[soycory]

Just like film, a digital camera benefits from lower ISO. So while we appreciate a high end of 500,000 ISO or whatever the highest ISO available for a certain camera is, the image degradation usually makes that undesirable. So, would the images be even better at 50 or 25 ISO if that were an option? Is the best ISO always the lowest? If not why wouldn't a camera manufacturer make up to a 1 ISO available. Even though that would not be useful for most things, if the ISO were to actually get better the lower you go, it seems like 25 would be awesome in the limited occasions you could use it. Or, is 100 really the best. why can't it be lower? Anyone know?
Thanks

[Spodeworld]

Really interesting question. What is the practical floor for ISO? I know that the K-5 family can go to ISO 80 and I've heard people say that is behind a slightly better dynamic range than the K-3.

[vonBaloney]

The sensor has a "native ISO" which is where you are getting an unboosted signal, and that will generally be in the 100-200 range for most sensors. (On our DSLRs, usually 100). Sometimes they do allow an ISO lower than the native, like the K-5 has an option to go to 80 even though the native is 100, but the difference is achieved in software rather than at the hardware level (I think). At some point the same is true of the higher sensitivities, i.e. past a certain point (1600 on the K-5) all the additional gain is from software boosting in the camera -- you could achieve a similar effect of shooting at 128,000 by shooting at 1600 and then just by boosting in Lightroom (but that makes in harder in the field because you are shooting dark images). But anything greater than 100 up to 1600, the signal is amplified via hardware gain, which will give somewhat better results than a pure software boost. But basically anything higher than the native 100 uses some "artificial" method to tease a greater signal out of what the sensor captures, at the expense of more noise to go with it.

Degradation also occurs when artificially lowering the ISO below the native, so everything is a compromise. I suspect that we will see sensors continue to be optimized for ISO 100ish until they figure out a way to have true varying native sensitivities or very very good methods of artificially boosting/lowering. What I would like to see is the sensors so they work like negative film, i.e. make the sensor refreshable/continuous so that the light-collecting "buckets" never get full (no blown highlights) and you suck up as much light in the dark areas as you can. Methinks this is a more natural way to collect light rather than having to cut-off your collection of photons from dark areas because your buckets collecting photons from the light areas are full...

[Bill2849]

In digital photography, the ISO measure is a measure of the camera sensor's sensitivity to light. While lower ISO images have less "noise" than the same images taken at higher ISO settings, it does NOT follow that you always want a lower ISO setting on your camera (although there are digital cameras on the market that are capable of ISO settings at least as low as 50).


The reason why you do not necessarily want the lowest possible ISO is because if you want to get a good exposure, the lower you take your ISO the more light you will need for your exposure. That means a wider aperture setting (which may compromise your depth of field) and/or a longer shutter speed (which can compromise sharpness in a wide variety of situations including photographing moving subjects, portraiture, and low light situations).


So, to answer your questions:
- lower ISO settings do not necessarily produce better images;
- the best ISO setting is not necessarily the lowest;
- reducing the ISO will not necessarily result in a better image.


At least, that is my understanding, based on what I have heard and read.


Best wishes,

[Rimfiredude]

To Bill 2829...I can't agree with your statements:
- lower ISO settings do not necessarily produce better images;
- the best ISO setting is not necessarily the lowest;
- reducing the ISO will not necessarily result in a better image.

The answer is simple a picture shot at ISO 100 using a shutter speed that is fast enough not to produce a blurred image and the best f stop of your lens will be better than the same image shot at the same f stop at 1600 ISO. PERIOD!

[ScooterMaxi Jim]

Native ISO, as far as I can tell, is a theoretical concept. That is, a manufacturer might have an ideal sensor voltage and amperage rating, so the sensor should perform best at that setting. However, modern technology makes this rating of marginal significance as near-maximum performance can be achieved over a broader range than in the past.

The importance of native ISO probably has dwindled as sensor and processing changes have reduced the relative impact of ISO changes, giving way to the ISO-less sensor idea. In fact, that concept doesn't quite yet exist in the real world as indicated by ratings measuring changes in dynamic range; no sensor loses a full three stops by changing from 100 ISO to 800 ISO (but most modern sensors lose around two stops). In other words, underexposing by three stops rather than changing ISO to 800 will likely lose you about one extra stop of dynamic range - and I suspect pulling up the midtones in post processing will show considerably more noise, as well.

My sense is that whatever ISO setting gives you the widest dynamic range should be considered as native because the noise differences in one stop are marginal in modern sensors. Typically, the best overall performance I have seen indicates that most Pentax sensors perform best when underexposed about one stop (i.e. 200 ISO), then lifted to allow for a smoother highlight roll off. It has been observed that the Pentax K-3 with HC applied at 200 ISO pretty much matches the Nikon 7100 (same apparent sensor) set to 100 with no compensation - providing similar dynamic range, exposure and noise characteristics. While this can result in better scores for the 7100 in sensor IQ comparisons, it somewhat violates photographic principles.

Bottom line, Pentax users should be setting HC (at least in Auto to avoid blown highlights) to get the most out of their cameras - and match or even exceed the performance built into the Nikon equivalents.

[6BQ5]

You can simulate a lower ISO with your in-camera multi-exposure function. My K-30 can stack 9 images. I usually stack multiple images of slow exposures to blur water. It's like an electronic version of a ND filter. However, it also lowers your effective ISO. Noise is averaged out. The image is averaged in. Two exposures brings your ISO from 100 down to 50. Four exposures will bring it down to 25.

If I'm way off base then please correct me.

