Dynamic Range - Underexposed to avoid High ISO

[John31000]

Hello everyone

I had some questions after looking some informations about dynamic range, iso and so on ...

My camera (olympus em1 mark ii) got a function to see if some area are overexposed or underexposed.

If I'm in bad lighting condition like night outdoor and at 0EV and I doesn't see underexposed area. Can I underexposed to gain some ISO stop right ?

Like i'm at : f 1.7 , ISO 6400, 0EV , SS : 1/10 -> f1.7 , ISO 3200, -1EV, SS : 1/10

In post processing, I can then overexposed by one stop to get the same image as I would have taken if i was at 0EV without losing details if I CANT see "underexposed area" at -1EV ?

I guess the same can be applied for overexposed area ?

Also, if the camera has a very good dynamic range (like some high end FF model) I would guess I can underexposed more than with my olympus ?

Thanks

 

[alcelc]

You are indeed underexpose the scene by -XXev, and then do a shadow push +XXev by PP (provided that you have protected the highlight section). Many shooters do it and basically it is the principal idea behind ETTR.

However when you think shooting with a lower ISO will keep noise down, when you do shadow push you actually have to face a more worse noise in the shadow area (result of underexposure, especially more serious on a smaller size sensor) such that heavier NR be needed (for the pushed area).

Applying the technique to protect highlight (underexpose the shadow) is better than on another way round because once highlight be overblown, it would be harder to recover in PP.

For very high contrast scenes I would rather use AEB shots and merge them into HDR. It can definitely help to widen the DR of end product at a lower cost on noise.

My 2 cents.

 

[Chris R-UK]

Hello everyone

I had some questions after looking some informations about dynamic range, iso and so on ...

My camera (olympus em1 mark ii) got a function to see if some area are overexposed or underexposed.

If I'm in bad lighting condition like night outdoor and at 0EV and I doesn't see underexposed area. Can I underexposed to gain some ISO stop right ?

Like i'm at : f 1.7 , ISO 6400, 0EV , SS : 1/10 -> f1.7 , ISO 3200, -1EV, SS : 1/10

These are actually the same exposure - the same amount of light hits the sensor in both cases. ISO affects the brightness of the image that the camera produces but not the exposure which is determined by the aperture and shutter speed only.

In post processing, I can then overexposed by one stop to get the same image as I would have taken if i was at 0EV without losing details if I CANT see "underexposed area" at -1EV ?

You finish up with the same amount of noise in the image. The noise is determined by the amount of light hitting the sensor, the exposure.

If you had underexposed by using 1/20s shutter speed, that would have halved the light hitting the sensor and would have increased the noise. Varying the ISO has no effect on the noise.

I shoot nighttime sports with an E-M1.2. I shoot in M mode with the aperture wide open, the shutter speed set at the minimum required to avoid most motion blur and auto ISO with a maximum of 5000. When the ISO hits the maximum and the EVF shows a negative EV value, I just continue shooting and correct the brightness in post processing.

 

[pocoloco]

In my experience, that is a bad idea!

Fixing underexposed images in post do not give you good results, they will show noise!

And... especially M43... one of the weaker areas of it is brightening darker areas in post processing, and if you underexposed... it will be really messy.

You might be able to get away with more using a FF, you have way more room to play with with images from a FF sensor.

 

[Iliah Borg]

Hello everyone

I had some questions after looking some informations about dynamic range, iso and so on ...

My camera (olympus em1 mark ii) got a function to see if some area are overexposed or underexposed.

If I'm in bad lighting condition like night outdoor and at 0EV and I doesn't see underexposed area. Can I underexposed to gain some ISO stop right ?

Like i'm at : f 1.7 , ISO 6400, 0EV , SS : 1/10 -> f1.7 , ISO 3200, -1EV, SS : 1/10

Please see the post made by Chris R, https://www.dpreview.com/forums/post/63686112

If the scene tolerates 1/10 of a sec., consider using a tripod and setting a slower shutter speed.

