Shooting skills of ISO sensitivity in digital photography

Lead: Please note that the documents about ISO sensitivity standard of film and digital sensor are protected by copyright. If you want to see these materials, you need to buy the corresponding ISO standards. In this paper, we only summarized some of the most important principles and conclusions.

Talking about the shooting skills of ISO sensitivity in digital photography

ISO sensitivity, aperture and shutter are called together? Exposure triangle? . Together, they decide the exposure of the photos.

If you have used film, you may notice the number marked on the outside of each film box, which indicates the sensitivity of the film to light. How much is this movie? ISO？ . The larger the ISO value, the more sensitive the film is to light.

ISO plays the same role in digital photography. That is, the ISO number measures the sensitivity of the camera sensor to light. Similarly, the larger the number, the more sensitive the sensor is to light.

The ISO value of the film or sensor is determined by the manufacturer. The International Organization for Standardization (ISO) has formulated a series of related standards, according to which manufacturers determine the ISO value of thin films or sensors. The function of these standards is to provide a unified framework for determining ISO values. For digital cameras, these standards ensure that sensors with the same ISO and film have the same sensitivity. This means that all exposure meters and exposure technologies suitable for film photography are also suitable for digital photography.

A brief historical introduction

The definition of ISO actually has two scales, one is linear scale and the other is logarithmic scale, that is? Asa? And then what? DIN？ . ISO sensitivity was established in 1987, which combined these two ancient scales into one. Linear ISO uses ancient? Asa? Scale, while logarithmic ISO is used? DIN？ Scale.

Do you need a slash when two numbers appear? /? Separate, you need to add the degree unit to the value. For example:? ISO 200/24？ . If only one number is written, it always represents a linear scale. For example:? ISO 200？ .

In the logarithmic ISO scale, the difference of sensitivity between every two levels is 3; However, in the linear ISO scale, the difference in sensitivity is twice that of every two levels.

In fact, logarithmic ISO is no longer used. So in this paper, we will ignore this scale and only discuss linear ISO.

We already know that the linear ISO scale changes exponentially. In other words, ISO 200 is twice as sensitive to light as ISO 100.

ISO values and ISO ranges

When it comes to ISO, there are actually two opinions: ISO value and ISO range. The former indicates a specific sensitivity, which is what we usually call it? How much is ISO? ; The latter represents a sensitivity range. What is the picture quality in this ISO range? Acceptable? .

The ISO range determines the range of ISO values that can be used.

For film, changing the ISO value requires changing the developer, temperature or development time. If the ISO value increases, it is called. Forced rush? ; If the ISO value decreases, it is called. Depressed? .

For sensors, ISO values are achieved by analog or digital amplification. Next, we will discuss this amplification method and its results in detail.

Film ISO

Before discussing the figures, let's briefly mention how the ISO of the film is determined.

The principles of defining ISO for movies are recorded in three independent standards: ISO 6: 1993, ISO 2240:2003 and ISO 5800: 1987. They are suitable for black-and-white film, color reversal film and color negative respectively.

The processing flow recorded in the standard is to irradiate the film to be tested with a series of specified lights, and then develop the film with a specified developer at a specified level. In this series of films, when the specified minimum density appears, the ISO value of the films is determined according to the standard. This is the benchmark ISO for movies.

The ISO of film can be increased or decreased by pressurizing or desensitizing. Sensitization will reduce the contrast of film and improve the dynamic range; Forced flushing has the opposite effect, and it will also increase visible particles.

Digital ISO

The film is exposed and developed to determine ISO. Unfortunately, this method cannot be used for sensors. The digital ISO must be determined by other methods. The specific method of determining digital ISO is recorded in the standard "Photography Technology". Digital camera. Determination of exposure index, ISO sensitivity calibration value, standard output sensitivity and recommended exposure index ". ISO 12232:2006

Like film, sensors also have it? Benchmark ISO. For modern sensors, it is usually ISO 100 or ISO 200. Is the benchmark ISO usually the sensor performance? The best? ISO means maximum dynamic range and minimum noise.

Canon has never published the benchmark ISO for its sensors. However, Canon users believe that some early Canon digital SLR have the largest dynamic range and the least noise under ISO 100, while later models have the best performance under ISO 200.

Nikon has published the benchmark ISO for digital SLR sensors in its early years. Sensitivity marks lower than the reference ISO are L0.3, L0.7 and L 1.0. Most of the benchmark ISO of Nikon's models is ISO 200. Starting with some new models (D3 100 and D7000), Nikon is no longer used? l？ Set, but just mark the lowest ISO as ISO 100. After testing, the dynamic range of these cameras under ISO 100 is smaller than that under ISO 200. It seems that Nikon has also started to learn Canon, and no longer publishes the benchmark ISO.

In a digital camera, the process of converting light into data is as follows: an electronic sensor collects photoelectrons (generated by photons hitting the sensor) through a photon trap. This process will accumulate charges in the photon trap. The amount of charge accumulated in the photon trap represents the amount of light collected by the photon trap during the whole exposure process.