[vonBaloney]

Native ISO, as far as I can tell, is a theoretical concept.

Although the rest of what you say is basically true, native ISO is the exact opposite of theoretical. It is the maximum point at which the analog signal from the sensor is not boosted before digital conversion. Remember the signal doesn't start out as digital information -- it is photons in the real-world, and the collection system (the sensor) has to collect those photons and then translate it into numbers. So its native capacity for collecting photons is anything but theoretical, it is real hardware with a real capacity that directly affects everything that can be done with that signal afterward.

[disconnekt]

You can simulate a lower ISO with your in-camera multi-exposure function. My K-30 can stack 9 images. I usually stack multiple images of slow exposures to blur water. It's like an electronic version of a ND filter. However, it also lowers your effective ISO. Noise is averaged out. The image is averaged in. Two exposures brings your ISO from 100 down to 50. Four exposures will bring it down to 25.

If I'm way off base then please correct me.

I think this would be good for stationary shots (like tabletop/product photography and to a degree landscapes), not so much for action shots.

[ScooterMaxi Jim]

Although the rest of what you say is basically true, native ISO is the exact opposite of theoretical. It is the maximum point at which the analog signal from the sensor is not boosted before digital conversion. Remember the signal doesn't start out as digital information -- it is photons in the real-world, and the collection system (the sensor) has to collect those photons and then translate it into numbers. So its native capacity for collecting photons is anything but theoretical, it is real hardware with a real capacity that directly affects everything that can be done with that signal afterward.

I'm pretty sure that the analog signal is always boosted before conversion. In any event, Roger Clark was the one who influenced my thinking on this, and he disputes the whole concept of "native ISO" here (see nearly halfway down the lengthy page):
Clarkvision: Digital Camera Review and Sensor Performance Summary

[TedH42]

I suspect that we will see sensors continue to be optimized for ISO 100ish until they figure out a way to have true varying native sensitivities or very very good methods of artificially boosting/lowering. What I would like to see is the sensors so they work like negative film, i.e. make the sensor refreshable/continuous so that the light-collecting "buckets" never get full (no blown highlights) and you suck up as much light in the dark areas as you can. Methinks this is a more natural way to collect light rather than having to cut-off your collection of photons from dark areas because your buckets collecting photons from the light areas are full...

I like this idea - adding a "shoulder" to the highlight side to keep the highlights from blowing out. Now if a "toe" can be added to the shadow side, then "HDR" can be done for single-exposure images of moving subjects. Just like in the good old film days, but without those refreshing chemicals. Well, maybe not to the level of those garish totally-HDR images that have been popular. But with better DR than with our linear sensors of today.

[vonBaloney]

I'm pretty sure that the analog signal is always boosted before conversion. In any event, Roger Clark was the one who influenced my thinking on this, and he disputes the whole concept of "native ISO" here (see nearly halfway down the lengthy page):
Clarkvision: Digital Camera Review and Sensor Performance Summary

Let me put it another way -- the buckets (pixels) can only collect so many photons. This is a hardware property. The size of a full bucket corresponds to the native ISO of the sensor system. If the native ISO were "only theoretical", well then we'd be able to change the native ISO with a dial on our cameras, and noise at 800 would be just as low more-or-less than at 100 because we'd have changed the "real ISO" of the whole system somehow. But hardware is hardware -- there are very real physical limits.

So yes, we can argue that today's sensors (and the processing of the signal they create) are much better than they were 5-10 years ago, and so the practical effect of native ISO isn't as great as it used to be GIVEN THAT WE DON'T ALSO WANT TO PUSH THE NEW BOUNDARIES CREATED but are merely happy that it is easier to stay within the old ones, i.e. at some point there will be so many megapixels and so much sensitivity for low light at low noise, etc etc that any improvements we make we won't be able to detect in any practical way with our human vision when looking at our images. But we are not there yet, so the limits still matter. (Of course, that's only if we restrict the discussion to "normal" types of photography -- once we get into the realm of imaging for scientific uses, well then they will always be looking to push the boundaries as far as they can.)

[jrpower10]

So, if I understand correctly, the base iso of the K5 series is 100 and hte ability to lower to 80 is a trick of the software? In previous threads I thought I read that many camera makers do, in fact, use software to achieve lower-than-base iso's, but that testing on the K5 sensor indicated a base of around 75ish, and that you actually do get gains by dropping down (my very unscientific eyeball tells me I'm not losing DR or contrast, and noise when boosting shadows looks cleaner, but maybe I'm seeing what I want to see).

[vonBaloney]

So, if I understand correctly, the base iso of the K5 series is 100 and hte ability to lower to 80 is a trick of the software?

Or maybe a trick of the hardware, not sure. Don't quote me on that one. It is not enabled unless you turn on expanded ISO, it stands to reason it is not "native", or maybe it is native but they turn off some of their other engineering miracles in the process so it is more or less a wash in effect...

[wombat2go]

Here is an article that gives some info on how the digital still cameras are rated
to the ISO 12232, and two other Japan standards used by the manufacturers.

dougkerr.net/Pumpkin/articles/SOS_REI.pdf

The rating is done at the bright end of the range, by all 3 methods.

It shows that the manufacturers have leeway to say what the base ISO of a camera is,
and scale the ISO ranges available, considering the camera's market sector etc.

There is no standard for signal to noise ratio either, so users with knowledge of the SNR
at base ISO, are free to set their camera to their own preferred EI.

It all sounds something like film was, some film types were more grainy than others at rated ISO,
and users were free, even encouraged, to run the film at their own preferred EI.