon USA product and project managers travelling with us hadn’t seen the production lines. 

If you’re imagining, as I did, rows of gleaming Terminator-style robots cranking out lenses without human interaction, you’d be wrong. Canon’s lens manufacturing is an incredibly hands-on process, a craftsman-centric approach to making the lenses that looks more like the shop of someone who lovingly restores vintage cars. Think Star Wars rather than Star Trek.

If you’re wondering like us why lenses like the Canon RF 100-300mm f/2.8 L IS USM are nearly impossible to find, we found out why.

The 100-300mm lens is backordered because the process of making them is laborious.. They’re assembled by hand, and the production team can only produce around nine of them a day and still maintain their quality standards.

Let that sink in for a moment. The $10,000 Canon 100-300mm lens has a production rate of under two hundred a month. 

Secrets from the Executives

We expected to see many PowerPoint presentations, which are a big part of meetings in Japan. But I was not expecting the openness from the most senior imaging executives when answering questions. Some of the answers were more frank than anything I’ve heard Canon say before.

In a sit-down interview with the most senior staff in the imaging divisions, I was allowed to ask any question I wanted, not just a pre-approved question. This is practically unheard of in the camera world; usually, the executives pre-vet the media questions. 

I started by asking if, being behind Sony in the mirrorless market, they’d seen any advantages in the delay. I wanted to know if there were things that they were able to bring to the R-series by taking time to catch up that benefited them when they entered the new mirrorless era. “Did that allow Canon to look farther down the road…to find an opportunity?”

“When we entered the full-frame mirrorless market,” said Mr. Manabu Kato, Chief Executive of IMG Business Unit 1, “it was in 2018. At that time, we were indeed behind Sony.” 

That might be the first time I’ve heard a Canon executive admit that they were caught unprepared.. Generally, Canon has talked about the strength of their DSLR system and their plans only to bring mirrorless to market when the time was right. 

Our lineup of cameras now stretches from the R1 to the R100,” he continued, “with more than 60 lenses. At this point, we don’t feel like we’re behind them.”

The most surprising answer came when I asked how they felt about stopping development on the EF lens mount, having been Canon’s standard for twenty years before the transition to mirrorless and the RF mount. 

“By evolving the EF mount into the RF mount, we gained advantages like large aperture, short back focus, and high-speed communication. Those opened up new worlds, so we saw it as a chance.”

“You learned what the EF mount could not do,” I asked, “and mirrorless gave you a chance to put all that into practice?”

Here’s the reply that surprised me. “Ultimately, we realized there were things the EF mount could no longer achieve. As we sowed those seeds, the mirrorless era arrived, and the opportunity became real.”

That’s the first time I’ve ever heard Canon say they were developing the RF mount before mirrorless cameras. Likely, we would have seen DSLRs move to the RF mount even if mirrorless cameras had not become the norm.

When mirrorless took over from DSLR, the older SLR cameras were reaching the end of their practical development life. Autofocus was limited by the need to use a separate focusing module, but perhaps the EF mount was a bottleneck in their development, too. 

By the Numbers

Through many PowerPoint presentations, Canon laid out its strategy for camera and lens development. As expected, they talked about how dedicated they are to innovation and detailed all the development and manufacturing processes they have pioneered. 

They talked about one of Canon’s key strengths being the in-house development and manufacturing of all of the key components of the cameras. It’s not the first time I’ve heard a camera manufacturer talk about this advantage. Sony has often said that their ability to develop camera bodies, sensors, and lenses gave them the ability to sprint into the mirrorless market, and to create bodies that would take advantage of future lens technology and lenses that could unlock the potential of the cameras.

This in-house development, or lack of it, is something that really hurt Nikon’s mirrorless plans. It was buying the sensors from Sony and designing the processors in-house. Buying sensors from your competitors is an overall bad business strategy. 

The Slow Pace of Incremental Improvements

Canon has been producing SLR cameras since 1937, and while some of the improvements in gear happen rapidly, most changes are incremental. The switch to digital occurred relatively quickly in the scale of photographic history, but most of the time, camera and lens updates have minor changes, though there might be technologies in updates that were a decade in the making.

Many of the presentations talked about the improvements new technologies are able to bring to Canon’s products, but it was surprising to see how much work goes into these often minor improvements. 

Canon spends an enormous amount of R&D on technologies that improve the shooting experience, though some of these improvements may not even be noticeable to the average user. They result in better images, if only marginally better, though the whole is greater than the sum of its parts. 

In one of the meetings, lens engineers described a technology they invented called Subwavelength Structure Coating, also called SWC, and it’s used in conjunction with another Canon technology, Air Sphere Coating.

The goal with SWC and the Air Sphere Coating is to reduce flaring and ghosting in lenses. 

Canon coats SWC lenses with a series of nano-scale spikes that sit between the light and the lens surface. Canon says these nano spikes ease the path of the light from the air to the lens. You can think of this like slowly getting into a cold pool versus jumping in headfirst. 

But can you see the difference in practice? Yes? Maybe? Sometimes? Anything that makes an image better is good, but in most images, they showed us the improvements are subtle. 

So Why Bother?

Everything in photographic gear is either an incremental change and improvement or the result of many incremental improvements. Nano spikes on a lens might not radically change images, but Canon didn’t spend a decade on the technology to revolutionize photography. They made SWC to improve overall image quality over lenses without it. At some point, Canon will combine Subwavelength Structure and Air Coating with some new technology, and image quality will improve again. 

Chris Niccolls of PetaPixel and I were chatting between meetings about the improvements we were being shown. ‘I think I get it,’ I said. ‘Lenses are good-lenses with SWC are a little bit better,’ and I made the gesture where you hold one hand flat and then hold another just above it to show a slight improvement. 

I would do this multiple times on the factory tour. While I found this running joke amusing, it’s the point of what most R&D does. 

Canon also showed us IBIS versus IBIS combined with the Periphrial Coordinated Control system in their Optical Image Stabilized (OIS) lenses. Perirperal control works to reduce the distortion in the very corners of an image that IBIS itself can’t correct for. It’s a subtle difference. 

We were shown issues, and while the corners were better, I would not describe them as radically different. The effect is similar to using lens correction profiles to adjust for lens distortion. 

“IBIS is good,” I said, making the gesture, “but IBIS with Periphreal Coordinated Control is better.”

I repeated the gesture after seeing how Canon uses deep learning to make auto white balance more accurate, especially for the blue cast that a clear sky often causes. “Auto white balance is good, deep learning auto white balance is better.”

And then I made it again when they showed the benefits of AI-based sport detection over traditional AI-based subject reaction. “Autofocus is good, sport-based autofocus is better.”

See the pattern here? While it was a good running joke, it turned out to be more accurate than funny.

Each time Canon showed us some advance in technology, it seemed minor in its own context, but Canon’s been at this for more than eighty years, and these little changes add up. 

Take a lens with Periphreal Coordinated Control, put a SWC coating on it, and mount it to a camera with action-based AF and better white balance control, and now these little steps add up to a much more powerful system. 

Inside Canon’s Off-Limits Lens Factory

The next day, we got a deep dive look, complete with clean suits, through Canon’s lens factory. Canon has never offered a tour of their lens production before, and having been on lens and camera factory tours with other manufacturers, I was surprised by how much access they gave us. 

Japanese culture is known for having people who dedicate their lives to a single pursuit. A Tōkō master creates gleaming and polished swords. A Geijutsuka makes the most intricate pottery. And, it turns out, there are artisans of lens polishing and lens assembly. 

Canon introduced us to their “meisters”, their collection of employees who have been working for decades to hone the craft of lens construction and assembly. Other companies might have similar craftspeople, but they’ve never been made available to the press. 

Mitsuharu Umei is Canon’s lens polishing meister, and its lens assembly Meister is Kazuyo Otsuka. Both have worked in Canon’s lens production facilities for decades, and both are artisans.

We followed the path of a camera lens from glass to shipping container, starting with the polishing of the glass.

The Art of Making the Lens

In its marketing, Canon talks about the idea of monozukuri. “It can be literally translated as ‘making things’ or ‘crafting things’ (‘mono’ meaning thing, and ‘zukuri’ meaning the act of making),” Canon Europe’s website explains. “But it is so much more than that. Conceptually and culturally, it reflects-and respects- the soul and art of the maker. 

Mitusharu Umei is Canon’s meister lens crafter, and Umei-san hand-polishes lenses with unbelievable precision. He is the embodiment of monozukuri. 

For some lenses, the tolerances are fractions of a millimeter. Umei-san told us that for a large-diameter TV broadcast lens, if the lens were the size of Dodger’s Stadium, the tolerance would be less than the thickness of a piece of paper. Umei-san can polish lenses to this tolerance by hand.

We walked through rows of equipment used to polish glass, boxes with nozzles that spray abrasive or water to take a lens from raw and opaque glass to a final piece of optics. These lenses work their way through a series of steps, including hand-polishing the glass on diamond-coated spheres. 

Umei-san has been honing his craft at Canon for thirty-seven years. Put another way, I’m 55 and Umei-san has been learning the art of glass polishing since I was in college. He told us it might take a decade or more for an apprentice to develop the skills needed to make the lenses that require the most accuracy. 

He had us test our skills on a large polishing machine, and I made the mistake of going first. A rotating base holds the glass, while a disc the size of a small pizza has to be pressed onto that rotating glass. The task takes two hands and is like patting your head while rubbing your stomach. I am, it turns out, never going to be a master craftsman at Canon as I nearly spun the disc off the rotating base.

Not all Canon lenses are made by hand, though. Aspherical lenses are shaped like a bell and can’t be made by hand. Kit lenses and any lens element that requires a sophisticated molding or shaping process are done by robots. 

Send in the Robots – Sophisticated Lens Shapes 

One of the few machines we were not allowed to photograph makes these aspherical lenses in a totally automated process. A slab of glass is loaded onto a platform, which is then heated to the melting point, and then a machine presses it slowly as the glass cools. Through the window in the machine, we could see molten glass being forced into shape. 

These slugs of glass go through polishing processes and then move to assembly. Another machine, which we also could not photograph, looked the most like the type of machine that might assemble a human-killing robot in a sci-fi movie. Robot arms turn, lift, and lower these lenses into the metal housings, pressing them into place and finishing them in a process that moves them automatically from robot to robot. 

We couldn’t photograph these because they’re made in-house by Canon and have proprietary designs. Like a lot of the equipment at Canon, they look both futuristic and like something made by a mad scientist. Arms swing, automated quality assurance systems flash the results of the optical tests they make, but the boxes clearly look like they were made by hand instead of purchased off-the-shelf from somewhere. 

Master Assembly – Every Lens Touched by the Same Person

Canon’s lens assembly Meister is Kazuyo Otsuka. She works inside the clean rooms of the lens assembly facility, guiding the production of every super-telephoto lens Canon makes. 

Otsuka-san has been a craftsman on this line for nearly forty years, and she said she considers each lens she has helped make one of her babies. I’ve been reviewing Canon lenses for more than twenty years, so I’ve unknowingly used dozens of lenses she made by hand. 

I find it incredible to think that one person has had a hand in producing so many lenses, which ties all Canon shooters together. That wedding photographer with a 200mm lens? Otsuka-san. The birder with a 600mm? Otsuka-san. 

Much of the camera and lens assembly is done by hand, mostly by women. When I toured Sony’s camera body facility years ago, and Nikon’s years before that, the staff explained that women tend to be more dexterous than men, and tend to have smaller hands that are more suitable for tasks like tightening millimeter-wide screws.

These jobs are not the monotonous assembly lines you might think of; the lines are small, making each member of the team more of a craftsperson than a widget. The meticulous attention to detail is what makes the production capabilities so small. You can have a handcrafted production team, or you can have an automated assembly line.

The Terminator Factory

Until this point, we had seen mostly processes carried out by hand. In the glass polishing area, tools were sophisticated versions of traditional tools. Diamond polishers sit next to manually operated washing and lubrication bays. 

Not every lens can, or should be made by hand. Entry-level and enthusiast lenses are produced by robots, with the assistance of humans, instead of humans with the aid of robots. 

One of the manufacturing lines we watched takes the aspherical blanks created in the automated lens creation process we saw earlier and fits them into their lens bodies. This robotic system is used for kit lenses and lenses with custom aspherics. 

This area looks like the factories in the Terminator movies, but only slightly more so. The robotic assembly lines are created in-house by Canon. They look more like a mad scientist created them in a lab. They’re not gleaming white like a car production plant; they’re physical, mechanical tools that have been assembled, also by hand, for specific tasks. You can tell they were made in a machine room somewhere in the factory, not at the type of plants that make those killer robot dogs. 

Quality Assurance

The most amusing part of the tour came in the quality assurance section. Here, prototype lenses and cameras are shaken, dropped, flung into a simulated wall, and subjected to extreme heat and humidity as well as extreme cold. 

All of the QA tools are automated and repeatable. If you want to see how a camera survives a fall from a meter above the ground, you have to be able to repeat the text exactly. Seeing machines designed to drop a camera on its attached lens precisely makes me feel a bit better about the times I’ve dropped cameras. Clearly, I’m not the only one. 

Canon also tests its shipping containers using these precise automated tools. If that $10,000 hand-assembled lens breaks on the way to B&H, all the effort is lost, as is the revenue. Seeing a box dropped onto a corner felt particularly comical. 

Canon invited us into the hot and the cold chambers, and I can tell you nothing feels quite as bad as being in a room in the high 90s with 100% humidity while wearing a business suit, except then entering a cold room where the sweat instantly freezes. 

The Takeaways from the Canon Factory Tour

I’ve been fortunate enough to have taken several factory tours over the years. In each one, the executives have been immensely proud of their processes and their approaches to producing the highest quality photography and video tools. 

Most camera users I’ve talked to have no idea how much hands-on design and production go into their gear. 

Every company has its own gleaming robotic production lines. Sony’s image sensor production lines are nearly completely automated and housed in a sparkling white clean room where only the team that maintains the equipment can be seen walking around. I’m sure Canon’s sensor facilities are equally gleaming. 

Yet all of the companies have production lines like the ones for the 100-300mm lens, where people work precisely and efficiently to bring the product to life. 

I have no idea if the other companies have the artisan “meisters” like Canon, partially because no company has given us such unprecedented access. 

What’s clear is that Canon takes great pride in what they do. Every executive, every product manager, every factory worker expressed the happiness they get in moving technology forward. 

Sometimes there are leaps in technology, and sometimes there are nearly imperceptible advances. If Canon had not shown us the nanotechnology they developed for their lenses, I might never have known about it, as the benefits are so subtle. 

In Shinto, both a religion and a practice that nearly fifty percent of the Japanese practice, there is the concept of “tamashii,” the spirit that physical objects possess. The translation isn’t perfect, but it’s akin to a product having a soul. 

To Canon, its products have tamashi, and everyone we met proudly talks about their commitment to bringing products to life. 

It might not be readily apparent when you’re shooting portraits or capturing wildlife, but Canon believes they are bringing gear to life. Canon talks about their commitment to your gear being part of your photography or videography experience.

Since photographers consider their cameras to be part of their creative process, this makes sense. Canon makes its cameras and lenses with a purpose and a dedication to the art of image creation, just like you use that gear to bring your vision to life. 

When showing off prototypes of the R1 and the C50, the team also showed a mint condition T90, the camera that formed the design directions for decades of Canon’s camera development. They handled it with care and precision, wearing white gloves to keep from marring the surface. But they also handled it with cotton gloves out of respect for the camera and how it launched Canon’s position in camera development. 

The takeaway from the Canon tour is that the small improvements that each new technology offers are part of a larger goal of constantly improving. Some of these improvements come to life relatively quickly, and some of them take decades to bring to life. 

The handcrafted nature of Canon lenses and bodies is something that connects all Canon shooters. The improvement designed at the Utsunomiya plant, and Canon’s many other design and manufacturing facilities, are tied together, making clear its goal of bringing its sense of creativity to all of its users. 

The post Exclusive Visit Inside Canon’s Lens Factory: A Surpising Look at Where Its Lenses Come To Life appeared first on Imaging Resource.

I recently got the chance to visit Sigma Corporation’s lens (and camera) factory in Aizuwakamatsu Japan again. As part of the big product announcement for the new Sigma fp full-frame mirrorless camera, they invited myself, Carey Rose from DPReview and Johnnie Behiri from Cinema5D to attend, and brought us to Aizu for a tour of their factory afterwards. You can read about the new Sigma fp camera on our review page; this story will be about the factory tour.

This was actually the third time I’ve been in this factory; in fact the article I wrote about it back in 2013 was the first really in-depth factory tour article I’d done. You can read that account here: A geek’s tour of Sigma’s Aizu lens factory: Precision production from the inside out. On my second visit a couple of years later, I got the low-down on their lens testing process, using their custom-built “A1” test systems; that story is here: Sigma’s key to quality? Our exclusive look at the A1 system reveals all!

My original tour article went through the entire lens manufacturing process at Aizu, from top to bottom. Sigma’s Aizu factory is one of the most vertically-integrated lens factories in the world; every part of the process apart from raw glass manufacturing and metal casting happens within a single facility. If you’ve ever wondered how lenses are manufactured, you should check it out! (It’s pretty … uh … “detailed”

Most of what’s in the Aizu factory is similar to what was there the last time, so I won’t repeat my previous tour, but just refer you to that earlier article instead.

What is new is that Sigma added an on-site facility for machining and processing magnesium just last Fall (2018). This is a big deal because magnesium literally can catch fire while you’re machining it if you’re not careful! If things do go south, magnesium burns very, very hot. Not only that, but pouring water on a magnesium fire only makes it burn even faster, and even CO2 fire extinguishers just fuel the blaze!

The potential risks involved meant that Sigma hadn’t previously machined magnesium in-house, but as they’ve used more and more of it in their lens barrels, and particularly with the fp camera bodies coming down the pike, it made sense to bring magnesium machining in-house.

Since they’d need more floor space to accommodate this, they decided to house their magnesium processing in a completely new structure, located a short distance from their main factory complex. That way, if something ever went seriously wrong there, they wouldn’t have to worry about it burning down their entire factory complex.

This story is mainly about Sigma’s new magnesium-processing facility, although I also talk some about plastic injection molding, which got rather short treatment in my earlier article.

Read on for an update on the new developments in the Aizu factory, including some shots of fp camera bodies being machined!

(By the way, special thanks to Carey Rose and friendly competitor DPReview for the photos below; I ended up losing all of mine from the tour through a piece of unbelievable stupidity that I won’t bore you with here; let’s just say I’ll never ever format a memory card on a trip, ever again :-0 Be sure to check out Carey’s own coverage of the factory tour.)

Â

Sigma’s Aizu factory has grown incredibly since it was first established in the early 1970s. (Actually, company founder Michihiro Yamaki once told me that the beginning of manufacturing in Aizu was just a few pieces of machinery in individual workers’ homes. Read my original article for the funny story of how that came about

The Sigma factory continues to expand; I saw many more production machines in various areas than the last time I visited a few years back. As mentioned above, they’ve built a whole new building just for magnesium processing, another new one is going up in a former parking lot, and a third is planned for construction by the end of 2020.

The new magnesium-processing facility presents an unimposing demeanor at its main-level entrance, but there’s actually much more to it than meets the eye.

This is a view of the rear of the magnesium-processing building. It’s built into a hillside, so the back is a good three stories tall. When I asked, I was told that the lower level was for chemical processing, for different surface treatments of the magnesium. (I gather it’s akin to anodizing for aluminum.) I don’t know how tall the chemical processing area is, but it either occupies two floors or is two stories high.

At high-enough temperatures, magnesium is extremely flammable in air. Don’t worry about your camera or lens spontaneously igniting, though; the autoignition temperature is 883°F/473°C The thing is, if it isn’t continuously flooded with coolant during machining, the chips can catch fire, and once ignited, it’s extremely difficult to put out. Water only makes matters worse, and even CO₂ fire extinguishers are useless, since magnesium can burn in pure CO₂.

While machinery used for processing magnesium completely floods the workpiece with coolant, there’s still some finite risk of things going awry — and if things go wrong in a factory full of magnesium castings, they could go very wrong indeed. For this reason, Sigma put their new magnesium processing facility in an entirely separate building, located some distance from the rest of the complex.

