Is it better to over or to under-expose

[Christof21]

Underexposing* will always result in a noisier image.

Overexposing* will sometimes result in unrecoverably blown highlights.

So, I'd suggest that if you can't meter for the highlights, but here are highlights in the scene whose detail you really want to preserve, underexpose. Otherwise, overexpose.

(*as you are using the terms)

Strictly speaking , overexposure always result in unrecoverably blown highlights.

I suppose this is why you added the asterix.

 

[Mark LW]

With the D3 I have been using for more than a decade I shoot RAW and set the camera to -0.3EV. Which gives a fuller saturation of colour.
Should I feel the need, the exposure can be modified in RAW Post Processing.
Judge for yourself from some of my images at https://www.flickr.com/photos/47885101@N07/sets/

--
All comments are purely personal and generally based on my experience.
What I state is an opinion; I may well be wrong.

 

[FingerPainter]

Underexposing* will always result in a noisier image.

Overexposing* will sometimes result in unrecoverably blown highlights.

So, I'd suggest that if you can't meter for the highlights, but here are highlights in the scene whose detail you really want to preserve, underexpose. Otherwise, overexpose.

(*as you are using the terms)

Strictly speaking , overexposure always result in unrecoverably blown highlights.

I suppose this is why you added the asterix.

For a reason close to that, I suppose. Unlike the term "exposure", for which there is a defined meaning, I don't think there is one specific meaning for "correct exposure", "overexposed" or "underexposed". If you define "overexposed" to mean "exposed so much that desired highlight detail is blown", then what you say is true. But I took OP to be using "over-expose" and "under-expose" to mean using more or less exposure than what will give desired lightness SOOC. In that case, it is entirely possible to overexpose without blowing desired highlight detail. Another possible meaning of "over-expose" and "under-expose" is to use more or less exposure than what ill give one desired DOF and motion blur or desired lens sharpness. In that case, it is also possible to overexpose without blowing desired highlight detail.

In general, "over-expose" means to expose more than correct exposure, and under-expose means to expose less than correct exposure, but what constitutes correct exposure depends on context and intent.

Since it is often entirely possible to overexpose without blowing desired highlight detail (unless we define "over-exposure" in terms of blown highlight detail), I consider an approach of always underexposing to be sub-optimal.

 

[Iliah Borg]

I don't think there is one specific meaning for "correct exposure"

Exposure is correct when all important elements of the scene are exposed in such a way that they fall on the linear part of the log-log characteristic curve, while the lightest important elements are within 1/3 of a stop from the beginning of a shoulder. That's from an old Kodak seminar

, "overexposed" or "underexposed". If you define "overexposed" to mean "exposed so much that desired highlight detail is blown", then what you say is true. But I took OP to be using "over-expose" and "under-expose" to mean using more or less exposure than what will give desired lightness SOOC. In that case, it is entirely possible to overexpose without blowing desired highlight detail. Another possible meaning of "over-expose" and "under-expose" is to use more or less exposure than what ill give one desired DOF and motion blur or desired lens sharpness. In that case, it is also possible to overexpose without blowing desired highlight detail.

In general, "over-expose" means to expose more than correct exposure, and under-expose means to expose less than correct exposure, but what constitutes correct exposure depends on context and intent.

Since it is often entirely possible to overexpose without blowing desired highlight detail (unless we define "over-exposure" in terms of blown highlight detail), I consider an approach of always underexposing to be sub-optimal.

 

[bobn2]

Underexposing* will always result in a noisier image.

Overexposing* will sometimes result in unrecoverably blown highlights.

So, I'd suggest that if you can't meter for the highlights, but here are highlights in the scene whose detail you really want to preserve, underexpose. Otherwise, overexpose.

(*as you are using the terms)

Strictly speaking , overexposure always result in unrecoverably blown highlights.

I suppose this is why you added the asterix.

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

 

[FingerPainter]

I don't think there is one specific meaning for "correct exposure"

Exposure is correct when all important elements of the scene are exposed in such a way that they fall on the linear part of the log-log characteristic curve, while the lightest important elements are within 1/3 of a stop from the beginning of a shoulder. That's from an old Kodak seminar

I suppose I should have added "in the context of digital photography, generally, as opposed to specifically RAW or SOOC JPEG shooting".

 

[Christof21]

Underexposing* will always result in a noisier image.

Overexposing* will sometimes result in unrecoverably blown highlights.

So, I'd suggest that if you can't meter for the highlights, but here are highlights in the scene whose detail you really want to preserve, underexpose. Otherwise, overexpose.

(*as you are using the terms)

Strictly speaking , overexposure always result in unrecoverably blown highlights.

I suppose this is why you added the asterix.

I'm french.... This means that first I like your reference and secondly that like all french people I am not an expert in english !!!

