When you learn digital photography one of the first things you are told is ETTR: Expose To The Right!
For those who may not know it, ETTR is the consequence of the way a digital image chip works. Those chips can be seen as containing million of “buckets”, one for each pixel. These buckets collect the incoming photons and once the shutter is closed (the exposure finishes) they are counted and their number is transmitted to the computer inside the camera that will create the JPEG image or, if you have decided to use the RAW file, will be associated to that particular pixel.
Since the digital chip has a linear response to the incoming photons, i.e. it keeps accumulating them during the exposure time, if you double the exposure you will get twice as many photons. Given that a digital camera can represent the number of photons using 12 bits it means that any given bucket can contain up to 4096 photons. By halving the exposure you will get a max of 2048 photons (assuming you would get exactly 4096 in the original exposure).
Now this is crucial because it means that increasing of one stop (doubling the exposure) you can capture 2048 more photons, ie the highest stop will get you 2048 degrees of precision. As yo halve the exposure you get half of that (1024) and if you go 5 stops down you get 128 degrees of precision. In other words: you get much more nuances in the highlights than in the dark areas. This is the reason why you need to look at the histogram and select the exposure that shifts it to the right. In this way you will capture as much latitude (nuances) as possible: ETTR! It will be up to the camera software (or Photoshop) to make the best use of that latitude.
Of course, the problem in ETTR is that you end up with an exposure that is flooding the bucket with more than 4096 photons you miss the extra ones, i.e. you blow the highlight. That is why you need to look carefully at the histogram to make sure you are not moving it too much to the right.
Not so anymore!
German researchers at the Institut für Mikroelektronik Stuttgart have published a paper reporting the construction of a CMOS digital image sensor that for each pixel (bucket) has a circuit that as the pixel has reached its maximum capacity (4096 photons) it is reset to 0, hence it is able to receive 4096 more photons, and so on. The computer processing the image will get the info of how many time the pixel has been reset, ie the number of photon actually received during the exposure, that can be much higher than 4096. This means that you will be able to expose even more to the right increasing the latitude of exposure.
Of course that will require an extension to the digital format (you need to include in the number associated to each pixel also the number of resets -that is equivalent to extend the 12 bits to 14 bits if you allow 4 resets, to 16 bits if you allow 16 resets and so on. More than that. The current implementation of the reset circuit, you need one for each pixel, is eating up space in the bucket (pixel) and makes the pixel smaller so that you are losing space. However, the researchers point out the possibility of using a 3D pixel architecture to stack the circuit in the pixel, to keep its original sensing size.
I don’t know when such a chip will become available but I found this an amazing news: it is another example of how ingenuity can push the boundary of what is feasible by taking an alternative approach. Notice that this is something you can only do in the digital world. In the physical world of atoms (silver atoms to be precise) this is not possible!