This shot shows just a portion of the magnesium machining area. The scale is pretty impressive, especially considering the cost of sophisticated CNC machining centers like the ones in the photo. :-0

The magnesium processing building is two (three?) stories, built into the side of a hill. The lower level is a chemical processing area, used for surface treatments akin to anodizing with aluminum.

This shot shows some of the components Sigma produces. Raw castings from outside providers are trimmed to size, holes drilled and tapped, etc. I was very surprised by how much material was removed from some of the pieces. The finished parts were quite a bit lighter than the original castings.

Most parts require multiple operations to reach their final forms, so there are stacks of in-process parts everywhere. This is a batch of lens-barrel components. They’ve already been turned on a lathe, and are now awaiting another process. (In this case, a milling operation on the machine in the next shot below.)

Did I say you need to use a lot of coolant when you’re machining magnesium? Most machining operations use coolant, often as much for lubrication to produce a nice surface finish as for cooling. The magnesium processing machines have intense jets of coolant, often striking the workplace and cutting tool from multiple directions. This is a CNC milling machine, where the part is held stationary while a rotating cutting tool removes material. (In contrast, a lathe rotates the workpiece and a stationary cutting tool shaves off the metal.)

Being able to do their own magnesium processing was an important part of cost control for the new Sigma fp camera, which has a central frame made of that material. This is a before-and-after shot of an fp chassis. If you look closely, you’ll see that the part on the right has a number of holes drilled through it that aren’t present on the raw casting on the left. Some are drilled all the way through; others only partway, with threads tapped to receive screws.

The fp has a magnesium central core, but aluminum front and rear panels. This is a front panel that’s used in the magnesium facility as a template to quickly check whether the chassis pieces are correctly to size or not.

Of course, Sigma doesn’t just rely on test-fitting to ensure things are in tolerance or not; they also use precise measuring instruments like the one in the picture. This is an electronic device, but the factory was virtually littered with conventional micrometers and calipers, and workers seemed to use them fairly regularly to spot-check their machines’ output.

This is the storage area for incoming castings. Given the pace at which the machines seemed to crank out parts, this must be only a day or two’s supply. That would make sense, both for just-in-time inventory control and for the sake of safety.

This is part of the former employee parking lot where the new assembly building shown on the site plan above is being constructed. It’s just at the stage of site preparation now, but will be finished and in operation before the end of next year.

This is back in one of the main factory buildings, and shows just a tiny portion of the number of metalworking machines in the factory. The scale is incredibly impressive.

Here’s an fp front panel after machining. You can see the holes that have been drilled and tapped around the flange mount.

Here’s the largest metal component made in the factory. It’s a lens barrel from Sigma’s enormous 200-500mm f/2.8 zoom lens. (That thing is huge; this is just a part of the barrel for it!)

In my original Sigma factory tour article, I had a shot of the special screw-manufacturing machines that Sigma uses to make custom screws for their lenses. They use standard off-the-shelf hardware wherever possible, but in many cases they need special screws with unusual head profiles or other characteristics that can’t be found in commodity parts.

What I hadn’t appreciated before was just how many different types of screws they need to inventory. The shots above show just a small portion of their screw inventory area. There’s drawer after drawer, packed with bags of tiny, tiny screws. They must have an inventory of hundreds of thousands of individual screws. Meanwhile, each of the banks of machines spits out a fresh screw every few seconds, non-stop.

Injection molding machines

This is an area that’s incredibly important to Sigma’s lens production, but one that got less than its share of attention in my previous tours of the Aizu factory. People often think that “plastic” means “cheap”, but it turns out that plastic has a lot of desirable characteristics, particularly some of the specialized structural plastics which Sigma uses in various places in their products.

On an earlier tour, Sigma CEO Kazuto Yamaki told me that their plastic molding can hold tolerances of just a few microns. That’s pretty incredible, given that a micron is 0.001 millimeter. (About 40 millionths of an inch!)

This time we had a bit more time around the plastic molding area, so I can share a bit more about that process with you here.

Plastic injection molding starts with the molds themselves. These are enormous blocks of stainless steel, usually a multi-layer stack of blocks that have to be carefully assembled before being loaded into the injection-molding machines. The worker here is assembling one such stack. The multiple layers are necessary both for the molding process itself, so plastic can be injected at multiple points, and for the sake of being able to separate and remove the part once it’s cooled and hardened.

When I first toured the factory, I was struck by the fact that injection molds seemed to be just everywhere; Sigma’s huge range of lenses require an even larger number of molds, given that any individual lens may have many individual molded parts making it up. This is a shot of just a tiny, tiny portion of the current mold inventory. Things didn’t seem as over-crowded this time as they were when I first toured; I think Sigma has added more factory space over the year since that first visit. But the storage problem is an enormous one. I don’t know how much one of these mold stacks weighs, but they’ve got to be at least 500 kilograms each. (Stainless steel weighs 8,000 kg per cubic meter; some of the larger mold stacks must be a half a meter on a side, so that’d be 2,000 kg.) :-0

This is a shot of one of the injection machines actually molding a part. I think the metal tubes you see here are for cooling water; the mold needs to cool the melted plastic after it’s injected, so it will solidify. The molten plastic is injected into the mold cavity under high pressure (anywhere from 4,000 to 16,000 psi). The hot plastic is injected at multiple points to insure that the liquid plastic flows into all the parts of the mold cavity evenly. Of course, those thousands of pounds per square inch of pressure are also pressing against the mold itself, trying to force the ends apart. To counter this, injection molding presses have huge hydraulic cylinders pressing the mold parts together. Depending on the size of the parts they’re set up to mold, injection presses can exert many tons of force to hold the mold stacks together.

Once the hot plastic has cooled and hardened, the mold stack separates and the part is either kicked out into a bin, or as shown here, picked up by a robotic arm to be transported and stacked into a parts carrier. The red dot you see here is from a light beam that’s used to make sure that the robotic handler successfully picked up the part.

Here’s a stack of parts that have just come out of a molding machine. I’m amazed by the complexity of parts that can be molded as a single piece. It’s not surprising that they can have a lot of fine detail; that’s pretty much a given with injection molding of plastic. The part I couldn’t figure out was how the heck they got some of these out of the mold once the plastic has solidified. Think about it; anywhere there’s a recess in the part means there’s a corresponding projection on the mold and wherever there’s a projection on the part, that means it’s fitted into a recess in the mold. So the pieces of the mold have to separate out from the part anywhere there are indentations or projections.

I’m honestly baffled by this. I can imagine that some of the pieces of the mold actually retract out from the molded part (the molds after all are huge chunks of solid stainless steel), but even allowing for that, I couldn’t figure out how some parts like the one above can be made. (Are there any readers out there who have experience with injection molding? If so, please chime in with your thoughts in the comments below, and share a little insight into how parts like the one shown above are created.)

Here’s just part of one row of injection molding machines in the Aizu factory. I don’t know how many of these machines Sigma has altogether, but just from what we saw, I’d estimate this is only a quarter to a third of the total. I didn’t think to time it while I was there, but I’d guess that a typical molding cycle takes 15-20 seconds from start to finish. So each machine kicks out several parts per minute, without pausing, for hours at a time. (They also run the molding machines across two shifts of human workers, so they have a lot of capacity to crank out molded parts.)

Summary

Sigma’s Aizu factory is an amazing place, in the most literal sense of the word. As far as I know, it’s the most vertically-integrated lens production facility in the world, incorporating essentially every part of the lens fabrication process, short of making the optical glass and metal casting. I could be wrong, but think the addition of a magnesium-processing facility on-premises is unique in the lens manufacturing world.

This article covers just a tiny part of the overall lens manufacturing process in Aizu; check out my original article about the factory, A geek’s tour of Sigma’s Aizu lens factory: Precision production from the inside out, from almost six years ago, for a complete start-to-finish tour of the entire production process.

Sigma has also invested a lot in final testing and quality control, with their internally-developed A1 testing system. Every single Global Vision lens — that’s their Art, Sports and Contemporary models — is tested as it comes off the production line, to make sure it meets specifications. You can read about that system in my article from 2015, titled Sigma’s key to quality? Our exclusive look at the A1 system reveals all Again, while I don’t know the intimate details of other companies’ final QC testing, as Sigma has given me unparalleled access compared to any other lens maker, I don’t think any other company’s final QC testing of lenses goes as deep as Sigma’s does.

Sigma is a truly unique company in today’s world. They grew from a very modest beginning to a global optical powerhouse, while still remaining essentially a family business. (After all these years and despite their size and global footprint, they’re still privately held.) This translates into a very different corporate personality and ethos than you’ll find in most public companies. I don’t think any other photo company has the kind of relationship with its employees that Sigma does: CEO Kazuto Yamaki and his father before him have gone to extraordinary lengths to keep production in Japan at their Aizu factory, resisting the lure of offshore production and the low labor costs associated with it. I can’t translate it directly into product characteristics, but it’s clear when visiting their facilities that Sigma’s employees appreciate and return that sense of loyalty to the company and the products they create. And before you say “of course, you were touring the facility with the Big Boss”, I can say that that’s also been true of tours I’ve had of other factories. While the loyalty of Japanese workers to their employers is legendary worldwide, the sense of partnership and personal relationship I sense between Sigma’s Aizu workers and the company itself really seems special by any measure.

As with any company, Sigma’s fate ultimately rests on the quality of its products, their pricing, and whether they can stay competitive in the marketplace. Fortunately for Sigma, their employees, and all the photographers who use and enjoy their products, Sigma’s unique approach seems to be accomplishing just that.

Â

The post Sigma Aizu factory update: Magnesium machining and Sigma fp body production! appeared first on Imaging Resource.

I’ve been on a lot of factory tours with various camera and lens manufacturers before, but had never had a chance to see how the optical glass was made that goes into the lenses we use every day. So I was really happy to receive an invite from Nikon to tour their Hikari Glass factory in Akita Japan, following the annual CP+ trade show in Yokohama this year.

This was a pretty special tour, as we got to see the whole process, from start to finish, hosted by three of Hikari’s top executives. Our hosts were Mr. Tatsuo Ishitoya, President-Director, Mr. Akio Arai, Corporate Vice President and Production General Manager, and Mr. Toshihiko Futami, Director and Management General Manager. Mr. Masaru Kobayashi, Assistant Manager of the Administration Section also accompanied us and contributed to the information we received. Arai-san is the person directly responsible for plant operations, and it was him who personally guided us on our extensive tour. All three executives briefed us before and after the tour itself.

Here are three of our hosts, Arai-san on the left, Futami-san in the middle, and Masaru Kobayashi, Assistant Manager of the Administration Section, on the right. It was pretty amazing, to have such high-level executives take us on a tour. I was in engineering-geek heaven; our guides were entirely up to fielding even the most technical questions I asked them. There were of course a lot of things that were too proprietary to share, but the knowledge (not to mention the degree of patience) they brought to the table was truly exceptional.

As I said, I’ve been on a lot of factory tours with various manufacturers, but this qualifies as one of the most interesting ever. I’d previously had only a vague idea of how glass was made; it turns out to be a lot more involved (and more fascinating) than I’d imagined. Here’s the story of our tour…

This the road between the Hikari factory and Motoyu Ryokan, the amount of snow apparently pretty typical for early March. The snow was even deeper, closer to the ryokan.

The Hikari factory is in northern Japan, in Akita Prefecture, just a little south and east of the capital city of Akita itself. Akita is on the western coast of Japan, so the prevailing winds blow for hundreds of miles across the Sea of Japan before hitting the coast, picking up a load of moisture along the way. These winds drop a lot of snow even in Akita proper, but when they hit the mountains, they really cut loose. There was a LOT of snow around the factory, and even more as we wound our way further up into the mountains, to spend the night at the very traditional Motoyu Ryokan, built over a free-flowing natural hot spring.

There were actually a couple of entrances to the factory complex; this one was convenient as we ended our tour. The plant covers quite a large area, with multiple buildings.)

Even though I grew up with serious winters in New England, the amount of snow in Akita was truly impressive. Our host Arai-san told us that this is actually a public road in the summertime. In the winter, the town has to pick their shots in their battle against snow, so this road is left unplowed. That’s a Japanese stop sign in the shot on the left, while the one on the right gives you some idea of how tall the snow piles were, compared to our guides. The two Nikon staff shown in this picture were pretty compact people, but the piles of snow towered over even my head, at 6′ 2

This was one of our first views of the plant, as we started our tour. It wasn’t terribly cold on the day that we were there, but it’s obvious it stays below freezing a lot of the time. The Hikari factory seemed so much more …

The basic recipe: Initial mixing and blending

Optical glass is a complex blend of ingredients, but some representative ingredients are quartz or silicon dioxide (SiO2). If you live near a beach, chances are a lot of the sand is quartz. All of Hikari’s glass begins life as a blend of several basic ingredients, the main one being quartz, although Arai-san politely declined to list what they were. (He also asked that we not take photos of the area where the sacks of ingredients were piled up.)

(Deep geekery: We don’t know the main components of Nikon’s optical glass, but glass generally consists of SiO2, some sort of an alkali flux to lower the melting point, and stabilizers to make it insoluble in water and increase corrosion resistance. Modern glass frequently uses sodium oxide, typically added as sodium carbonate (Na2CO3) and a tiny amount of potash (K2O), added as potassium carbonate (K2CO3) for the flux. Finally, Lime (CaO) and Magnesia (MgO) are added as stabilizers, to increase corrosion resistance.)

It’s important that the ingredients are blended thoroughly, which is the job of a pair of giant mixers like the one shown in the video below.

Ingredients enter the mixer from hoppers on the floor above, via the pipe you can see sticking down from the ceiling. The mixer handles batches of about 500kg at a time, or about 1,100 pounds. Once the ingredients have been delivered to the mixer, the operator sets it rotating for a fixed amount of time. The powdery mix is discharged from the bottom of the V-shaped mixing barrel into plastic bins, to be transported to the melting furnaces, in a nearby part of the facility.

Here’s what the mixed raw-ingredient powder looks like, while waiting to be melted. It’s pretty plain-looking, with a texture somewhere between table sugar and flour.

First Melt

We couldn’t show the furnaces used to melt the mixed powers, as some parts of them were proprietary. (It’s too bad, they were pretty dramatic structures!) The furnaces were quite tall, with steps providing access to a platform for servicing the dosing mechanism.

This was interesting: I’d expected that the melting process would just consist of dumping the mixed power into a crucible of some sort, then shoving the whole thing into a furnace. It turns out that this wouldn’t work very well, as the unmelted powder doesn’t conduct heat very well. So a huge crucible full of it would take a long time to fully melt, working from the outside in.

Instead, they start with the crucible empty, and a mechanism drops small amounts of powder into a metal box on the end of a long mechanical arm. A hatch opens in the side of the furnace, the arm extends and the box flips upside down, to dump a small amount of powder into the hot crucible. This small amount of powder melts relatively quickly, after which the next allotment can be dropped on top of it.

In this way, the crucible gradually fills with molten glass, in a process that Arai-san said can take several hours or so.

(Image from Falorni Glass Furnaces)

The photo on the right has nothing to do with the furnaces at Hikari Glass, but it’s at least some sort of a batch charger, albeit a part of a very high-volume commercial glass production facility (the kind that makes glass for windows, auto windshields, etc). The general idea is the same; a hopper above feeds the glass mix down to the charger, where a bucket slides in and out of the furnace periodically, to deliver doses of the mix into the furnace interior.

The primary melting furnaces were fascinating contraptions, but unfortunately, we weren’t allowed to take photos of them. I found out why this was probably the case, when I went looking for an illustration image to use to break up the text here: It seems to be an unusual arrangement, or at least everyone else who uses it considers their solution proprietary was well: I couldn’t find a photo of a similar dosing or “charging” system anywhere, despite a lot of Google-searching. The image at right was the closest I could come.

The crucibles used for this melting process is made of fused silica, or … quartz! But wait a minute, didn’t we just learn that glass is made from quartz? What keeps the quartz crucible from melting as well?

We weren’t allowed to photograph the crucibles, but they were massive, a good couple of feet in diameter by perhaps three feet long, and with walls more than an inch thick. They looked like they’d be very expensive consumable items! The photo above is from a web page by a maker of high-end kilns for glass and ceramics hobbyists. It gives you the general idea of what the crucibles look like; just imagine something that looked like the above, but was 2-3 feet tall. (Image from Paragon Kilns)

It turns out the quartz crucible does melt with each firing, but only a little bit, and the Hikari Glass engineers take this into account in their formulas for the powdery ingredient mix. They basically assume that they’ll end up with a bit more quartz in their final glass than was present in the mixed powder.

Once the batch of glass has fully melted, a worker melts a hole in the bottom of the crucible, letting the molten glass (~1200C) pour into a large water tank. (~6 x 6 x 4 feet?). Note that once he’s got the glass flowing, he turns on a water jet that sprays across the stream of glass, just as it hits the water surface in the tank. This fractures the glass into tiny, snowflake-like shards, called “frit”. Having the glass in the form of frit helps the next step, of homogenizing the glass that’s been produced.

Here’s a shot of the frit, scooped up in a bucket by the worker running the operation, for us to look at. You can see how it’s in the form of many fine, fractal-looking shards.

I always assumed that optical glass was made by just mixing together the various component, melting it, and pouring it out. It turns out though, that the composition of the glass can vary, depending on where it was in the crucible during melting. Parts that were up against the quartz crucible walls will have more SiO2 in them, and parts near the surface will have less of some more volatile components.

Something else I never knew about glass-making: Some of the compounds used have a higher vapor pressure than others at the melting temperature, so they actually evaporate away during the process. (Hence the need for the exhaust-gas scrubbing equipment shown at the beginning of this article.) So depending on the temperature cycle, you’ll end up with less of some components than you initially mixed in, in parts of the melt that were near the surface.

If these look like cement mixers, it’s because that’s what they are! They’re used to mix batches of frit, to make sure each batch is completely homogenous.

Between the crucible melting slightly each time and the evaporation of more volatile elements near the surface, there can be quite a bit of variation in frit coming from different parts of the melted glass. Consequently, after each melt is completed, the frit is tumbled for a while in a converted cement mixer, to homogenize the mix. The shot above shows two of the three huge mixers we saw in the room. (I’d estimate that the barrels were about 2 meters/6 feet in diameter.)

The problem with a stock cement mixer is the steel from the barrel would contaminate the frit, changing the glass’ properties. Thick rubber liners prevent this from happening.

The problem with an off-the-shelf cement mixer is it has a steel barrel, and steel would contaminate the glass and change its optical properties. To avoid this, Hikari Glass fits them with thick natural-rubber liners, as shown above. Any tiny bits of rubber that abrade off into the frit mix end up burning off in the final melt so they have no effect on the glass itself.

I’ve shown the frit-mixing process as the next step, immediately following the melt and frit-production stage, but there’s actually a step in between, where they melt a sample of the frit into a block of solid glass and measure its optical properties. Depending on where the refractive index of each batch ends up, they’ll combine the output of different melts, to be able to hit the target refractive exactly on the money. (Although it occurs to me that there might be two mixing stages, one to make sure the frit from a given melt is homogenized, then a sample of it is melted and tested, after which frit from different batches is mixed together before the final melt.

This shows a generic three-zone glass melting furnace, similar in general concept to the proprietary and highly specialized ones used by Hikari Glass in their final melting process. This isn’t what a furnace at Hikari Glass looks like, but it gives the general idea of a furnace with three temperature zones in it. (Image from British Glass & not from Hikari Glass)

Once the frit has been made and blended, it’s time for the final or fine melt. The details of this are very proprietary, as it’s the key to obtaining uniform, defect-free optical glass. Arai-san explained that every company makes its own final melting furnaces, and the details of them are very proprietary.

Unlike the initial melt, the final melt is done in platinum(!) crucibles. These must be extremely expensive, although the batch sizes for final melts are usually somewhat smaller than the initial melt. Still, a platinum crucible able to hold a couple hundred kilograms of molten glass must be a pretty pricey item! (In practice, I think they’re platinum-lined rather than solid platinum. Still, they must be very costly!)