 

[Iliah Borg]

I don't think there is one specific meaning for "correct exposure"

Exposure is correct when all important elements of the scene are exposed in such a way that they fall on the linear part of the log-log characteristic curve, while the lightest important elements are within 1/3 of a stop from the beginning of a shoulder. That's from an old Kodak seminar

I suppose I should have added "in the context of digital photography, generally, as opposed to specifically RAW or SOOC JPEG shooting".

Here is another one: exposure is correct when all important elements of the scene are reproduced satisfactory on the selected media, displayed at the selected size, and viewed at the selected conditions.

I use something very close to the first one for my purposes, but I recognize that much depends on purposes

 

[bobn2]

Underexposing* will always result in a noisier image.

Overexposing* will sometimes result in unrecoverably blown highlights.

So, I'd suggest that if you can't meter for the highlights, but here are highlights in the scene whose detail you really want to preserve, underexpose. Otherwise, overexpose.

(*as you are using the terms)

Strictly speaking , overexposure always result in unrecoverably blown highlights.

I suppose this is why you added the asterix.

I'm french.... This means that first I like your reference and secondly that like all french people I am not an expert in english !!!

Not my experience of French people at all. I find that their knowledge of English is often very good indeed, but they certainly won't admit to it. My experience of travelling in France is that those that speak English will often get a non comprehending look, whilst those that speak French with an atrocious English accent will be answered in excellent English. - which to be honest is fair enough. It is a common English speakers' conceit that the whole world understands English as long as you say it very loudly.

BTW, the substitution of the little warrior for 'asterisk' is extremely common amongst native English speakers. Probably the majority get it wrong these days.

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

 

[Michael Fryd]

I don't think there is one specific meaning for "correct exposure"

Exposure is correct when all important elements of the scene are exposed in such a way that they fall on the linear part of the log-log characteristic curve, while the lightest important elements are within 1/3 of a stop from the beginning of a shoulder. That's from an old Kodak seminar

I suppose I should have added "in the context of digital photography, generally, as opposed to specifically RAW or SOOC JPEG shooting".

Here is another one: exposure is correct when all important elements of the scene are reproduced satisfactory on the selected media, displayed at the selected size, and viewed at the selected conditions.

I use something very close to the first one for my purposes, but I recognize that much depends on purposes

Of course, there are many other issues that can lead to poor quality reproduction. For instance a print that is too light might be the result of using the wrong printer profile, or clogs blocking ink flow.

I suspect we could contrive a situation where there were multiple issue that canceled each other out. We might get a good looking print, even though the exposure was not correct.

 

[Iliah Borg]

I don't think there is one specific meaning for "correct exposure"

Exposure is correct when all important elements of the scene are exposed in such a way that they fall on the linear part of the log-log characteristic curve, while the lightest important elements are within 1/3 of a stop from the beginning of a shoulder. That's from an old Kodak seminar

I suppose I should have added "in the context of digital photography, generally, as opposed to specifically RAW or SOOC JPEG shooting".

Here is another one: exposure is correct when all important elements of the scene are reproduced satisfactory on the selected media, displayed at the selected size, and viewed at the selected conditions.

I use something very close to the first one for my purposes, but I recognize that much depends on purposes

Of course, there are many other issues that can lead to poor quality reproduction. For instance a print that is too light might be the result of using the wrong printer profile, or clogs blocking ink flow.

I would understand if you'd brought DR issues up, but not this.

I suspect we could contrive a situation where there were multiple issue that canceled each other out. We might get a good looking print, even though the exposure was not correct.

If the print is satisfactory, exposure was correct.

 

[Michael Fryd]

I don't think there is one specific meaning for "correct exposure"

Exposure is correct when all important elements of the scene are exposed in such a way that they fall on the linear part of the log-log characteristic curve, while the lightest important elements are within 1/3 of a stop from the beginning of a shoulder. That's from an old Kodak seminar

I suppose I should have added "in the context of digital photography, generally, as opposed to specifically RAW or SOOC JPEG shooting".

Here is another one: exposure is correct when all important elements of the scene are reproduced satisfactory on the selected media, displayed at the selected size, and viewed at the selected conditions.

I use something very close to the first one for my purposes, but I recognize that much depends on purposes

Of course, there are many other issues that can lead to poor quality reproduction. For instance a print that is too light might be the result of using the wrong printer profile, or clogs blocking ink flow.

I would understand if you'd brought DR issues up, but not this.

I suspect we could contrive a situation where there were multiple issue that canceled each other out. We might get a good looking print, even though the exposure was not correct.

If the print is satisfactory, exposure was correct.

It seems like you are saying that there is a range of exposures that are "correct", as there are a range of exposures that produce satisfactory prints.

I am not disagreeing, just trying to clarify. Some people suggest that there is only one single "correct" exposure for a situation, and others are in the multiple correct exposures camp.

