Why don't digital cameras have better dynamic ranges?

Question

ISO in digital cameras is a digital amplification of the signal. If your camera's native ISO is 100, setting the ISO to 400 would make the sensor amplify the brightness 4x.

So if ISO is digital, does that mean cameras artificially limit the dynamic range?

Let's say you shoot a photo at f/8.0, 1/500th sec, and ISO 200. The shadows are clipped. Now you shoot a photo at f/8.0, 1/500th sec, and ISO 400. The highlights are blown out. But ISO 400 just amplifies the signal, and the highlights are well below the absolute maximum limit of the sensor since they look good at ISO 200. Why does the camera cut off the whites early then?

What's stopping camera manufacturers from mapping a wider range of sensor values to RGB(0-255)?

Answer 1

ISO usually is analog amplification, not digital. It is a multiplication factor applied to all sensor data equally.

Imagine you shoot a photo at ISO 100. You get values from 1 to 150. Fine, lets up the ISO to 200. Now you might get low values starting with 2, but the highs will be 300, ie. clipped at 255. (if your lows really are blocked at 0, higher ISO won't help, but, let say 1-10 look completely black to you as well, and expanding that to 2-20 may let you see graduations [numbers freely invented])

The point is, your highlights fit into the range when recorded at 100, but amplified they don't anymore.

If the dynamic range of your scene is low enough to fit in your range (<= 8 EV if we're talking of 8 bit), you will indeed not get clipping if you expose ideally. And in reality, sensors have more DR than 8 EV and have to apply some contrast curve when converting to JPEG. so there is wiggle room with contrasty vs. more flat curves. And when you record RAW you can manipulate the data later to your hearts contend, even applying locally different adjustments (see tone mapping, HDR).

Answer 2

A part of what enables a sensor to have a large dynamic range is the physical size of the pixel. During an exposure, light hits the pixel and creates an electrical charge which is stored by the pixel. The larger the pixel, the more charge it can store during the exposure. The larger storage capacity, the less likely the pixel will reach it's full capacity during any given exposure. This is what causes the highlights to clip even when the exposure is spot on.

There are plenty of cameras that have more than 15 stops of dynamic range. However what's stopping manufacturers from say 25 stops or 100 stops? It is because as I read in an article on dpreview.com "...since dynamic range - at the pixel level - cannot exceed the bit-depth of the ADC." The ADC is the analogue to digital converter.

Answer 3

Without entering into physical limitations and manufacturing processes; there's a practicality issue: the resolution in the luminance domain.

Imagine that instead of measuring light intensity, we are measuring distance with a ruler. We can make that ruler as long or short as we want (that is the dynamic range), but we are limited on how many ticks we have marked on it (say 1024 marks for a 10 bit color depth).

So if we make the ruler longer, the ticks become more and more spread apart. That is, if you want an unconditionally greater dynamic range you are going to miss finer details (you'd see posterization in what otherwise would be smooth gradations).

Answer 4

What's stopping camera manufacturers from mapping a wider range of sensor values to RGB(0-255)?

Nothing stops them — in fact, some cameras do exactly this. For example, Fujifilm has "DR 200%" and "DR 400%" modes. The nominal ISO is 400 or 800 but the image is actually exposed at ISO 200 and then "pushed" in a way that brightens the overall exposure but preserves highlights. Pentax calls the exact same thing "highlight correction".

Why isn't this the default? Well, in order to treat the highlights specially, the amplification needs intelligence. It's much easier to do that later in the pipeline, after analog/digital conversion when you're just working with bits. On the other hand, increased ISO is usually implemented as analog amplification earlier on. This generally results in cleaner results. So, "DR extending" modes often show more noise in the shadows than taking the same exposure with the same nominal ISO and the feature off — let alone taking a longer (or faster aperture) exposure at lower ISO and letting the highlights go.

Answer 5

You are limited to 0-255 values because you use an 8-bit storage format. If you record RAW images, you will have access to sensor's full ADC resolution.

But then, why is my ADC, for instance, 10-bit (0-1023) instead of, let's say, 32-bit (0-4294967295)?

Camera manufacturers could use ADCs with more bits of resolution, but it would be useless. The lower digits of the values you get would be just random noise anyway.

Here's an attempt at a specific example:

You can cover your lens so your pixels should be all zero. Unfortunately, they won't be because they will have some noise. Let's say their values vary randomly between 0 and 4. This means that, when you use the camera in practice and you got a value of 2, you won't be able to tell if there's actually some light there or if it's just random noise. Effectively, your dynamic range would be your max image value (4095 for 12-bit ADC) divided by 4, which is 1023. This means that your manufacturer could have just used a 10-bit ADC without any loss of useful information.

Answer 6

Mind that typical camera sensors work in a cumulative/integrating way - think of either a tiny capacitor either being charged by a voltage source with a tiny photocell (photodiode, phototransistor, CDS cell, whatever!) in series, or a pre-charged capacitor shorted by a photocell. IIRC the second way is actually what is happening in typical sensors.

This structure will stop integrating correctly when the capacitor is empty (or charged to the available voltage). The meter needle is pinned, so to speak.

On-chip capacitors are relatively large, so you cannot make them arbitrarily high in capacitance; also, you would end up with smaller voltage differences for the same amount of light integrated, and thus with more readout noise.

A typical sensor pixel is not a non-integrating photocell with a buffer and a sample and hold circuit (activated at readout) attached - such an arrangement would likely be much noisier and would also sabotage the meaningful usage of a traditional shutter. Also, would be much more complex (see below).

One might be able to put some kind of logarhithmic amplifier between a low capacitance photocell and the actual capacitor ... complicating the circuit much. Also, you'd need to shield any non-trivial analog circuitry in the sensor pixel from light very thoroughly: Silicon semiconductors are both somewhat translucent to IR, some of the structures used in making especially MOS/CMOS parts are actually quartz (which is translucent to a lot more), and all that stuff is photosensitive even if not intended as a photodiode or phototransistor. Log amplifiers especially are very dependent on circuit balance, thermal or optical interference is very good at upsetting it...

A path that maybe hasn't been explored much yet would be to use a different filter structure on top - for example, an R, G, B, R+ND8, G+ND8, B+ND8 mosaic... Blooming might ruin the fun, though.