The reason they use platinum crucibles for the final melt is because the platinum won’t dissolve into the molten glass and change its characteristics, the way the fused-silica crucibles do that are used for the initial melting.

Despite the use of platinum-lined crucibles, though, the composition of the glass will still change slightly due to the evaporation of some of its components, especially in the central, higher-temperature part of the furnace (see below). So this has to be taken into account, and the mixture adjusted to get the right final result.

Arai-san explains the thermal cycle in their final-melting furnace. The actual temperatures are proprietary and different than those shown, but the general idea is that the glass passes through three temperature zones, an initial melting, a higher-temperature zone, and then into a cool one before final casting.

Bubbles are trouble

One trick in final melting is making sure there are no bubbles in the glass as it’s cast into its final form. A bubble in the middle of a lens element would obviously be a problem, so great care is taken to eliminate them.

I was curious how they did this. I thought they might perhaps use a vacuum furnace, so any bubbles would expand and come to the surface. The actual solution is a bit more clever than that, taking advantage of the natural properties of hot glass.

It turns out that air and other gases dissolve in hot glass, in much the same way that air dissolves in water (which is why fish can breathe underwater; they rely on the dissolved oxygen). As with water, cooler molten glass can hold more dissolved gas than hotter glass can. Hikari Glass takes advantage of this fact to eliminate dissolved gas, with a three-zone temperature profile in their final melting furnace.

(Note: The temperatures shown here are just for the sake of discussion; the actual temperatures are different and proprietary.)

The setup is shown above in the rough diagram Arai-san has drawn on the whiteboard. The temperatures shown aren’t the ones Hikari Glass actually uses, but they serve to illustrate the concepts involved. On the left, glass is initially melted in an input chamber to a temperature of about 1,200C, a similar temperature to that in the first melting crucible we talked about earlier. From there, the glass flows to a second chamber, held at ~1,400C. Because it is so much hotter, dissolved gas is driven out of the glass, into the surrounding atmosphere. Passing out of the high-temperature chamber, the glass flows into a final crucible that’s held at ~1,100C. At this cooler temperature, any bubbles left in the melt from the higher-temperature chamber are dissolved back into the glass, leaving behind perfectly clear, bubble-free glass that’s drained from the bottom of the crucible onto the continuous casting conveyor.

As we’ll see, the process isn’t 100% perfect, because some bubbles and other defects still make it through, and are caught by a subsequent visual inspection.

Casting

The final casting process is pretty amazing; the glass flows very slowly from the bottom of the final melting crucible onto a conveyor belt in a long, long oven, where it’s gradually cooled. The casting process is continuous, lasting until the batch of glass in the final melting furnace is exhausted. Arai-san was deliberately vague about specific details of the final casting process, as it is heavily proprietary.

Here’s a long ribbon of glass, exiting the casting oven. The final melting furnaces are behind us in this shot, up on a second-floor mezzanine level, above the casting ovens. The details of those furnaces are so proprietary that we weren’t allowed within 50 feet or more of them, and couldn’t take any photos facing in that direction. It was kind of amazing to see finished glass creep out of the oven like this, a process that continues 24/7 until the entire batch of glass has been cast.

The glass is cast into ribbons of different widths and thicknesses, depending on the size of the lenses it will eventually be made into. We saw samples from ribbons that ranged from perhaps 125-150mm across and 15mm thick, down to maybe a 50mm across and 6-8mm thick.

At the end of the cooling tunnel, the glass ribbons very slowly inch along, propelled by an open-grid metal conveyor belt. When I asked how long it takes to complete the casting for one batch of glass, I was amazed to hear that it can take anywhere from a couple of days to a full month(!)

At the output end of the final casting line, a worker is waiting to label the glass and break it into 30cm-long chunks. He uses a small hammer and chisel to break off the pieces. (We were a little surprised that something as crude as a hammer and chisel would produce such clean breaks, without danger of cracking the slab into shards.)

Sometimes a slab of glass doesn’t fracture all the way through from the chisel strike, so the worker uses a padded post to complete the break.

Visual inspection for defects

After the strips of glass come off the casting line, they’re inspected visually for defects. This step involves checks for two different types of defect; bubbles and inhomogeneities.

Bubbles are spotted by shining a strong light through the glass, peering through it at a dark background. Even tiny bubbles show up as bright specks within the glass ingot. (If you look very closely at the image above, you can see a few bright points of light within the glass that are the bubbles.) Bubbles are apparently a fairly rare occurrence, thanks to the special design of the final melting furnaces; the sample in the image above was one that Hikari Glass staff selected so we could see the defects clearly.

Once a bubble is identified by looking through the glass lengthwise, the worker turns the ingot 90 degrees and finds and marks each bubble’s x/y position with red marker. This way, the defect-free parts of the ingot can be used, and the parts containing defects discarded.(This is the same sample a shown above, specially selected because it was easy to see the defects in it.)

Inhomogeneities in the glass are more subtle and a bit harder to detect. Once again, a bright light and human eyeballs do the trick.

The other thing to watch out for in optical glass is inhomogenieties caused by changes in the refractive index, resulting from the evaporation of component substances during the high-temperature portion of the three-step thermal processing used to eliminate gas bubbles. (Actually, evaporation occurs in all three thermal stages, but it is obviously most severe at the highest temperature. As mentioned earlier, parts of the melt close to the surface can become depleted of the more volatile components, and if that sort of glass makes its way to the final casting, its refractive index will be different than the rest. (Arai-san didn’t give any but the most basic details of the final melting process, for obvious reasons of proprietary information, but I assume there must be some sort of mixing taking place within the three crucibles involved in the final-melt furnace. If the glass wasn’t mixed, I would think there’d be a lot of homogeneity problems, or they’d have to waste a good portion of each melt, to avoid parts that had lost too much of their volatile components to evaporation.)

With a little Photoshop work, you can see what the inspection worker was looking for in the previous image. Note the curving line across the width of the slab here. That’s an example of “striae”, caused by variations in refractive index.

In the visual inspection, inhomogeneities are found by projecting light through the glass ingot onto a screen, and observing the light/dark patterns as the sample is rotated slightly about its long axis. The telltale optical artifacts are pretty subtle, so we cropped the image and radically adjusted the tone curve to highlight them. You can see the “striae” that the technician is looking for in the image above, as light/dark horizontal lines.

I asked whether glass ingots containing defects could be recycled by re-melting them, and was told that it depends on the type of glass involved. Some can be recycled, but my impression was that most could not. (I wonder if Hikari Glass could earn some additional revenue by selling rejected slabs of glass? I’d certainly pay a fair amount to have one as a keepsake/conversation-starter on my desk!)

Refractive index and light-transmission measurement

It’s probably become clear from the preceding that refractive index is a key parameter that’s controlled very precisely. It’s no surprise then that it would be measured at various points throughout the production process. There’s a separate room with precision optical instruments in it that measure both refractive index and optical transmission (how transparent the glass is).

This is one of the refractive-measuring machines; glass samples are loaded inside, via door on the right side (you can see the grey handle sticking up). This machine measures refractive indices at several different wavelengths, so they can tell the dispersion of the glass as well as its overall refractive index. As mentioned earlier, dispersion is a measure of how much the refractive index changes as a function of light wavelength/color. (If you’re wondering about the wonkiness in the computer screen, it’s because I blurred it in Photoshop, to avoid revealing any proprietary info.

There were two different machines used for measuring refractive index, one somewhat more sophisticated than the other. Both bounce light through a square block of glass, and read-out the refractive index, but one of them measures refractive index at a single wavelength, while the other measures refractive index at several different wavelengths, to also measure dispersion. These refractive-index measurements are performed on small test blocks melted directly from the raw frit we saw earlier, as well as on blocks cut from the continuously-cast glass slabs. The final check on refractive index is performed after the annealing step (see below), to make sure it precisely matches specifications.

This is the business end of the other refraction-measuring instrument, with a block of glass mounted in it. Both of us unfortunately missed getting a more interesting shot with the light shining through the sample :-/ As described throughout this article, glass samples like this may be collected at several points in the production process.

Light transmission measurement

In the same room with the two refractive index measurement instruments was another one that measured light transmission. I’m not sure what would cause glass to pass more or less light, but it’s obviously an important parameter. The transmission-measuring instrument was just a large grey metal box, but there were three sets of glass samples sitting on top of it for us to see.

The problem with measuring light transmission is that light reflects off the surfaces of the glass you’re trying to evaluate. And it’s not just the front surface, where the light first strikes the glass; it’ll reflect internally from the back surface, some of the internally reflected light will then bounce back off the front surface, etc, etc. When you’re looking for very small differences in light transmission, any reflection will disturb the measurement.

Here Arai-san holds a couple of samples prepared for light-transmission measurement. By testing two different thicknesses of the same glass simultaneously, they can cancel-out the effects of reflection, and measure just the light lost to absorption within the glass itself.

The solution turns out to be pretty simple: Just prepare two identical pieces of glass, differing only in their thickness, and measure them both. The surface reflections will be the same between the two samples, so any differences in transmission will be due to the difference in thickness between them.

Three sets of glass samples ready for transmission measurement. The thinner ones are 2mm thick, while the thicker ones are 10mm. The colored marks on them are just for keeping track which batch of glass they’re from.

Cutting into “dice”

After the continuously-cast slabs of glass come off the line and are quality-checked, they’re cut into chunks before being pre-formed into lens shapes. This was another surprise for me, in that the glass is fractured, rather than being cut.

Slabs of glass supported above strip-heaters, waiting to be fractured in two lengthwise. We were surprised by how perfectly clean a cut could be obtained so quickly and easily with this method.

The first step is to split the glass ingots in two lengthwise. (At last I assume they’re always split just into two halves, as that’s what we saw being done. I guess it’s possible larger slabs might be split into thirds, but it looked like the slabs were always sized laterally to be twice the width needed for the final preforms.)

Depending on the thickness of the glass, it takes a little while for each piece to heat up to the point that it’s ready to be fractured. There were two workers doing this task, each running about six stations simultaneously. They were constantly in motion, turning out a pair of glass strips every minute or so. (I’m sure they were also working very hard, with the big boss Arai-san looking on!

The lengthwise splitting was done using thermal shock. The slabs of glass were laid on some sort of temperature-tolerant substrate, with a coil of nichrome resistance wire running along a slot in the middle. The slabs of glass rested over this coil for a matter of a couple of minutes, until the heat from it had had time to work through the thickness of the glass immediately above it. The worker would then just touch one end of the piece of glass with what looked like a pointed wooden stick that had been dipped in water. The sudden shock from contacting the tip of the cool stick would make the glass crack at that end, the crack instantly propagating down the length of the slab. It happened in the blink of an eye, producing smooth, straight edges every time.

Once the glass ingots had been split in two, each half was chopped up into little chunks or “dice”, each approximately the right size to create a lens preform.

Amazingly, the lengths of glass were chopped into smaller pieces using a completely smooth steel “blade”, with no abrasive on it, let alone teeth of any kind. As the workers held the pieces of glass against the spinning blade, friction would heat up the point of contact, producing a clean fracture in just seconds.

Here again, the chopping process was incredibly efficient, and relied on thermal shock to do the work. What looked like saw blades that the workers pressed the strips of glass against were actually smooth steel disks, with no teeth or abrasive on them at all. (Arai-san demonstrated how harmless they were by holding his hand against the edge of a “blade” while it was spinning.)

The video clip above shows a worker chopping the longer strips of glass into dice. It was interesting to watch, you could tell the moment that the fracture first formed, as a little line would suddenly become visible inside the block of glass, and the sound of the blade against the glass would change slightly. Very shortly after, the glass would split cleanly into two pieces.

Rather than using abrasive to cut the glass, pressure against the spinning steel disk produced heat from friction, concentrated at the point of contact. Thermal expansion of the glass in that immediate area resulted in a crack that then propagated almost instantly through the thickness of the block. As you can see in the video above, the process was pretty quick, with no kerf loss, powdered glass or expensive diamond blades required.

The little dice of optical glass were very pretty, glistening and sparkling in the light. Smaller scrap pieces might make nice earrings for female photography geeks

The little glass “dice” sparkled in their trays, thanks to their high refractive index. (The lead glass or “crystal” used in fine-dining glasses and chandeliers sparkles as it does because its high index of refraction bends light more, resulting in more internal reflections. The very high refractive index of cubic zirconia is why that gem sparkles so intensely as well. In fact, the too-high “fire” of CZ relative to diamonds is one give-away that this popular diamond substitute isn’t the real thing.)

Weight-adjusting and rounding

While thermal and friction-cutting are very efficient, there’s some variation, due both to the manual procedures involved, and slight variations in the width and thickness of the original glass ingots.

A row of large vibratory tumblers, used for grinding the glass dice down to final size. You can see a row of large rotary tumblers in the back, and there were a number of much smaller rotary-drum tumblers out of the shot to the right. There was a lot of tumbler capacity in this room; you’re seeing only about a third of it here.

The final lens preforms have to be held to a fairly close weight tolerance, though, so there’s a weight-adjusting step between cutting and pre-forming. This is done by grinding the glass dice in large tumblers, filled with smooth rocks, abrasive and a little water. The video above shows a large vibratory tumbler at work, performing this operation. As its name suggests, a vibratory tumbler uses vibration to grind its contents against each other, with the abrasive grit gradually abrading away the work pieces. These were pretty big units, with barrels that I’d estimate to be 2-3 feet (~70-100cm) across, and perhaps a foot (~30cm) deep. The smooth rocks that seemingly fill the barrel are just there to carry the grit and rub up against the optical glass dice that are being ground down.

Here’s a shot showing a close-up of a tumbler barrel, letting you see a few of the squarish-looking dice that are being ground down, mixed in among the rocks.

Most of the tumblers being used were the vibratory types shown above. There were a number of smaller ones that we don’t have pictures of, that were the more conventional rolling-drum type familiar to rockhounds, often used by hobbyists to smooth and polish colored glass, agates and semi-precious gemstones. (I had a couple of smaller versions of these as a boy, and have a tin of polished amethyst, quartz, jade and tigereye somewhere in the piles of detritus stashed in my basement. Vibratory tumblers were available even then, and were way faster than the drum-type ones that I had, but were priced way beyond my budget.)

Arai-san explains how the raw glass dice are sorted into four weight categories, which are then ground progressively, to bring them all within the necessary weight tolerance for preforming.

Arai-san told us that this tumbling process was used to adjust the weight of the dice prior to preforming, but I didn’t see how just tumbling a bunch of glass dice together would work to homogenize their weights. As I suspected, it turned out that the dice are pre-sorted into weight groups. The heaviest group is loaded into the tumbler first, and once the average weight has reached that of the next-lighter group, that group is added. This process continues until all four weight groups have been added, and the lot of them reduced in weight to bring them within the final size tolerance.

Here’s what the glass dice look like after they’ve been tumble-ground. Note the rounded edges and soft matte-finish.

Post-tumbling QC and repair

The smoothed dice are visually inspected after tumbling, to check for defects. It seems that common faults are chips, where a small fracture along an edge resulted in a chunk flaking out of the die. Provided the defects aren’t too large, some of these chipped dice can be recovered by grinding-out the edges of the chip, as shown in the shot below. (I’d think that this would result in the die involved ending up under-weight, but perhaps there’s enough slack in the tolerance that this sort of post-facto repair can still be applied.)

I was surprised to learn that chipped dice could be repaired to some extent, by manually grinding away the sharp edges of the chip, to prevent them from cracking during the preforming process. You can see the chip on the top edge of the die here, circled in red marker, and with its sharp edges ground into a smoother curve. Apparently, there’s enough room in the weight spec to permit this sort of minor adjustment.

Preforming

Optical glass is delivered to lens companies as “preforms”, chunks of glass having the general shape of the lenses they’re to become. Without thinking much about it, I’d always just assumed that these preforms were made from rectangular chunks of glass by rough-grinding them.

Of course this is a glass factory, though, whose stock in trade is molten glass. So the preforms are of course made by pressing heated, softened glass into rough molds. (I mean duh, right?)

To keep the optical glass from sticking to the preforming molds, it’s first coated with fine boron nitride powder. As you’d expect, this is a dusty operation, as the BN powder seemed to be about the consistency of coarse flour. The floor all around this area was a little slippery, thanks to a thin film of the powder that was ground into the cement. (They obviously kept it well-swept, but the powder settled into the fine pores of the concrete itself, making for a slick surface.

Soft glass can be kind of sticky, though, easily attaching itself to molds or tools. To prevent this from happening, the smoothed dice are covered with boron nitride powder, which acts as a mold release agent. Since preforming is carried out at a relatively low temperature (the glass is only soft, not molten), the BN doesn’t contaminate the glass, and the outer layer containing it is ground away in the first stage of lens grinding.

Preform-pressing looked like something out of the Industrial Revolution, with huge, glowing ovens, open gas flames heating the preform press and molds, and buff, muscled workers laboring stripped to the waist. (Well, actually not the latter, but it certainly wouldn’t have seemed out of place. While some areas of the factory were open to the outside air and quite cold, none of us felt a need for our heavy, Nikon-issued jackets in this section!)

Preform-pressing is done either automatically, by machines, or manually. Since the volume of finished preforms is made in Akita itself is lower than in the sister factory in China, most of the forming done there is manual. (Operational details of the one automated pressing machine we saw working while we were there apparently involved proprietary elements, so we weren’t allowed to photograph the machine in operation.)

The preform pressing process was a closely-orchestrated dance between teams of either two or three workers. There’s apparently some skill involved on the part of the press operator, who needs to judge how long to press each blank, depending on visual cues and monitoring the press operation itself.

As seen in the video above, the manual-pressing operation evoked images of early-industrial metal foundries or the like: Dull-red chunks of glass were flopped into a mold/carrier, and pressed by a pneumatic ram, which was surrounded by gas flames to keep it hot.

Here, Arai-san is showing the start of the preforming process. The tan blocks are ceramic holders that carry the boron-nitride-coated glass dice through the long oven you see the exit end of in the background. Pacing for the whole process is governed by the worker doing the pressing. After each die exits the mouth of the furnace and is handed off to the press operator, a worker puts a new die into the carrier and adds it to the head of the line, shown here. The dice travel in their carrier blocks up a conveyor, get pre-warmed along the way, then turn a corner at the far end of the furnace and proceed back down to the mouth.

Glass dice are placed in ceramic carriers, that cycle up and through a long heat-treating furnace. They exit glowing a dull red and are visibly soft and pliable as they’re dumped into the bottom half of the mold. One worker pulls each softened die from the furnace and dumps it into a mold held by a second worker, who then places it beneath the heated ram, hits a treadle switch to trigger the ram, and then waits a little while, the duration determined by visual cues he’s learned to judge, based on years of experience.

One of the preform-pressing stations was making two smaller preforms at a time. It may not have been clear in the movie or other shots, but lower halves of the molds that were attached to the handles had a gas tube running to them, and a ring of burners around each mold half, to keep it at the necessary temperature at all times.

I asked Arai-san how the press-operator knew how long to apply the forming pressure, and he replied that there were three factors: 1) the softness of the glass, based on its appearance, 2) how the glass feels while pressing it, and 3) how the glass feels when it drops from the mold. I can imagine that it takes a lot of experience, to be able to take all these cues into account, to produce perfectly-pressed preforms!

Here, a worker inspects just-pressed preforms after they’ve exited the press mold.

Once the appropriate amount of time has passed, the press-operator triggers the ram’s retraction, removes the now-preformed lens element and transfers it to a third operator. From there, it seems it briefly goes into an intermediate-temperature furnace, to relieve the worst of its internal stresses, then is transferred to another, longer-cycle oven, where it’s gradually cooled to room temperature.

Annealing

In metallurgy, “annealing” generally refers to a thermal process that reverses the effects of hardening. Annealing can also mean a process that relieves internal stresses caused by too-rapid cooling.

This is the row of huge annealing ovens at Hikari Akita. They’re stacked two-high; the tops of the bottom level are perhaps 6 feet (2 meters) tall. Notes on the front of each furnace tell: 1) What the type of glass is that’s being annealed, 2) What its glass-transition temperature is, 3) What the cooling rate is, and 4/5) two other things I forget :-0

In optical glass manufacturing, though, annealing has a much different purpose, namely adjusting the refractive index. (Stress-relief is important as well, but that would occur with much shorter cooling cycles. The most important function of annealing is to change the refractive index.)

