There’s no rule that says when a technology becomes mature, but one might think that image sensors would be by now, 40 years after their conception. Nothing could be further from the truth. After decades of use in fax machines, scanners, and camcorders, image sensors are now in the largest growth period in their history, and much of the innovation is still happening.
Let’s step back and look at the overall growth in the business over several years. The accompanying chart shows the total revenues in this market since 1996, with a forecast out to 2011. We count here the value of all image sensors manufactured in the year, whether for the merchant market or for an internal customer. Commercial image sensors were in the market for 26 years before the revenues passed the $1 billion mark. Now they are approaching $7 billion. How did it happen? Why now?
The Road to Success
Many readers may be surprised to find out that CMOS image sensors were actually conceived before CCDs. CMOS dates to 1967; CCDs to 1969. There have always been applications for these early CMOS arrays, but for the most part, the lithography of the time was too coarse to take full advantage of the design. CCDs offered a simple and robust alternative, and were the dominant design from the 1970s until recently. CCDs are still very good for video frame rates, specialty designs like astronomy, and as general purpose designs. There will always be a market for CCDs.
From the 1970s through the 1990s, the resolution of lithography improved by orders of magnitude, thanks to investments in the semiconductor industry. The designers of CMOS image sensors also got more clever and were able to improve the fill factor (the percentage of pixel area used for photon capture) and reduce the noise. Today, the minimum pixel pitch of CMOS arrays is nearly the same as for CCDs.
But the improvement in CMOS arrays by itself was not enough, because the image sensor applications had not grown much beyond early applications like camcorders, security cameras, scanners and fax machines, and machine vision. What helped was that CMOS arrays improved at about the same time that digital cameras and mobile phones were taking off.
The CMOS arrays had two advantages that fit the new applications. Technologywise, CMOS has lower power consumption and can integrate more features on one chip than CCDs can, and are therefore better suited for portable appliances like camera phones that always need to extend battery life and minimize module size.
Market-wise, CMOS suppliers were able to get their arrays designed into new products before CCDs became too entrenched. Moreover, handset makers are a different breed from traditional imaging or electronics vendors, and product models have a short lifetime. This makes for a very dynamic market, open to new competitors. Thus, unlike digital camera makers, most of the leading names in mobile phones are not from Japan and haven’t made their own image sensors: LG Electronics, Motorola, Nokia, and Samsung (although Samsung is now also making image sensors).
The challenge now is to make a transition to slower growth in the camera phone business without significant disruption to the suppliers. Handset sales have passed 1 billion units per year and are still growing, but at a slower rate. The adoption of cameras in handsets is also growing, oftentimes including a secondary camera, but that too is slowing. At the same time, the migration to higher pixel counts is slowing. Meanwhile prices continue to decline, so the revenue growth will be slower.
The same slowing trend is seen in other applications. Digital cameras are now reaching market saturation, with the remaining growth coming mainly from expansion of the digital SLR segment. And while image sensors for optical mice were an overnight success for Avago (then Agilent) and STMicroelectronics, that market is now saturated.
The growth may be slowing, but the year-to-year gains will still remain substantial. For example, in 2005 and 2006, the industry added over $1 billion each in revenues. This is phenomenal growth for an industry of this size. Much of this is, again, thanks to camera phones and digital cameras since even small growth in a big market amounts to tens to hundreds of millions of dollars in new sales. A new generation of laptops with embedded webcams will also provide substantial growth. But there is also a lot of opportunity in what we call the specialty markets: security cameras, automotive, and digital radiography.
The security camera business is experiencing a convergence of multiple trends. First, after many years of analog closed-circuit system designs, vendors are now selling more digital and network-friendly systems. Japanese CCD suppliers Sony, Matsushita, and Sharp still strongly dominate the sector, but CMOS suppliers like Pixim and Omnivision are there too. In security cameras, Pixim takes advantage of several features of CMOS arrays to provide very wide dynamic range. As the security market goes through a growth spurt, CMOS suppliers stand to do especially well.
Automakers have looked at using imaging in cars since the early 1990s, but the industry is very conservative and deliberate, favoring suppliers with a lot of patience and the means to last through long design cycles. There are now several new automotive applications that are appearing in vehicles and seem promising. These include: rear and corner cameras as monitors for the driver, blind spot warning, lane departure warning, automatic headlight dimming, driver alertness monitors, and video-based event recorders. Other applications that are less promising for imaging, but are still being evaluated, are: collision avoidance systems, night vision monitors for the driver, and air-bag occupant sensors. Some companies to watch are Canesta, Omron, and Sensata Technologies.
Digital radiography is one of those applications that is superior to its predecessor in every way. It replaces film x-rays for applications from mammography and dentistry to industrial inspection. Many years ago, I heard doctors claim that digital x-rays could never be as good as the film x-rays, which were first developed a century ago. Now x-ray imaging is the last medical imaging technology to be digitized.
Why so long? Once again, the reasons are due to both technology and markets. There were some breakthroughs by companies like Canon, dpiX, and Perkin-Elmer in the 1990s that helped perfect the amorphous silicon technology for large format arrays. The quality of displays also improved substantially. But another challenge is the way that radiography is financed. Large hospitals can make the switch to digital imaging, and this is being extended to more networking of image files. But the individual medical or dental practitioner finds it less compelling to switch from film to digital. It is not just a matter of the large capital costs for a small business, but the digital systems add to the software that an office must manage, on top of the other software and administrative overhead that takes away from a practitioner’s core practice, which is to treat patients.
Scientific and military applications were among the first for image sensors, but there continue to be advances even there. CCDs continue to dominate for the lowest noise characteristics. In 2006, Dalsa delivered the world’s largest single-chip image sensor, at over 111 million pixels, for the U.S. Naval Observatory. e2v technologies was awarded a contract last year to deliver a mosaic of CCDs to form a 1 billion pixel array for the European Space Agency, to be launched aboard the astronomical satellite Gaia in 2011. Low light level CCDs are also being applied in a variety of biological applications related to fluorescence microscopy.
CMOS arrays, on the other hand, are well suited for high frame rates, and especially those with large formats. These arrays can now provide stunningly high frame rates of up to 2,000 frames per second for 1.3 megapixel formats, and up to 200 million frames per second using multiple arrays. The high bandwidth of CMOS arrays is also used in digital cinematography; such as using three 8 megapixel arrays (one for each color) at 24 frames per second. In fact, Panavision SVI claims the highest bandwidth for any application, at about 1.5 gigapixels per second, for its CMOS array aimed at cinematography.
Key Image Sensor Suppliers
The Japanese CCD suppliers dominated the image sensor market for many years. Their production is still growing, for the most part, but they have had to share the spotlight with the larger of the new CMOS image sensor suppliers. Most notable are Micron, Omnivision, and ST. But there is a lot of shuffling going on as the volumes ramp and prices fall.
An interesting trend is that many new CMOS companies now use a fabless strategy and it is not limited to small, start-up companies. Omnivision started that way, and is now a leading supplier. Kodak is a longtime CCD supplier, but has a fabless approach for its new CMOS effort. And this summer, ST looked to 12-inch foundries to take over production from its own 8-inch image sensor fabs, to help remain competitive.
The next several years will be very exciting, but also challenging, for the suppliers of image sensors. Slowing growth will force some hard decisions, but continuing innovation and advancing applications suggest that the story is far from over.