In post processing, I can then overexposed by one stop to get the same image as I would have taken if i was at 0EV without losing details if I CANT see "underexposed area" at -1EV ?

It's conversion, not post-processing.

I guess the same can be applied for overexposed area ?

Depends on the converter, if the ISO adjustment in the converter (often called "Exposure" slider) is not preserving highlights by applying a proper curve, highlights will suffer clipping.

Also, if the camera has a very good dynamic range (like some high end FF model) I would guess I can underexposed more than with my olympus ?

To benefit from a larger sensor you need to collect more light with that sensor than with your current sensor.

--
http://www.libraw.org/

 

[bclaff]

Hello everyone

I had some questions after looking some informations about dynamic range, iso and so on ...

My camera (olympus em1 mark ii) got a function to see if some area are overexposed or underexposed.

If I'm in bad lighting condition like night outdoor and at 0EV and I doesn't see underexposed area. Can I underexposed to gain some ISO stop right ?

Like i'm at : f 1.7 , ISO 6400, 0EV , SS : 1/10 -> f1.7 , ISO 3200, -1EV, SS : 1/10

In post processing, I can then overexposed by one stop to get the same image as I would have taken if i was at 0EV without losing details if I CANT see "underexposed area" at -1EV ?

I guess the same can be applied for overexposed area ?

Also, if the camera has a very good dynamic range (like some high end FF model) I would guess I can underexposed more than with my olympus ?

f/1.7 1/10s is exactly the same exposure regardless of the ISO setting.
You are not underexposing; you are under lightening.

Provided the results at ISO 6400 is not clipped you would do (slightly) better than at ISO 3200. You can examine a chart like Shadow Improvement of Photographic Dynamic Range versus ISO Setting to see how much it matters.

For example ISO 6400 is 0.94 and ISO 3200 is 0.83 so you sacrifice 0.11 stops by using too low an ISO setting and lightening later.

In this case you would be OK but in general it's a bad strategy.

First, gather as much light as you can (aperture and exposure time) at base ISO setting.
Then, raise ISO as needed (without clipping) to achieve the correct lightness.

In this case I would suggest Manual mode with ISO Auto turned on.
Use exposure compensation if necessary.
The camera is much better at choosing the ISO setting than you are.

--
Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

 

[j3ffw]

You are indeed underexpose the scene by -XXev, and then do a shadow push +XXev by PP (provided that you have protected the highlight section). Many shooters do it and basically it is the principal idea behind ETTR.

Hi Albert, maybe I'm reading you wrong, but isn't ETTR the exact opposite? Expose To The Right as much as possible without clipping the highlights and then pulling the shadows back down in post. To clarify overexpose the image beyond your taste (not clipping the highlights) then tone down your shadows (normally with a curves adj) in post not pushing up the shadows.

Again I may be reading your sentence wrong???

 

[alcelc]

You are indeed underexpose the scene by -XXev, and then do a shadow push +XXev by PP (provided that you have protected the highlight section). Many shooters do it and basically it is the principal idea behind ETTR.

Hi Albert, maybe I'm reading you wrong, but isn't ETTR the exact opposite? Expose To The Right as much as possible without clipping the highlights and then pulling the shadows back down in post. To clarify overexpose the image beyond your taste (not clipping the highlights) then tone down your shadows (normally with a curves adj) in post not pushing up the shadows.

Again I may be reading your sentence wrong???

 

[Razor512]

When you adjust the ISO in a camera, what you are really adjusting is the behavior of the LNA and ADC. For example:

While ISO doesn't impact anything about the physical CMOS sensor, it doesn't change the limitations of how the front end and back end usually works. Even for super expensive lab equipment, the signal range of an analog sensor typically far exceeds the working range of the ADC, thus the analog signal needs to be processed in a way that puts the most relevant parts of the analog signal within the working range of the ADC.