This was the first time that I’d heard that refractive index could be adjusted by thermal processing, so I asked Arai-san how it works. He deferred the question until we could be back in the conference room, with a whiteboard available for him to diagram the process for me.

The shot above shows Arai-san and the diagram he drew for us to explain annealing. The critical temperature involved is Tg, the “glass transition temperature”. In simple terms, this is the point at which glass goes from being a hard, brittle substance to one that’s pliable and can flow. (The full definition of Tg is beyond both the scope of this article and my own understanding, but the preceding is close enough for this discussion.)

In an annealing cycle, the preforms are heated to Tg, held at that temperature for some period of time, and then slowly cooled to some temperature beyond which no further changes in optical characteristics would occur. During this process, the refractive index will change, depending on how quickly or slowly the cooling occurs. The density of the glass is the ultimate controlling factor, and different cooling cycles affect the refractive index because of the influence they exert on density. Slower/longer cooling cycles result in more dense, higher refractive-index glass, while faster/shorter ones produce less-dense, lower refractive-index glass. The annealing cycle needed for each batch of glass is determined by the refractive-index measurements made after the melting process.

The annealing ovens were pretty big, towering over our small group. (I think two or three of us could have comfortably sat inside one, without feeling too claustrophobic. The green panel on the front is a chalkboard, where operators would make notes about annealing temperature, cooling schedule, etc, so anyone could tell at a glance what was going on. This was an empty furnace, so had no notes inscribed on it, and hence was one of the only two we were allowed to take detailed photos of.

It’s not clear to me just why slower cooling cycles result in greater density; the details of that were beyond the scope of questions I was able to ask during the tour.

Take it as given, though, that slower cooling = higher refractive index, and that annealing gives Hikari Glass very fine-grained control over refractive index.

OK, so much about refractive index, but what about dispersion?

I was struck by how much emphasis was placed on fine-tuning refractive index, and how little discussion we had about dispersion. When I asked later, it turned out that this was because dispersion is a much more complicated topic, and a full discussion of it wouldn’t remotely have fit in the time we had. (Especially given how many questions I ask

Dispersion refers to how much the refractive index varies based on color/wavelength. Dispersion is why a prism projects a rainbow from white light; all else being equal, high dispersion means you’ll get a very wide rainbow, low dispersion means you’ll see a much narrower rainbow. So-called “ED” glass is characterized by low dispersion.

This is the standard “Abbe Diagram”, showing dispersion vs refractive index. It looks a bit like a map of the Japanese islands, so optical engineers will often talk about “Hokkaido”, “Tokyo” or “Nagasaki” glass. The different regions labeled on the graph give some idea of how different ingredients affect the glass’ properties, with barium and lanthanum appearing in the names of some glass types.

The image above shows the standard graph of Abbe number (a measure of dispersion, the thing that makes ED glass “ED”) vs. refractive index that will be immediately familiar to any optical engineer. As you can see, there are a lot of different glass types, and this only shows the major categories! As an interesting side note, the general shape of this diagram calls to mind the shape of the Japanese archipelago, so Japanese optical engineers will often refer to a type of glass as a region of Japan. For instance, if an engineer is looking for a glass with low Abbe number but high refractive index (the extreme upper right of the diagram), they’d say they’re looking for a “Hokkaido” glass. (Hokkaido is Japan’s northernmost island.) On the other hand, a glass from the lower left-hand side of the diagram would be referred to as a “Nagasaki” glass. (Nagasaki is the capital of Kyushu province, and located at the far southwestern tip of Japan.)

When I asked about dispersion, it sounded like it’s a fairly basic quality, affected only slightly by process variations. Apparently, dispersion is somehow set by the overall mix of components in the glass recipe, while the overall refractive index is subject to fine-tuning, by mixing different batches of frit, or (seemingly more routinely) by adjusting the annealing cycle. As noted above though, even a basic understanding of how dispersion is controlled would have required far more discussion than we had time for.

(Dispersion really does seem like a very deep subject, I wasn’t able to find much in Google searches beyond dozens upon dozens of pages with the same basic description of what it is, vs how different glass ingredients affect it. I’d really like to learn about it and write up an article on it at some point; maybe I can convince a glass engineer to teach me about it, on another visit to Japan someday

Visual preform QC inspection

After pressing, the lens preforms go through a visual inspection. Using a bright light in fairly dark surroundings, the workers look for any chips, cracks or other flaws. The shot below shows a preform for a binocular prism that has a crack on one side of it. (Hikari makes optical glass for all Nikon products. Camera lenses are a big part of that, but they also make glass for everything from huge semiconductor stepper lenses to microscope lenses to prisms for binoculars.)

After the preforms are pressed, they all go through a visual inspection stage, with a very bright light shone through them against a black background. This will highlight any cracks or imperfections. We have some shots of flawed lens blanks as well, but this photo of a binocular roof prism blank with a crack in it was the best example of how defects might appear.

Packing and shipping

At the end of the whole process, the finished preforms are packed for shipment to Nikon lens factories. It’s a long, complex and fascinating process (at least if you’re techno-geeks like us), and very impressive that Nikon maintains this entire operation, just to satisfy their internal needs for optical glass. (Or at least 90% to satisfy their internal needs; as mentioned earlier, 10% of Hikari’s output is sold on to other manufacturers.)

The finished preforms are packed in vinyl trays, with a number of preforms in each carrier – at least for smaller lenses like these. I imagine that the huge front elements for the like of a 600mm f/4 might call for more robust packaging. Then there’s glass for the enormous lenses inside Nikon’s semiconductor “stepper” exposure systems. Those are so large they must have to be crated individually!

Epilogue: Motoyu Ryokan

It was a really long day by the time we were done; we’d departed from Tokyo’s Haneda Airport fairly early in the morning (at least for a jet-lagged night owl like me), and it was a pretty intense day of asking questions and absorbing information. So we were pretty happy to roll into our overnight digs that evening, especially since it was a pretty long bus ride to get there from the Hikari factory.

Japan is on the Pacific “Rim of Fire”, and many of its mountains are of relatively recent volcanic origin. (Recent in geologic terms, at least.) So there’s a lot of magma close to the surface in many places, and hence a lot of natural hot springs as well. So Onsen (hot spring spas) are a significant cultural thing there, and a lot of ryokan (traditional Japanese inns) are built around them. This was probably the fourth or fifth time I’ve stayed in a traditional ryokan (yes, I’m truly fortunate, and realize it :-), but this was a particularly nice one. I don’t know its history, but it’s apparently a pretty well-known one, and as always was a great experience.

Although I don’t have many photos to show from the dinner that evening, it was an unusually lavish affair, with more than a dozen little courses/small plates, all delivered more or less simultaneously to the table. We also had some of the best sake that I’ve ever tasted. The best was a “raw” sake, meaning it still had live yeast in it. It was one of the tastiest liquids I’ve ever put in my mouth, and I so wanted to bring a couple of bottles home with me. Unfortunately, though, the live yeast meant that it had to be drunk within a week or so of its manufacture, so it would have been past its prime before I even left Japan. (And I had a series of meetings scheduled for a full week following our visit to Akita, so even with my prodigious sake capacity, I wouldn’t have been able to do justice to it from my hotel room in Shibuya :-/ )

Our not-so-humble abode for the evening, Motoyu ryokan.

All in all, our tour of Nikon’s Hikari glass factory was an extraordinary experience. As I said at the beginning, it was easily one of the most interesting factory tours I’ve been on, and that covers a lot of ground. It was certainly a tour in which I learned an incredible amount, about things that I had no knowledge of previously.

Nikon obviously wanted us to draw from the experience the impression that they have a unique ability among camera manufacturers, in that they exercise an unparalleled amount of control over the quality of the most basic material that goes into their lenses, namely the glass itself. Even allowing for the obvious PR intent for the trip, though, we came away very impressed with just that: Nikon really does have a unique ability to control their own destiny and optical designs, all the way from the raw materials to their finished lenses.

Entirely apart from the intended PR message, this was a remarkably informative tour, that left us knowing far more than we did before it began. Many thanks to Nikon Tokyo and Hikari Glass for their hospitality and patience, in answering all our many (many!) questions!

A finished lens blank from the Hikari factory, ready to be polished and built into an amazing Nikkor lens. (We know it’s destined for an amazing lens, because of how big this blank is; this is probably the front element for a long, large-aperture tele

Thanks for reading! See more of our recent factory tours by clicking here.

The post Glass for geeks: An in-depth tour of Nikon’s Hikari Glass factory appeared first on Imaging Resource.

Back in early August, I had the very rare opportunity to tour Sony’s sensor factory in Kumamoto, Japan. We had only limited access to the factory itself, but I learned a lot about Sony’s sensor technology and fabrication process while I was there, and will be sharing more on the topic in other forthcoming articles. In this first piece, we’ll look at a few of the intriguing parts of the sensor production process.

We weren’t allowed to take photos ourselves of the factory floor, given the extremely proprietary nature of the wafer processing. I can understand the need for such caution, given my own time in the semiconductor industry, albeit many years ago. Even seemingly innocuous details in a photo could reveal important aspects of Sony’s process to competing production engineers who know what to look for. Instead, Sony provided us with a collection of their own photos, that had been carefully screened and cleared for public dissemination.

Even with the limited access, it was a very interesting glimpse into how the sensors in so many of our cameras are made. (Sony holds a dominant position in the global sensor market. There’s a better than 50% chance that the sensor in your current camera came out of this factory.)

This is where the whole process begins, outside of Sony’s factory at a company that grows the raw silicon. “Grows” is a literal description, because the silicon wafers are sliced from massive, single crystals of silicon.

These large lumps of silicon are called boules, and at the current state of the art are absolutely enormous. The finished wafers are about 12 inches (300 mm) in diameter, by perhaps 4 to 6 feet long (1.4 – 2 meters). The large boules are produced by melting a large amount of very pure silicon, holding it very close to its freezing (crystallization) point, and then dipping a small, precisely-oriented seed crystal into the top of the melt.

With careful control of the temperature, the molten silicon will begin crystallizing around the seed crystal, growing a single crystal with the same orientation as the seed. The seed crystal is slowly pulled up and away from the melt, while being rotated to help maintain uniformity as the crystal grows.

This technique is called the Czochralski process after the Polish scientist Jan Czochralski, who invented it in 1915, and is how most semiconductor-grade silicon is made today. The process is much the same as you might have observed when crystallizing sugar to make rock candy. There, multiple nucleation sites make for a multitude of crystals; when growing silicon, a single seed crystal and careful control results in just one monster crystal.

Silicon wafer sizes have gotten a lot larger over time, because larger wafers make for more efficient processing and thus cheaper chips. In the early 1970s, 3-inch wafers were standard, now 12-inch (300mm) wafers are commonplace, and manufacturers continue working to increase sizes even further. At less than a millimeter thick, today’s wafers are pretty fragile, requiring careful handling.

All this happens at the silicon supplier, who then slices the boules into thin wafers, again unspecified by Sony, but the standard thickness for a 300mm wafer is 0.775mm. (This gives you some idea of how carefully they have to be handled. Imagine a 12-inch disk of brittle material, less than 1mm thick. Oh – and you can only grab it by the extreme edges!)

The huge (roughly 12 inches in diameter) silicon wafers arrive at Sony’s Kumamoto factory in special carrier boxes that protect them during transit from the silicon provider, avoiding breakage and keeping them reasonably clean relative to the surroundings.

“Reasonably clean” isn’t anywhere near clean enough to be allowed to enter the factory environment, though, so the first step in their journey is the liquid cleaning station shown above. Here, they’re removed from the incoming carrier box, and placed on a spinning stage. As the wafer spins, high-purity water is squirted onto it, and a rotating scrubber moves from the center to the edge and back again. When done, the scrubber retracts, a final jet of water is applied, and then the wafer is spun at high speed to sling all the water off the edges, leaving a pristine surface.

A mechanical handler then picks up the wafer and transfers it to a different, ultra-clean carrier box that is used to carry batches of wafers from one station to the next, via the factory’s automated material-handling system.

Circuit patterns are made on the wafers using a photolithography process. This happens dozens of times throughout the sensors’ production cycle, to define patterns for different parts of the process. In some cases, the patterns control where “dopant” atoms will go, to turn the raw material into N- and P-type silicon. Sony’s sensors have quite complex structures, and each structure requires its own photolithography step.

In photolithography, the wafers are coated with a thin film of photo-sensitive material known as a “photoresist”. An image of the desired pattern is projected onto the wafer’s surface and the image is developed, leaving some areas covered by the resist and others clear. Interestingly, there are both “positive” and “negative” resists, meaning that the exposed portion either stays or is removed during the development process. Way back in semiconductor pre-history when I was in the IC business, positive resists had higher resolution, but I have no idea what resist technology is like these days. Sometimes the resist acts as a mask for the ion-implantation process (see below), while other times, the resist is applied over a continuous metal or layer, with the resist patterns determining what metal/oxide will be etched away and what will be left behind.

The first step in the photolithography process is to apply the photoresist in an incredibly thin, uniform layer, all across the wafer’s surface. That’s what’s being done above, and if the setup looks a lot like the liquid cleaning station we saw first, it’s because the main piece of equipment is the same sort of spinning stage as we encountered there. Here, a small amount of liquid photoresist is applied to the center of the wafer, which is then spun at very high speed to produce a microscopically-thin film.

This is a view down one aisle of the Kumamoto factory’s massive clean room area. We don’t have a number from Sony for the total square footage of clean rooms, but the factory buildings are massive, and two entire floors are clean room space.

That’s not all, though, because it takes three other floors to support one clean room level. An entire floor above the clean room is dedicated to air supply, and an entire floor below the clean room is air return, to be cleaned and recirculated again. Below the air-return floor, on the first floor of the building is a machinery level, with pumps, water filtering, etc, to support the clean rooms above. (I’m not sure, but I think that the first-floor machinery supports both the 3rd and 6th-floor clean room areas.)

It’s no surprise that clean rooms are all about dust control. The main clean room area here is actually fairly “dirty” as such standards go, being a Class 1000 environment. That means there could be up 1000 particles larger than 0.5 microns per cubic foot of air. That sounds like a lot, but it’s about 35,000 times cleaner than typical city air.

The really clean areas are inside the processing machines themselves, and inside the so-called FOUP wafer-carrying boxes that are automatically carried throughout the plant via a track system hanging from the ceiling. These smaller areas are rated Class 1, meaning no more than 1 particle larger than 0.5 microns in size per cubic foot of air.

I’m not sure, but I think that this particular aisle is one that we looked down when we were at the factory, and it contains ion implanters, diffusion furnaces and photolithography systems. Basically, it’s an area where the dopant atoms are implanted in the silicon to make electrically-active areas with different characteristics.

Here’s a shot of another clean room area; I think this one is mainly photolithography, or perhaps etching of some sort. You can see one of the wafer-carrying robots in this shot, protecting a load of wafers in their ultra-clean internal environment as they’re transported between machines.

This shot and the one above are actually a little unusual, in that there are people present. The Kumamoto factory is almost completely automated, with humans rarely even being in proximity to the wafers. The factory workers’ duties seemed to be almost entirely limited to servicing, tweaking and repairing the production equipment itself. Other than that, most operations in the factory run 24/7 without human intervention.

The extreme level of automation is key to keeping costs down and production high. Humans are problematic in semiconductor production. Even outfitted in “bunny suits”, they’re incredibly dirty, shedding dust and skin cells everywhere. They also get tired, get sick, have bad days and just by their nature make mistakes from time to time, no matter how careful they are. Oh, and you have to pay them, give them vacations and benefits, etc, etc.

By contrast, the robot assembly line cranks day and night, never gets tired or sick, and always does things exactly, precisely the same way. (Thanks, robots, for our incredibly affordable sensors, relative to what they can do!

This shot clearly shows a FOUP wafer container being carried by one of the overhead robotic trolley cars. When a trolley reaches its destination machine, the FOUP is lowered down on what looked like steel rods, but which obviously must have been telescoping tubes, to have fit within the body of the carrier robot. It was surprising to see how far down the FOUPs were lowered; depending on the machine involved, I’d estimate that they were lowered down a good 6-8 feet below the trolley car. It seemed a little precarious, but I guess with no clumsy humans around to blunder into a dangling FOUP, the setup didn’t need to be especially robust.

During the huge Kumamoto earthquake, many FOUP-trolleys and the FOUPs themselves came crashing to the floor, and in some areas, the entire ceiling collapsed. It’s almost unbelievable that the entire factory restarted for limited production in just a month and a half, and was back to full operation just seven months after the quake.

You can also see the alternating grid of air outlets, covering the entire ceiling of the cleanroom area. As mentioned earlier, the entire floor above the cleanroom is dedicated to air supply and filtering. Given the area involved and the need to main a continuous flow of ultra-clean air, the total airflow must be absolutely enormous.

You can also clearly see the grid of holes in the floor that the air exits through. Any dust particles kicked up from the floor the few times humans are present will just be sucked right down into the air-return system.

I don’t know what’s going on in this shot, as Sony was extremely guarded and cryptic about anything having to do with the production details. (Although I suspect any semiconductor process engineer would immediately recognize these machines for whatever they are.) Sony just described this area as being for the “photolithography process”.

Once the sensor chips have been fabricated, they’re subjected to wafer-level tests, then sliced into individual chips. This is called “dicing”, and because silicon surface area is such a precious commodity, it’s done with extremely thin-kerf diamond saws. (The “kerf” is the width of the slit they leave behind.) The specs for these kinds of saws and blades are pretty astonishing. A typical blade might be only 20-30 microns thick (0.02-0.03 mm or just 0.008-0.012 inches), and spin at 30,000 – 60,000 rpm to maintain stiffness and minimize silicon damage. Multiple nozzles spray high-purity coolant at the blade and wafer during cutting, to keep the saw and wafer cool, and also to carry away the microscopic silicon particles that are cast off.

For ordinary sensors (that is, ones which aren’t stacked with memory/processor chips the way that the ones in cameras like the Sony A9 and RX100 V are), the next step is to mount them in packages and then bond incredibly fine wires to them, electrically connecting the contact pads on the chip to the pins of the package. The shot above shows a wire-bonding machine in action, but it hardly gives an appreciation for how fast they are. Time is money, so the more chips per hour a wire-bonding machine can process, the better.

We don’t have any details on the type of wire-bonding machines Sony uses in their Kumamoto plant, but the video above gives you some idea of just how quickly they can move. Besides the direct point-to-point wiring the machine in the video is doing, some applications and packages require the machines to perform a complex dance after the wire is initially bonded to the chip, bending it into the right shape to clear chip or package geometry before bonding it to the package. Keep in mind, too, that even at these incredible speeds, the wire bonder needs to hit the pads on the chip with an accuracy on the order of 0.001 inches (~50 microns or less). The combination of temperature and pressure the machine uses to make the bonds also has to be very tightly controlled; even a failure rate of 0.001% would be disastrous for sensor yields.

This shot shows a loooong double-row of Sony’s in-line sensor assembly machines. Processed sensor chips and empty chip packages go in one end, and finish-packaged sensors come out the other.

There are a number of steps involved in packaging sensor chips, all done within these integrated systems. Once the wafers have been diced, each sensor die is glued (yes, glued) into a package, using four drops of special resin. (Some kind of epoxy, I think?) Next, it’s connected to the pins of the package via the wire-bonding step just described above. Then the package is filled with sealing resin, a cover glass is placed atop it, and the sealing resin is cured with a blast of ultraviolet light. Then a laser-engraving system marks part numbers, etc., onto the backs of the packages, they’re sent for a final image-quality test, packed in shipping containers and warehoused before shipping.

The first-stage tests performed at a wafer level checked for raw electrical characteristics, while the second test after packaging verifies each sensor’s imaging capability. I don’t know how long the imaging test takes, but we were shown a huge room that was filled with second-stage image testing machines. Once again, we weren’t allowed to share any photos, but the sheer scale of Sony’s Kumamoto plant is mind-boggling.