 

[Iliah Borg]

I don't think there is one specific meaning for "correct exposure"

Exposure is correct when all important elements of the scene are exposed in such a way that they fall on the linear part of the log-log characteristic curve, while the lightest important elements are within 1/3 of a stop from the beginning of a shoulder. That's from an old Kodak seminar

I suppose I should have added "in the context of digital photography, generally, as opposed to specifically RAW or SOOC JPEG shooting".

Here is another one: exposure is correct when all important elements of the scene are reproduced satisfactory on the selected media, displayed at the selected size, and viewed at the selected conditions.

I use something very close to the first one for my purposes, but I recognize that much depends on purposes

Of course, there are many other issues that can lead to poor quality reproduction. For instance a print that is too light might be the result of using the wrong printer profile, or clogs blocking ink flow.

I would understand if you'd brought DR issues up, but not this.

I suspect we could contrive a situation where there were multiple issue that canceled each other out. We might get a good looking print, even though the exposure was not correct.

If the print is satisfactory, exposure was correct.

It seems like you are saying that there is a range of exposures that are "correct", as there are a range of exposures that produce satisfactory prints.

That's the premise for the second definition. This definition exists, and if you are reading, I said "I use something very close to the first one for my purposes".

 

[Razor512]

The way to get the best overall image quality, is to saturate in hardware and get the overall look you want in post. While there are situations where proper saturation may not be possible, it is good to strive for where both possible and practical.

There is no benefit to undersaturating the sensor, Outside of that, there are times where clipping can happen and it is not practical to completely avoid clipping, e.g., a really bright light bulb visible in the scene, or the sun.

With that in mind, saturating as much as possible while avoiding unwanted clipping, is effectively the best balance you can get with the very limited dynamic range in current sensor technology.

 

[photonut2008]

The other "ETTR" is when you are above base ISO.

Good that you put ETTR in quotes, because, by definition, ETTR is necessarily accomplished at base ISO. If you're above base ISO, you cannot ETTR.

I don't think this is technically correct.

 

[photonut2008]

I'll say that digital film is free so bracket your heart out. The only cost is the time to sort through them.

I find that often what is at stake is capturing the moment.

 

[Michael Fryd]

The other "ETTR" is when you are above base ISO.

Good that you put ETTR in quotes, because, by definition, ETTR is necessarily accomplished at base ISO. If you're above base ISO, you cannot ETTR.

I don't think this is technically correct.

Assume you are getting a histogram that is snug against the right when shooting manually at 1/100, f/4 and ISO 400. If reduce the ISO to 200, but leave everything else the same, your histogram will shift to the left. However, as the exposure (light reaching the sensor) was the same, the noise and dynamic range in your RAW files will be the same.

This demonstrates that when you are above base ISO, there isn't a significant advantage to adjusting the ISO to push the Histogram to the right.

ETTR is really about maximizing exposure. At base ISO ETTR is simply a useful technique for maximizing exposure. Above base ISO, the histogram only informs you how the ISO relates to the exposure. This is useful when shooting JPEG, but not important when shooting RAW.

 

[photonut2008]

Maybe I should just rephrase this. If one is slightly uncertain of the best approximate exposure should one, over or under, expose.

The short answer is that by definition you would rather underexpose than overexpose. The longer answer is that when you shoot Raw that ideally you want to make a negative adjustment to the brightness of your files, and that involves understanding how your camera meters and being able to interpret your histograms accurately. For the really long answer, you will want to discuss ETTR and UniWB with knowledgeable Canon users.

 

[photonut2008]

The other "ETTR" is when you are above base ISO.

Good that you put ETTR in quotes, because, by definition, ETTR is necessarily accomplished at base ISO. If you're above base ISO, you cannot ETTR.

I don't think this is technically correct.

Assume you are getting a histogram that is snug against the right when shooting manually at 1/100, f/4 and ISO 400. If reduce the ISO to 200, but leave everything else the same, your histogram will shift to the left. However, as the exposure (light reaching the sensor) was the same, the noise and dynamic range in your RAW files will be the same.

Assume that the OP uses a Canon 5D Mark IV. See here:

http://www.photonstophotos.net/Charts/PDR.htm#Canon EOS 5D Mark IV

Bill says you get 10.36 stops of DR at ISO 200, but in the scenario you offered we take one stop off of that for a total of 9.36 stops of DR. According to Bill, at ISO 400 you get 9.74 stops of DR. This looks to me like you get nearly 40% of a stop more DR exposing to the right at ISO 400 than using the same EV at ISO 200.

This demonstrates that when you are above base ISO, there isn't a significant advantage to adjusting the ISO to push the Histogram to the right.

That depends on the camera and how far you are pushing the development.

--
Internet Interlocuter

 

[photonut2008]

If a photographer relies on camera feedback such as the JPEG-based histogram for ETTR, they are going to invariably leave at at least 2 stops of dynamic range on the table.

That hasn't been my experience. It's typically closer to one full stop and depends on the scene and the metering method.