31

Recently a very professional photographer who shoots stunning pictures told me that the best images can be obtained from digital cameras at ISO 200. He also stated that this is what all the manufacturers (Nikon, Sony, Canon, ...) say.

He explained it something like this (I heavily paraphrase as my amateur brain does not remember the correct technical terms used): The light is taken in by the sensor as is at ISO 200. At higher values the signal is amplified (which I understand is correct). At lower values like 100 however, the signal is artificially diminished, which leads to bleeding(?) of the pixels. Therefore ISO lower than 200 is only to be used for extreme lighting situations like shooting a solar eclipse.

However, I could not find any information corroborating these claims. I also checked a couple manuals of high-end DSLRs. They did not contain any mention of such a manufacturer's recommendation. I also took some comparison shots with my own digital camera (Lumix) and found ISO 100 to produce better, lower-noise results.

Now I'm at a loss as what to make of that man's statement?

  • 9
    You are about to open a can of worms with that. It will differ between manufacturers and sensor type (ccd vs cmos) and I very much doubt that "always 200" is a useful answer. You have always two mechanisms to obtain the final ISO. One is analogue amplification in the sensor, second is maths with the raw values. There are usually only so much analogue amplification steps available and the rest in between is done with maths. This may mean multiplying with numbers greater, equal or smaller than 1. I would suspect that there is lots of differences around. – PlasmaHH Oct 8 '17 at 14:34
  • 3
    Define "best images". – FarO Oct 9 '17 at 14:43
  • 5
    Anecdotal: I've made my worst photos at ISO 200 and some of my best ones at 3200. – Pavel Oct 9 '17 at 17:19
23

The information your friend gave you was essentially correct for most digital cameras, particularly compact digital cameras with very small sensors, made about 15-20 years ago. Digital imaging sensors were more primitive and noise reduction techniques were less sophisticated. By placing the native sensitivity of a sensor at one stop higher than what otherwise might be desired (typically ISO 100), the camera could have a wider 'usable' ISO range at very little detriment to the ISO setting one stop below the native sensitivity of the sensor. In exchange for a little less dynamic range at ISO 100 a sensor with a native sensitivity of ISO 200 could gain a stop of sensitivity on the higher end with regard to dynamic range versus the noise floor.

Such is not the case with most current dedicated cameras. There is more variety with regard to native ISO among current cameras, but most of them have a native sensitivity somewhere in the neighborhood of ISO 100. This is particularly the case with the current crop of APS-C and full frame cameras.

Once ingrained in a particular culture, such as professional photographers, some ideas are hard to modify when the current state of technology makes those ideas outdated and in need of modification if not outright obsolete.

Look, for example, at Image Stabilization. When the IS lenses for interchangeable lens SLR cameras first began appearing in the second half of the 1990s, there was an issue with vibration feedback loops when the camera and lens were mounted on a tripod. Vibration from the IS unit would cause the camera to vibrate, which would induce correction from the IS unit, which would cause vibration, which would induce correction from the IS unit, which would cause vibration... By the year 2000 pretty much every lens manufacturer had upgraded their IS technology to automatically turn off IS when the camera was detected to be mounted on a tripod. Some lenses have even appeared on the scene that have IS profiles specifically designed to be used with a tripod. Yet more pros than not will still tell you in 2017 to ALWAYS turn off IS when you are using a tripod.

  • 13
    Photography is an area particularly susceptible to old wives' tales. – Agent_L Oct 9 '17 at 9:57
  • I don't think the tripod/IS case is the best example, it is still good default advice if you aren't familiar with using that particular lens and camera in that situation. Exceptions made for tripods on vibrating/shifting surfaces of course. For instance, the Sigma 180mm f/2.8 APO Macro EX DG OS HSM lens was released around 2012 (from review and preview article dates) and definitely should have the OS turned off while on a tripod. – ttbek Oct 9 '17 at 14:51
  • 4
    The point is, it is a lens specific issue, not one that universally applies to all examples. There are many more examples of lenses that turn off or shift to a tripod optimized mode than there are of lenses that require one to manually turn them off to prevent an issue for which a very few first generation IS lenses were known. And it isn't good default advice if one is using a lens such as the Canon Super Telephoto IS II series that perform better on a tripod with it turned on. – Michael C Oct 9 '17 at 14:59
  • I've accepted this answer, because it gives the historical background. @Itai's answer was very helpful to understand the technical details. – PhillipM Oct 9 '17 at 19:20
  • The way I remember it is that 15 or so years ago, you could actually get a 100 that was worth something, and then over the few years following that, native 200 started to predominate. – hobbs Oct 9 '17 at 20:40
41

There is an ISO which is not necessarily 200 that is the native sensitivity of the silicon from which the sensor is made. That sensitivity depends on the sensor itself, so will vary between cameras, but it is almost always between ISO 100 and 200.