The charge accumulated in the photon trap can be converted into voltage. These voltages generate raw data through an analog-to-digital converter (A/D converter), and finally become pixels to form a photo.

The above picture shows that in a digital camera with 12 bit A/D converter, a photon trap outputs data during the whole exposure process. The left side of the graph starts from 0 (no photons hit the sensor), and the right side indicates that the sensor feels a lot of photons. The blue area in the figure represents the effective signal and the red area represents the noise signal. Note that the abscissa and ordinate are logarithms.

The working process of photon trap is linear, but there is a limit. When the accumulated charge of a photon trap exceeds the limit (the position of the green line in the figure), if more light continues to be felt, it will overflow and even affect the adjacent pixels (sensor halo). The reference ISO is determined by the current collection capacity of the photon trap and the reference quantum efficiency (the time ratio of incident photons to photoelectrons) * * *.

In addition to the effective signal, noise signals will be generated in this process. Are these noises included? Dark current? Noise (even without light, there is current in the sensor), output noise and thermal noise. The noise is shown in the red area in the above picture. In the darkest shadow position (on the left side of the figure), noise has appeared and drowned the signal. As the signal became stronger and stronger, it began to drown out the noise in turn. What do we call it in the area where the signal is submerged by noise? Below the noise floor? .

Note that the linearly varying photon well is covered at both the left and right ends. Deadline? Yes On the left, the signal is intercepted by the noise layer, and on the right, it will overflow. This is very different from the movie. The response of film to incident light is an S-shaped curve. This means that the film is quite tolerant to both underexposure and overexposure. RAW conversion programs usually use S-shaped curve tone mapping by default to simulate film performance, but tone mapping can't recover signals lost in nonlinear areas (such as signals below noise level or overflow).

Dynamic range and hue range

The dynamic range of the sensor refers to the range from the left signal to the right signal. Tone range refers to the number of available digits representing different brightness levels.

Dynamic range and tonal range are related and sometimes interchangeable. However, as shown in the figure below, they are independent of each other. We can get high or low dynamic range image data and high or low tone range data, and any two combinations can appear at the same time (high dynamic range and high tone range, high dynamic range and low tone range, low dynamic range and high tone range, low dynamic range and low tone range).

Dynamic range measures how much information can be recorded from darkest to brightest. It is usually expressed by EV, and the adjacent 1EV indicates the brightness increase 1.

When the real-world tones are mapped to the recording medium, the tone range measures the transition. The transition of treble range data is very smooth; The transition of the bass range is very stiff. Can you see it clearly in the photo? Ribbon? .

Adjust digital ISO

In order to change the ISO of the film, the whole film must be exposed to the same ISO.

Every digital camera user knows that the ISO of a digital camera can be adjusted individually for each photo. In fact, most digital cameras allow photographers to use ISO below or above the benchmark ISO. When you change the ISO, the camera will also adjust its internal handler to adapt to the new ISO.

When you set an ISO higher than the benchmark, the camera's metering system will determine the exposure based on this. Digital raw data will also be affected. (Usually it is realized by an amplifier circuit, which is located in front of the A/D converter, that is, the voltage signal is amplified first and then digitized. For example, the voltage of ISO 100 is1000 mv. When it is adjusted to ISO 200, the voltage becomes 200mV, and the voltage of ISO 1600 becomes 1600mV, and so on. )

The design idea of allowing photographers to adjust ISO is to make full use of the complete input voltage range of A/D converter. The highest range can be obtained by making full use of the bit depth of A/D converter. However, the sensitivity of the sensor to light has not been substantially improved, and the camera only amplifies the signal generated by the sensor. When we amplify the signal, the noise is also amplified, so we will lose some image quality. This is somewhat similar to the result of film pressing.

Although setting a high ISO will not change the reference ISO of the sensor, it will also affect the raw data generated by the A/D converter. The specific impact varies with different models, but for most cameras, improving ISO shooting can get less noise and higher dynamic range than using standard ISO shooting and then increasing exposure through software.

The figure below illustrates this problem. Both photos were taken with Nikon D700, using the same aperture and shutter (1/40s, f/5.6). The photo on the left was taken using ISO 3200 without post-processing (directly converted from RAW to JPG using ACR). The picture on the right was taken with ISO 200 (the benchmark ISO of D700, with exposure less than 4 files), and then ACR was used to increase exposure by 4 files. Below are 100% screenshots of two photos.

The comparison of this set of photos shows that improving ISO collection? Right? Exposure, less noise than insisting on shooting with benchmark ISO (brightness will be improved later).

When we develop film, the added noise sometimes produces better visual effects. But the noise generated by improving the ISO of digital cameras will not. However, the latest digital cameras have excellent noise control performance and better high-sensitivity image quality.

It is basically meaningless to use ISO below the benchmark. Due to the wide demand for low ISO in the market, most cameras provide ISO below the benchmark. However, we do not recommend users to use these settings, which will lead to the loss of dynamic range.

