Expose Left, or Expose Right (or neither?)

[Scott Stephen]

I see some people advocating that you deliberately either over-expose or under-expose shots (in RAW), so that you can go back and either recover more shadow detail or cheat the sensor out of an extra stop of Noise Reduction, or to get more dynamic range, or some reason. Does anyone favor shooting either over or under-exposed on purpose? In what situations if any?

Also, I shoot Canon, but do I understand correctly that on Nikon, the light/exposure meter is reversed? (Left is bright and right is dark?)

[RustBeltRaw]

I'm a bit of a stickler for detail, so I tend to expose left (under-expose). My reasoning is that PP can recover surprising amounts of shadow detail, but highlight recovery is at least trickier, and if a section's completely blown-out, impossible. Usually I shoot around -2 to -2/3 stop by my camera's meter. If I'm shooting a non-manual mode (Tv or Av), I'll usually set the camera for about the same amount of under-exposure via exposure compensation.

But it's worth mentioning that "correct" exposure doesn't have much to do with what the camera says, so my tendency to expose left applies only to the object(s) at the focal point. Generally, that's where you want to maintain detail, which may mean blowing out some areas or leaving others dark. The camera's metering system doesn't know what you're shooting, unless you're using center-spot metering and your subject's in the middle of the frame.

[DanK]

There are really two separate questions here. One is over and underexposure. The other is whether, without over or underexposing, one should expose to the right or left, particularly if the histogram is fairly narrow. I think the answer is the conventional ETTR. The reason for this is that sensor noise is constant, but signal is not. Therefore, if you expose to the left, the S/N ratio is lower. You will generally get less noising pulling a bright image down in post, compared to pulling a dim image up.

[bambleweeney]

Depends, on landscape I tend to expose to the right. Spot metering on the brightest part and overexposing 2 1/3 to 2 2/3 stops. I then adjust back in ACR and continue processing 'normally'. The difference in shadow detail is marked and it's something I first noticed this whilst calibrating my spot meter to my incident light meter a while back. Otherwise I use my incident light meter and do whatever that tells me.
I shoot Nikon and I've never particularly noticed any reversal ? I think you can (un)reverse it in the menu system anyway.

[Dave Humphries]

I just expose so nothing in my composition is clipped on the (admittedly 8 bit jpg derived) histogram or blinking - unless a specular reflection of the sun in something like polished chrome, copper or brass.

I probably should push it more to the right for better noise ratio, because I shoot RAW and so I have at least a couple of stops highlight recovery in hand, but I hate losing highlight detail in clipped channels.

On a Nikon, the histogram isn't the other way round, but the meter is + on left. Since I use the histogram ETTR still works for me as a concept.

Cheers,

[Donald]

If you have not already done so, I would strongly recommend reading this CiC tutorial.

[RustBeltRaw]

I suspect most of these gents are on better cameras than I am. The 60D's high-ISO performance is pretty poor, so it's crucial for me to maintain the lowest ISO possible. Anything over 640 makes me nervous, so in many situations, underexposure is almost inevitable, and something I've learned to work with. That's probably why my approach is a little different than most.

[John Morton]

I definitely tend to expose toward dark: nothing more annoying than a blown highlight. Blocked in darks are fine with me because dark areas are suppose to be dark; but detail in highlights is a 'non negotiable' item for me (except specular highlights such as sun on chrome, for instance).

[Inkanyezi]

I think it is important to differentiate between "overexposure" and "plus compensation", as they are entirely different concepts.

When taking a reading with spot metering from a highlight, it is essential to adjust exposure to the value of that highlight in the intended image. If you want that particular highlight to be on the verge of blowing out, but still showing at least its right colour, you must adjust compensation toward plus by an amount that you have found out by experiment with your system. Mostly this will be somewhere between two and three stops to plus. So when you set your metering to compensate +2 and spot meter a highlight, you do not over-expose, but you expose for the highlight. In fact you do the ETTR raindance in a simplified way.

The zone system is alive and kicking, although it needs a bit of adjustment in the digital age. Instead of exposing for the shadows, we expose for the highlights - i.e. for the hightlights not to be blown out. Then just as in the film era, the other tones fall wherever they belong in relation to that first highlight reading. The dynamic range will in some way be governed by the ISO setting, although it is unwise to always try to use all your available dynamic range. Sometimes a low ISO setting makes sense also for a low contrast subject, because it is better to expand the tonal range, raise the contrast, from something on the right side than the middle or left. That is because there is more data there. But often, when a high dynamic range is not necessary, ISO may be increased to gain other benefits, as shorter shutter time or a smaller diaphragm.

So, at least when saving RAW and converting afterwards, there's no sense in keeping the histogram in the middle. All the way to the right is where it should go, regardless of which ISO you have chosen, but it should never climb the right wall, not even the least little bit, unless you want some highlight blown out.

My approach is mostly using a real time histogram and push it as much to the right that I dare. I carefully decide whether to accept any blown highlight area, and act accordingly. On occasions, I use spot metering, with plus compensation, to expose for the highlight, but mostly the real time histogram tells the story well enough.

