ETR Metering

[Steve]

I've been thinking about this conversation a bit and one thing I think is happening is that ISO is treated as if it's an exposure control when it's not. I think there's an idea that ISO, Shutter Speed, and F/Stop are all equals in the exposure triangle and the truth is, they're not.

I think it helps to consider the order of how an image is created. Shutter speed and aperture are first and expose the sensor. This is the ONLY place where you can improve your signal-to-noise ratio and lower your noise levels.

So, if you shoot a properly exposed image at base ISO, no ISO gain is added - it's done. However, if you overexpose to the just before point of clipping at base ISO, that can only be done with shutter speed and aperture - so you can increase the SNR here.

However, if the image you have captured at base ISO is too dark, then you use the "volume knob" of ISO to turn up the brightness. The thing is, this is done after the shutter speed and aperture have done their job and the noise patten is already fixed. ISO can't change what's there and it can't improve the SNR - that's already set my the shutter speed and aperture. All ISO can do is can only make what you already have brighter.

That's why I keep saying ETTR is really only effective at base ISO (or, in the case of dual gain sensors, you can see an advantage once the secondary amp kicks in). That's also why I say that, unless you're at base ISO (or at the ISO just before the second gain circuit kicks in), there's nothing to be gained by deliberately overexposing your shot. Just shoot a proper level of brightness and it's all good

 

[DRwyoming]

I think what most are describing in this thread is simply adding a little extra exposure as needed, not ETTR as I understand it. I think that's why there are so many differing thoughts

Agreed, introducing positive exposure compensation to bring out the important shadow details like eyes or facial features isn't really the same as shooting with an ETTR philosophy where you'd consistently go beyond ideal shadow or midtone exposure and to the edge of hard clipping the raw highlights with the intention of darkening the image in post.

The former I do all the time, actual ETTR not so much.

 

[Butlerkid]

For wildlife I shoot manual exposure. I try to set my exposure to capture all the data in the image that I want at the lowest ISO possible once I've set the shutter speed and aperture. Generally, but not always, the data I want includes shadow and highlight details. Thus no clipping of the blacks or white. The shape and position of the histogram will vary for each image, but will contain no clipping of blacks or highlights.

Guess I don't shoot ETTR. Never bothered me or slowed me down in the field. Or caused my head to burst from considering all these technical analyses ................ LOL!

 

[DRwyoming]

However, if the image you have captured at base ISO is too dark, then you use the "volume knob" of ISO to turn up the brightness. The thing is, this is done after the shutter speed and aperture have done their job and the noise patten is already fixed. ISO can't change what's there and it can't improve the SNR - that's already set my the shutter speed and aperture. All ISO can do is can only make what you already have brighter.

FWIW, that's exactly why the only automated exposure mode I use with modern cameras is Manual/Auto ISO. I want direct control over shutter speed and aperture for creative reasons including not only stopping or blurring motion and controlling DoF but also for noise management reasons.

Once those are set the specific ISO chosen, whether in-camera or in post doesn't change the fundamental image nearly as much as tradeoffs in those manual settings and most modern digital cameras have fairly broad regions of ISO invariance so small corrective ISO tweaks in post aren't substantially different than making those small tweaks to ISO in-camera.

I know, that's thread creep and folks are talking about ETTR, but to me your post nicely sums up the advantages to shooting Manual/Auto ISO as an automated exposure mode, control the things that matter and let the ISO chips fall where they may within reason knowing that something will have to give on the Shutter Speed/Aperture front if the ISO is just too far from ideal.

 

[Wes Peterson]

I find ETTR doesn't really work as most hope. Once upon a time those tricks helped with older sensors, but with the newer sensors it really doesn't make a difference. ISO is a brightness control, nothing more. It doesn't add light, cranking it up doesn't make your sensor more sensitive to light (dual gain sensors can help, but there's still no extra light added). It's like if you're in Lightroom and crank up the exposure and then bring it back down. As you crank it up, the noise gets worse, as you bring it back it gets better, but the level of noise you have at say ISO 800 doesn't get any better by cranking it up and back down.

I've tested this and I think the exposure and metering book has some samples. There's no benefit to over exposing and then pulling the exposure down in post unless you are doing it at base ISO (for increased DR).

The way I find it can help is if you already have at high ISO but with too much dynamic range... I think its a little cleaner to lower the highlights than to raise the shadows. Its a rare situation with a minimal gain tho.