Imaging parameters that are tested include dark current (basically, leakage current in the sensor’s photodiodes that contributes to fixed-pattern noise in very dark conditions), low light performance, high-brightness performance and overflow handling (checking how the sensor behaves in the face of severe light overloads). And of course, they check for dead or stuck-high pixels as well.

Conclusion (for now)

That’s it for this mini-tour of Sony’s sensor production line. Next up will be a similar look at the production process for micro-OLED displays, of the sort used in the electronic viewfinders of cameras like the Sony A9, which are also produced in this factory. It’s coming up on crazy-season for me here at IR, but I hope to also write a more detailed description of the processes involved in making a sensor. So many parts of it are familiar to me from my own time in the semiconductor business, even though that was decades ago now. Today’s technology is enormously more advanced, but the basic steps are pretty much the same, and I’d love to be able to share some of the magic of how it all happens.

Meanwhile, do you have any questions about any of the above? Leave them in the comments below, and I’ll try to circle back over the next few days to answer any that I can.

The post Inside Sony Kumamoto: A rare glimpse inside Sony’s super-secret sensor factory appeared first on Imaging Resource.

Topics for discussion included the differences between Sony’s strategy and that of its rivals; the future of interchangeable-lens cameras and whether we might one day see full-frame and sub-frame cameras with similar pricing; the difficult problem of competing not just with rivals but also with one’s own camera models from past years; possible areas for improvement in future interchangeable-lens cameras; and the surprising rise of the Chinese market for full-frame cameras, among others.

Without any further ado, let’s get down to the interview!

Dave Etchells/Imaging Resource: Thank you so much for meeting with me, Tanaka-san. We heard in a presentation yesterday about Sony’s strategy of innovation. How do you apply that to your own business in a concrete way? Are you allocating more money to R&D than other companies, do you think?

Kenji Tanaka
Senior General Manager
Business Unit 1
Digital Imaging Group
Imaging Products and Solutions Sector
Sony Corp.

Kenji Tanaka/Sony: As you know well, our key driver is the image sensor, and we already invested a lot of money for the image sensor development. And the sensor is a custom [design, meaning that] only Sony can use these sensors, and our strength is our in-house technology. So I invested in that and we will keep investing in the in-house technology like image sensors.

[An extended editorial note from Dave Etchells: This has been the subject of a *lot* of misinformation, misinterpretation and speculation on the ‘web, so let me attempt to set the record straight…

This at first glance was new information for me; since as far as I had been aware, other manufacturers always eventually got access to the same sensors that were used in Sony cameras. Instead, Tanaka-san was saying that there would be some developments that were co-developed by Sony’s camera division, and that those would of course remain proprietary.

(I’d initially said that I thought there was a time delay involved, between when Sony could use the most advanced sensors in their own cameras vs when competing camera makers had access to the tech, but they informed me that there was no public position to that effect, and particularly no timeline for that sort of thing. – I’d said that other makers got access to the most advanced sensors a year after the Sony camera engineers did, but there appears to be no such official timeline.)

What clearly seems to be the case, though, is that Sony camera engineers are able to work with Sony’s sensor engineers, to develop unique system architectures that take advantage of their unique integration capabilities, when creating their own cameras.

It’s important to note that this isn’t in any way a matter of holding back “higher-quality” sensors to disadvantage competitors. When it comes to quantum efficiency and fundamental noise levels, we’re up against fundamental physical limits, with all the manufacturers. The limiting factor in extremely high-ISO noise levels isn’t an issue with the silicon or a particular fab process, it’s the “shot noise” that results from random variations in how many photons arrive at each pixel in a given time. If Sony were to somehow dial back specs in the sensors they offer competing camera manufacturers, other sensor makers would rapidly step in to fill that gap.

Instead, Sony’s advantage is that their camera engineers can collaborate directly with their sensor-engineering team, to jointly develop new system architectures that fully leverage their exceptional integration technology.

Completely apart from this conversation with Tanaka-san, I’ve long been aware that Sony has integration technology driven by the smartphone market, that goes well beyond commodity sensor tech in the camera space. In the camera world, it takes significant changes in system architecture to take advantage of this. We’ve recently seen Sony’s collaborative advantage in this area with the A99 II and A6500. (Read my interview with Kimio Maki and Takashi Kondo from Photokina 2016, for more about their “front-end LSI” technology that’s deployed in those cameras.)

Recently, Sony’s gone well beyond what we saw in the A99 II and A6500, though, with an announcement of three levels of circuit-stacking in the smartphone sector, now incorporating massive amounts of DRAM directly integrated with the sensor chip and logic. In the small-sensor world, this now allows full-HD resolution video capture at 1,000 frames/second, and full-resolution still image capture at 19.3 million pixels per frame, at 120 fps. (The announcement linked above not only discloses 1,000 fps FHD recording, but also the ability to intelligently switch between normal recording and super slow-mo, when the sensor system detects rapid subject motion.)

So let’s be clear: Sony isn’t somehow holding back their “good” sensors for use in their own cameras; it’s instead more about taking advantage of the fact that their camera and sensor engineers are able to collaborate with each other, developing new camera architectures that take advantage of their unique integration capabilities. This isn’t news, it’s just that current advances mean that overall system architecture plays a bigger role than it did in the past.]

DE: So Sony’s invested billions of yen in sensors…

KT: Yes.

DE: …that applies to many areas, and now for cameras, you can draw on that. So it’s sort of like that investment has already been made or is being made, and you can draw on that and have proprietary sensors just for Sony’s cameras.

KT: Mmm-hmm.

DE: Yes.

KT: And the sensor is one aspect. Another one is the lens, yeah? And the same as the sensor, we invested a lot of money for the development of our lenses. So the point is that we have a key technology, so we invest in-house. We don’t just think to buy something [from another company].

DE: Mmm, mmm. You want to develop all of the capability in-house. So I would think that as a percentage of revenue, you probably spend more on development than most companies do?

KT: I think so…

DE: Maybe some of it is the direction in which you’re allocating resources. I admit to being surprised that the enthusiast and pro section of the camera market is holding steady. And I’m not sure if that was units or revenue…

KT: Revenue.

DE: It was revenue, yeah. What do you think has been contributing to that strength in that segment?

KT: This is quite a difficult question. I think visualizing an attractive image is a kind of a human instinct, so that kind of instinct never disappears. So in the sense that the full-frame market is quite stable and never disappears, that is because many enthusiast wants to create something beautiful, something high quality. So that [full-frame market stability is thanks to] the human eye, human nature I think, you know?

One of Sony’s image sensors. The company sees its sensor technology as one of its key strengths, allowing it to differentiate itself from rivals with custom, Sony-only sensor designs.

DE: Mmm. And so regardless of other technology, there’s always people that have the personal passion for imaging. And then as you say, the full-frame is kind of the ultimate expression of that.

KT: Yeah.

DE: Do you think we will see growth in that segment, or is it just a certain percentage of society that is that way?

KT: I want to increase this segment, but you know, globally speaking I think that this segment is stable. If the segment increases, [it will do so] very slowly. In the US, we expect it to grow 10% this year.

DE: One thing that I see all of the manufacturers facing is that the biggest competitive problem isn’t necessarily the other manufacturers, it’s everybody’s cameras that they’ve already made. I call it the sufficiency problem, meaning that for a lot of users, a three or four or five-year old camera is just fine.

KT: That is enough, yeah.

DE: And as the technology continues to advance, each new year’s cameras are being that much more of a problem for the cameras that follow them. And so the answer’s obviously to innovate even more, but what do you see as being the key area in which to innovate? You know, if people have a 24-megapixel camera, do they really need 50 megapixels, or if they have 24 frames per second do they really need 60 fps? You know, at some point you reach a level that it’s OK, where people don’t really need more in a given market segment.

KT: First of all, please think back to the 4K TV, yes? The initial attempt at the beginning, everybody said the HD, 1080 is OK. We don’t need 4K. But right now, people want to have a 4K TV. So the demand looked mature, but technology [overrode it], overachieved the demand. But that kind of mindset change happens gradually. But if people look at the situation, some day will [be] today’s 4K TV era. So kind of the same thing has happened to the ILC I think. Before 42 megapixels… nobody needed such a resolution…

DE: Ah, yeah.

The 42-megapixel Sony A7R II and its sibling the A99 II offer resolution that would have been unheard of not so many years ago.

KT: …but right now, people enjoy 42-megapixel cameras. So our mission is to envision new innovations, new experiences for the people. If the experience is beyond the users’ imagination, they [will] want to have it.

DE: Mmm, mmm-hmm.

KT: So the point is how to go beyond the imagination.

DE: So where do you think the areas are that are ripe for innovation? We have 42-megapixel cameras, and people now think “Oh!”, they would like to have 42 megapixel. But do we go to 100 megapixel, and then will people really want that? Or what are the other areas where there’s an opportunity to exceed expectations and then create demand?

KT: Now we are focused on three essential factors. One is of course the resolution, and another is the sensitivity, and third is the speed. [For example] 20 frames per second and 30 frames per burst or 100 frames per second, everybody wants to [have] that kind of functionality, but I can’t say yes. Because they don’t know how to use this kind of technology, but if we can show the benefit for that, people [will] enjoy the new technology, I think.

DE: Mmm, yeah. So once people see 100 frames per second or 120 and slow motion or whatever, that that will create new demand for that.

KT: I’m talking about the professional sports shooter. For example they are shooting tennis or badminton, they are the kind of professional targeting the one millisecond [in time that they want to capture].

DE: Mmm.

KT: [That] basically means we need more frames per second continuous shooting.

DE: Mmm, yeah.

Thanks to its Translucent Mirror design, the Sony A99 II allows for impressively swift 12 frames per second capture with autofocus and autoexposure adjustments between frames. But that could one day be eclipsed: Sony is now aiming for even greater performance from its mirrorless cameras using intelligent sensor technology, it seems.

KT: So for that kind of performance, the hardware demands are very high. But some day, we can do that.

DE: Yeah, interesting. One area that seems very important to me is autofocus, and it feels like that’s an area where there’s still a lot of room to rise to meet not even expectations, just needs for consumers. But it’s also extremely complex, and looking at it from our end it’s very difficult to test. There are so many different scenarios that you might have, and so then it’s, it’s difficult to communicate actual progress that you make. You can say “This camera’s better than our last one”, but how do you communicate significant improvements in autofocus to consumers without it just being kind of empty branding, and giving it a new market name?

KT: First of all, I agree with your opinion that autofocus is very important, and we have room to improve this technology. And one classic area for potential improvement is speed. Another thing is how to integrate intelligence.

DE: Yeah, intelligence for subject recognition.

KT: Yeah, for example we already provide the eye focus function. That is one example of intelligence. Many customers find it difficult to focus because they want to focus on the eye, or focus on some other object, so we need to solve this complexity, we need to improve. So if we are to solve this very difficult problem, the image sensor is very important. Because Canon and Nikon have a separate AF sensor, that kind of structure cannot use intelligence. [With a] combined AF sensor and the image sensor, that kind of innovation can use the intelligence. So we’re now developing that kind of a thing. So some day I think that if the user wants to focus on the hand, your eye, your computer, or… That kind of function.

DE: It’s like object-recognition for the camera to be able to “see”. Like I was just watching two boats going past on the river; my eyes have no problem realizing this is the boat on the back, and this is the one on the front.

KT: Because you are thinking.

DE: I’m thinking, exactly, yeah. And you mentioned with the sensor too, in the human visual system there are many layers of neurons that figure out, see different things. And so I’ve felt that intelligence is really the next big thing for computers. But very difficult, too. It takes a lot of very advanced processing, and it probably takes different sensor structures.

KT: Not only on autofocus, I think for example the autoexposure [could benefit from intelligence].

DE: Mmm, yes.

KT: With the intelligence we can create something new, but you know, many people say it happens from here, from image sensors, I think, yeah.

Yojiro Joe Asai
Senior Manager
Area Marketing Section 2, Department 1
Marketing Division
Digital Imaging Group
Imaging Products and Solutions Sector
Sony Corp.

Joe Asai/Sony: I think that Sony has been repeatedly saying about the advantage of mirrorless, is that not having a mirror, having direct information…

DE: Mmm, the fact that it’s, it’s all integrated immediately.

JA: Yes.

DE: Yeah. And one limitation early on was just to get the data off of the sensor quickly enough to be able to do autofocus cycles. But now with the stacked sensors and local memory and local processing, you can get the data through that quickly, so now you can realize the benefit of having the integration of distance and image information.

KT: You’re right.

DE: I’m very excited, I think we’re going to see some really interesting things, and you’ve hinted some that you’re working on some very advanced things in that area. I think there’s a lot of exciting things coming.

And so another big message we heard yesterday was that you don’t see your goal as growing by taking a bigger share of the same pie, but rather by expanding the pie itself by drawing in new ILC photographers. And you showed some very impressive spikes in your market share, like when the A7R Mark II came out it was a big jump. But that’s a very high-end model, it’s not something that new people coming into the market would be buying. Do you have any stats on how many A7R II users were actually new to ILCs, or you view it more that sub-frame products, A6500, A5100, that kind of product are growing the market?

KT: In the case of our Alpha 7R Mark II, honestly speaking that spike comes from just a shift from the other ILC area.

DE: That’s from other ILC products, yeah.

KT: So in that case, it’s not growing the market, just a shift. So we are thinking how to invite the new customer into our market in a different way?

DE: Yeah. I have said, too, that amateurs, people who don’t consider themselves photographers, like moms and dads, they need more advanced technology than the pros, because they’re the ones who more than anything need smart cameras and intelligent sensors. So maybe that will come. Related to that, looking at the full-frame format we saw a lot of emphasis in the presentations on your strength in the full-frame market. I’m sure you don’t intend to ignore the sub-frame market, but is full-frame where you really see the main focus of your efforts? Would that be a primary direction?

KT: Yes. Full-frame is our primary area. But as you said, we don’t ignore the APS-C market, and we can also grow the APS-C market.

DE: Mmm, yeah, it seems APS-C is where you would be able to draw in new people, potentially.

KT: And especially the younger people, I think [they would] easily jump into the APS-C market.

Sony’s sub-frame cameras such as the A6500 (shown here) have proven to be extremely popular thanks to a combination of great image quality, solid performance and affordable pricing.

DE: I was surprised to learn that China and the US together account for more than 50% of the full-frame market, and I was astonished that China is actually twice the market size of the US. That’s just amazing to me. Has that phenomenon of China accounting for so much of the full-frame market been something that developed recently, or have you seen that from the beginning of your entry into the market?

KT: I don’t know Canon and Nikon’s data, but in the case of Sony, these last three years, things changed.

DE: Hmm! Just in the last three years. So it used to be that the US was buying more full-frame, and now it’s switched?

KT: Yes.

DE: Wow.

KT: And so after we launched the Alpha 7-series, I think things gradually changed, and of course [with the current] economic situation the Chinese people can [now afford to] buy valuable, full-frame cameras, so many factors are related to this phenomenon, I think.

DE: Yeah, yesterday Asai-san explained that within China, that consumers will actually consider a sub-frame A5100 or a full-frame A7R and ask “Which one should I buy?”. That’s really kind of amazing. So part of it is that the difference in price between them isn’t an issue. As technology advances and full-frame prices come down, do you think we will start to see that kind of behavior in the US, where people will actually be actively considering a full-frame and an APS-C camera on an equal basis?

JA: Maybe not in the same way as we’ve been seeing in the Chinese market, because I still think that customers in America will probably want to know what they want first. I think value for money is on a higher concern amongst American customers, so I think buying behavior won’t drastically change. I think it’ll be stable but it probably won’t grow rapidly from customers changing their purchase behavior. That’s kind of my personal view on this.

DE: As sensor manufacturing becomes more and more efficient, will the cost difference between full-frame and APS-C decrease over time, do you think? I’m thinking that the percentage of cost that the sensor accounts for will decrease, and the mechanical construction is going to be a similar price [to what it is currently], so I’m wondering if we will see the cost of full-frame cameras approaching [those of sub-frame cameras]. What do you think the cost difference between the two different formats might be five years from now?

KT: For example, the mirror for the full-frame and the mirror for the APS-C, the cost is actually different, and other mechanical devices are totally different. But in the future if there’s no mechanical parts, I think that the price could be getting close, that is the truth. But the customer is different.

DE: Mmm, mmm!

KT: So to satisfy the customer, many things like the body or durability could cost more on the full-frame camera, so that the customers’ demand cost [will] still exist, I think.

DE: Mmm, there will always be some people who will want a sub-frame or that full-frame owners will want more rugged than APS-C.

We’re coming to the end of our time together, and I want to give you a chance to stand on your soap box (that’s an American phrase for having a prominent place in the public square, to talk to people). What’s the one message that you would most like to communicate to Imaging Resource’s enthusiast readers? What’s the one thing you would want to tell them about Sony and about what you’re doing?

KT: I want to show our technology. The technology can change the future. Right now, I think American people still think the DSLR structure with mirror and shutter is best. But the technology can change that kind of way. And I think only Sony can do it. So I want to explain why the future is changing. And sensor is one aspect, and the lens is another one. Of course, intelligence is another aspect, but there are many technologies we have, so these things I want to tell your users. And of course [by combining] these technologies we create new cameras. So I want to show you the new camera.

DE: Yeah, yeah. So people have an automatic assumption that a professional camera, a high-end camera, is an SLR. And your message is that, no, actually, in the future higher-end means mirrorless.

KT: Yeah. And I think a mirrorless can take a photo that the even professional DSLRs couldn’t capture. So we want to give a new experience not only [for] the amateur [but also] the professional.

DE: A great summary. I see that we are exactly out of time, so I’ll let you go to your next appointment. Thank you very much for spending the time to talk with me!

KT: Thank you!

JA: Thank you!

The post Sony Thailand factory tour Q&A: Mapping out the future of the interchangeable-lens camera (updated) appeared first on Imaging Resource.

Shortly after the completion of this year’s installment of the annual CP+ tradeshow in Yokohama, Japan, IR founder and publisher Dave Etchells and Senior Editor William Brawley headed a couple of hours north to Utsunomiya, the capital of the Tochigi Prefecture, and home to Canon’s primary lens factory. The reason: Both gents were fortunate to be able to tour the facility for a behind-the-scenes look at how Canon lenses are designed and manufactured. But that wasn’t all — the cherry on top of the visit was an opportunity to sit down for a Q&A session with senior executives from the company to pick their brains about the nuances of lens design and manufacture, and how cutting-edge technology promises to bring us even better glass with yet more affordable pricing.

For the session, Dave and William were joined by Canon Inc.’s Masato Okada, Executive Officer and Deputy Chief Executive of Image Communication Product Operations; Shingo Hayakawa, Deputy Group Executive, ICB Optical Business Group; and Kenichi Izuki, Plant Manager, Utsunomiya Plant. Since the interview was conducted through an interpreter, our transcriptionists were unable to ascertain precisely whom was the source of any given answer, and so we have marked all responses as coming from Canon throughout.

Without any further ado, let’s get down to the interview!

Dave Etchells/Imaging Resource: Thank you very much for the tour, for lunch and for this opportunity to talk with you. It’s a little hard to think of questions so quickly after seeing it!

<laughter>

Canon: Could we start with a question? What did you think of the tour?

DE: Ah, we liked it very much. We found the automated assembly and testing was interesting…

William Brawley/Imaging Resource: Very fascinating, yeah. The complexity required to go from the blanks and then through that whole automated process, and just the intricacies of that process were very interesting.

DE: I had seen normal lens normal lens polishing before, mechanical machines, but never automated like that. And it was very impressive to us that the head optical meister, Saito-san, would manually make the masters. Regular assembly was not as interesting to me personally, because I’ve been to two other lens factories in the past, but I think for people who have not seen it…

WB: Yeah, I’ve never seen it before!

HUGE news – test data stored in every lens!

DE: And then finally, the automated testing was very interesting to us. I was especially interested in what you told us about how the testing data is going to be stored in the lens. And the storing of the testing data is happening now for the 16-35mm Mark III?

Canon: Yes, it is.

DE: Are there other lenses that are also currently having that data stored in them?

Canon: The new lenses we’ve introduced in the last five years all have this chip data.

DE: In the last five years?!

Canon: In [these] five years. The ones before that do not, but the ones released in the last five years [do].

