Native ISO range vs. Lo 1, Hi 1, 2, etc.

[dslater]

So when you push your ISO outside of the "native" range into either the Lo or Hi range, does the camera do something different from what it does within the native range, or is it doing the same thing and these are just arbitrary labels used by Nikon to indicate you're going outside the recommended range?

[Roland Vink]

Not absolutely sure ... base ISO is where the sensor performs optimally - greatest dynamic range, lowest noise.

As you increase the ISO you are basically telling the camera to accept a lower exposure and amplify the signal to compensate. At some point the signal gets so low that noise becomes significant. That point where that happens is pretty arbitrary, I guess manufacturers choose a point where they feel the output is acceptable, but allow you to go beyond (Hi range) if needed. It is possible that at some point (not necessarily the same as the top of the "native" range) that signal processing changes to help suppress noise. I guess that is done on a case by case basis for each camera.

Going the other way into the Lo range is different because the signal needs to be reduced instead of amplified.

[Bjørn Rørslett]

Basically the native range is what Nikon considers comply with ISO standards and their own quality criteria. They designate this range as 'calibrated'. Meaning, not only will the usual relationship apply between light levels, aperture, and shutter speed, but colour rendition, noise, and dynamic range, all are within set quality limits.

Whether you go 'Lo' or 'Hi' you are entering uncharted territory and thus are on your own. Strictly speaking we shouldn't even express those 'Lo' or 'Hi' settings in ISO [speed] terms as they no longer follow the standards. You can no longer rely on doubling the exposure time and stopping the aperture down one notch will produce the same exposure to the file.

In the film days, lower ISO meant more fine-grained films and thus less 'noise'. As the input signal to the digital camera has to be reduced on the 'Lo' settings, noise concomitantly is lowered. However, instead of being purely advantageous from a quality point of view, image quality actually suffers because dynamic range also is reduced, plus colours may no longer be exact. Thus only under low contrast conditions is it advisable to drop into the 'Lo' range and expect improved quality. One should always shoot a colour checker card to ensure any colour deviation can be detected and corrected for.

The highest quality all parameters considered corresponds to the base ISO number, usually around ISO 100 these days. When you push beyond the calibrated range and into the 'Hi' figures, you gradually lose track over what happens. Dynamic range declines, noise increases, and colours suffer. The current better cameras will do fine on Hi-1 and perhaps even Hi-2 if you can accept some loss of quality. For video use the useful envelope is broader.

[dslater]

I see. So rather than being arbitrary, Lo & Hi represent the point where you've pushed things to the point that colors may shift and a linear relationship is not necessarily maintained.

[Bjørn Rørslett]

That is an entirely correct interpretation. Do keep in mind dynamic range is being compromised as well.

However, although stuff might not come over perfectly at 'Lo' or 'Hi' settings, they still can be useful, for example, to capture video. Or if you are into a nostalgic mood and fondly recall grainy 'pushed' Tri-X, by all means explore the 'Hi'  alternatives.

[Michael Erlewine]

I don’t care about the “Hi” setting, but the ”lo” might be useful, but I have never used it... yet. The fact that the Nikon D810 has a base ISO of 64 means the world to me and I am tempted to buy a second D810, just in case Nikon drops that feature in the next high-MPx camera they bring out. Here is my question:

Let’s say Nikon comes out with a 50 MPx (or greater) camera, with an ISO of 64. How does more megapixels affect the base ISO? If they brought out a 100 MPx camera, but had a higher base ISO of, say, 100, how would that related to the D810’s 36 MPx and 64 ISO?

[Bjørn Rørslett]

The base ISO is the maker's recommendation for best quality when all performance parameters are factored into the equation. Its actual number as such is irrelevant, as the makers define their cameras' response slightly different.

However, you cannot compare ISO figures for one camera against another, at least not directly. The ISO comparison is longitudinal not lateral to speak in statistical terms.

[Akira]

What confuses me is the surprisingly large discrepancy of "base ISO" between still and video.  Take Sony alpha 7S II for example, the base ISO for stills is 100 while that for video is 1600 or 3200.

The test performed by Cinema 5D website concludes that 7S II shows the lowest amount of noise in its video images at ISO 1600 or 3200.

https://www.cinema5d.com/ultimate-sony-a7s-ii-vs-a7s-test-difference/

[simsurace]

There is some arbitrariness to the ISO ratings for digital cameras. Still, a lower sensitivity can only be achieved by either losing data or increasing the full-well capacity. That's why going below base ISO reduces the dynamic range.

Imagine that you have set your exposure such that at base ISO, the brightest pixels are fully saturated. Now you would like to double your exposure time (say, to blur something). You are now sending double the light to the sensor, and the brightest pixels receive double the number of photons that would normally saturate them. By choosing one half of the base ISO, you are scaling down the signal, so the wells that are fully saturated are now output as a raw level which is one stop below saturation. Doing the same RAW conversion as you would be doing with the first exposure at base ISO, you get an image which has the same overall brightness. But since you clipped the top stop of light (all the saturated pixels are mapped to one value irrespective of the actual amount of light that hit them), you lost one stop of information.