 

[bleirer]

I've been thinking about this conversation a bit and one thing I think is happening is that ISO is treated as if it's an exposure control when it's not. I think there's an idea that ISO, Shutter Speed, and F/Stop are all equals in the exposure triangle and the truth is, they're not.

I think it helps to consider the order of how an image is created. Shutter speed and aperture are first and expose the sensor. This is the ONLY place where you can improve your signal-to-noise ratio and lower your noise levels.

So, if you shoot a properly exposed image at base ISO, no ISO gain is added - it's done. However, if you overexpose to the just before point of clipping at base ISO, that can only be done with shutter speed and aperture - so you can increase the SNR here.

However, if the image you have captured at base ISO is too dark, then you use the "volume knob" of ISO to turn up the brightness. The thing is, this is done after the shutter speed and aperture have done their job and the noise patten is already fixed. ISO can't change what's there and it can't improve the SNR - that's already set my the shutter speed and aperture. All ISO can do is can only make what you already have brighter.

That's why I keep saying ETTR is really only effective at base ISO (or, in the case of dual gain sensors, you can see an advantage once the secondary amp kicks in). That's also why I say that, unless you're at base ISO (or at the ISO just before the second gain circuit kicks in), there's nothing to be gained by deliberately overexposing your shot. Just shoot a proper level of brightness and it's all good

That certainly has to be true for photon noise, but some cameras have in camera magic after exposure to help with other sources of noise that is not simply digital amplification. So for some cameras it seems there is something to be gained with iso, as in the photons to photos for the A1 I linked above, but less to gain as cameras handle other than photon noise better and better. In the A1 case only 1 stop of extra help.

 

[EricBowles]

With any of these techniques, just keep in mind what you can do with the software you use now - and in the future. ACR/Photoshop/Lightroom, Capture One, NX Studio, CDDP4, and others each have pretty good tools to manipulate RAW exposures. But each tool has it's sweetspot and its limits. Steve described testing to understand he can recover a specific amount of blown highlights. There are going to be other situations where you can recover the exposure, but colors may be tricky. Add to that everyone has a skill set with the tools they use, and proficiency with using tools can make a difference.

 

[bleirer]

One other thought is that ettr is not always or even usually brightening the image. I know from testing my camera in Rawdigger that whatever the meter places at the middle I only have 3.5 stops to the right to work with, so a lot of times it's still ettr even if I'm clawing back a needed highlight with negative compensation. Also from Rawdigger I know that at the highest non-blinkie exposure in camera, I still have 2/3 stop before I blow out, so it's possible to place the brightest important part of the scene pretty precisely, and that often means negative compensation to keep it within that 2/3 stop leeway.

 

[John Woodworth]

I've been thinking about this conversation a bit and one thing I think is happening is that ISO is treated as if it's an exposure control when it's not. I think there's an idea that ISO, Shutter Speed, and F/Stop are all equals in the exposure triangle and the truth is, they're not.

I think it helps to consider the order of how an image is created. Shutter speed and aperture are first and expose the sensor. This is the ONLY place where you can improve your signal-to-noise ratio and lower your noise levels.

So, if you shoot a properly exposed image at base ISO, no ISO gain is added - it's done. However, if you overexpose to the just before point of clipping at base ISO, that can only be done with shutter speed and aperture - so you can increase the SNR here.

However, if the image you have captured at base ISO is too dark, then you use the "volume knob" of ISO to turn up the brightness. The thing is, this is done after the shutter speed and aperture have done their job and the noise patten is already fixed. ISO can't change what's there and it can't improve the SNR - that's already set my the shutter speed and aperture. All ISO can do is can only make what you already have brighter.

That's why I keep saying ETTR is really only effective at base ISO (or, in the case of dual gain sensors, you can see an advantage once the secondary amp kicks in). That's also why I say that, unless you're at base ISO (or at the ISO just before the second gain circuit kicks in), there's nothing to be gained by deliberately overexposing your shot. Just shoot a proper level of brightness and it's all good

I contend that there is a SNR difference between adjusting aperture versus adjusting shutter speed.

Increasing aperture size lets in more light and increases the SNR proportionally.

Increasing shutter duration doesn't change the SNR at all. It just reduces the proportional statistical variation of noise and signal, making it look smoother.

 

[bleirer]

I contend that there is a SNR difference between adjusting aperture versus adjusting shutter speed.