DE: Wow! Has the amount of data from five years ago to today changed? Is there more data stored now than when you first started doing it?

Canon: I can’t go into the details of how much, but we are saving more and more data.

DE: More and more over time, yes. So you said you are working on having cameras able to read that data and use it at some point. Do you have any idea how much longer it will be before we can do that?

Canon: <laughs> So the actual data that’s stored in the [lens] chip, that’s probably not actually something that the end users will be able to tap into, but what we are actually trying to introduce is for that data to help the users to get the last part of the image processing. So that’s the sort of the service we are looking to develop and introduce.

DE: Yes. Not that the users would take that data, but I’m thinking that the camera would be able to use it to correct the lens.

WB: Like real-time lens corrections?

Canon: So it’s not so much that the data will work to help with the lens adjustment [during manufacture], per se, but it’s about each lens [having] its own unique data. In other words, it would have its unique, quirky characteristics, which might actually have an impact on the end result. And what that data will help [us do] is to actually take this quirkiness and the characteristics of that particular lens into account to produce the optimal image at the end. So it’s about helping us with the image processing of the data.

DE: Yeah, that’s what I was thinking it would do; I was thinking that it could correct distortion…

WB: …specifically for that lens.

DE: …or even maybe like spherical aberration or something. It could apply different sharpening operators in the center of the lens versus the edges; or different amounts for different focal lengths or apertures or whatever.

Canon: At Canon, we always feel that we don’t have to rely on image processing technology or applications. We are always trying to make [it so] that our products can produce and take the photos at the highest quality at the initial stage. So obviously, if we can produce each lens to its ideal spec, to that very point, it would be perfect, but it’s not the case in an imperfect world. We do have a slight aberration breath, if you will. So what we will do is make sure that the lens will try to meet that ideal quality, but we intend to [account for] the aberration breath. We will have the data processing to make sure that it does reach the fine quality that we’re known for.

DE: So small variations, but then you can tighten that in the camera.

Canon: So, in other words, it’s about taking that production [which] always has a slight degree of variance in what it gives us in the end product. So what we will do is incorporate this image processing technology and application to allow us to achieve that end result that’s to a high quality that we always look for.

DE: And will that be done in-camera or in an application like Digital Photo Professional?

Canon: So yes, at the moment, most of our cameras would actually rely on the DPP application as you mentioned to actually do that final image processing. But eventually, in the future, we would like to have the function inside the camera. And some of the latest ones already actually do have this function.

DE: They have the processing in the camera for these lenses? And so that data’s actually being used right now by some cameras?

Canon: In terms of having that quirkiness to be incorporated, [no] that is not actually incorporated at the moment. We aren’t able to actually use that exact unique data for each lens for the [image quality] adjustments that we were just talking about. But we do have a sort of similar system in place where we do take into account certain unique characteristics of that particular [model of lens] to do that adjustment. So we’re on the way to actually making it to uniquely adjusting future lenses.

DE: And you say some of the characteristics of that line, so it would be data that’s characteristic for instance of all 16-35mm [lenses], or all 400mm f/2.8’s, or whatever?

Canon: There is data that’s actually saved, for example for the 16-35mm, that is saved onto that. The last inspection that we saw at the end of the tour, that data actually is set for something that’s already being stored for the moment. But that particular unique data for that lens is actually not being used for that adjustment process, per se. But we are using the stock data for particular [products], for example the 16-35mm, to do the [image quality] adjustment. So it’s like a gradual step of getting to that point of being able to adjust for that [individual lens].

DE: So there’s sort of generic or average data that’s an average of all the 16-35’s, and that data is being used to correct?

Canon: So [the] production site [has] a [potential range of inaccuracy]. So we’re doing adjustments at that level [to correct for generic defects] and then we will move into the future of where we do the adjustments for each unique lens. This is where they are now.

DE: Very interesting. You said so some cameras do that — which cameras in the line now use that data?

The Canon 5D Mark IV is the first Canon camera to offer real-time JPEG processing with Digital Lens Optimizer adjustments.

Canon: So Digital Photo Professional supported the Digital Lens Optimizer [data] in software in the past, and still we support that. The 1DX Mark II achieved in-camera RAW conversion to JPEG [using Digital Lens Optimizer corrections/adjustments], but not real-time JPEG shooting. But the 5D Mark IV enables the real-time JPEG [compression] applied to the [results of the] Digital Lens Optimizer [algorithm]. That’s the first one. But as we explained, it’s not [using the] unique, [individual] lens data.

DE: It’s just “This is what a 16-35mm does”, or “This is what the theoretical 16-35mm does” and so it’ll adjust based on that, just not on the individual [lens]. Yeah, that’s extremely interesting. It seems to me like this is a good argument for people to buy Canon lenses to use with Canon cameras!

So we were very interested in the automated testing process that we saw at the end. How long ago did that start? Was that also about five years ago; did that coincide with when you started being able to save the data?

Canon: A little [earlier] than that.

DE: So you were automating, and then it was “Aha, we can save the data!”?

Canon: Yes.

This is a big, BIG deal…

(An extended editorial note, by Dave Etchells)
This is potentially HUGE news; the end of sample variation? Every lens will have minor quirks and variations, even within a tightly-controlled production process. As we’re all aware, some samples of any given lens are better out of the box than others. With the highest-quality lenses from the best manufacturers, this “sample variation” will be small, bit it’s always there; even the best-controlled production process has to have at last some tolerance built into it, or the yield would be zero.

As Canon points out, they don’t intend to rely on software to correct problems after the fact, but there’ll always be some “quirks” (as they call them) in each lens, relative to the theoretical ideal of the lens design itself.

Related to this, some lens designs can be extremely difficult to manufacture, because they push production tolerances to the limits of what’s attainable, so yields on them are low. (I’m aware of at least one advanced lens by another manufacturer where this is the case; it’s a difficult design to manufacture, so theyr’e having a hard time keeping up with demand.)

It turns out that Canon has been storing testing data in each lens’s internal memory for years now, adding more and more detail over time, as the capabilities of their automated testing have evolved. While this “quirks” data isn’t being used yet, the potential is there to correct for the minor quirks and manufacturing variation from lens to lens, either in host software like DPP (Digital Photo Professional, Canon’s RAW-processing software) or even the camera itself. Fully implemented, this could let Canon Lenses on Canon bodies approach perfection relative to the lens designs themselves. (That is, when used on a Canon body or with DPP RAW conversion, every lens off the line could perfectly match the theoretical ideal of its optical design.)

While they were very clear that they don’t intend to rely on image processing to correct for basic lens performance, it does seem that this kind of per-lens processing might make it easier for Canon to push the boundaries of what’s possible in the fabrication and assembly process itself, allowing for even more advanced lens designs that aren’t possible today.

It’s important to note that this goes far beyond current “lens corrections”, as provided for by generic profiles for sharpness, distortion or CA. As we’ve seen in our own lens testing, things like CA and sharpness can vary unevenly across the frame, with some lenses being off in one direction, and others off in the opposite. (That is, one lens might show decentered sharpness favoring the right side, another of the same type might favor the left.) While generic lens profiles can improve average performance, with a decentered lens, they could make one side a bit better, but the other side actually worse.

While I can’t talk in detail about how Canon tests lenses, let alone what their test targets look like, it doesn’t require any special knowledge to realize that the test software will be looking at and recording data for the entire frame, so it can tell if one corner is a bit softer than the others, if the CA variation across the frame is different from what’s theoretically predicted by the optical design, etc, etc.

While Canon’s lens-manufacturing prowess is literally second to none, the ability to compensate for these minor variations between lenses and within the frame could bring another whole level of optical performance.

Canon gave no indication of when this kind of per-lens image processing might make its way into DPP, let alone the cameras themselves, but I bet we’ll see it sooner than later, given that they actively disclosed it in this tour and interview.

Fully implemented, this will be a powerful argument for Canon camera owners to stay loyal to the Canon brand for lenses as well.

One of Canon’s automated assembly machines, which in this case is configured to assemble parts of the EF 16-35mm f/2.8L III USM lens.

Testing individual lens elements vs assemblies vs final test

DE: And so the automated testing that we saw was one complete, assembled lens. What do you do for quality control to characterize the individual elements as they come off of the polishing line? How do you measure those to make sure that they’re in-spec?

Canon: So the last part that we saw was the “final-final” check of how it would work. So we would have another stage before that where we do a slight adjustment for each of the lenses, where we actually have a certain adjustment specification that the lens should meet. So that sort of quality control assurance is done at that stage, before going into that final check.

DE: And is that testing sub-assemblies or do you optically test the individual lenses before they go in? Or do you do a statistical sample of a batch that comes off [the production line]?

Canon: We would have a certain optical adjustment that would happen beforehand for each of the lenses. So it won’t be just a sub-sample from the batch. We would do it for all of the lenses that come out to make sure that the optical adjustments are to the spec.

DE: So you’re 100% testing each individual lens to make sure that it meets spec?

Canon: Yes.

DE: And is it an optical test, or is it a profile test with a probe to check the lenses?

Canon: So all the assembly of the floating part would happen before that, and then you would actually go into the phase of doing the actual optical adjustments.

DE: Oh, so you’re testing assembled lenses, not individual elements?

Canon: In terms of the optical inspection, it would happen as we do the assembling itself. For example, if the lens is decentered and we need to adjust the centering of the lens, we would actually do the adjusting in the process of the assembly and then actually test it as it is assembled.

WB: So you do probe profiling of the individual elements and then once it passes that, then they assemble and they test it optically, and then make adjustments as necessary?

DE: So, individual lenses that we’re going to stack up, but the individual lens is tested by probing?

Canon: Aspherical lenses, yes.

DE: Aspherical lenses you do probe, but non-aspherical lenses, you can assume from the production process that they are going to have the right shape, that they don’t need to be tested?

Canon: So we do light testing [with] an interferometer.

DE: Oh, an interferometer! So you have like a master and then an interferometer to check. So has the automated testing really affected just the final assembly? With the grinding process, it’s always just been interferometric measurement, and so there hasn’t been a change over time, or has there been any sort of quality control improvement that could feed back into the conventional grinding process? Where that question came from was thinking that there’s the automated grinding and it would check each one and if it was a little bit off one way, then it could adjust the grinding for the next one.

Canon: The second gentlemen who we saw downstairs, when he was talking about the automated process, said that when we actually see that the [divergence from] the specification, that the adjustment is reflected from the next lens. That’s how the process is learning to kind of reflect that.

DE: Can you do something like that with the conventional, manual grinding process? If you grind a batch of lenses, look at it with the interferometer and it’s maybe a little one way or the other, do you adjust then, the same as the automated machine does?

Canon: In a way, we’ve actually always been having those step-by-step inspections with an amount of human process, if you will. What we’ve been doing in the automation is one thing, but in this process of how we create the lenses by hand, we’ve always had the inspection stages of each step by step to make sure that it does reach that level of specification that it requires, and if it’s not to the bar, we would actually go back to adjust it accordingly.

DE: So [it’s] really the same process, it’s just done manually by a human as opposed to the computer figuring it out?

Canon: Yes.

Automation: Not just mass-production, but mass customization

WB: I have a related question to that: Will there be a point in time when all of Canon’s lenses undergo an automated production or is there always something that’ll require a human touch? It sounds like the optical checking process might still need to be manually checked [for some lenses].

Canon: We are looking to actually introduce more automation; we’d like to go fully automated. And what we would mean is that not just for the mass-produced lenses. We mean mass-customization. For mass-produced lenses, for like 3000-5000 [units] per month, that would be something we could accommodate in the near future. But we are looking to introduce that [even for lots with just] 10-20 units per month. We’d like to create a machine where it can cater to the different lens specifications according to the lot that would come in. So it’s a small lot, but still we’d like to go for full automation. But having said that, the 0.3 micron precision that we were talking about with the Takumi experts, that’s actually something for the 4K or 8K level of resolution. But obviously technology is evolving, so we might see [higher] resolutions coming in [the future], and more precision required. So in other words, we need to expect the Takumi experts to up their game, if you will, to maintain the demands that come into the future. And in the factory next door, we have something called the 30-meter telescope that we’re producing, which has an incredibly large lens.

DE: Oh, yeah.

Canon: Obviously for things like that, we would still need to have that Takumi expert to come in.

WB: Yeah, there’s no machine for that yet!

<laughter>

DE: That’s very interesting. So what that means is that if you can achieve sufficient automation, it will become efficient to produce specialty lenses that people really want, but it’s [only] a very small group of people. With the full automation, you can produce lenses like that efficiently, and so for a good price?

Mass customization? Another Really.Big.Deal.

(Editorial note by Dave Etchells)
This is another really big deal. Extreme automation doesn’t just mean cheaper mass-produced lenses, but also that it will be practical for Canon to produce lenses with much smaller market demand. This would open the way for lots of really interesting, niche-market lenses that just aren’t feasible today, and over time could radically expand the range of EF lenses. Rather than needing to produce tens of thousands of a given lens per year in order for it to be profitable, they could make lenses that only a couple of hundred people might buy in a given year.

How often have we heard virtually every manufacturer say “Yeah, that would be an interesting lens, but there’s just not enough demand to justify us making it”? There may not be big markets for any given speciality lens, but the existence of a range of them in EF-mount could be a strong reason for people to choose Canon over other lens/body systems.

Takumi Toshi Saito demonstrates the process of hand-shaping a master template used to make the diamond-grit grinding tools for lenses.

The role of the Takumi experts

Canon: If the Takumi experts have developed the exact specification [for manufacturing the lens] for the exact rotation, the exact pressure, the exact ways to do the polishing and grinding, and that’s the knowledge and intelligence that we have developed, which is a strength of Canon. That learning is something we are actually incorporating into our machines, and to help to maintain that quality that Canon is known for, to be able to produce those lenses at more of a greater scale and more efficiency. In terms of just materials, we use over 100 types of material [with] different levels of hardness of the actual glass.

DE: 100 different glasses?!

Canon: …We’re talking about quite a vast amount of variety in terms of how we do it. But obviously, different lenses would require a different curvature, different [element] sizes, etc. And so we would have a set of standards that we would actually allocate for each of the lenses, and what we are looking for in the future is for the machine to be able to have this in itself as a [stored] data. So when this lens comes, [the machine] knows how to do this kind of curvature, this kind of size. So that’s the sort of automation we are looking to create.

DE: That was actually my next question, was how does the human skill get incorporated? And so it’s things like “For this shape, this size, this glass, how much pressure, how much do I move like that or like this, or how fast it rotates”, all those sorts of things.

WB: Yeah, how do you translate that [into something a computer can perform]?

DE: Well, I’m kind of amazed that even the humans can know that in the first place, if you have 100 different glasses and you know lenses of all these different sizes and shapes, and they somehow know how to polish each one! It’s amazing to me.

<laughter>

Canon: So I guess that’s a testament to our 80 years of history of building those skills over the years. And Saito-san, who you met, he joined Canon when he was 18 years old. At that time, he didn’t know [anything about lens manufacturing], but then he started learning from his bosses, going up the hierarchy. So we have a hierarchy of experts that we build up. Somebody who is talented will get that track to be able to develop and ascend to the next level, and the next level, and then they finally reach this Takumi expert level. And we’ve always had that as the skills that we pass on through generations of employees that we have had over the years, as well as keeping the data of that. So [we are] making sure that we have human side of skills, continuing on to build Canon skills as well as the data side to help up make sure we can actually leverage this as we go forward.

WB: And on average, how long does it take for a meister to be completely sufficient in their area?

DE: Depends on how smart the meister is…

<laughter>

Canon: 30-35 years. It took Saito-san about 36 years for him to become a meister, he just became a meister maybe in the last year or two years ago.

DE: Oh really, just a couple of years ago? Wow, that’s amazing!

Is Fluorite still relevant?

DE: Now, this is completely nothing to do with what we saw today, but it was just something that occurred to me as we talked about grinding lenses. I guess with modern glass, you have less need for fluorite lenses now? But I’m curious though, fluorite is so different from glass, what do you have to do to grind it? Is it a very different process, and if so what’s different about the processing? Fluorite lenses are very expensive — is that because of the raw material, or is it because of the processing is so difficult?

Canon: The actual raw material is not that expensive, but it has to do with the molecule complex of the fluorite as opposed to the glass. The molecule complex of the fluorite, the actual molecules are actually aligned nicely as opposed to glass where it’s all scattered about. And the processing stage is quite different. With glass you have the powder that you heat, as opposed to fluorite, it takes more of a slower process of making sure to treat the molecules properly. So in other words, because that process takes a longer time, that’s why it turns out more expensive. But it’s not actually the raw material that’s actually different in pricing.

DE: Hmm.

Canon: You saw the lens processing today, but in terms of fluorite, the processing that follows, there’s another material that’s similar to fluorite, which has a similar characteristic quality, and for those kinds of materials, the processing is a bit more complex, hence the reason for the price difference. But we have the technology inside Canon that we can actually use the fluorite as well as this similar-quality material to be able to produce it into a glass that can be turned into a lens. So we have this technology in-house.

DE: You can create the fluorite material yourself from the raw material?

Canon: So like the glass sample that we saw earlier, where we mentioned that we buy [raw blank glass] from Hoya and Ohara, [it’s] similar to that. We buy [fluorite] in a similar format as well, but the processing technology that follows from that base is something that we have in-house.

A fully-automated lens grinding/polishing machine.

How the heck do you automate something as complex as lens assembly?

DE: When you began to automate the assembly process,what did the beginning of that [process] look like? Did you have a machine that could just do one very simple operation and you began using that? Or did you wait until you had figured out a number of operations that could be done together before you first deployed it?

Canon: I would say from 10 years ago we started actually automating different small stages, modules or cells, you could call it, and what we did was develop those different little stages and sort of put that together.

DE: So you might have a machine that would just put the top on, and then another machine that would just do something else. And now one machine can do all of those.

Canon: Right.

DE: And in the tour [there was a] woman who, I think, was assembling a stack of lenses into an assembly. But our guide at that point said that that’s the next job that they’re going to automate. I’m wondering what was that job specifically that’s the next step, because I couldn’t really see what she was doing? And, what are the challenges in managing to take that human job and automate it?

Canon: So what she was doing was [attaching] the flexible circuit board and putting it into the connector. When the material is quite flexible and soft, it’s actually quite difficult for a machine to operate that. At the moment, we can do it with the machine, but it’s more costly and so it’s cheaper to do it with human hands. Hence the reason [why we haven’t automated it yet]. But still, that’s something we’re looking into.

DE: What’s next after that? Certainly it seems like there’s a large human element in assembling the individual lens elements into the subassemblies. Is that the next thing that comes, to be able to do that with machines?

Canon: We are working to make sure we can automate that next year.

DE: Next year, wow! Because there’s a lot of human labor in that. That’s a big part of the process in terms of time, I think.

Canon: It is, but we are looking to automate that next year.

Three current lenses are assembled by machine

DE: How many lens models use automated assembly now? We saw the 16-35mm f/2.8L III being assembled, but I’m wondering are there other stations for other lenses in the product line?

Canon: So the exact levels of automation that would happen vary, but in terms of full automation [there are] three types [of lenses with fully-automated production]

DE: And what are the other two types?

Canon: The 24-70mm f/2.8L II and 11-24mm f/4L.

DE: Ah, all wide-angle.

Canon: Yes. So for all the other lenses we have, they would have certain levels of automation, as well. Some have 30%, some have 80%. So there are elements of automation that are happening in our lens production across the board.

DE: It’s just a matter of how much in each one. And I guess, the goal is to eventually automate everything?

Canon: So in addition to that, obviously, we have mass customization for small lots of lenses. So in other words, one machine [is required] to be able to do different sorts of specifications, depending on what lens comes through. So that’s something worth looking into and we’re developing [it] as we speak.

DE: And we saw super telephotos being made, and also the 16-35mm. And obviously you have many other areas producing other lenses. I’m curious though, how do you organize the production? Is it that there is one group that just continuously produces one lens, or is it that you will take an area, configure it for making a particular lens, build a large number of them, and then reconfigure for a different lens? Do you do them in batches, or do you always have separate lines for them?