For amplifiers there are no noise free amplifiers, thus more amplification will introduce some additional read noise, but it is far less than the noise floor of the ADC, and any analog signal within or below the noise floor will simply be lost.



When sensor saturation is low, not only is some of the low gain information being lost in the noise floor and below the working range of the ADC, but you are also not utilizing the full range of the ADC. since it can't adjust its own working range. e.g., if its working range is between 300 and 1500mv, then to get the best sampling accuracy of what little signal is input, you want to have the most relevant parts of the signal in the ideal range of the ADC.

The balancing of the various gains will not mean a massive boost in SNR, but it is all about making the most of the signal that the sensor is outputting.

As you boost the ISO, since the gain is global, any portion of the sensor that is saturated, but not clipping at the analog level, will end up being clipped when the LNA then goes and adds like +20dB to the signal and sends that aspect above the working range of the ADC, thus highlights that would be perfectly fine at ISO 100, will be clipped beyond recovery at ISO 3200.



In your situation, you will have better results by using the proper ISO for your given light level, the only worthwhile exception is if you are dealing a situation where you would be saving a histogram where you have 2 aspects of the scene that you must keep, e.g., a bunch of shadow detail but also a few bright highlight that you do not want clipped, where you are willing to trade a higher noise floor for retaining a specific set of highlights. Outside of an edge case like that, just use the proper ISO for the scene if you are unable to get away with using the base ISO.

 

[bobn2]

In my experience, that is a bad idea!

Fixing underexposed images in post do not give you good results, they will show noise!

Very likely they will if your adjust in 'post', because the in processing all the information below the output black level will have been thrown away. To do this kind of thing you need to work from a raw file and set the output lightness as you want in 'processing', not 'post-processing'. It is in precisely this kind of situation that the distinction matters.

And... especially M43... one of the weaker areas of it is brightening darker areas in post processing, and if you underexposed... it will be really messy.

'Brightening darker areas' is really not the way to look at it. What you're dong in processing ('development', 'conversion') is determining the output lightness for the input exposure. You're not 'brightening' because there was no determined 'brightness' before you processed. In this regard, mFT is no worse than any other format. It's sensors tend to have low electronic noise and quite nice noise characteristics.

You might be able to get away with more using a FF, you have way more room to play with with images from a FF sensor.

That depends entirely on how you expose them. Some FF cameras will give you distinctly 'less' to play with than some mFT or APS-C cameras.

 

[bobn2]

When you adjust the ISO in a camera, what you are really adjusting is the behavior of the LNA and ADC. For example:

While ISO doesn't impact anything about the physical CMOS sensor, it doesn't change the limitations of how the front end and back end usually works. Even for super expensive lab equipment, the signal range of an analog sensor typically far exceeds the working range of the ADC, thus the analog signal needs to be processed in a way that puts the most relevant parts of the analog signal within the working range of the ADC.

For amplifiers there are no noise free amplifiers, thus more amplification will introduce some additional read noise, but it is far less than the noise floor of the ADC, and any analog signal within or below the noise floor will simply be lost.

When sensor saturation is low, not only is some of the low gain information being lost in the noise floor and below the working range of the ADC, but you are also not utilizing the full range of the ADC. since it can't adjust its own working range. e.g., if its working range is between 300 and 1500mv, then to get the best sampling accuracy of what little signal is input, you want to have the most relevant parts of the signal in the ideal range of the ADC.

The balancing of the various gains will not mean a massive boost in SNR, but it is all about making the most of the signal that the sensor is outputting.

As you boost the ISO, since the gain is global, any portion of the sensor that is saturated, but not clipping at the analog level, will end up being clipped when the LNA then goes and adds like +20dB to the signal and sends that aspect above the working range of the ADC, thus highlights that would be perfectly fine at ISO 100, will be clipped beyond recovery at ISO 3200.

