Questions about ISO

[TravelingChick]

For some reason, the universe has seen fit to share a lot of ISO-related YouTubes with me lately.

ISO Invariance: This has come up a couple of times, and if I understand it, if a camera is iso-invariant you can for example underexpose and image at ISO 400 and boost the shadows in post, OR properly expose an image at any ISO above that (or whatever the ISO invariant ISO is, I guess?) and the results as far as noise go would be the same. Is that remotely correct and/or is the OMD em1mkiii iso-invariant? In my experience it is not. I think the noise is obviously worse at higher ISOs. I think the speaker was reference a full frame Nikon, but I'm confused about the concept in general.

Base ISO for a given camera: Is there a reference or base iso for a given camera model? In my instance the OMD em1mkii. What does this mean in practical terms? Would that ISO give the best combination of dynamic range and noise? And does dynamic range actually change with ISO changes?

I'm just curious. Obviously, composition and quality of light are much more important.

[pdk42]

For some reason, the universe has seen fit to share a lot of ISO-related YouTubes with me lately.

ISO Invariance: This has come up a couple of times, and if I understand it, if a camera is iso-invariant you can for example underexpose an image at ISO 400 and boost the shadows in post, OR properly expose an image at any ISO above that (or whatever the ISO invariant ISO is, I guess?) and the results as far as noise go would be the same. Is that remotely correct and/or is the OMD em1mkiii iso-invariant? In my experience it is not. I think the noise is obviously worse at higher ISOs. I think the speaker was reference a full frame Nikon, but I'm confused about the concept in general.

Very close. Let me slightly re-phrase it:

- If a camera is ISO invariant, then you can boost the brightness of an underexposed shot in post to achieve the same brightness and noise as you would if you'd taken a "correctly" exposed shot at a higher ISO.

- In both cases the shutter speed and aperture will be the same, so the light captured is the same. If you boost in post then you're adding the brightness in Lightroom (or whatever). If you boost by cranking up the ISO then you're adding the brightness in the camera (perhaps by electronic amplification, perhaps by digital processing). If the camera is ISO invariant then the result will be the same so far as noise is concerned whichever route you take. If the higher ISO shot is cleaner, then the sensor is not ISO invariant.

So, for example:

- Let's say you meter a scene at ISO 800 and it's 1/125 at f8.
- You can take the picture with the camera at ISO 800 and the end result will be "correctly" exposed.
- If you keep the same shutter speed and aperture but reduce the ISO on the camera to 200, then when you look at in Lightroom (or whatever) it will look dark.
- However, if you then add +2 EV correction in LR you'll get the same brightness.
- If the sensor is ISO invariant then the two images will look identical in both brightness and noise.

In all this discussion note that:

- THIS ONLY APPLIES TO RAW
- It really only makes sense to use this information when applied as adjustments to base ISO

This last point is because ISO invariance is very useful if you plan on making a lot of shadow corrections later, or need to selectively boost certain parts of the image (e.g. if you've exposed to retain highlights and hence have deep shadows that are too dark). This applies in spades to landscape photography so an ISO invariant camera is very useful for that. It's really not of such benefit in BIF etc where straight-up high ISO performance is what you need.

In my experience, the Sony sensors used in most of the Olympus OMD cameras are almost, but not quite ISO invariant.

Base ISO for a given camera: Is there a reference or base iso for a given camera model? In my instance the OMD em1mkii. What does this mean in practical terms? Would that ISO give the best combination of dynamic range and noise? And does dynamic range actually change with ISO changes?

I'm just curious. Obviously, composition and quality of light are much more important.

Base ISO on all the 16 or 20Mp Olympus cameras is 200. That's where you'll get the best DR and lowest noise. The extended low ISO values work by overexposing at capture time and pulling back in post (via in-camera JPEG processing, or if you're shooting raw by setting the grey level value in the raw file that is then interpreted by LR (or whatever you're using). It can deliver lower noise, but only when you can be sure not to blow highlights.

What base ISO means is that when the camera is set for that ISO and you give a calibrated exposure corresponding to that ISO (using the ISO standards), then the sensor charge wells will be filled to an optimum level. In other words, you'll get the best signal to noise ratio on the image. If you think of the sensor as lots of little buckets that get filled up with light, then when you are at base ISO, the metered exposure will fill the buckets, but they won't overflow. If you crank up the ISO above this then the metered exposure will be less in absolute terms (faster shutter speed and/or smaller aperture), and the result will be that the buckets will now be only partly full. To get the right brightness later, you need to amplify the buckets and that means noise and hence lower signal to noise.

Finally, if you want to see the effect of ISO invariance, you can do this using the DP Review studio test scene (links below). Here's an example of shooting the same scene with the EM1.3. One at ISO 3200, the other at ISO 200 and boosted four stops:



It's close, but not perfectly ISO invariant - the ISO 3200 is a little better (but not really much).

Here's the corresponding example with the Panasonic S5:



You'll see that the two shots now have very similar noise.


