Review of the Light L16 computational camera
If you read my article about computational photography you will know I am very interested in the Light L16 camera which uses 16 small cameras (with cell-phone level sensors and different focal length lenses) to produce an image they hope will rival high end cameras like DSLRs.
The plan is an excellent one. I purchased the L16 but must sadly report it is "not yet the camera of the future" though I feel the general idea points the way there.
Some of the problems I have had with the L16 will be fixed in time. As a computational camera, they are doing software updates from time to time, improving the images and adding features. All cameras should be this way.
The L16 has lenses set for the field of view of 28mm, 70mm and 150mm on a full frame camera. A single shot uses up to 10 of these mini-cameras. Each camera has a mirror which both folds the lens to make it compact and offers the ability to slightly steer the aim of each one. The sensors are small 13 megapixel as might be put in cell phones.
(The lenses are much shorter than 28mm, 70mm and 150mm, but Light and I will refer to them by those numbers for simplicity, as is often done in small sensor cameras.)
Most of the expected benefits I wrote about for computational photography come from getting more than one image of your scene, either by taking multiple images at the same time (as the L16 can) or by taking images in quick succession of a scene that isn't changing too fast.
The L16 chooses however, to put an emphasis on tiling the images. When you set it wide (28mm) it takes the 5 "70mm" cameras to tile the center of the image -- 4 of them in a square and one in the middle for extra overlap, and it has 4 "28mm" cameras that mostly overlap. From this it is able to calculate am image of around 60 megapixels. I am guessing that it does the following:
- For detail, it uses the 70m frames. The center of the photo gets each pixel from 2 different frames, as do regions of small overlap. Most pixels in detail on the outside come from just one 70mm frame.
- The 28mm frames mostly overlap, and I presume they are set to different exposure values, in order to do high dynamic range. While each pixel's sharp value comes from the 70mm frame, you can learn about overall light values and focus from the overlapping frames. Stereo is used to build a depth map of the image for later simulation of shallow depth of field.
- Tiling should not be that useful for noise reduction, but overlap even with a wider angle shot can identify pixels that are clearly out of range. Most computational noise reduction makes use of multiple overlapped frames shot at very slightly different times, I don't know how the L16 does it.
- If you zoom all the way in to 150mm, it can't tile any more, it can only overlap, to produce a smaller but better image.
The Light team is earnest and keen to improve their camera, and certainly have a better attitude to innovation than the large and slow moving incumbents.
So let's look at the issues:
The L16 is quite bulky for a pocket camera. Especially when in its protective sleeve. It fits in a large men's pants pocket or jacket pocket, but barely. It's the size of an XL cell phone but an inch thick. Everybody wants every camera to be smaller, but since that's not possible, photographers change their goal to "The best camera for a given size."
That's why many serious photographers have a full-frame "DSLR" camera which is the best camera they can get, but is bulky and heavy and requires a bag of bulky heavy lenses, and they also carry smaller cameras, including a smaller mirrorless camera, a "bridge" camera (long zoom single lens camera), a high-end pocket-sized point-and-shoot (like the Sony RX100) and finally a cell phone.
There are a few other sizes out there -- including digital versions of medium format, APS-C and Micro 4/3s DSLRs and more. The lower end point and shoot market has been drying up because serious photographers never wanted those, and casual photographers find their cell phones doing a large fraction of what those can do.
When I'm on a serous shooting trip, particularly a car-based shooting trip, nothing but the big SLR and bag of lenses will do because I want their impressive quality and features. But they are too heavy to carry for just walking around, and are even too big for air travel with just a carry-on bag.
For lightweight travel, popular options have included the mini SLRs for those who want the serious features of being able to change lenses, or the bridge cameras which can "do it all." These don't fit in your pocket -- you have to make the decision to be wearing a camera on your neck.
Sometimes you want to be able to just walk around casually, with no camera on your neck, not looking like a tourist, so you buy the best camera that fits in your pocket. For I and many others, this has been the Sony RX100 line for the last few years.
The cell phone camera is special. It is "free" -- as though it has no weight or size -- because you are always carrying it, to be a smartphone, not a camera. But as the saying goes, "The best camera is the one you have with you" and it's very nice that the photo abilities of phones has been going up.
The L16 hopes to be something as good as a DSLR that can still be pocketed. It's not there yet, and it may not get there for some time. Until it gets there, it has to be "the best camera I can pocket" and the problem is it may not be there either. Price plays another role -- at $2,000, the L16 is much more than almost all the non-SLRs out there. For some photographers, they want "the best" in a size class regardless of price, but most do care about price, especially in their #2 camera.