Increasing aperture size lets in more light and increases the SNR proportionally.

Increasing shutter duration doesn't change the SNR at all. It just reduces the proportional statistical variation of noise and signal, making it look smoother.

I'm not sure how that works. To my thinking more photons are more photons, whether it's from longer exposure or larger aperture or better light in the scene. Going back to the photons dislodging electrons in the pixel/photosite, a longer duration has to dislodge more pixels than a shorter exposure until the 'well' is too full to hold any more. Would you mind clarifying?

 

[Woodyg3]

ETTR is an interesting concept, and I can see times when it might make sense to consider. I never think of anything like that in the field, though. I am simply exposing for the subject. For me, keeping it simple gives me the best picture files.

 

[Nimi]

I contend that there is a SNR difference between adjusting aperture versus adjusting shutter speed.

Increasing aperture size lets in more light and increases the SNR proportionally.

Increasing shutter duration doesn't change the SNR at all. It just reduces the proportional statistical variation of noise and signal, making it look smoother.

I'm not sure that's the case with modern sensors. Sure, there'll be the occasional hot pixel, but I am under the impression that there is no "reciprocity failure."

 

[John Woodworth]

I'm not sure how that works. To my thinking more photons are more photons, whether it's from longer exposure or larger aperture or better light in the scene. Going back to the photons dislodging electrons in the pixel/photosite, a longer duration has to dislodge more pixels than a shorter exposure until the 'well' is too full to hold any more. Would you mind clarifying?

It's because increasing the shutter duration changes neither the rate of photons falling on the sensor, nor the rate of the noise. It merely collects both for a for a longer period of time. In this case, the ratio of signal to noise doesn't change.

When changing the shutter duration, the amount of signal and noise changes, the ratio doesn't. Increasing the amounts means that statistical variations are reduced, so that both signal and noise are "smoother", and that's good. But decreasing statistical variations in this way is not as efficient as directly increasing the signal (increasing the light on the sensor) or decreasing the noise (e.g. cooling the sensor).

 

[John Woodworth]

I'm not sure that's the case with modern sensors. Sure, there'll be the occasional hot pixel, but I am under the impression that there is no "reciprocity failure."

Jack your ISO up to 12,000, put the lense cap on, and give it a try.

 

[Replytoken]

I've been thinking about this conversation a bit and one thing I think is happening is that ISO is treated as if it's an exposure control when it's not. I think there's an idea that ISO, Shutter Speed, and F/Stop are all equals in the exposure triangle and the truth is, they're not.

I think it helps to consider the order of how an image is created. Shutter speed and aperture are first and expose the sensor. This is the ONLY place where you can improve your signal-to-noise ratio and lower your noise levels.

So, if you shoot a properly exposed image at base ISO, no ISO gain is added - it's done. However, if you overexpose to the just before point of clipping at base ISO, that can only be done with shutter speed and aperture - so you can increase the SNR here.

However, if the image you have captured at base ISO is too dark, then you use the "volume knob" of ISO to turn up the brightness. The thing is, this is done after the shutter speed and aperture have done their job and the noise patten is already fixed. ISO can't change what's there and it can't improve the SNR - that's already set my the shutter speed and aperture. All ISO can do is can only make what you already have brighter.

That's why I keep saying ETTR is really only effective at base ISO (or, in the case of dual gain sensors, you can see an advantage once the secondary amp kicks in). That's also why I say that, unless you're at base ISO (or at the ISO just before the second gain circuit kicks in), there's nothing to be gained by deliberately overexposing your shot. Just shoot a proper level of brightness and it's all good

Very well summarized. I think that the one point I would add it that "ETTR" is really the maximizing in the capture of data, which can really be looked at as a continuum. As a lot of people have stated in a variety of ways, we tend to optimize our exposures, which is also on that continuum, jut not necessarily at the maximum point. I think that the looking at this as a continuum rather than an either/or choice might help some folks understand the concept (maximizing the the capture of data) a little better.

--Ken

 

[Replytoken]

It's because increasing the shutter duration changes neither the rate of photons falling on the sensor, nor the rate of the noise. It merely collects both for a for a longer period of time. In this case, the ratio of signal to noise doesn't change.

When changing the shutter duration, the amount of signal and noise changes, the ratio doesn't. Increasing the amounts means that statistical variations are reduced, so that both signal and noise are "smoother", and that's good. But decreasing statistical variations in this way is not as efficient as directly increasing the signal (increasing the light on the sensor) or decreasing the noise (e.g. cooling the sensor).