In your situation, you will have better results by using the proper ISO for your given light level, the only worthwhile exception is if you are dealing a situation where you would be saving a histogram where you have 2 aspects of the scene that you must keep, e.g., a bunch of shadow detail but also a few bright highlight that you do not want clipped, where you are willing to trade a higher noise floor for retaining a specific set of highlights. Outside of an edge case like that, just use the proper ISO for the scene if you are unable to get away with using the base ISO.

A nicely informative post. Just one quibble - it's quite important here to differentiate what the 'ISO control' does, which is unstandardised and varies from camera to camera, and what 'ISO' does. The 'ISO control' should change the ISO (the relationship between input exposure and output lightness) as per the ISO standard, but it will typically do other things as well, as your post says.

--
...because you know, sometimes words have two meanings.

 

[Razor512]

There are other things ti can change, but typically when you get to a point where it stops adjusting analog gain, the manufacturer will list those options as expanded ISO ranges or bury it in some menu.

A crude way of looking at it to take a camera, even ones that some may consider ISO invariant , and with the lens cap on and in a dark environment where you can ensure no light is entering, take a few images at a low ISO, then a few more at a high ISO (limiting the high to the end of the analog gain range in the case of cameras with multiple smaller range LNAs (some companies do that to reduce the read noise). (works on cameras that can do uncompressed raw)

Then just scrub through the raw files using a decompiler such a IDA pro. No need to try and make sense of any of the raw data (not designed to even try and make sense of it) the actual raw data will look a lot different comparing a low and high ISO file compared to just comparing a few of the same ISO images of complete darkness.

But once you get to an expanded range, things will start to behave more like if you are using the same ISO, basically just more variation, randomness thrown into the mix.

Outside of that, for lossless compressed raw, you will notice that at high ISO, the file size increases.

PS, the first few hundred lines will be the EXIF data as well as other fixed data that the device maker wants to include such as embedded lens profiles, where unused address space will just be left blank or show up as a bunch of 20s in a hex view. After that, you will get into the actual raw image data, (while you can't make sense of it, making random changes to portions of that will actually cause visible corruption in an image, thus it is the pixel data.

If you were only dealing with digital gain, values would be offset but the overall look of the binary would be about the same

The issue is that ADCs cannot adapt to a constrained input, they simply have a working input range, and depending on the bit depth, in equal increments that range will be sampled at that bit depth.

While each camera maker has their own special way of handling ISO, where each company is trying their best to get the maximum amount of usable info from the sensor). All still try to keep as much of the analog signal as possible above the noise floor of the ADC. If the signal is above the noise floor, then the noise can effectively be isolated, leaving just the noise introduced by the amplification process. That kind of stuff will then be handled by the SNR firmware processing the raw files somewhat to deal with dome of the read noise.

Outside of that, Most ADCs also allow for digital gain where they essentially offset the values up though I don't think it is used very often unless you are trying to further process an ADCs output before finally saving it.

 

[bobn2]

There are other things ti can change, but typically when you get to a point where it stops adjusting analog gain, the manufacturer will list those options as expanded ISO ranges or bury it in some menu.

A crude way of looking at it to take a camera, even ones that some may consider ISO invariant , and with the lens cap on and in a dark environment where you can ensure no light is entering, take a few images at a low ISO, then a few more at a high ISO (limiting the high to the end of the analog gain range in the case of cameras with multiple smaller range LNAs (some companies do that to reduce the read noise). (works on cameras that can do uncompressed raw)

Then just scrub through the raw files using a decompiler such a IDA pro. No need to try and make sense of any of the raw data (not designed to even try and make sense of it) the actual raw data will look a lot different comparing a low and high ISO file compared to just comparing a few of the same ISO images of complete darkness.

But once you get to an expanded range, things will start to behave more like if you are using the same ISO, basically just more variation, randomness thrown into the mix.

Outside of that, for lossless compressed raw, you will notice that at high ISO, the file size increases.

PS, the first few hundred lines will be the EXIF data as well as other fixed data that the device maker wants to include such as embedded lens profiles, where unused address space will just be left blank or show up as a bunch of 20s in a hex view. After that, you will get into the actual raw image data, (while you can't make sense of it, making random changes to portions of that will actually cause visible corruption in an image, thus it is the pixel data.

