New Sigma FP

[BJL]

Thanks for the reference to the standard... Is EI maybe the term we should be using for "the sensitivity value we set on our camera that tells it how to expose, and what amplification to use on JPEGs"? It sounds like the value we are setting is not actually aligned with the ISO standard?

Further confusion is added by the fact that many cameras today are "ISOless" or nearly so - the actual gain applied to the output of the sensor (which would be the closest thing to the old ISO film speed) either doesn't change or changes only in one or two steps. What happens when you set a higher "ISO" value is often that the camera applies a digital exposure modification to the (underexposed) sensor data after the image is captured.

This is more equivalent to what photojournalists used to call "push-processing" and Zone System devotees called "+1 or +2 development" than it is to using an inherently more sensitive film, no?

Dan, I think the best way to view it is that digital cameras simply do not have a single “sensitivity” the way most films have a single ASA/ISO speed number; they are more like some special cases where the ASA speed of a film can truly vary with different development. And the dial on a digital camera genuinely varies that “Standard Output Sensitivity” or “Recommended Exposure Index”, to use the names from the ISO standard.

The ISO standard also explicitly tries to give measures for upper and lower limits on that range of “output sensitivity”, with an upper limit based on SNR of 10 or 40 in the mid-tones, and the lower limit (SSat) based on barely adequate roughly 2.5 stops of headroom between mid-tones and where highlights get blown due to overfilled photosites or such.

As far as I can see, cameras usually offer a range of “normal” output sensitivity settings on the “ISO” dial that lies within this ISO-defined exposure latitude, and then sometimes additional “High” and “Low” choices that fall outside it. However, the SNR-based upper limit is fuzzy and not so strictly followed; after all, resolution trade-offs like down-sampling or noise reduction can improve that mid-tone SNR, so it can often make sense to use a higher exposure index (outside the ISO-suggested latitude) to get a high enough shutter speed and deal with the noise later.

TL;DR I’d love to see a name like “output sensitivity” used for what the ISO setting does, but “exposure index” has more chance and is consistent with both traditional meaning and the language sometimes used in the ISO and Japanese CIPA standards.

[Guillermo Luijk]

Dan, I think the best way to view it is that digital cameras simply do not have a single “sensitivity” the way most films have a single ASA/ISO speed number; they are more like some special cases where the ASA speed of a film can truly vary with different development. And the dial on a digital camera genuinely varies that “Standard Output Sensitivity” or “Recommended Exposure Index”, to use the names from the ISO standard.

The ISO standard also explicitly tries to give measures for upper and lower limits on that range of “output sensitivity”, with an upper limit based on SNR of 10 or 40 in the mid-tones, and the lower limit (SSat) based on barely adequate roughly 2.5 stops of headroom between mid-tones and where highlights get blown due to overfilled photosites or such.

As far as I can see, cameras usually offer a range of “normal” output sensitivity settings on the “ISO” dial that lies within this ISO-defined exposure latitude, and then sometimes additional “High” and “Low” choices that fall outside it. However, the SNR-based upper limit is fuzzy and not so strictly followed; after all, resolution trade-offs like down-sampling or noise reduction can improve that mid-tone SNR, so it can often make sense to use a higher exposure index (outside the ISO-suggested latitude) to get a high enough shutter speed and deal with the noise later.

TL;DR I’d love to see a name like “output sensitivity” used for what the ISO setting does, but “exposure index” has more chance and is consistent with both traditional meaning and the language sometimes used in the ISO and Japanese CIPA standards.

IMO all the ISO story deserves much less attention than it receives. A good RAW shooter should simply study carefully how his camera responds to every amount of light, and forget about the ISO conventions. For me ISO values are just amplification gains, and I know I'll get the best overall DR and SNR captured if I can manage to do ETTR at the lowest real ISO. The ISO100, ISO200, ISO Low,... conventions are pretty useless, the only important thing is to know what's going on with the captured signal at every ISO setting.

For example: it's usually said (in the first place by the camera maker itself) that Olympus and Panasonic cameras lack ISO100, starting at ISO200. Well, I'd rather say Olympus cameras really HAVE ISO100 than saying they don't have it. When set at ISO200, a Olympus camera behaves closer to any other camera's ISO100 (specially old Canons) in terms of RAW highlight headroom, which in the end is what really matters. This means an Olympus RAW file from my E-P5 at ISO200 will have roughly the same RAW values than a RAW from my Canon 350D at ISO100 when both cameras receive the same amount of light per sensor surface unit (i.e. when the same aperture and shutter are set on both). If so, how the hell can you say the PEN doesn't have ISO100? or how can you say the Canon has ISO100?. When used with camera's suggested exposure at ISO200, Olympus cameras produce really bad SNR (an example of this is seen in so many users complaining at Olympus noisy skies), but the reason is not so much the M4/3 sensor performance, is that compared to other cameras you are sistematically underexposing by one stop!

Therefore I tend to ignore the "ISO100" naming convention, I just know what ISO setting I need to adjust to get the most DR of my captures and that's it.

Regards

[shadowblade]

IMO all the ISO story deserves much less attention than it receives. A good RAW shooter should simply study carefully how his camera responds to every amount of light, and forget about the ISO conventions. For me ISO values are just amplification gains, and I know I'll get the best overall DR and SNR captured if I can manage to do ETTR at the lowest real ISO. The ISO100, ISO200, ISO Low,... conventions are pretty useless, the only important thing is to know what's going on with the captured signal at every ISO setting.

