CNET también está disponible en español.

Ir a español

Don't show this again

Cameras

Light's bizarre -- and a little bit brilliant -- take on camera tech

This will either be the camera of the future or another toy to send to the island of misfit cameras.

There have been a lot of interesting takes on redesigning the camera for the 21st century, from the dual-lens Kodak V570 in 2006 to Lytro's original Light Field camera in 2012 (which got a slightly more traditional overhaul this year) to phone attachments such as the DxO One . But two-year-old startup Light, headed by Dave Grannan (former CEO of the company whose speech recognition eventually became Siri) and Rajiv Laroia (former CTO of the company that developed key LTE technologies), has come up with one of the most novel I've seen in a while.

Though the company showed off the technology earlier this year, it's now officially announcing it as a product and hoping to generate some advance sales. Light bills the L16 as the "first multi-aperture computational camera that packs DSLR quality and capability into a device that fits in your pocket."

As a camera, it has decent specs: a zoom range of 35-150mm (4.3x zoom), with a completely touchscreen-driven interface on the 5-inch LCD. It does HD video and has built-in Wi-Fi, outputs raw files in DNG format, has only internal storage (128GB) and a nonremovable battery (charged via a USB Type-C connector) it rates at 400 shots. It has an LED light for continuous, rather than strobe, flash. There's a microphone and headphone jack, too.

It's not small -- it'll fit in a pocket, as long as it's a loose jacket pocket -- but as with most large enthusiast compacts, that makes for a more stable, comfortable-to-hold camera. The price? $1,300 (directly converted, £850 and AU$1,804) if you order in advance, $1,700 (£1,110, AU$2,360 converted) if you wait till it's fully baked.

Note that I haven't metioned two key specs on the camera, resolution and aperture. That's where things get complicated.

The technology

A "computational camera" refers to the way it generates the photos. The L16 uses an array of 16 tiny, individual camera modules. Each module is comprised of two primary components. The first is a lens composed of molded-plastic elements that uses not-so-new folded-lens technology. Rather than emerging from the front of the camera in a straight-through optical path, like a typical lens, folded lenses use a mirror or prism to reflect/refract the light at a 90-degree angle, sending it into optics that reside within the body of the camera.

Each lens in the L16 has an f2.4 aperture and the angle of view of 35mm, 70mm or 150mm lenses (35mm-equivalent). There are five 35mm lens modules, five 70mm modules and six 150mm modules.

Each of these lenses is pointed at a 1/3.2-inch sensor with 13-megapixel resolution and a pixel pitch of 1.1 microns. Those are the same specifications as the sensor in a phone; in this case, like the sensor that's in the Samsung Galaxy S4 .

When you press the shutter button, the L16 takes 10 simultaneous shots with algorithmically selected camera modules, then tiles five of them into a single image: four quadrants plus one in the middle overlaying the center. (Each lens takes a photo of the full image -- there's no cropping involved.) When you "zoom," it combines the tiles from different lenses to simulate the angle of view of the zoom you've chosen -- that's computational zoom.

And that's why there's no fixed resolution; the resolution varies from 13 megapixels to 52 megapixels, depending upon what focal length you've chosen.

The redundant photos help the camera compensate for the slightly different perspectives caused by the offset among the lens modules. They can also be used for noise reduction the way a multishot HDR layers different exposures.

Each lens focuses at a different distance, and those tiny sensors mean each camera has big area of sharpness (depth of field) so that the final image is uniformly sharp from front to back. The camera then uses a Gaussian blur algorithm to defocus to the depth-of-field of the aperture you've specified; that's how it turns fixed-aperture lenses into a multiaperture camera.

My take

This architecture gives Light a real flexibiilty for future products. Want a pocket-sized 400-600mm camera? Use different camera modules. Want a smaller 35-50mm camera? Drop some of the modules from the L16. It's not quite that simple, but you get the point.

There are some ways in which the company can improve on it, too. Using sensors without color filter arrays would be one. That way every sensor would operate at its maximum dynamic range rather than a fraction of it. (For example, with a Bayer CFA, the most common type, only 25 percent of the pixels capture red wavelengths, 25 percent blue and 50 percent green.) Light says it's "working with a sensor manufacturer to procure sensors that do not have CFAs, but in the current prototype all sensors have CFAs." It's not clear what will be in the final product.

Because you never need a small aperture to get everything sharply in focus, you don't have to worry about diffraction limiting -- the physics that make photos get softer once you've stopped the aperture down past a certain point (like f16). And the molded-plastic lenses are certainly cheaper to produce than glass; whether or not they're better as Light claims will come down to testing.

There are some potential weaknesses. The camera also uses rangefinder-type (distance-to-subject) autofocus, which I think is less effective for focusing on moving subjects during continuous shooting, and Light hasn't released a target burst rate spec yet.

Quality, of course, will be the real test. I've had the knee jerk reaction of "how can those tiny sensors deliver what Light claims is the same photo quality as a full-frame camera?"

They're asking people to take a lot on faith for the preorders, of which they'll require a $200 (£130, AU$280 converted) deposit. And the prices seem awfully high to me. At that price, the appeal is for professionals who can spend some money on another camera to play with, but I'm not sure consumers will be willing to fork over that much money unless it really gets raves. And even then, I think it's more likely to eat sales away from cameras like the Sony Cyber-shot RX10 II.

I'm pretty confident that the product will ship, in some form or another; the founders have a good track record and have contracted with Foxconn for manufacturing. And I can say for sure that this will be one of the most interesting cameras to test and shoot with that I've seen in a long while.