Links:

- DPR studio scene: https://www.dpreview.com/reviews/ima...son/fullscreen
- DPR studio scene with EV comp in post: https://www.dpreview.com/reviews/ima...en?&widget=327

[TravelingChick]

Thank you for your detailed response, Paul. After reading through, my initial thought would be that from a practical standpoint (i.e. in the field) trying to stay between 200-1600 ISO is kind of the sweet spot. I only really go lower than 200 if I need to for shutter speed. Likewise, I'm not adverse to pushing the ISO way up if I need the speed (BIF, esp early or late in the day). I know the noise gets quite bad especially in the extended high range ISO, but sometimes you have to do it.

I'm a generalist as far as my photography goes. I lean to natural history - landscapes, animals, etc - but I travel a lot so there's a fair amount of 'whatever is in front of me.' I always shoot RAW. Now that I have a better understanding of ISO, I feel like I've been on the right track: I sometimes do exposure blending, and sometimes use the in-camera HDR; But MOSTLY, don't blow out the highlights while exposing to the right. If you need ISO, you need it, and fix it in post.

Thanks again. The physics behind these sensors is fascinating.

[Box Brownie]

If I recall, in regard to ISO invariance, your description is about right.

Plus like you, of things I saw/read about ISO invariant camera bodies, it seemed that only(?) Nikon exhibited such such behaviour with IIRC specific reference to the D800 body (though I think at least one other body was mentioned?)

[pdk42]

Lots of cameras are ISO invariant, it's not just Nikon (who in fact use sensors from Sony so anyone who uses these sensors gets the same ISO invariance characteristics). In my experience, most Oly 16 and 20 Mp cameras are ISO invariant to a point. I reckon that up to about ISO 1600, underexposed shots at ISO 200 can be boosted to give results about the same as those taken at ISO 1600. Above that, it's better to boost the ISO on the camera.

[pdk42]

For some reason, the universe has seen fit to share a lot of ISO-related YouTubes with me lately.

ISO Invariance: This has come up a couple of times, and if I understand it, if a camera is iso-invariant you can for example underexpose an image at ISO 400 and boost the shadows in post, OR properly expose an image at any ISO above that (or whatever the ISO invariant ISO is, I guess?) and the results as far as noise go would be the same. Is that remotely correct and/or is the OMD em1mkiii iso-invariant? In my experience it is not. I think the noise is obviously worse at higher ISOs. I think the speaker was reference a full frame Nikon, but I'm confused about the concept in general.

Very close. Let me slightly re-phrase it:

- If a camera is ISO invariant, then you can boost the brightness of an underexposed shot in post to achieve the same brightness and noise as you would if you'd taken a "correctly" exposed shot at a higher ISO.

- In both cases the shutter speed and aperture will be the same, so the light captured is the same. If you boost in post then you're adding the brightness in Lightroom (or whatever). If you boost by cranking up the ISO then you're adding the brightness in the camera (perhaps by electronic amplification, perhaps by digital processing). If the camera is ISO invariant then the result will be the same so far as noise is concerned whichever route you take. If the higher ISO shot is cleaner, then the sensor is not ISO invariant.

So, for example:

- Let's say you meter a scene at ISO 800 and it's 1/125 at f8.
- You can take the picture with the camera at ISO 800 and the end result will be "correctly" exposed.
- If you keep the same shutter speed and aperture but reduce the ISO on the camera to 200, then when you look at in Lightroom (or whatever) it will look dark.
- However, if you then add +2 EV correction in LR you'll get the same brightness.
- If the sensor is ISO invariant then the two images will look identical in both brightness and noise.

In all this discussion note that:

- THIS ONLY APPLIES TO RAW
- It really only makes sense to use this information when applied as adjustments to base ISO

This last point is because ISO invariance is very useful if you plan on making a lot of shadow corrections later, or need to selectively boost certain parts of the image (e.g. if you've exposed to retain highlights and hence have deep shadows that are too dark). This applies in spades to landscape photography so an ISO invariant camera is very useful for that. It's really not of such benefit in BIF etc where straight-up high ISO performance is what you need.

In my experience, the Sony sensors used in most of the Olympus OMD cameras are almost, but not quite ISO invariant.

Base ISO for a given camera: Is there a reference or base iso for a given camera model? In my instance the OMD em1mkii. What does this mean in practical terms? Would that ISO give the best combination of dynamic range and noise? And does dynamic range actually change with ISO changes?

I'm just curious. Obviously, composition and quality of light are much more important.

Base ISO on all the 16 or 20Mp Olympus cameras is 200. That's where you'll get the best DR and lowest noise. The extended low ISO values work by overexposing at capture time and pulling back in post (via in-camera JPEG processing, or if you're shooting raw by setting the grey level value in the raw file that is then interpreted by LR (or whatever you're using). It can deliver lower noise, but only when you can be sure not to blow highlights.

What base ISO means is that when the camera is set for that ISO and you give a calibrated exposure corresponding to that ISO (using the ISO standards), then the sensor charge wells will be filled to an optimum level. In other words, you'll get the best signal to noise ratio on the image. If you think of the sensor as lots of little buckets that get filled up with light, then when you are at base ISO, the metered exposure will fill the buckets, but they won't overflow. If you crank up the ISO above this then the metered exposure will be less in absolute terms (faster shutter speed and/or smaller aperture), and the result will be that the buckets will now be only partly full. To get the right brightness later, you need to amplify the buckets and that means noise and hence lower signal to noise.