Thank you for this additional information. I had the similar questions to @bleirer about why aperture and shutter are not the same. Can you possibly explain a bit more why opening the aperture is not the same as slowing down the shutter with an emphasis on the aperture side of the equation? I am just trying to better understand your point and this part is not coming together for me.

Thanks,

--Ken

 

[bleirer]

It's because increasing the shutter duration changes neither the rate of photons falling on the sensor, nor the rate of the noise. It merely collects both for a for a longer period of time. In this case, the ratio of signal to noise doesn't change.

When changing the shutter duration, the amount of signal and noise changes, the ratio doesn't. Increasing the amounts means that statistical variations are reduced, so that both signal and noise are "smoother", and that's good. But decreasing statistical variations in this way is not as efficient as directly increasing the signal (increasing the light on the sensor) or decreasing the noise (e.g. cooling the sensor).

Interesting. I clearly don't have the math to say you are right or wrong. Intuitively to me it seems if noise is capturing too bright or too dark pixels, more light from either source would make the distribution take on its characteristic and correct curve.

 

[Replytoken]

Interesting. I clearly don't have the math to say you are right or wrong. Intuitively to me it seems if noise is capturing too bright or too dark pixels, more light from either source would make the distribution take on its characteristic and correct curve.

That's similar to what I was thinking and why I asked for a bit more information in hopes of better understanding they underlying physics of it.

--Ken

 

[DRwyoming]

When changing the shutter duration, the amount of signal and noise changes, the ratio doesn't. Increasing the amounts means that statistical variations are reduced, so that both signal and noise are "smoother", and that's good. But decreasing statistical variations in this way is not as efficient as directly increasing the signal (increasing the light on the sensor) or decreasing the noise (e.g. cooling the sensor).

That's of course technically accurate but to some extent is splitting hairs for the field photographer. Yes the reduction in the AWGN component of noise happens through averaging but the net result is that longer shutter speeds do increase the final SNR in the image. Anyone running a tracking mount or doing multiple exposure stacking for astro photography can attest to the value of much longer integration times for bringing out color and dynamic range in astro images. Yes, the mechanism is averaging and it only applies to those noise components that are zero mean but the image quality improvement through longer shutter speeds can be substantial.

Sure if bringing in more light through a wider aperture is an option that's an even better way to go but once the aperture is maxed out as far as practical taking into account lens quality vs aperture and DoF issues, shutter speed is a viable way to increase both signal and noise integration times which can be very helpful. But yeah, if you delve into the physics of it, aperture and shutter speed are not identical nor perfectly reciprocal in terms of how they impact SNR.

 

[bleirer]

That's similar to what I was thinking and why I asked for a bit more information in hopes of better understanding they underlying physics of it.

--Ken

This article covers a lot of ground, but they also emphasize using shutter speed for the ettr part.

How to Maximize Image Quality in Photography

In the past, I’ve written about camera settings in terms of optimization - pushing your gear to the limits in order to maximize image quality. Today, I’ll revisit those advanced techniques and explain how to combine them to capture the highest quality images you possibly can.

photographylife.com

 

[John Woodworth]

Thank you for this additional information. I had the similar questions to @bleirer about why aperture and shutter are not the same. Can you possibly explain a bit more why opening the aperture is not the same as slowing down the shutter with an emphasis on the aperture side of the equation? I am just trying to better understand your point and this part is not coming together for me.

Thanks,

--Ken

Noise is inherent to the sensor (the sensor system) and doesn't depend directly on the incident light. Of course, it does depend on ISO which depends on incident light...Let's call this indirect. For a given ISO, you are stuck with the steady rate of noise the sensor gives you.

For a shutter duration of t, the amount of noise from the sensor is N x t and the amount of signal is S x t / A^2, where N and S are the rates of each contribution to the signal and A is proportional to aperture.

So, the scaling of SNR = S / (N x A^2) and time (shutter duration) factors out so SNR doesn't change with shutter duration. It does depend on aperture.

You can increase the knowledge of S + N and N with longer shutter durations. With assumptions, S+N will deviate about its true (average) value as 1/Sqrt(S+N) x t), approaching its true value slowly as 1/Sqrt(t).

Note that the alternative of changing aperture increases the SNR as 1 / A^2 . So a change of aperture by 1 stop doubles the SNR.