The L16, while trying to be small, feels big and bulky because you immediately compare it to the high end P&S cameras it competes with and the cell phones it reminds you of. It has no grip, and you can't let your fingers slip around to the front while gripping it for reasons listed below. As such it simply isn't a pleasant camera to hold and shoot with.
This is the primary consideration for most. The answer is "it depends." Some images are quite good, and with tons of pixels and the benefits of computational photography. Unfortunately, not all those pixels are of the same quality, as you might expect when combining different cameras like this. Images have sharp sections and fuzzy spots. In some cases this can be fixed manually, but nobody wants to have to do that.
In some ways, the quest for megapixels may be seducing the L16 to the wrong choices. The problems is that 13MP is not enough for most people today, when you can get very good 18-20mp compact cameras. This means there is no choice but to do the tiling, and this gives you too many pixels with flaws.
To fully judge the image quality I need to shoot in more conditions. Light knows that the biggest complaint is on low-light image quality. Shadow regions are noisy, as you expect when your source image chips are so tiny. Indoors, it wants to shoot at 3200 ISO all the time, and even then often has a slow shutter speed that leads to blur. In daylight, it's better.
Ease of Use
The L16 fails in too many ways on ease of use. They decided to go with a cell-phone like UI, which has advantages and makes use of the touchscreen. In the end it faces too many problems, and too many functions are "too many clicks." Worse, they are touchscreen clicks, which require the photographer focus their attention on the UI, not the scene. One reason traditional cameras have so many buttons and dials is that even though they take more time to learn, photographers soon learn to use them without looking at them or even thinking too much about it. GUIs and touchscreen interfaces are great for learners but not so great for the experienced.
GUIs of course sometimes produce low-thought interfaces, and one example is the use of the touchscreen to indicate where to focus, which the L16 does. (Recently, Light added a way to lock focus which I have not yet tried out.)
In general the touchscreen interfaces are slow to use. Setting exposure time or ISO with a virtual slider is not nearly as sure and fast as doing it with a dial.
The physicality of many regular cameras is not a bug, it's a feature. While on phones replacing all with a touchscreen has been a good choice, it is not in cameras. People want to make certain changes fast or their miss their shot.
The greatest issue, though, is an ease of use issue related to the process for getting photos and the speed. All you see when you take a photo is the preview. To do a basic processing of an image so you can look at it with pinch and zoom requires that you manually request processing, and wait about 6 to 8 seconds! And you're not done. To really see the images, you have to transfer them to a computer, and do full processing on them which takes around a minute per photo. I'm one of the people who shoots mostly in jpeg because the fact that processing RAW takes a few seconds extra per photo is too much.
The speed issue is going to be a tough one. Most big camera vendors design their own custom chip for their camera line. They do the slow things in hardware. That lets them do everything in camera and fast. Once you have fast you can't go back, but the L16 has no way to get it a custom chip any time soon. Perhaps someday when it or something like it becomes a computational camera.
In addition, the post-processing UI is far too complex. To do it right, you need to set a variety of manual settings on each photo. In particular, you choose the main focal point and depth of field for each photo during post-processing. To some people that sounds great. The ability to change these things after the fact is of course useful. But it should not be required.
It is not strictly required, you can use the defaults and get a photo with large depth of field. Because the camera takes photos with small sensors, the native images have a large DoF. Shallow DoF is done artificially, using depth information learned from the multiple cameras. But you can't set your desired DoF when shooting. I want to set my DoF when I am composing the image. If I want to change it later, great, but I want to be able to pick it, or understand the constraints the light puts on it when I shoot.
Setting the focus spot and DoF is cumbersome and time consuming, and worse, the software gets it wrong sometimes and you have to manually tweak it.
Setting desired F/stop (really DoF) is in the roadmap.
Custom settings memories
Cameras today do have complex interfaces. That's why all the high end ones come with "memory" settings which let the user create their own custom environments and then quickly switch between them. Frankly, the big players also get this wrong, giving the user a few numbered memories. Custom configurations should be unlimited, and named, especially on a touchscreen camera. Then I should be able to quickly switch among my favoured modes. To get named modes, the UI to build them should be on the web, and then the result transferred into the camera. All the common "scene" modes found on cameras should be done this way, letting the user pick what they like.
As noted, manual focus has just been added, or rather a focus lock. Face detect focus was also added. They could do a true manual focus but it would never be the same as a focus ring. If you have ever seen a pro photographer who is good at manual focus, it is quite impressive.
Another option might be triple-tap for "hold focus on the object I triple-tapped on" when it is moving in the scene. Sports and action shooters like that.