So going below the base ISO is not really recommended unless you either don't care about the clipped highlights or you are well under saturation at base ISO. But in the latter case you could just try to expose to the right at base ISO and lower the brightness during RAW conversion, which should give equal or better results.

A lower base ISO is correlated with a larger dynamic range, all else equal. E.g. from the D800 with base ISO 100 to the D810 with base ISO 64, we measure a 0.43 stop increase in DR at their respective base ISOs, which is slightly less than the 0.64 stops difference between ISO 100 and ISO 64.

Since there are ways to increase DR without decreasing base ISO, not all cameras with, say, a base ISO of 100 have the same DR. This is obvious if we contemplate the multiple-stop increases in DR over the last decade while base ISO has at most decreased from ISO 200 to ISO 100 or 64. The lower noise floor of newer sensors is responsible for that feat.

Hopefully, some day in the future we will get sensors which do not have a saturation limit. You could theoretically count the number of times saturation has occurred for each pixel, the limit to this being the speed at which you can read out data from individual pixels. I believe I once read about a patent for such a technology. Given such a sensor, the biggest challenge would still consist in making this dynamic range visible to the eye in some way. 

[simsurace]

What confuses me is the surprisingly large discrepancy of "base ISO" between still and video.  Take Sony alpha 7S II for example, the base ISO for stills is 100 while that for video is 1600 or 3200.

The test performed by Cinema 5D website concludes that 7S II shows the lowest amount of noise in its video images at ISO 1600 or 3200.

https://www.cinema5d.com/ultimate-sony-a7s-ii-vs-a7s-test-difference/

This is just speculation because I'm not much into video, but I would presume that the high number of frames and the serial correlation of frames makes it possible to apply noise reduction algorithms already at the level of the ADC which are not available for static images. In addition, for sensors whose resolution is a multiple of the output resolution (even 4K is small compared to some sensors), binning of pixels is required, which means that the signal level required from single pixels is decreased. I don't think these two things would account for the entire 4-5 stop discrepancy, but maybe for 2-3 stops.

[Bjørn Rørslett]

Recasting my mind a little back in time, I did test video shooting with D4S at night using "ISO" 102400. The scene was so dark I hardly could see anything with the naked eye, but the footage came out great and the noise, although present, did not intrude too much in the output clarity.

When I make UV videos with my Panasonic GH-2, I mostly use its ISO 3200 setting. The footage is excellent, even though stills shot at the same ISO are very noisy. For UV stills with the GH-2 I normally stick to ISO 160 or so.

[charlie]

Due to the max sync speeds of 1/250s I use the ISO 'Lo' setting fairly regularly when working with strobes outdoors in bright sunlight. 'Lo' provides the opportunity to set the aperture to f/8-f/11 instead of f/16-f/22, or to really darken the background ambient light if the aperture is set to f/16-f/22.

This is particularly valuable on a camera like the D700 with a base of iso 200, but I use it on my D800 as well. I can't say I've ever noticed a color shift or dynamic range issue that wasn't worth the trade off of going into 'Lo' iso's.

ND filters are of course an alternative to 'Lo' iso settings in this circumstance, but they have their own drawbacks.

[dslater]

Recasting my mind a little back in time, I did test video shooting with D4S at night using "ISO" 102400. The scene was so dark I hardly could see anything with the naked eye, but the footage came out great and the noise, although present, did not intrude too much in the output clarity.

When I make UV videos with my Panasonic GH-2, I mostly use its ISO 3200 setting. The footage is excellent, even though stills shot at the same ISO are very noisy. For UV stills with the GH-2 I normally stick to ISO 160 or so.

When you're shooting video, aren't the frames down-sampled from the sensor's native resolution? Wouldn't the resulting pixel combining account for a lot of the reduced noise compared to stills?

[Akira]

Gentlemen, thank you for sharing all your wisdom.

The pixel binning seems to be the key to the higher ISO value for the base ISO of movies.  That said, some cameras use only the central portion of the pixels for the necessary resolution of the video format.  In that case, this method may not apply...

[Andy]

What confuses me is the surprisingly large discrepancy of "base ISO" between still and video.  Take Sony alpha 7S II for example, the base ISO for stills is 100 while that for video is 1600 or 3200.

The test performed by Cinema 5D website concludes that 7S II shows the lowest amount of noise in its video images at ISO 1600 or 3200.

https://www.cinema5d.com/ultimate-sony-a7s-ii-vs-a7s-test-difference/

To add to Simone's explanation one more component:
If the output fileformat chosen for the video is not able to record i.e. a dynamic range of 14 such a camera is capable of at base ISO, then the manufacturer has one additional attribute to optimize. For instance, to reduce the actual exposure time per frame of the 30fps movie to reduce motion blur (which is harder to fix in SW than noise).

rgds,
Andy