Finally, if you want to see the effect of ISO invariance, you can do this using the DP Review studio test scene (links below). Here's an example of shooting the same scene with the EM1.3. One at ISO 3200, the other at ISO 200 and boosted four stops:



It's close, but not perfectly ISO invariant - the ISO 3200 is a little better (but not really much).

Here's the corresponding example with the Panasonic S5:



You'll see that the two shots now have very similar noise.


Links:

- DPR studio scene: https://www.dpreview.com/reviews/ima...son/fullscreen
- DPR studio scene with EV comp in post: https://www.dpreview.com/reviews/ima...en?&widget=327

[TravelingChick]

Thank you for your detailed response, Paul. After reading through, my initial thought would be that from a practical standpoint (i.e. in the field) trying to stay between 200-1600 ISO is kind of the sweet spot. I only really go lower than 200 if I need to for shutter speed. Likewise, I'm not adverse to pushing the ISO way up if I need the speed (BIF, esp early or late in the day). I know the noise gets quite bad especially in the extended high range ISO, but sometimes you have to do it.

I'm a generalist as far as my photography goes. I lean to natural history - landscapes, animals, etc - but I travel a lot so there's a fair amount of 'whatever is in front of me.' I always shoot RAW. Now that I have a better understanding of ISO, I feel like I've been on the right track: I sometimes do exposure blending, and sometimes use the in-camera HDR; But MOSTLY, don't blow out the highlights while exposing to the right. If you need ISO, you need it, and fix it in post.

Thanks again. The physics behind these sensors is fascinating.

[pdk42]

If you want to see a camera that has poor ISO invariance, then look at the Panasonic GH6. These two shots are at the same exposure, but one is at ISO 100 and pushed four stops. The other is at ISO 1600 but not pushed at all. The pushed image has way more noise. The noise on the ISO 1600 shot is competitive though so for BIF etc where you're just needing a higher ISO, it will perform much the same as an Olympus camera; but for landscape where you need to push shadows, it'll be much worse.

[pdk42]

For comparison, here's the Olympus EM1.3 with the same bit of the image enlarged.

[TravelingChick]

Wow. Quite a difference in those Panasonic shots. The Olympus looks amazing (from the standpoint of iso invariance) in comparison.

[MikeOxon]

I have a slight quibble with pdk42 - it's not just down to the sensor but to how the signals are processed in the camera.

The original signal from the sensor is analogue and at some stage has to be converted to digital in an 'A to D' converter. Some cameras use more 'analogue' amplification, whereas others multliply the digital numbers, to brighten an image. If the camera relies mainly on 'digital' amplification then it makes no difference if you do the same multiplication in a computer on the raw digital data - these camera are 'ISO independent'. If the camera makes changes to the analogue signal, as many Canon cameras do, then post-processing is not the same as changing ISO in the camera. It's a complicated matter but it's not just down to the sensor - the design of the electronics in the camera plays a part as well.

Mike.

[pdk42]

Yes, I agree. I was being a bit loose with my explanation!

[MikeOxon]

I used to use Nikon cameras and discovered, long before the term 'ISO independent' appeared, that it made little difference if I increased 'exposure' in my raw converter, instead of turning up the ISO setting on the camera. Friends with Canon cameras found that they didn't behave in the same way. It seems to work with my Olympus e-m1-ii, which the OP referred to.

[gazza95]

To follow up in Mike point about ampler steps. The infamous photosonphotos web site shows that the em1-ii has two step points at 200 and 1600 iso;



So academically this suggests that for the em1-ii iso 320 MIGHT give better response than 200. While the OM-1 has these points set to 800 and 12,800.


Second point to make is that there is a lot more noise processing being done now so the apparent effects of using higher iso or underexposing are being much reduced. This means I shoot in manual and auto iso instead of the shutter priority that I would have done on older cameras.

The noise reduction process's we all use totally changes the dynamic range we see in reality compared to the graph above an means we are getting acceptable images at much higher iso's that we used to.

It is therefore easy to to get lulled into the thinking that iso does not matter much.In reality shooting lower iso is way better as long as shutter speed high enough to avoid blurring.

Gary

[MikeOxon]

As Gary says, camera manufacturers tweak things to make their product look better than the competition.

If customers want to take black cats in coal cellars at ISO 128,000, then the manufacturers respond by changing the internal processing in order to make such images look 'better'. There's an assumption that users turn to higher ISO in poor light, so the processing parameters are changed to favour that type of photo. We cannot therefore assume that there's a simple proportionality between sensor (analogue) output and raw (digital) data.

[tmp]

if i understand it right...choose a higher, appropriate, ISO at time of shooting instead of using too much exposure 'slider' at computer afterwards?
same as what Thomas Eisl explained here https://www.youtube.com/watch?v=SXm3peOg5rc