Canon: We have a flexible factory, you can say. We don’t have a certain section just dedicated to creating just that one [lens model], per se. We have the machines to be able to cater to more, we have a section [catering] to the different models that would come in [and] out.

DE: And I could imagine the telephoto section might always produce telephotos of some kind, because they would need special fixturing for that size lens, whereas another area might usually produce wide angle, etc.

Canon: For the lens processing that’s not the case, but in terms of assembling, obviously as you mentioned, they require different fixtures to be able to cater to the different sizes.

Assembly precision and Intranet of Things

DE: I’m curious about how much the assembly precision has increased, compared to say five years ago, in terms of the closer tolerances you can hold? And to what extent has that helped previously-existing lens designs. Have you been able to reduce sample variation due to the new measurement techniques, etc?

Canon: So we’re actually incorporating Internet of Things (IoT) in our factory, and what that would mean is that we have sensors built into the lenses as well. So that would allow us to really do the feedback from each lens as to what sort of adjustments are required, etc. And that actually helped us to reduce the variance of the different lenses. So yes, the precision is rising, and that’s based on the IoT that we’ve introduced.

DE: Interesting that it’s IoT. Everything is IoT, right?

<laughter>

Canon: I said Internet of Things, but it’s actually “Intranet of Things”.

<laughter>

DE: Intranet, haha, we get it. No internet!! Ah, that’s interesting.

Canon: So this final inspection data that we were talking about, we saw earlier and that’s actually obviously done with the glass processing as well as the assembly side of things. So whatever data that we can pickup at that point in terms of assembly would come from a slight variance in the parts, as well, and how it has been assembled. Those kinds of data are something that can tell us if there is a certain tendency for this kind of assembly. So we would actually have a feedback system where we would have that learning to be fed back so that we can make adjustments accordingly.

Not just final test data, each lens carries its full production history

DE: Wow, that’s very interesting. So the lens itself is carrying with it a history of the adjustments that were made as it was built, and then at the end you can download that data and see what needs tweaking in your process?

Canon: Yes, that lens would actually have a history of what kind of adjustments have been made, but the machines will also carry that as well. So what it would do is that it would be able to understand the history of lenses that have gone through, and if there’s any sort of variance with the parts and we pick up on some tendencies, we would feed [that] back as well. And as you mentioned, it does have that history, so it’s a matter of being able to keep that together. And also during those stages, we would have optical inspections that would come into play which would sort of help to keep to that level, so that at the end of the day the final product would be to the specification that we’re looking for.

Even older designs are more consistent now

DE: And so, the Intranet of Things, the machines that you’re using to calibrate the lenses and measure them, those are also reporting back to the cloud, to the mothership? So that means that it’s helped older lens designs too, not to perform differently, but to be more consistent?

Canon: We wouldn’t be able to say that for all [lenses], but yes, for some existing models, that’s the case. The ones you buy today [have more of a] slighter variance.

DE: I’m curious too, because here you have production and design and everything together [in one location]. How have advancements in your manufacturing capability affected your lens design? Can you build lenses today that you wouldn’t have been able to build before, or that you couldn’t control the tolerances on well enough? And has that allowed the lens designers more latitude?

Canon: Because we have the lens production site here and we have R&D on the other side, and we have a separate high-standard production site on the other side, we have these three parties in this proximity, which actually helps in all aspects. Because there is technology that gets built with [each] plant that we can actually learn from. For example, treating some hard materials like quartz, [we might have] processing techniques that we could actually incorporate into all lenses, for example. We are the largest optics business to have production sites as well as the design and R&D sites all together, and what we would always do is collaborate with each other to make sure that we can actually optimize our capabilities and whatnot. I strongly believe that we shouldn’t let the technology lead into the product. What I was taught [is that] it’s a dream. Let’s say we have this ideal product that we want to create: We bring the teams together to make sure that [they] rise up to that level to realize that dream. It’s not the technology that defines us, it’s the dream that defines us. So in other words, if we have a certain ideal spec of a product that we envision, we would actually get the teams together to work towards that dream to make it happen. So that’s the whole process that we have here.

The nano-scaled pyramidal structures of Canon’s “Subwavelength Structure Coating” (SWC).
Image courtesy of Canon, Inc.

FINALLY! Someone tells us how nano-coatings are made!

DE: I’m very curious about your nano-coating technology, because from illustrations I’ve seen it appears fundamentally different than how other people’s nanocoatings are made. Other nanocoatings [have] sort of clumps of stuff that gradually get more space in between them, whereas the illustrations I’ve seen of Canon’s nanocoats [show] spikes that have grown. Are these diagrams accurate, or is that just an artist’s [interpretation] showing the surfaces as little microscopic spikes?

Canon: It’s easier to understand, so that’s why we [showed spikes in the diagrams]…

DE: Ohhhh!

<laughter>

DE: I was thinking, “How do they grow the spikes? Is it an electrochemical or something…?”

Canon: It’s a comparative density. The base is quite a high density, but as you go up it gets lighter…

DE: And you showed [the spikes] to illustrate. I got it. <laughs> Are nanocoats generally an organic material, or are they inorganic of some kind? I’m not asking what it’s made out of [exactly] but I’m just curious, because I know they are very, very soft.

Canon: [It’s alumina] (Al2O3). Not organic.

DE: Oh, alumina, interesting. How do you make it vary in density… Is alumina [using] vacuum evaporation?

Canon: [It’s] spin-coated. This alumina is really small, 5-10 nanometers. It’s actually liquefied and we put it into a spin coating [machine], and then that gets dried, and then that get put into hot water. Then when it dries, it crystallizes, and when it crystallizes at that stage, the density level changes from the base to the top, so it’s quite highly dense on the bottom side, and the density gets lower on the high side. So that’s how we create that.

DE: So it’s spin-coated and dried, and there was some kind of carrier material, a liquid that it was in…

Canon: The initial liquid that you put the material in hasn’t been…

DE: That’s proprietary, yeah.

Canon: … and then that gets spin-coated, and then we dry it.

DE: So the spin-coating is a uniform thickness, and it’s a uniform density at that point…

Canon: Yes, it’s uniform.

DE: And then you put it in the hot water, and…

Canon: It gets dried, and then it gets put into hot water. And in the process of crystallization, the density varies as you go higher.

DE: So the water is helping it re-crystallize, and it sort of consolidates towards the bottom and that’s how the density changes?

Canon: The re-crystallization starts from the surface closer to the hot water.

DE: Huh, closer to the water… I would think that would be higher density then, if it’s recrystallizing…

Canon: If you look at the patent documents you could find more.

<laughter>

DE: Good, good, that’s what I will do then, yes. And you must also have US patents, so I can read it in English then.

Canon: Yes, we have US patents too. And that’s our time. Thank you!

(See much more in our Canon Factory Tour article!)

Canon Factory Tour

The post Canon lens factory tour interview: Better lenses yet lower prices, new details about nano-coatings! appeared first on Imaging Resource.

The Kodak factory in Harrow, England used to be the largest manufacturing facility in Britain, but diminishing demand for photographic paper over the years has led the company to cease production at the factory, in operation since 1891. Before the factory shuts its doors in October, James Casha, photographer and former Kodak employee, managed to take a photographic tour of the facility.

Making Tracks – One Last Visit to the Kodak Factory in Harrow

Reading the article, it’s hard to imagine the scale at which the factory operated to produce a huge output of paper: for example, “parent” rolls four kilometers long passed through an eight-story building at a rapid pace. Mr. Casha’s article brings the production process to life, and the accompanying photographs give you a real sense of what it would have been like to produce the massive quantity of paper.

(Seen at colorgrinder.com)

The post Some photos before the doors close: a photo walk through the Kodak Factory in Harrow, England appeared first on Imaging Resource.

According to the Image Sensors World blog, Sony will be receiving the plant less some of its equipment, which has been transferred to another Renesas location to manufacture the lesser quantity of chips Nintendo now needs to meet its revised sales forecast. This equipment will have to be replaced before Sony can begin using the fab, which has a 40nm process. Image Sensors World, citing an earlier report from Semiconductor Portal, also suggests that there is a split in opinions inside Sony as to whether the acquisition makes sense, given that the production line at Tsuruoka is not fully automated, and may thus not be able to churn out BSI CMOS sensors with the required efficiency.

Sony’s new BSI CMOS sensor fab in Yamagata will supplement its three existing sensor fabs, all located in southern Japan.

Be that as it may, Sony’s press release states that a “definitive agreement” has been reached and executed with Renesas. The company plans capital investment at the site over the next two fiscal years that should see it meet its goal of increasing total production capacity from 60,000 to 75,000 wafers per month. The new facility, to be renamed the Yamagata Technology Center, will join existing Sony image sensor fabs in Kagashima, Kumamoto, and Nagasaki.

More details can be found in Sony’s press release.

(Silicon wafer image courtesy of .RGB. / Flickr. Used under a Creative Commons CC BY 2.0 license.)

The post Sony to buy CMOS fab used by struggling Nintendo, refit it for BSI sensor production appeared first on Imaging Resource.

On a trip to Japan this past August, Sigma Corporation’s CEO Kazuto Yamaki asked if I’d like to tour their factory in Aizu Wakamatsu. Engineer/geek that I am, I couldn’t possibly resist. Sigma Corporation has been a leading player in the third-party lens business since there was a third-party lens business. Long seen primarily as a source of inexpensive add-on lenses, they’ve recently been reshaping themselves into a purveyor of true high-end, no-excuses glass that’s in the top tier of optical performance. (In our recent SLRgear review, we called the new Sigma 18-35 mm f/1.8 DC HSM “A” “by a wide margin, the best constant-aperture/wide-aperture zoom lens we’ve ever tested”.) With this as background, the opportunity to have a true no-holds-barred geek’s tour of their factory was a chance not to be missed.

So it was that I found myself on a short flight from Osaka to the Fukushima airport, where I was met by Shinji Yamaki, Sigma’s manager of European marketing. A pleasant 90-minute car ride through the mountainous countryside brought us to the Sigma factory in Aizu. The Aizu site has been home to Sigma’s manufacturing from very early on, only a few years after Kazuto-san’s father founded the company. Aizu is an interesting region historically, with a somewhat sad tale of betrayed loyalty to the Emperor during the Civil War period, which led to a long period of relative isolation from the rest of Japan. In some part, that isolation and the culture it produced is what led to Sigma locating their factory there.

In the early days of Sigma, Michiharo Yamaki (Kazuto-san’s father, and founder of the company) faced stiff competition for skilled workers in the Tokyo area, as it was then the boom time of the Japanese optical industry, with not enough skilled workers to go around. Through a worker from the Aizu region, Michiharo-san learned of the local population’s reputation for diligence and ability to concentrate on fine details. A recruiting trip to the area led to a particular night that involved a great deal of sake, in which the local farmers pressed Mr. Yamaki to build a factory in the area. They were hoping for jobs that they could work at in between their agricultural duties. In a moment of inebriated agreeableness, Michiharo-san half-jokingly said “sure, I’ll build you a factory.” The next morning, he discovered to his dismay that, while he had been joking, the farmers were completely serious. After a not-insignificant amount of badgering, Mr. Yamaki relented, rented a single room in someone’s home, stuck a metal lathe in it and hired a person to run it. That was the beginning of Sigma’s manufacturing in Aizu.

Mr. Yamaki soon learned that the citizens of Aizu had an exceptional work ethic, and an unusual capacity for maintaining focus on details that would be mind-numbing for more ordinary folk. One thing soon led to another, culminating in construction of the first building in Sigma’s current complex in the early 1970’s. Today, that original building has been added onto a number times, with total area across all buildings now totaling more than 50,000 square feet of high-tech manufacturing space. That’s a bit over one and a half American football fields, most of it densely packed with manufacturing equipment. It’s a seriously impressive operation. The diagram above shows the overall layout.

One of the things that’s remarkable about Sigma’s manufacturing operations is the extent to which they’re vertically integrated: Almost all production steps for their lenses occur in-house, including lens grinding, aspheric lens molding, metal machining of lens barrels and other components, injection molding — including making their own injection dies — optical coating, final assembly, and more.

Besides its sheer scale, the other thing that immediately impressed me about the Sigma factory was how clean it was. Especially remarkable for a lens-manufacturing operation, where the polishing process involves slurries of polishing compound and glass residue, the Sigma plant was absolutely spotless, from top to bottom. Even aisle space right next to the polishing machines themselves looked like you could eat from it. (Sushi, anyone?) It may not matter much to the polishing process itself, but the attention to cleanliness throughout spoke to me of a degree of meticulousness that was evident at all levels of the operation.

Optics production
As you’d expect, lens manufacturing begins with grinding and polishing raw optical glass into lens shapes. At Sigma, this occurs in three stages for conventional (non-aspheric) lenses; CG or curve generation, smoothing, and polishing. Aspheric elements are formed by pressing the glass in precision molds at high temperature.

The CG machines pick circular glass blanks out of trays, press them against a rotating spindle with a grinding head angled and shaped to produce the desired profile, then return the ground blanks to the tray. The shot above shows the input side of a CG machine, with the mechanical arm having just picked up a fresh blank from the tray. (It’s the white disc you see suspended from the arm just right of center.)

Here’s a look at the business end of that same CG machine. The glass blank is mounted on the end of the white spindle at center (the one with the grey-colored coupler between the two white pieces. The grinding head is inside the white box at right, apparently housed that way to contain the spray of coolant with glass dust suspended in it.

The shot above shows a “smoothing” machine; think of this as a fine-grinding operation. It uses a finer grit than the CG machine, removing scratches left by the CG machine, but doesn’t significantly change the curve shape.

The final step is polishing, carried out on machines like the one shown above. This machine is designed to polish large-diameter lenses. The particular machine shown in this shot is producing a concave profile on several lenses. Polishing “fine tunes” the lens shape, taking the shape produced by the preceding combination of curve generation and smoothing, and bringing it to its final shape. In many cases, polishing is a process of progressive refinement, with some lenses receiving as many as three or four polishing operations. (Note how spotlessly clean the floor is here, even immediately adjacent to the machine. I’ve seen other lens-grinding facilities, but never anything approaching this level of cleanliness.)

You’d expect lens production to involve lots of polishing, but the sheer scale of it is pretty amazing. The shot above shows about half of one room of polishing machines. There’s another entire room of these machines, perhaps twice as large as this one, so what’s shown here is less than 20% of the total polishing capacity in the Sigma factory.

Conventional polishing operations can only produce lens profiles that are sections of a sphere. Aspheric lenses help lens designers eliminate spherical aberration, and also make lens systems much more compact, in many cases replacing a complex, multi-lens assembly with a single element. While it’s possible to grind and polish an aspheric surface, it’s vastly more efficient to mold it. Machines like the one above do this, pressing pre-formed glass blanks into ultra-precise molds made of refractory metals (tungsten or similar), at elevated temperatures. This is an incredibly tricky process, as the stresses and distortion of the glass as it moves through the temperature cycle need to be taken into account, and the final surface geometry needs to be controlled with sub-micron accuracy. There isn’t an awful lot to see in the shot above, as everything happens inside an enclosure, with the lens passing through multiple heating/cooling zones. (I thought I had a wider shot of one of these machines, but couldn’t find it on my SD card.) As with everything else in the Sigma factory, there’s a huge roomful of these machines. Quite an investment at a cost of about $300,000 – $400,000 each!

After the lenses are ground, the very important step of “centering” is performed. The lenses are clamped in precisely-contoured jigs, spun on a lathe-like device, and the edges are ground by a rapidly-spinning head, running in an oily coolant liquid. An air-handling system connected to each machine vacuums away the coolant mist, but a very small amount still escapes, making the floor in this area a little slippery. (This was the only part of the factory where there was any hint of a foreign substance on the floor.) When they leave the centering machine, the lenses are in their final shape, with smooth, beveled edges concentric with the lens’ optical center.

As with the polishing machines, there’s a huge roomful of centering machines whirring away. The shot above shows a fraction of the total.

There was a lot of evidence of quality control and checking of the various components within the plant. Here, a worker is checking individual lens elements to ensure that they’ve been ground to the correct profile. The instrument he’s using projects a laser light source through the lens and bounces it off an optical flat. (Actually, I’m not sure whether it’s a flat or an optical surface with a profile polished into it.) The resulting interference pattern (for photographers with darkroom experience, think “Newton’s Rings”) shows whether the profile is correct or not.

Coating is the final optical step for the individual lens elements. Optical coatings reduce internal reflections in complex lenses, reducing flare and ghosting in your images, and improving contrast. The polished lens elements are loaded on big carousel discs, which in turn are loaded into huge vacuum evaporation machines. The shot above shows a worker loading the polished, centered lenses onto one of these carousels. I was surprised to see the workers handling the lenses without gloves, because finger grease would be death to either a high-vacuum system or the coatings themselves. It turns out they’re wearing thin, transparent plastic sleeves over their fingers, rather than gloves. I imagine these provide much better dexterity, and mean that the workers don’t have to work with their entire hands in latex or neoprene gloves all day.

My apologies here, though – I neglected to get a shot of the vacuum evaporators themselves. They were pretty obscured through a window, though, so wouldn’t have been a great shot anyway. They’re huge machines, with a large, central high-vacuum chamber that always stays pumped down, and airlocks on either side of it. A carousel is loaded into the left airlock, it’s pumped down to high vacuum, then the carousel is slid over to the evaporation chamber. The coating material sits in the bottom of the vacuum chamber, where it’s heated to the evaporation point, usually by either a tungsten filament or an electron beam. (I didn’t ask, but as far as I know from prior experience in the semiconductor industry, big production machines like this are more likely to use electron beam technology.) Since there aren’t any air molecules to get in the way or react with the evaporant, the molecules or atoms fly in straight lines, and are deposited on the lenses in the carousel. The carousel rotates during the deposition process, to make sure the coating is deposited evenly. Mr. Yamaki-san told me that the evaporators were the most expensive machines in the factory, costing about a million dollars apiece.

Metalworking
While all this has been going on with the lens elements themselves, the lens bodies have been taking shape in other parts of the factory.

There are a lot of metal components in Sigma’s lenses, so there’s quite a bit of manufacturing capacity devoted to machining, with rooms full of vertical machining centers of varying degrees of capability. The shot above shows a huge, multi-axis machine. The most sophisticated units are so-called 9-axis machines. (I’m not sure if the one above is one of these or not.) The 9-axis systems can basically position a part at any angle relative to the three coordinate axes, can have the milling cutter approach at any angle, and also have the ability to spin the part to do lathe operations. To give you an idea of scale, the machine above is about 15 feet long, 14 feet wide, and 12 feet high. From the look of it, it’s been making some base plates for injection molding (the metal rectangles with many holes in them at the bottom and left of the frame.)

Here’s a view looking down about half of one room of vertical machining centers. The machines are two deep on each side of the aisle. I have no idea of the exact count of the machines in this room, but would guess there are somewhere between 50-100 of them, of various types.

Here’s the output end of a machine making lens mounts. The shiny pieces in front are the finished parts, the dull golden ones in the background are the raw, die-cast blanks. You can just see the robotic arm at upper left that loads and unloads the machining center, which is out of the picture, to the left.

Not everything is CNC-driven — some operations are still performed manually. This gentleman is running a manual lathe, turning some sort of flanged tube from aluminum stock.

Some very hard/tough alloys are difficult to machine with conventional cutting tools. The electrical discharge machine (EDM) above is one solution to that problem. Dies are made in the shape of the metal that’s to be removed, loaded into a head that travels up and down, and both die and part are submerged in an insulating liquid. High voltage/high current electricity is applied to the die, and the machine lowers it onto the part. At some point, an arc will jump from the die to the part, eroding metal wherever it hits the raw stock. Every time an arc is struck, the machine pulls the die back a little, then advances it again to strike another arc. The fine metal particles removed from the part are carried away by the recirculating dielectric fluid.

It’s a laborious process, but has the advantage that it doesn’t matter in the slightest how hard the stock being machined is; if it conducts electricity at all, it can be machined.

Sigma also operates a number of wire discharge machines, which operate on a very similar principle, but the working electrode is a string of wire, stretched taut between a pair of electrodes. These machines are basically used like giant jigsaws, cutting through sheets of tough metal from top to bottom, sectioning them or creating edge profiles. The wire is fed through the machine continuously, so erosion of the wire doesn’t affect the kerf width. As you might expect, wire EDM machines use a lot of wire. The wire exiting the machine is chopped into fine pieces and sent in bulk to metal recyclers.

