We’ve seen huge breakthroughs in the computational photography techniques in the latest smartphones, as well as the launch of some excellent small sensors in more traditional cameras. Does that mean that bigger is no longer necessarily better?
To answer that question, let’s look at why big sensors tend to produce better image quality and what smaller-sensored cameras and smartphones are doing to close the gap.
Benefits of larger sensors
The three main benefits that prompt people to move to larger sensors are the ability to blur the background with shallow depth-of-field, exhibit less noise in low light situations and the ability to capture a wider dynamic range between the brightest and darkest areas of the image. Since these all primarily stem from getting more light, it presupposes that you have a lens with a wide-enough aperture to let in enough light, but this is usually possible. Collectively, these three factors mean that large sensor cameras can usually produce better-looking images. They can’t make up for the photographer, of course, but if used by the same person, the bigger sensor usually ‘wins.’
Computational photography allows small sensors to imitate the effects of a larger sensor, while leaps forwards in sensor performance can help small sensors produce better-looking images. But do these advances mean you no longer need a larger sensor?
The most visible example of computational photography in widespread use are the portrait modes in the latest smartphones. These use a variety of techniques but fundamentally they make some attempt to assess the depth in the scene, then apply differing amounts of blur at different depths, to simulate shallow depth-of-field.
The results aren’t always totally convincing and won’t necessarily satisfy the kind of demanding photographer that looks for particular character to the out-of-focus rendering (bokeh) of their lenses. However, as processing power and machine learning continue to improve, the results will only get better. And for many applications, will quickly exceed the threshold of being considered ‘good enough’ for an increasing number of people.
|This image was taken using Portrait mode on a modern smartphone. As well as simulated shallow depth-of-field, computational photography has added artificial lighting effects to the subject’s face.|
Computational photography can also help with low light and high dynamic range scenes. Again, it’s increasingly common for smartphones to be able to shoot multiple shots, then combine them. The processing is getting sophisticated enough that this can even be done with moving subjects, without the user even necessarily realizing that this is what their phone is doing.
Because most noise in photography comes from the randomness of light, shooting the same scene again and again helps to average out this randomness, just as rolling a die repeatedly helps you get a clearer picture of whether it’s weighted.
The results will start to rival the output of
The same can be done in good light, sampling the scene several times (sometimes with different exposures), to extend the dynamic range beyond what could be achieved in a single exposure.
Unlike the shallow depth-of-field effect, which is simulated, these multi-shot techniques directly compete with larger sensors, since they allow the image to be constructed from more light. At which point, the results will start to rival the output of larger sensors, so long as the combination of images and movement cancellation is clever enough.
More advanced sensors
We’ve seen some very good small sensors in recent years, whose performance can narrow the gap with those above them. The adoption of technologies such as back-side illumination have allowed sensors to become more efficient (capturing more of the light that’s shone on them). Again, since most of the benefits of larger sensors come from them generally receiving more light during any given exposure, an increase in efficiency can help smaller sensors narrow this gap.
|G1 X Mark III
F3.2 (F5.2 equiv)
|G7 X Mark II
F2 (F5.5 equiv)
The Canon G1 X Mark III offers an APS-C sensor to the G7 X Mark II’s 1″ sensor, but the latter uses a more efficient BSI design.
This only goes so far. Partly because these technologies are also likely to have some benefit when they ‘trickle up’ to larger sensor sizes. This widens that gap back to the extent you’d expect, just based on the size difference. Also, the gaps between most popular sensor sizes are proportional to at least two-thirds of a stop, which is more of a difference than technology advances tend to compensate for.
Small sensor advantages
There are also areas in which small sensors offer an advantage over large ones, with readout speed being the most obvious one. In general it’s quicker to read out the data from small sensors, which brings a series of benefits. The most obvious is that it allows improved video quality, either through being able to read all the pixels, enabling more detailed capture, or simply through a reduction in rolling shutter.
|The RX100 V’s stacked CMOS sensor with DRAM incorporated into the chip allows faster readout, which boosts video and autofocus performance.|
The other advantage, of course, is that it makes it quicker and easier to offer many of the computational photography benefits discussed earlier in the article.
There are ways to improve the readout speed of large sensors too, such as the stacked CMOS sensor that ‘trickled up’ from Sony’s smartphone and 1″-type sensors to the full frame a9. This incorporates readout circuitry and RAM into the sensor to allow faster readout. But this comes at a considerable cost premium, as these chips are extremely complex and time-consuming to make, making it difficult for large sensors to match some of the capabilities of smaller chips.
Convenience of small sensors
The other potential advantage of small sensors is that they allow cameras to be smaller. This tends to means lenses with smaller physical apertures, which is one of the main things that hold the image quality of small sensors back, but the trade-off is cost and convenience. So, although you can buy mirrorless cameras with large sensors, it’ll be the ones with smaller sensors that will provide the smaller overall package, most of the time.
For all the bluster you sometimes encounter with people claiming to be committed enough that they always carry a full DSLR kit around with them, most people find it’s easier to carry a small camera with you. And you know what they say about the camera you’ve got with you…
Is bigger still better?
Overall then, for conventional, single-shot photography, there’s no substitute for making a photograph using more light, and it’s usually easier to give a large sensor more light since it has a larger area to capture light. In that sense, bigger sensors are still better. However, that’s not the same as saying ‘you need a bigger sensor.’
Even though a larger sensor may be able to produce a better result, smaller sensors are getting better and better all the time
Improvements in sensor tech, the availability of large sensor compacts with bright lenses (that help get more light to their sensors), and advances in computational photography allow better images than ever before from small sensors.
At which point, we come back to the question of what’s ‘good enough.’ Which isn’t so much about accepting limitations, but more about being able to attain a quality you’re happy with for what you’re trying to achieve. So even though a larger sensor may be able to produce a better result, smaller sensors are getting better and better all the time, exceeding ever higher ‘good enough’ thresholds such that you may reasonably conclude that you don’t need any further improvement.