If you were only dealing with digital gain, values would be offset but the overall look of the binary would be about the same

The issue is that ADCs cannot adapt to a constrained input, they simply have a working input range, and depending on the bit depth, in equal increments that range will be sampled at that bit depth.

While each camera maker has their own special way of handling ISO, where each company is trying their best to get the maximum amount of usable info from the sensor). All still try to keep as much of the analog signal as possible above the noise floor of the ADC. If the signal is above the noise floor, then the noise can effectively be isolated, leaving just the noise introduced by the amplification process. That kind of stuff will then be handled by the SNR firmware processing the raw files somewhat to deal with dome of the read noise.

Outside of that, Most ADCs also allow for digital gain where they essentially offset the values up though I don't think it is used very often unless you are trying to further process an ADCs output before finally saving it.

All good stuff and true, but not addressing the point that I was making. Clearly you have an electronics/ signal processing background, and use the terms that you would from that background. The problem is, here on a beginner's forum, is what lay people, particularly beginners, make of those terms. In my experience, what they make of the discussion is that ISO is 'gain'. which is fundamentally untrue. From that idea follow a whole load of misconceptions. Your summary of how the ISO control works on most cameras, and what is the effect of making changes in ISO setting is all good, but probably outside the scope of 'beginner's' concerns, as is the discussion about why looking at 'ISO' as 'gain' is fundamentally wrong. If you want to follow up on pm or the PS&T forum, I'm quite happy to do so.

 

[ThrillaMozilla]

Guys, this is the Beginners' Forum, and the OP is confused. You might want to think of a simple answer.

 

[Brian Kimball]

Hello everyone

I had some questions after looking some informations about dynamic range, iso and so on ...

My camera (olympus em1 mark ii) got a function to see if some area are overexposed or underexposed.

If I'm in bad lighting condition like night outdoor and at 0EV and I doesn't see underexposed area. Can I underexposed to gain some ISO stop right ?

Like i'm at : f 1.7 , ISO 6400, 0EV , SS : 1/10 -> f1.7 , ISO 3200, -1EV, SS : 1/10

These are actually the same exposure - the same amount of light hits the sensor in both cases. ISO affects the brightness of the image that the camera produces but not the exposure which is determined by the aperture and shutter speed only.

As others have noted, this is a great response.

In post processing, I can then overexposed by one stop to get the same image as I would have taken if i was at 0EV without losing details if I CANT see "underexposed area" at -1EV ?

You finish up with the same amount of noise in the image. The noise is determined by the amount of light hitting the sensor, the exposure.

Shot noise, yes. Not necessarily total noise.

If you had underexposed by using 1/20s shutter speed, that would have halved the light hitting the sensor and would have increased the noise.

Definitely.

Varying the ISO has no effect on the noise.

A reasonably practical approximation for OP's camera only. Read noise can vary with ISO setting and does with many cameras. I know OP's EM1 is relatively ISO invariant, but it's still important to know that some other cameras' higher ISO settings produce less noise in photos than images shot with the same exposure at lower ISOs and lightened in post.

I shoot nighttime sports with an E-M1.2. I shoot in M mode with the aperture wide open, the shutter speed set at the minimum required to avoid most motion blur and auto ISO with a maximum of 5000. When the ISO hits the maximum and the EVF shows a negative EV value, I just continue shooting and correct the brightness in post processing.

 

[Brian Kimball]

[...]

if the ISO adjustment in the converter (often called "Exposure" slider)

I was taken to task recently in the PS&T forum for insisting that, for all practical purposes, raw files had an ISO value even if technically the ISO standard doesn't cover raw files.

I definitely don't want to re-argue that; I got bobn2's point (the nuanced difference between "this raw file has an ISO value of XXX" and "this raw file was shot with the camera set to ISO XXX") and backed off. I get it.