Did you ever wonder where screws come from? Some are stamped, but the better ones come from machines like the one above, called, appropriately enough, screw machines. This is the back end of one, with a bar of feedstock running into the maw of the machine, towards the top middle of the frame. Operation is completely automated; once programmed, the machine will happily digest bar after bar of raw stock, with the finished screws tumbling out in a flood of coolant on the other end, to be caught in a steel-mesh strainer.

Here’s the finished output of one of the screw machines. The shot at left shows a batch of finished screws, caught in the strainer that separated them from the coolant. The shot at right shows a couple of them in Yamaki-san’s hand.

As you’d expect, Sigma purchases a lot of their screws from suppliers, wherever a standard configuration can be used. In many situations, though, only a custom fastener will do, so Sigma keeps a bevy of screw machines pretty busy.

After metal parts have been fabricated, they go the surface finishing area. Surface finishing may include painting, bead-blasting, plating, or anodizing. The shot above shows a portion of the anodizing operation, where aluminum parts are dipped into acid baths and current passed through them to create a layer of super-hard aluminum oxide on their surfaces. Once the oxide is created, it can be dyed to whatever color is desired, and then passivated in a process that closes the pores in it, so it’ll be impervious to staining out in the field. The anodizing room was hot, humid, and generally unpleasant. Excellent ventilation meant there was no hint of chemical smell or fumes, despite the large tanks of acid bubbling away. Still, my hat’s off to the workers who labor there.

I was surprised that some parts of the process that I’d assumed were probably automated were actually manual. In the shot above, the worker is injecting paint into engraved/recessed letters on bezels for AF/MF switches. I had expected that this sort of lettering would involve something like a squeegee blade passing over the recessed letters, depositing white paint in its wake. Thinking about it, though, that would be tough to make work, given that the part here has other recesses that would collect the paint, as well as the lettering it was intended for.

The “painting” being done here is interesting too, in that it doesn’t involve a brush. Rather, the worker is wielding something that looks like a hypodermic syringe, minus the plunger. Paint is supplied continuously through a tube, and injected down a narrow tube that the worker uses to fill the recessed lettering. Every part looked perfect, I can’t imagine sitting there 8 hours/day doing this. Hats off to this worker, too…

Plastic fabrication
Leaving metalworking for a bit, we’ll turn to plastic fabrication, specifically injection molding.

As the name suggests, injection molding consists of injecting the raw material in a molten state and under high pressure, into precisely-machined molds. Chances are that every piece of consumer electronics or household appliance in your home has at least one injection molded part somewhere inside it. In modern photo equipment (cameras and lenses both), a majority of the structural elements are injection-molded.

Injection molding has the advantage that it can hold very tight tolerances on parts, yet the incremental cost of production (the per-part cost, once the molds are paid for) is very low. The shot above shows one of the huge injection-molding machines, with a tray of finished parts in the foreground. A robotic arm (very top/center of the frame; the grey thing projecting downward) removes parts after molding, clips and removes the sprues (extraneous pieces of plastic, from channels used to carry the molten plastic to the interior of the mold) and drops them in the output bin.

Having always associated plastic with cheap construction, I was surprised to learn that Sigma’s injection-molded parts are actually considerably more precisely dimensioned than their metal counterparts. This is because, once the master mold has been created, there’s essentially zero wear from part to part, so consistency is extremely high. By contrast, when machining metal, the cutters are always subject to wear, in ways that’s hard to compensate for — even with CNC machine tools.

This shot is a closer view of the parts exiting the machine above. These are actually pretty simple compared to many of Sigma’s injection-molded parts (some can be very complex), but even here, there’s a lot of detail that would require a lot of machining time to create. What’s great about injection molding is that even the most complex parts can be cranked out like jellybeans, once the molds have been made.

In earlier days, plastic parts in lens assemblies were an issue because the thermal expansion properties of the plastic were different from the aluminum parts surrounding them. At high or low temperatures, tolerances could shift, parts could bind, or optical quality could suffer due to poor alignment of elements. Addressing this, Sigma has developed what they call Thermally Stable Composite (“TSC”) plastic, which has the same coefficient of thermal expansion as aluminum. Composed of a mixture of polycarbonate resin, glass fiber, and metal fibers, TSC not only matches the thermal characteristics of aluminum, but is considerably stronger and more elastic than conventional glass-loaded polycarbonate formulations. The improved mechanical characteristics mean that parts can be made smaller, while retaining the strength and toughness of their larger predecessors.

Injection molding can also combine plastic and metal parts into a single assembly. The parts shown above have three metal tabs that are molded right into the plastic. The machine that inserted the tabs into the mold cavity was quite clever in how it picked up and precisely arranged the metal parts, but Mr. Yamaki was understandably concerned about me videotaping within the factory, so I don’t have a movie to show you of it. (I greatly appreciate the access Mr. Yamaki gave me for the still shooting: It’s very unusual to be allowed the level of access and freedom to photograph that he extended to me on this tour.)

Here’s a surprising stat: Sigma says they can maintain dimensions to less than +/- 1 micron on their injection-molded parts. (A micron is one thousandth of a millimeter, about 39 millionths of an inch.) Of course, in order to produce parts to those kind of tolerances, you need to be able to measure them that accurately as well. That’s the job of the machine above and its sibling in the background. A ruby-tipped stylus can trace the outlines of a part, measuring to an accuracy of less than a micron, anywhere within what looked to be a working volume of a meter on a side.

Sigma is unusual, in that they maintain their own mold-making shop. Molds (actually called “dies” in the trade) are massive blocks of high-grade stainless steel, with cavities in them the shape of the parts to be made. Mold-making (at least to the tolerances required for lenses) is an exacting, iterative process. A major obstacle is the thermal properties of the plastic itself. As the plastic cools from its melting point, it shrinks considerably. The size you want is obviously the size it will be when it’s cool, but that has to be a good bit smaller than the size of the cavity in the mold.

A worker moves a mold “stack” in the mold shop, as part of a trial run of a new pattern.

The CAD software used to design the molds can compensate for thermal expansion, but it’s a highly complex situation, with different regions of the part shrinking more or less, depending on how thick they are, what stresses are acting on them from adjacent regions, etc.

So, while CAD will get you most of the way there, there’s ultimately some amount of trial and error involved. (Noting that “error” here means a dimension off by a few microns.) Successful mold making often requires two, three, or even four iterations, with molds made, parts cast and then measured on the behemoth measuring machines shown above. Any discrepancies are recorded, a new mold is made, new parts cast and measured, and the cycle continues.

As you might expect, this is an expensive process. The molds have to be machined to incredibly tight tolerances, carved from a very tough stainless-steel alloy. Surface finish is also critical, which adds considerably to the difficulty and time required for the machining process.

As expensive as it is to maintain their own mold-making shop, Mr. Yamaki said that without it, the development costs for many new lenses would be simply prohibitive. Development cycles would also be considerably lengthened, to accommodate parts shipping to/from die-making companies, scheduling with those vendors, etc. By keeping the mold-making in-house, costs are reduced and iteration cycles are shortened, so Sigma can produce more and better lenses each year.

Once the plastic, glass, and metal parts have all been fabricated, the next step is assembly. This is essentially a manual process, because it’s important to inspect each lens element to make sure there are no minute flecks of dust on it that would interfere with the lenses’ mating properly with each other. Note the “bunny suit’ the worker above is wearing: All optical assembly is performed under clean-room conditions. The nylon suit prevents things like loose hairs or shed skin cells from contaminating the optics. The air in the clean-room assembly areas is also filtered to remove dust particles. Each assembly workstation has a bright light that the workers use to inspect the lenses, ensuring that no dust or other contaminants are present on the surfaces of each lens element.

The worker above is assembling lens elements into an optical group. A small drop of adhesive is placed between each pair of lens elements to hold them together. The adhesive is completely fluid until the assembly is exposed to UV light, which hardens the resin and bonds the elements together.

There’s an intermediate step between optical-group assembly and final assembly of the completed lens, which I don’t have documented. That’s edge-painting, where the rims of the lenses and optical groups are coated with flat black paint, to reduce internal reflections. This is a manual process, with workers applying the black paint with a brush.

Final assembly and testing

Once all the optical groups, metal, and plastic parts are finished, it’s time for final assembly of the finished lens. The shot above shows a final-assembly station. I don’t know what lens was being assembled here, but from the look of the unit that’s on the work surface (just to the left of the worker), it’s probably some sort of wide-angle prime.

When everything has been assembled and the lens is complete, the final step is testing. Testing and QC have always been key parts of Sigma’s process, but there’s been a recent & significant upgrade in this area. For a long while now, Sigma has spot-checked lenses on an MTF tester of their design, which employed a Kodak image sensor. As digital SLRs and CSCs have evolved to ever-higher resolutions, though, the old system began to show its age, in that it wasn’t able to discriminate critical focus as well as some of the camera systems Sigma’s lenses were being used on.

Searching for a better solution, the Sigma engineers had an epiphany when they realized that they already had an excellent, very high-resolution sensor at their disposal — namely the Foveon chip used in their flagship SD-1 camera. The only catch was that that sensor was an APS-C sized one, smaller than the full 35mm frame that many of Sigma’s lenses cover.

The solution, of course, was to cover the frame piecemeal, using four individual, overlapping exposures to achieve full-frame coverage. One of Sigma’s clever young engineers developed a combined software/hardware solution that did this (sorry, I didn’t get the engineer’s name; he surely deserves credit), the result being Sigma’s new A1 MTF-measuring system. A housing encloses the sensor assembly and the motorized stage that moves the sensor either successively to the four corners of a full-frame area, or to the center of the frame, for testing lenses designed for cameras based on either APS-C or Micro FourThirds sensors. A universal lens-mounting system on the front of the housing accommodates lens mounts for Sigma, Canon, Nikon, Pentax, Sony (both A- and E-mount) and Micro Four Thirds systems, precisely matching the flange-sensor distance spec for each.

Unfortunately, I wasn’t allowed to take photos in the area where the A1 systems were being built and used; the work being done there was a bit too proprietary for public dissemination.

What’s particularly remarkable about the new A1 MTF-testing systems is not so much the systems themselves (which are exceptional enough in their own right), but the way they’re being used. All of Sigma’s new “Global Vision” lenses are being tested 100% on A1 systems, with any that don’t pass muster across the entire frame being rejected. As far as I can tell, this is an unprecedented level of QC for the lens industry, and sample variation among the new Global Vision optics should be dramatically lower than we’ve seen thus far from the industry as a whole. The engineer behind the A1 system has built 30 of these systems to date, for use in production testing, with more in the pipeline.

Mr. Yamaki noted that the new A1 testing system gives them a much finer ability to discern minor variations in sharpness, and smaller amounts of decentering of a lens’ elements. Sigma’s lens designs have always been characterized by much better than average sharpness at a given price point, but the new A1 testing system ensures that all of the Global Vision lens models will perform closer to design specs than has previously been possible. Mr. Yamaki said that using the A1 system and the more stringent quality controls it enables has reduced their manufacturing yield somewhat, but he’s happy, knowing that they’re producing lenses with an unprecedented level of quality and consistency.

Take a video tour
While he gave me unprecedented access to the factory floor, and permission to shoot stills there, Mr. Yamaki-san was understandably reluctant to do the same with video. Sigma themselves have produced a film about the Aizu factory, though, and it’s easily the most beautiful factory-video I’ve ever seen. (The Aizu area of Japan has impressive scenery and lots of natural beauty.) When you watch the video, you may recognize some of the factory areas and processes described in the article above.

Sigma’s hospitality
Mr. Yamaki and Sigma were also exceptionally gracious hosts during my visit. After touring the factory, they treated me to dinner at the very traditional (and excellent) Ashina Ryokan in Aizu. Ryokans are traditional Japanese inns, originating in the Edo period of Japan (1603 – 1868) as lodging for travelers along Japan’s highways. Many modern ryokans have been westernized to a fair extent, but Ashina is very traditional, with tatami-matted rooms, no furniture but low tables and floor-chairs, and a traditional onsen (communal, segregated by gender) bath. The lack of a conventional bed caused me a little consternation when I arrived after a long day of travel and wanted to lay down – no problem, though, I just gathered the cushions from four of the floor-chairs and lay down on those.

While we were at dinner, Ashina’s staff set out traditional futon bedding for us in our rooms. Spread on the floor, the futon mattress was quite firm, but I found it comfortable, and slept well. (Helped, I’m sure, by the prodigious quantities of sake I had with dinner.)

Ryokans distinguish themselves by their food, and the food at Ashina was really exceptional. I honestly don’t care for a lot of Japanese fare, finding it rather bland, despite its freshness. The food at Ashina was very different. Cooked over an open charcoal fire by a gracious and accommodating young woman named Alona, it was truly delicious. Alona was a wonderful hostess, and it was nice that she spoke such excellent English, so I actually knew what I was eating – not always a given for me in Japan. The service at dinner that night was some of the best I’ve ever experienced in a restaurant: Alona was constantly and almost invisibly attentive to our needs; we never had to ask for anything, it just appeared when we needed it. It’s a shame that Japanese tradition forbids tips, she certainly would have earned a generous one by American standards. (Maybe Ashina’s owners will read this, though, and give her a raise

The shot above shows the first course of our dinner, with some mushrooms on the left, edamame (steamed soybeans) in the orange bowl, something delicious that I can’t remember in the center (sliced octopus in a savory sauce?), and sashimi and raw meat in the upper right. The two dishes in the right foreground have sauces in them for dipping the fish and meat in. Anyone want to guess want sort of meat that is? It’s … horse(!) Once I overcame my reticence over eating horse meat, I found that it was arguably the best meat I’d ever tasted. Seriously! OK, I know that most of our US readers are probably completely appalled at this point, but I have to say that, if I could ignore the fact that it was a horse I was eating, it was incredibly tender and delicious. It turns out that a lot of horses are raised as meat animals in the Aizu area, through long tradition. I tend to think of horses more as companions than dinner, but have to admit that they taste wonderful.

Here’s our hostess/chef Alona, serving some roasted, salted fish to Shinji-san, Sigma’s manager of European marketing. The fish were the first thing we saw cooking when we came down to dinner; you can see them in the background of the previous shot above. They were local trout, skewered on thin bamboo spits, which were stuck into the sand of the cooking pit, leaning over the charcoal fire. They were delicious, very slightly on the dry side, but with excellent, delicate flavor. The Japanese way is to eat the entire fish except the fins. I did, and the head was actually quite tasty.

Dinner also involved a number of rounds of local sake. I’m a huge sake fan, it’s one of my favorite drinks. Here in the US, only a very few varieties are imported, and they’re almost by definition from huge mega-brands, who are equipped to handle international import/export. In Japan, there are hundreds (thousands?) of sake producers, with a number of them in each local area. This might compare to the US phenomena of local brew pubs, each making their own artisan beers, although I believe the sake breweries generally sell their output to restaurants or at retail, vs on-site. We probably had five different sakes this night, all from small, local suppliers, and it was all truly excellent, every one better than any I’ve had in the US. Each had a unique flavor and mouth feel, distinct from the others; one was even very mildly carbonated.

During dinner, we were also attended by two geisha-san, Mamewaka-san and Tsukino-san. In the US, our image of geisha is rather exotic, perhaps with sexual overtones. In practice, many real geisha-san more closely resemble a favorite aunt than an object of desire. Their social function is to provide interesting conversation and entertainment, although admittedly in some circumstances with flirtatious overtones. In the case of Mamewake-san and Tsukino-san, they were pleasant and engaging dinner companions, and I learned a lot about the lives of modern geisha through them.

After dinner, each performed a traditional Japanese dance. I have to admit that this was a very foreign form of entertainment to me. The dances are very highly structured and symbolic, acting out the stories told in traditional Japanese ballads. It might have helped if I’d known more of the stories that the dances corresponded to, but as it was, it was all pretty opaque to me. I could see, though, that Mamewaka-san and Tsukino-san had devoted a great deal of effort to learning the art form; I appreciated the precision and grace with which they executed their dances. That’s Tsukino-san above, performing her dance. (No she doesn’t look like someone’s aunt; she’s quite a young geisha, compared to many.)

The whole experience at Ashina ryokan was exceptional – I highly recommend it if you’re ever in the Aizu/Mt. Bandai area of Japan, it’s a great example of very traditional Japanese hospitality. If you stop in, tell them that you heard about them from Dave-san at Imaging Resource.

Conclusion
From start to finish, my visit to Sigma’s Aizu factory was an extraordinary experience. As I said at the outset, I love factories, and the tour Mr. Yamaki-san took me on was far more in-depth than any I’d previously experienced. As I mentioned at the outset, in the early days, Sigma’s stock in trade was inexpensive kit and accessory lenses for film-based SLRs. Following a path he and his father first charted together, though, Kazuto Yamaki is transforming the company into a producer of truly world-class optics, with performance fully meeting the demands of the digital era. We saw some of the first fruits of this in the truly exceptional Sigma 18-35mm f/1.8 DC HSM “A” lens, arguably the best wide-angle lens we’ve ever tested on SLRgear.com, regardless of price — primes included. Subsequently, the Sigma 120-300mm f/2.8 DG OS HSM “S” tele zoom turned in a stellar performance as well.

I’ve been on many factory tours where access was very tightly controlled, to the extent that you wondered whether you were perhaps seeing a Potemkin village version of the production process. In my tour of Sigma’s facility, though, I had access to literally every corner of the factory and every element of the process – the only areas I didn’t venture into were those inside clean-room environments, and the offer was open to don a bunny suit and tour those as well. As you can gather from all the above, I was highly impressed with what I saw: I think we’re going to see a lot more lenses as exceptional as the Sigma 18-35/1.8 and 120-300/2.8 as time goes on.

Many thanks to Mr. Kazuto Yamaki and Mr. Shinji Yamaki for the time, effort, and care they devoted to my visit, and for their gracious hospitality.

The post A geek’s tour of Sigma’s Aizu lens factory: Precision production from the inside out appeared first on Imaging Resource.

It’s not just our cameras that are high tech these days — in an increasingly digital world, lens development has come on in leaps and bounds, too. An announcement today from Nikon sees even loftier levels of tech deployed in its development process — and that should mean better lenses for its customers.

It’s only a few decades since lenses were developed by rooms full of people manually performing ray tracing calculations ad infinitum. Today, computers do the heavy lifting, but it’s still not until you hit the prototype stage that anything resembling real-world photography happens. Even once you’ve got the prototypes ready, it’s a complex task to judge their performance. It’s not just about sharpness — all sorts of other factors come into play, and they have to be measured and quantified.

For Nikon at least, the task has apparently just gotten easier thanks to a new lens testing system dubbed OPTIA, as well as new lens simulation software. An abbreviation of “Optical Performance and Total Image Analyzer”, Nikon’s OPTIA system allows it to quantify how a particular lens design performs for a wide range of metrics. According to the company, these cover “nearly all aspects of optical performance”, including resolution, aberrations, bokeh, texture reproduction, and sense of depth.

OPTIA is said to be based on argon fluoride immersion scanner technology already used in developing projection lenses for integrated circuit steppers and scanners, where optical performance is key. It’s been adapted for the much wider range of wavelengths in visible light, though; IC production uses a much narrower range of wavelengths.

And even before lenses reach the prototype stage, new simulation technology deployed by Nikon should bring better results. The newly-developed software will now allow simulation of real-world photography, letting Nikon gauge how a particular lens design will perform and refine its creation before it invests the time and expense required to create prototype optics.

If you’re shooting with a Nikon DSLR or mirrorless, perhaps it’s time to start saving pennies. According to the company, development of the OPTIA system was completed in May 2012, and as of this month, it has now fully deployed both OPTIA and the new simulation software in its development process. Good things could be on their way in the not-too-distant future!

(Nikon logo image courtesy of Joe Shlabotnik / Flickr. Used under a CC-BY-2.0 license.)

The post Aberrations begone: Nikon’s new OPTIA analyzer should yield better SLR and mirrorless lenses appeared first on Imaging Resource.