Hi all,
recently I read an article about astrophotography that explained a controintuitive aspect of photography, namely higher ISO=less noise.
The author obtained that result by increasing ISO keeping diafram and shutter speed fixed, then he took multiple pictures with different ISO, and he matched the exposure in post via software, and compared the level of noise, and indeed there was much more noise in the low ISO ones! The same behavior can be observed in the input referred read noise chart in photos and photons web site for pretty much all cameras.
So my question is: How does this happen from a technical point of view? From my understanding of ISO, it is basically analog gain, if this is true it does not multiply the amount of photons captured, but it should just multiply the power of the already captured signal which also embeds the noise, so how is it possible that multiplying the signal and the noise analogically and reducing it digitally later gives less noise? I would have expected the same noise.
Broadly speaking, there are two types of noise we deal with in digital photography. The first and most prominent is photon noise. It's sometimes called shot noise and is determined by the total light used to make a photo. The more light used, the less prominent shot noise will be.
The second is read noise and is a product of in-camera image processing. Historically, read noise is worst at the lowest ISOs and decreases as ISO increases. It's lowest at higher ISOs. At very low ISOs, the associated exposure is typically so strong that signal - the total light used - overwhelms noise (both shot and read noise) to deliver the highest quality images a camera is capable of producing.
However, as exposure and total light used become diminished, shot noise starts to become an issue. Photographers mistakenly associated this increase in noise with an associated increase in ISO and jumped to the incorrect conclusion that increasing ISO causes more noise to be added to photos. In fact, nothing could be further from the truth.
Shot noise (photon noise) decreases as exposure decreases. However, signal (total light) also decreases and, as a result, shot noise becomes more prominent. It's the ratio of signal to noise that determines how visible noise will be. Signal-to-noise ratio (SNR) decreases as the total light used to make a photo decreases and noise becomes annoyingly more visible.
As mentioned, read noise historically decreases as ISO increases. This read noise profile was the catalyst for the best practice of using the highest ISO that paired well with the exposure delivered to the sensor. If a photo was made with too low an ISO and needed to be lightened further in post - exposure cannot be altered once the shutter actuation ends - the increased read noise associated with that lower ISO would become more prominent.
In astrophotography and other genres of photography in which it's common to work with minimal light, making photos with the same exposure (same light level, f-stop and shutter speed) but different ISOs could reversal this behavior in older cameras.
However, it's been common for several years for digital cameras to be built around dual-gain sensors. Nikon, Sony and Fujifilm, for instance, all make cameras featuring dual-gain sensors. The advantage of this architecture is that the read noise at ISO 500, for instance, in a Nikon camera is no worse than at ISO 6400. The dual-gain noise profile kicks in at different ISOs, depending on the brand and model camera. It's typically somewhere in the ISO 400 to 800 range. When the dual-gain threshold is reached, the penalty of low ISO read noise is no longer an issue.
How does this benefit the photographer? Well, suppose you're making a photo of a scene with bright highlights and rather deep shadows. You might select exposure settings (f-stop and shutter speed) to maximize the light delivered to the sensor without blowing out the highlights at ISO 500. The resulting photo would have fairly dark shadows showing minimal detail. However, you could then bring the photo into Lightroom Classic (or your photo editing app of choice) and increase the lightness of the shadows to reveal detail without suffering any penalty in increased read noise or compromising the final image quality.
While f-stop (lens aperture) and shutter speed (exposure time) determine how much light from the scene is delivered to the sensor - or film negative, or glass plate - and what exposure the camera works with, ISO is used to manage image lightness. It communicates a relationship to the camera between the exposure being used and the lightness of the output photo. The camera uses this information to process the recorded data and produce an image of a given lightness.
ISO is not an exposure setting. (It never has been.) It has no direct affect on how much light is delivered to the sensor. Nor is ISO a source of noise in a photo. ISO merely communicates an exposure value to the camera. That's its defined role in the digital imaging industry.
My second question is. If increasing ISO would decrease noise in astrophotografy, one could think that it is always better to use higher ISO settings, however with high ISO everybody teaches that we loose dynamic range, so how to know the sweet spot to shoot at?
You will always be well-served by maximizing exposure - the amount of light from the scene per unit area as projected by the lens upon the sensor - within your creative goals for an image. If you use the widest lens aperture that produces an adequate depth of field and the slowest shutter speed that acceptably renders movement in the frame without blowing out important highlights, you've optimized exposure. At that point, use the ISO that results in a photo having a pleasing lightness.
In night sky photography, you'll balance the ability of a lens to capture light (its widest available aperture) against optical aberrations that compromise image quality away from the center of the frame. You'll balance a long exposure time and the benefits of putting more light on the sensor against the ability - or lack thereof - of the tripod to track and compensate for Earth's axial rotation and mitigate star trailing.
There are advanced techniques, such as stacking multiple shorter exposures to produce a composite image representing more total light and, therefore, having less visible noise. These are considerations that you'll need to weigh, that you'll need to prioritize according to your image aesthetic. If you seek out the good folks in the "Astrophotography Talk Forum," they'll be able to provide expert guidance on specific questions you may have.
Last question, from a technical point of view why do we loose dynamic range by increasing ISO?
Dynamic range in a photo is maximized when the camera works with the greatest exposure and most total light it's designed to capture. Our photos lose dynamic range when we work with diminishing amounts of light.
Think of ISO as the proverbial canary in the coal mine. It's not an exposure setting, doesn't control how much light is used to make a photo or the dynamic range your finished image will have. But it is often a reliable indicator of a strong or weak exposure and the dynamic range that will be present in the final image.
--
Bill Ferris Photography
Flagstaff, AZ
!