But why then is it acceptable to call an "Exposure" slider in post processing software an ISO adjustment? I thought the ISO standard applied to images created by DSCs. PP software running on a computer is not a DSC.

I'm really trying not to sound snarky here. Genuine question.

 

[Iliah Borg]

[...]

if the ISO adjustment in the converter (often called "Exposure" slider)

why then is it acceptable to call an "Exposure" slider in post processing software

It's not a post-processing software, it's a converter - a digital developer, in other words.

an ISO adjustment?

ISO speed is "born" in the development, be it in-camera converter or a self-standing converter.

--
http://www.libraw.org/

 

[D Cox]

[...]

if the ISO adjustment in the converter (often called "Exposure" slider)

I was taken to task recently in the PS&T forum for insisting that, for all practical purposes, raw files had an ISO value even if technically the ISO standard doesn't cover raw files.

I definitely don't want to re-argue that; I got bobn2's point (the nuanced difference between "this raw file has an ISO value of XXX" and "this raw file was shot with the camera set to ISO XXX") and backed off. I get it.

But why then is it acceptable to call an "Exposure" slider in post processing software an ISO adjustment? I thought the ISO standard applied to images created by DSCs. PP software running on a computer is not a DSC.

I'm really trying not to sound snarky here. Genuine question.

The Exposure slider is not an ISO adjustment. It's a Midtones adjustment, to go with the Shadows and the Highlights sliders. Look at a Levels control, and you will see all three together.

The "ISO" control on a camera is best thought of as a control that biases the meter to give less exposure (in auto modes) and increases noise reduction to reduce the bad effects of underexposure. You most often want to reduce exposure in order to use a faster shutter speed. You can do this also with the Exposure Compensation control, but does nothing to the signal-to-noise ratio.

Noise reduction includes the analog amp between the sensor and the analog-to-digital converter, and any software noise reduction in the JPG engine or the raw converter.

 

[petrochemist]

[...]

if the ISO adjustment in the converter (often called "Exposure" slider)

I was taken to task recently in the PS&T forum for insisting that, for all practical purposes, raw files had an ISO value even if technically the ISO standard doesn't cover raw files.

I definitely don't want to re-argue that; I got bobn2's point (the nuanced difference between "this raw file has an ISO value of XXX" and "this raw file was shot with the camera set to ISO XXX") and backed off. I get it.

But why then is it acceptable to call an "Exposure" slider in post processing software an ISO adjustment? I thought the ISO standard applied to images created by DSCs. PP software running on a computer is not a DSC.

I'm really trying not to sound snarky here. Genuine question.

The Exposure slider is not an ISO adjustment. It's a Midtones adjustment, to go with the Shadows and the Highlights sliders. Look at a Levels control, and you will see all three together.

In levels you may get 3 sliders but there are plenty of other places where the brightness can be controlled. Some of these are simple 'Brightness' controls, usually sitting alongside a contrast & possibly Gamma slider.

 

[petrochemist]

Hello everyone

I had some questions after looking some informations about dynamic range, iso and so on ...

My camera (olympus em1 mark ii) got a function to see if some area are overexposed or underexposed.

If I'm in bad lighting condition like night outdoor and at 0EV and I doesn't see underexposed area. Can I underexposed to gain some ISO stop right ?

Like i'm at : f 1.7 , ISO 6400, 0EV , SS : 1/10 -> f1.7 , ISO 3200, -1EV, SS : 1/10

In post processing, I can then overexposed by one stop to get the same image as I would have taken if i was at 0EV without losing details if I CANT see "underexposed area" at -1EV ?

I guess the same can be applied for overexposed area ?

Also, if the camera has a very good dynamic range (like some high end FF model) I would guess I can underexposed more than with my olympus ?

Thanks

Underexposing will give just the same noise issues as using the same exposure & brightening in post. With some cameras it will reduce dynamic range more than using a higher ISO will. I don't know of any camera where underexposing to keep the ISO down will help.