For example: it's usually said (in the first place by the camera maker itself) that Olympus and Panasonic cameras lack ISO100, starting at ISO200. Well, I'd rather say Olympus cameras really HAVE ISO100 than saying they don't have it. When set at ISO200, a Olympus camera behaves closer to any other camera's ISO100 (specially old Canons) in terms of RAW highlight headroom, which in the end is what really matters. This means an Olympus RAW file from my E-P5 at ISO200 will have roughly the same RAW values than a RAW from my Canon 350D at ISO100 when both cameras receive the same amount of light per sensor surface unit (i.e. when the same aperture and shutter are set on both). If so, how the hell can you say the PEN doesn't have ISO100? or how can you say the Canon has ISO100?. When used with camera's suggested exposure at ISO200, Olympus cameras produce really bad SNR (an example of this is seen in so many users complaining at Olympus noisy skies), but the reason is not so much the M4/3 sensor performance, is that compared to other cameras you are sistematically underexposing by one stop!

Therefore I tend to ignore the "ISO100" naming convention, I just know what ISO setting I need to adjust to get the most DR of my captures and that's it.

Regards

Exactly. ISO-invariant sensors only have a few real ISO levels, with the rest being no different to pushing a lower-ISO shot in postprocessing. The Sony 42MP sensor only has ISO 100 and ISO 640. Nikon's D850 only has ISO 64 and 400.

Which leads to a particular quirk regarding the way non-base ISOs are treated. Basically, they're throwing away highlights. There's no disadvantage in shooting at USO 100/640 (Sony) or ISO 64/400 (Nikon), then pushing to a higher equivalent ISO, as opposed to shooting at a higher ISO in the first place. Conversely, many scenes where you'd want to shoot with a higher ISO are mostly dark, but with a few, small bright areas, which are captured well at ISO 100 (for argument's sake) but which are blown out at ISO 400. So it actually works out better to shoot at the base ISO and push, sparing the bright areas, rather than shooting at ISO 400, getting the rest of the image at the right exposure, but blowing the highlights. It would be great if there was a camera setting to only shoot at the base, 'real' ISO levels, only using the other levels for the purposes of exposure simulation in the viewfinder. For instance, on the Sony, if you shot a scene at ISO 2500, you would see the image at ISO 2500 in the viewfinder, but the actual exposure would be captured at ISO 640, allowing for selective pushing to bring the image back to proper exposure, while sparing the highlights. At this time, there's no option to do this - if you want ISO 640, you have to shoot at ISO 640, which results in a completely black viewfinder (if using exposure simulation) or you having no idea what ISO you actually need (if not using exposure simulation).

[BJL]

@Guillermo, agreed that makes sense for photographers like you who learn enough about sensors to then expose for an optimal raw file and adjust levels in raw conversion. On the other hand, the great number of photographers who aim for settings that make the in-camera preview informative and the default JPEG conversion useful much of the time, and then maybe use raw+JPEG so the hard cases can be improved by custom raw conversion, the cameras’ ISO “output sensitivity” settings do the job quite well. The extra stop or so of highlights in raw files vs default JPEGs is useful insurance.

(Do you all remember the bad old days when people criticised digital cameras for their very limited highlight headroom? Overcoming that is why the minimum “normal” ISO output sensitivity is so often a stop or so above ISO SSat “base ISO speed”.)

[BJL]

@shadowblade: indeed when sensors gain almost nothing in shadow handling and noise by using other than the two base ISO speeds, it might be best to not use intermediate degrees of analog amplification, but instead essentially give raw photoelectron counts in raw files, with metadata indicating the intended exposure index, to guide amplification used in raw conversion. (Some MF CCD cameras with good discrete ADCs did this, and it made sense, and some CCD cameras at least did it for their “High” output sensitivity settings.)

[shadowblade]

Today's cameras are all extremely sensitive, and the competition, at least for my type of photography, is more about who has a low ISO setting with exceptional image quality. Nikon's ISO 64 is actually useful - I shoot my Z7 down there all the time, and am rewarded with images that look like 4x5" film in many ways - in a big print, the extra DR and utter lack of noise shows, even compared to ISO 100 from the same camera. One of the real drawbacks of many smaller than 24x36mm sensors is that they not only don't have 64 as a "real" setting, they even lack 100. A camera that starts at ISO 200 can't (with today's sensor technology) match the low noise and wide DR of one that starts at ISO 64.

I would mostly argue the opposite. We need cameras with several stops more sensitivity - not for 19 stops of DR at ISO 100, but for 13-14 stops at ISO 3200-6400, with noise and image quality similar to what we currently get with ISO 100-200.

A lot of landscape and cityscape situations are taken in dim lighting and/or at narrow apertures. Exposure times of 1/4s to 4s are fairly typical. But landscapes and cityscapes aren't entirely still. Leaves flutter, branches sway, pedestrians and cars move. Exposures in this range don't freeze motion sufficiently. Shooting at a longer exposure can blur cars out if there's little traffic, but pedestrians have a nasty habit of standing in place and loitering, while the technique does nothing for moving leaves, grass or branches. If I could shoot the same scenes at 1/20-1/80s or so, it would eliminate this unwanted motion blur and make for far less frustrating shots and improved shooting opportunities, much more so than getting a bit less noise in an already-clean ISO 100 shot by being able to shoot at ISO 6.

If I really want ISO 6-25, I can get that simply by taking 4-16 ISO 100 exposures and averaging them. But I can't generate extra sensitivity that the camera doesn't already have.