Buttons and dials
The L16 would be improved with at least a couple of physical buttons and at least one physical dial. In many cases the button might be used where you hold it down and then use the touchscreen as a wheel or slider. The "wheel on touchpad" interface made famous on the iPod has merit. The main reason for a physical wheel is to offer an interface where you don't need to look at the screen to make an adjustment. Or rather, you are looking at the screen or viewfinder, but at the scene. You are not attempting to put your fingers in the right place for a GUI menu. The buttons could be soft buttons, put next to the screen (including on the top of the camera) so they can get a label that changes.
Consider that a high end DSLR like the Sony A7 has 5 wheels, 2 switches and 13 buttons, plus 2 wheels and a couple of buttons on many of the lenses. You might think this is overkill, but there is a reason customers want this. Only physical controls can get muscle memory and use without thinking. And speed is everything in some kinds of shooting.
The camera has a built in battery you can't change. If it drains, you can't shoot, other than by hooking up an external battery via the USB-C port. That's ugly -- you need a way to mount that external battery right on the camera if you can. You can't hold it in front of the screen.
The camera runs Android, which I am fine with, so long as I can always just press the shutter to shoot no matter what I am doing. But Android does not suspend to a zero power mode. Most cameras will suspend to zero power mode after you leave them for just a couple of minutes. The L16 goes to a mode where it drains about 5% of the battery each day. This is not going to kill you day to day, but if you leave the car in the house or back of the car without turning it to full off, you can be in real trouble.
If you turn it full off, it takes 20 or more seconds to boot. Forget about grabbing it from the back to take an urgent picture.
One option might be to have a fast standby written to the buffer flash memory, or even a fast-boot to a recording of the post-booted state. The other answer is to have a physical way to clamp a USB-C backup battery onto the camera.
A more serious problem occurred to me on a recent trip. I took a flight and near landing pulled out the camera to shoot a nice view. The camera was warm and showed 8% battery. It then declared low battery and refused to shoot. Even after plugging it in to a portable power pack, it showed the charging light but refused to shot. The shots were lost! That's seriously unacceptable, though we have all seen it happen on phones. It should both detect if something is draining the battery and stop it or alert about it, and it should be able to shoot even if low, and should definitely be able to shoot once plugged into a power pack.
The camera does not take memory cards. Presumably this is to help it be waterproof in future. My past history shows a real problem here. You must travel with a laptop, and you must pull images off the camera often. And then you must copy them to memory cards, if, like me, you want to have 2 copies of every photo. If you have had a device or camera stolen or lost, you know why. I always travel with 3 256gig cards, which is enough for a trip. I copy the cards to my laptop and keep the images on the cards too. When I get home I copy the images to 3 different disks -- only then do I re-use the cards. (Cards are so cheap that some just never erase them, they remain as backups.)
It does not shoot video. It will shoot 1080p soon and 4K later, but just from one internal camera, so it will probably be similar in quality to a cell phone -- though with the ability to shoot at 70mm and 150mm fields of view, which the cell phone can't. It is uncertain if it will surpass the 4K performance of the high end P&S cameras with so-called 1" sensors.
Once they do video I will suggest what I suggest to all cameras -- record multiple audio tracks. If I plug in an external microphone, record both that microphone and the internal one. If you support wireless mics do that to. Don't throw away information.
Would be nice to add one. You need it for shooting in bright sunlight, and for certain types of action shooting. Unfortunately in a camera where all is done with touchscreen controls, you would have to not look through the viewfinder for any adjustments. In addition, looking through a viewfinder usually will plant your nose on the touchscreen -- a problem most cameras also face.
See below on USB-C
Fingers in front of camera
The front of the camera is all lenses. That means you can't put your fingers around the camera so that some come to the front. They will block a lens. The camera notices this and gives you a warning. The warning is good, but the idea of a camera you must hold only on the sides is hard to get used to.
This causes a couple of other problems. When you put the camera down, you usually want to put it on its face so you can look at the screen. Of course it then complains about a lens being blocked.
There is also a problem shooting through "holes." I could not shoot with this camera through leaded glass. Any other camera in the world could shoot through it but the L16 complains about it. It's like you have a 4" wide lens on the front and most have a hole that big to shoot through.
Other cameras solve this in a couple of ways. Small lens cameras usually just find a hole large enough to shoot through in dirty windows and screens. If you have a large lens, you can actually shoot through even close mesh screens by opening the camera up wide. You lose a bit of contrast but the mesh does not show in your image.
The L16 could solve this, by accepting the mesh and blockages, and combining images in a way to remove them if it can, and warning you only if it can't. Even fine mesh might work that way, but it would be difficult.