Let's explain here that when you set an ISO below the benchmark, the charge generated by the photon trap is still determined by the amount of incident light, but in order to simulate the low ISO effect, it is necessary to weaken the signal. This does not change the noise layer, so the noise generated by the sensor is the same as the reference ISO, but the effective signal is weakened. In other words, using ISO below the benchmark won't bring any benefits (but it gives the shutter speed more choices).

The choice of ISO provided by digital cameras is decided by manufacturers, who will try their best to control the noise within an acceptable range. Reference to ISO usually has the least noise.

Regarding the ISO range, the upper limit of the ISO value is determined by the maximum acceptable noise level, and the lower limit is determined by the dynamic range loss that the manufacturer considers acceptable.

Because exposure time, temperature and humidity will affect the image quality, each item is standardized in ISO standard. In practice, we will shoot in a very changeable environment, regardless of the experimental conditions controlled in the standard, but the ISO value (obtained in the laboratory) is still valuable for image quality.

Use high-bit A/D converter.

For a long time, 12 bit A/D converter has been widely used in digital cameras. Besides analog amplifier circuit, higher digit A/D converter can be used to improve ISO. 1 bit in A/D converter stands for 1 multiplication. If we have a sensor based on ISO 100, the sensitivity can be improved to ISO 1600 by using16-bit A/D converter instead of amplifier circuit. Some cameras use this technology to provide high ISO, while others combine a high-level A/D converter with an amplifier circuit.

Using 16-bit A/D converter will not reduce the noise, because the noise will be doubled when the numbers are multiplied, just like the result of the amplifier circuit. However, other noise reduction techniques can be used for the unamplified digital signal stream, and some people think that this method can achieve better noise control effect.

Digital ISO and noise

Just as some ISO 400 color reversal films have better image quality than other ISO 400 films, some sensors will also have better image quality than others under the specified ISO. Generally speaking, a large-sized photon trap can generate a strong signal (to obtain a better signal-to-noise ratio).

In the above picture, the photon well is compared to a bucket for collecting rainwater. Empty barrels represent all black, and full barrels represent all white. When exposed, the shutter opens and the bucket begins to collect rainwater. By measuring the amount of rainwater collected in the bucket, we can determine the gray level of the bucket/pixel.

As shown in the above picture, the speed at which a small bucket receives water is the same as that of a large bucket. But the larger the surface area of the bucket, the more rainwater will be collected in a specific time.

When collecting photons, a photon trap with a small surface area is like a small bucket, and a photon trap with a large surface area is like a big bucket. The amount of water represents the signal intensity generated by each photon trap. Larger buckets collect more water, which means that larger photon wells can generate stronger signals at the same time and have better signal-to-noise ratio.

The physical size of photon wells is determined by two main factors: the physical size of the sensor and the number of photon wells on the sensor. Large-size sensors obviously have a larger surface area. The surface area of each photon well can also be increased by placing fewer photon wells on the same size sensor. For Bayer array sensor, each photon well can be roughly regarded as a pixel.

With the same pixels, the larger the physical size of the sensor, the larger the photon trap. This is why Quan Huafu SLR cameras can get clear photos in ISO 1600, while the image quality of portable cameras in ISO 1600 is very poor.

Under the same sensor size, fewer pixels can also obtain larger photon traps. This is why Nikon D3S (12 million pixels) has better picture quality than D3X (24 million pixels). Looking at the original image of 100%, the noise of D3s at ISO 1600 is equivalent to that of D3x at ISO 400 level.

But if we reduce the photo of D3x from 24 million pixels to12 million pixels, which is the same as the resolution of D3s, then the noise of the two cameras will be similar under the same ISO. Reducing the resolution reduces the noise because it enhances the information (effective signal) that determines the brightness and color in each pixel.

Although the sensor size and the number of pixels are two main factors that determine the noise level, there are other parameters that can also affect the noise level.

For example, sensors need electricity to drive. If the electronic components are located on the sensor surface, they will occupy a part of the surface area, thus reducing the space occupied by the photon trap. By designing these circuits on the back of the sensor or reducing the circuit volume, the surface area of the photon trap can be increased and a stronger signal can be obtained. On the premise of not increasing the physical size of the photon trap, the use of microlens makes the surface area of the photon trap have the same effect of collecting light (this is like adding a funnel with a large opening on a small bucket). In addition, designing a better signal processing system can also improve the signal-to-noise ratio. At present, the design level of sensors and signal processing systems is improving, and we will see better noise control ability on new sensors.

Test ISO image quality

You may want to test your camera to see the image quality under different ISO standards.

Shoot the same scene with different ISO settings to see how high the ISO is, and still get acceptable picture quality.

Many digital cameras have built-in high-sensitivity noise reduction function, but at the same time, they will lose details and make photos look fake. If so, the test should be carried out with the ISO noise reduction function turned off and on.

Knowing the limitations of your camera can prevent you from taking a bunch of invalid photos because of poor quality after an important activity.

Using high ISO shooting will increase noise, using noise reduction function will lose details, and using shutter for a long time will increase the risk of jitter. You need to find a balance point that suits you best through testing.