Hope this helps.

 

[Replytoken]

Noise is inherent to the sensor (the sensor system) and doesn't depend directly on the incident light. Of course, it does depend on ISO which depends on incident light...Let's call this indirect. For a given ISO, you are stuck with the steady rate of noise the sensor gives you.

For a shutter duration of t, the amount of noise from the sensor is N x t and the amount of signal is S x t / A^2, where N and S are the rates of each contribution to the signal and A is proportional to aperture.

So, the scaling of SNR = S / (N x A^2) and time (shutter duration) factors out so SNR doesn't change with shutter duration. It does depend on aperture.

You can increase the knowledge of S + N and N with longer shutter durations. With assumptions, S+N will deviate about its true (average) value as 1/Sqrt(S+N) x t), approaching its true value slowly as 1/Sqrt(t).

Note that the alternative of changing aperture increases the SNR as 1 / A^2 . So a change of aperture by 1 stop doubles the SNR.

Hope this helps.

Thanks. I'll need to chew on it a bit more before I can better digest it, but I do appreciate the additional information. Always good to keep an open mind about learning.

--Ken

 

[Ed Erkes]

I've been thinking about this conversation a bit and one thing I think is happening is that ISO is treated as if it's an exposure control when it's not. I think there's an idea that ISO, Shutter Speed, and F/Stop are all equals in the exposure triangle and the truth is, they're not.

That's also why I say that, unless you're at base ISO (or at the ISO just before the second gain circuit kicks in), there's nothing to be gained by deliberately overexposing your shot. Just shoot a proper level of brightness and it's all good

I agree with you 100% about the first point you make above. But I really don't understand the last point at all.

What do you mean by "proper level of brightness"? When I'm capturing an image, my goal is generally to capture as much data, or "signal", as I can. I determine proper exposure during post-processing when I adjust highlights, midtones, and shadows to get the "look" I want for the image. If your definition of proper exposure leaves a good bit of room on the right of the in- camera histogram (whether at base or higher ISOs), then you are not capturing as much signal as you could.

If I try to the place the data as far to the right of histogram, using the zebra stripes to prevent clipping of highlights, how is that not not ETTR? My definition of proper exposure is ETTR. My definition of overexposure is the clipping of important highlights.

Now if I am photographing action against changing light levels and background tonalities, then I may hedge my exposure to the left to avoid clipping highlights. And to counter another argument for ETTR, I don't recall ever noticing a problem with poor color fidelity during post-processing of ETTR images.

 

[Replytoken]

I agree with you 100% about the first point you make above. But I really don't understand the last point at all.

What do you mean by "proper level of brightness"? When I'm capturing an image, my goal is generally to capture as much data, or "signal", as I can. I determine proper exposure during post-processing when I adjust highlights, midtones, and shadows to get the "look" I want for the image. If your definition of proper exposure leaves a good bit of room on the right of the in- camera histogram (whether at base or higher ISOs), then you are not capturing as much signal as you could.

If I try to the place the data as far to the right of histogram, using the zebra stripes to prevent clipping of highlights, how is that not not ETTR? My definition of proper exposure is ETTR. My definition of overexposure is the clipping of important highlights.

Now if I am photographing action against changing light levels and background tonalities, then I may hedge my exposure to the left to avoid clipping highlights. And to counter another argument for ETTR, I don't recall ever noticing a problem with poor color fidelity during post-processing of ETTR images.

Its only ETTR if you are adjusting the exposure at base ISO because changing ISO does not add any data to the sensors, it only amplifies what was captured. A proper level of brightness can be obtained with any of the three controls, but you are still only working with what was captured by the aperture and shutter settings (and the latter is even in question form the posts above). People shoot for proper exposure for a variety of reasons rather than ETTR, especially of they do not plan on doing post processing.

--Ken

 

[John Woodworth]

Its only ETTR if you are adjusting the exposure at base ISO because changing ISO does not add any data to the sensors, it only amplifies what was captured. A proper level of brightness can be obtained with any of the three controls, but you are still only working with what was captured by the aperture and shutter settings (and the latter is even in question form the posts above). People shoot for proper exposure for a variety of reasons rather than ETTR, especially of they do not plan on doing post processing.

--Ken

Just to clear, increasing shutter speed does increase image quality, but not in the same way, and with the same efficiency, as widening aperture.