Physical ports and mounting
The camera has a tripod screw hole like all cameras. Unfortunately, the USB-C is right next to it. Almost all serious photographers put a quick release plate on the camera, and this would possibly block that port and certainly block the accessory port. The port is on the bottom, not the side, which means you can't charge it without taking it half-way out of its protective sleeve.
Speaking of which, I wish that cameras just modified their bottoms to have the dovetails of Arca Swiss plates on them. On the bottom and on the side, since that's better than having an L bracket. An L bracket would be pretty challenging for this camera as well. (A tripod screw hole on the side would be a passable fix.)
There is no flash shoe. That could be fixed possibly with some sort of radio controlled flash in future. As the old saying goes the factors, in order, that make a great photograph are 1: Photographer, 2: Lighting, 3: Lens, 4: Camera. A serious camera needs ability to control lighting. I would recommend just using the protocol of some existing radio or IR controlled flash, and let people buy flashes for that camera line.
The USB-C port should be exploited as a way to add all sorts of hardware features. To do that, though, you need points on the camera that you could mount things to. Ideally a way to mount them and automatically plug in to the USB, without a wire. On the existing L16 the tripod screw is the best bet, not really good for a viewfinder or flash.
Next to the screw is also an 11 pin accessory port, currently unused. It's not well placed (a mounting plate will surely cover it.)
Think of the things which could be added with the port and a place to clamp or screw in: * A viewfinder * A tilt/swivel screen * A grip for better holding * A grip or plate with extra buttons, dials and switches. Or more than one. * A flash shoe, or flash remote ports * Other lens/camera combinations. You could even design the L16 as a smaller L9, with a clamp-on module which adds extra lenses/cameras to get longer zoom range or other abilities at the cost in bulk. * Video and audio inputs and outputs (there is one on the base at present.) * Extra batteries * Extra storage or flash-card adapter * Shutter release on cable
Being the best
As noted serious photographers own multiple cameras, each one "the best camera for its size class." Here are the things the L16 might be able to claim that title for.
- It's the highest megapixel barely pocketable camera.
- It may be the only pocketable camera that can deliver 50 megapixels with 70mm field of view
- High-pixel camera with computational benefits -- change focal plane and depth of field after shot
All cameras should understand the value of being a platform, which is to say allowing users and developers to build apps which can improve the Camera. Since this is very rare, it is not a flaw in the L16 compared to other cameras, but it is a bigger issue because the L16 is an android device. All the mechanisms are in place to allow an app ecosystem.
Apps for cameras could do several things. They could create shooting modes customized to particular needs. They could run shooting scripts for everything from time lapse to new ideas in computational photography. They could potentially offer new UIs if given deep enough APIs into the system. They could also open up new technologies like wireless microphones and more.
Apps let other people improve your product. You will never have all the best ideas -- if you let the community in, you win. Admittedly to do some things requires fairly low level APIs which require work to support, but it's worth it.
The long-term roadmap calls for this.
The L16 is not waterproof, though its design (with no slots for battery or storage) suggests that it might be in the future. The nice thing is that, presuming it is strong enough for the pressure, this camera could be taken underwater with a very simple case which is just a plexi box, as long as capacitive touch goes through the back plate. Underwater bags for cell phones exist which just have a flexible plastic sleeve that wraps tightly around the phone -- it must be able to take the pressure -- and you can use a touchscreen that way.
Another alternative is to make a waterproof control box with the necessary buttons and controls (shutter, power and touchscreen) which talks to the camera via USB-C or bluetooth. This allows the camera enclosure to be simple and cheap. Camera enclosures for SLRs tend to be very expensive -- $500 to thousands, which could make the L16 an interesting solution. Strobe control is needed on high end enclosures.
Of course, if they made the future L16 waterproof down to even a modest distance it might gain some traction in that space.
There are things that I want all cameras to have, and the L16, which embraces being a computer as much as a camera, should be where these features arrive, or get made possible through an API. Here are some features that would be nice to see:
- Bluetooth flash control
- Bluetooth microphones (high fidelity, and more than one) for recording audio on video
- Audio annotations of photos, automatically turned into text in post-processing by today's good voice recognition
- Voice commands as an alternate UI. "Shooting mode nightclub!"
- Remote shooting application (control from phone).
- Quickly show me the best-focused part of that picture I just took, so I can figure if it's blurry. Or perhaps just a blurry alert.
- More of the computational tricks I outlined in my article including those that come from shooting several shots in time as well as with multiple imagers.
- Accelerometers and gyros for image stabilization and also for panorama modes (as found in my cell phones.)
Let me say that the Light crew has been exemplary in reaching out to me about this and other feedback. They are very motivated to improve their camera.