The camera amplifies the signal to get higher sensitivities. It scales down the signal to get lower ones. Amplification leads to more noise, scaling down leads to loss of gradation which I guess is what your photographer meant by bleeding.

Basically the silicon cannot be changed, it's a physical substance that releases electrons when light lands on it. The amount of electrons determines how illuminated that pixel is but if you must simulate a lower sensitivity, all you can do is divide the number of electrons by some factor.

A simplistic example could be: Imagine a camera with ISO 200 native sensitivity. Each photosite can hold a certain amount of electrons which is read using a 12-bit read-out. That means it can output any value from 0 to 4095. To simulate ISO 100, you just read the photosite and divide by two. This gives you numbers of 0-2047, so there is 1-bit loss of nuances.

The process of scaling down the signal also reduces noise which is why cameras often have a Low ISO. On an Olympus mirrorless for example, it is clearly labelled. When you select ISO 100, it says Low 100 Expansion. This also means that there is sometimes a valid reason to use ISO 100 but usually staying at the native sensitivity is best. As a matter, again on an Olympus, selecting the ISO 200 says ISO 200 Recommended.

  • 5
    Wait, what? ISO 100 does not do analog attenuation but instead just throws away one bit of ADC resolution? Is this true for most/all cameras? – AndreKR Oct 9 '17 at 6:41
  • 3
    So, if you go below native sensitivity in an attempt to avoid overexposure, this won't help you since the overexposed pixels will just be downscaled after they have already been blown. Is this right? – Ruslan Oct 9 '17 at 6:52
  • 2
    The opposite of amplification is attenuation, in case you felt like replacing "scaling down" in your answer. – Marc Dingena Oct 9 '17 at 9:29
  • 1
    @AndreKR - For the sake of example, this is a simplistic view. Reality is more complex and varies by sensor but the example makes it easier to understand. – Itai Oct 9 '17 at 13:50
  • 2
    Just to add to the confusion, Olympus RAWs that were taken in "low" ISO mode record the "fake" ISO value in exif, even though the output is identical to the same exposure at ISO 200... this can result in some head-scratching later on if you aren't aware of the situation. – junkyardsparkle Oct 9 '17 at 15:16
1

The sensitivity to light of photographic film is determined by preforming test exposures and measurements as specified by the International Organization of Standards (ISO). In recent times, the equivalent the sensitivity of the digital image chip closely parallels the sensitivity of film. However, test methods for the digital imaging chip continue to evolve. Currently digital there are three categories for ISO testing of the digital image chip. Native ISO – a sensitivity setting that does not require that the software apply increased voltage (amplification) to the imaging chip.

Amplified ISO – a sensitivity setting that that requires that an increased voltage and amplification be applied to the imaging chip. As the amplification is upped, some static creeps in. The higher the amplification, the higher will be the ratio of static to good signal. In the realm of digital, we call this static “noise”. Noise is perceived as a granularity (lack of uniformity) akin to grain that associated with photographic film images.

Simulated ISO – higher sensitivity settings achieved by raised voltage induced amplification and software algorithm that attempt to mitigate the ill effects of this action.

As to using an ISO setting below the Native ISO – a little degrading might be observed however more likely this will not be noticeable.

Allow me to add, significant improvements in the digital image chip allow higher and higher ISO setting with little downside unless the setting near its maximum.

  • 1
    "In the realm of digital, we call this static “noise”." - Welllll... more so outside the realm of digital, which is where the signal that you're describing is... in the digital realm there's quantization error and stuff. :) – junkyardsparkle Oct 8 '17 at 18:17
  • @ Junkyardsparkle --- The “noise” in an image, we are taking about is comes about when a higher signal to noise ratio occurs with the use of an elevated ISO setting. We could inform, this is the analog signal being amplified vs. a digital signal but I doubt if anyone cares. Now we are taking piddling stuff! – Alan Marcus Oct 8 '17 at 19:31
1

He appears to be conflating ISO and exposure time.

There is a well-recognised problem with digital cameras and long exposure times. Running a CCD sensor takes a certain amount of current, and reading the voltages back for each pixel takes a bit more. Ohm's Law says that this dissipates power as heat, and of course you're not sat in a vat of liquid nitrogen so there's more heat around you. The problem then is that thermal noise is a fact of the universe, and the more heat you've got, the more noise you've got on your image. If you're measuring over a long time with relatively little light coming in, the noise can become a relatively significant part of the signal.

I used to do a lot of night photos with my Canon bridge camera back in the early 2000s, and this was very noticeable indeed. The good news is that camera manufacturers have greatly improved on their sensor designs over the last 20 years or so, and this is greatly reduced on modern cameras. It's still something to be aware of though, if you're planning on taking particularly long exposure images. If you're planning on doing an all-day long-exposure shot (which traditionally would have a heavy filter and slow ISO setting), you may find you get better results by taking multiple "normal" shots during the day and digitally summing them.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.