And DSLR:s don't have a real time histogram, so the spot meter approach makes sense, unless you have the time to chimp and have another shot at it. There are also other approaches, and your mileage may vary. But essentially, the histogram should always end just down in the right corner, never climbing up the wall, and not ending far before the right side. The left side can be left alone as long as your dynamic range is sufficient. When you need more dynamic range, lower ISO is the answer, still keeping exposure to the right.

[John Morton]

I definitely use exposure compensation quite a bit, but more often than not it is a "-" value to move my histogram to the left and toward a darker image. Probably, the camera one uses has a lot to do with their best exposure strategy; I'm shooting with a Nikon D700 so it does hold shadow detail very well without an abundance of noise; but no camera deals with blown highlights well.

And blown highlights can sneak into any image; I am not at all convinced that they can be completely dealt with using a camera's histogram and spot metering (not at longer distances, anyway). If you are shooting anything like landscapes with white flowers in sun, or where there is snow in the scene, trying to shoot to the right of the histogram toward highlights is just asking for some highlights to be blown. There is always something in the image a little brighter than what you thought was the maximum white value...

[Inkanyezi]

I didn't mention colour saturation, which is the trickiest part of exposure. Often, particularly with red colours, but also with some bluish tints, one or two colour channels may reach saturation without the highlight warning kicking in.

It is a matter of experience what to do about it. Depending on camera, you may have an overhead of maybe one stop, but for some cameras more, in the RAW file versus the jpeg from which the histogram is derived. Then you must try it out just how much you really may have, there is no other way to know it than experience. When shooting jpeg only, there is no other way than decreasing exposure. For a red flower sometimes as much as two stops less exposure is needed to maintain structure over the petals, and some violet flowers, as rhododendron or clematis may saturate the blue channel too. Yellow flowers also may saturate red, as there is no yellow in the sensor, but it mixes red and green. Green is a colour I never saw saturated like that.

White is seldom a problem, although small white spots, as flowers a bit distant, often cannot be properly spotmetered, so it may be necessary to err to the safe side.

A live histogram, which might be available in live view mode, often will show whether there is a risk of burning a highlight or a colour channel. A live histogram often can replace spot metering, but as anything else, there's a learning curve. I never learned to use the live histogram in my Panasonic camera, although both my compact camera (Canon G7) with CHDK and my Olympus OM-D clearly show when there is a risk of blowing highlights. The drawback is that the histogram obstructs part of the viewing area, so I use it only to determine exposure and then turn it off for subsequent shots, keeping the same compensation (often negative) with evaluative metering.

Subjects with large bright areas can be done with spot metering and plus compensation without any problem with most cameras, although small highlights as flowers are mostly smaller than the meter spot, so other approaches are reasonable.

[global18]

Thanks to Scott for the question and to all the responders for a fine coverage of the topic !

[Glenn NK]

Expose as far to the right as possible without losing significant highlights.

Significant: having importance, being of consequence.

For example, shooting over water virtually guarantees there will be specular highlights. It is very unlikely that no matter how "dark" one exposes, these highlights will be blown. So for practical purposes, they become non significant (why fight what you can't win?).

I rather like this article:

http://www.luminous-landscape.com/tu...se-right.shtml

For some reason, many prefer the look of an underexposed image on the camera's LCD. This is the worst possible reason for exposing darker, as the LCD can be quite useless for judging exposure.

For an example, refer to post number 17 by GUI (Guillermo Luijk):

http://photography-on-the.net/forum/...d.php?t=411402

I can't relate to other cameras, but on the POTN forum (Canon), it's been demonstrated that if one sets the in camera contrast to minus 2 or minus 3, the camera histogram will come closer to the real histogram than without this setting. I've been using this method from the beginning using digital. Some even suggest a contrast setting of minus 4.

There are very few highlights I cannot recover in Lightroom - the most notable being specular highlights.

Glenn

[ajohnw]

I think this can vary between one make of camera and another, Looking at reviews on dpreview for instance they usually show that canon tend to extend into the high light area and nikon into the dark end of things. Later graphs usually show things which confuse the issue. Probably a numbers game really. I prefer Canons approach where highlights are less likely to be lost as detail in dark areas can always be bought out. If the highlights are over exposed they are gone for good. On the other hand many under expose on purpose to allow shots to augmented later with software. This effectively means that the shot isn't making full use of the dynamic range of the camera., It works because there will invariably some light caught in pixels at the dark end of things that can be bought out but obviously it's better if light levels are such that there are no really black areas in the shot. The best option really depends on what's in the shot and the end result that's wanted. A nice sky with clouds for instance might be exposed perfectly and use up all of the cameras hight light capture dynamic range. It could equally well be under exposed to allow software to enhance it's appearance - the dynamic range is enhance with software and the under exposure leaves room to do that without clipping the highlights or only marginally so if that isn't a problem.

John

[Glenn NK]

The references I quoted make it pretty clear that the answer to Scott's original question is to ETTR. It DOES NOT depend on the scene/image being captured.

Quoting directly from the LL article:

Let's assume for the purposes of illustration that a digital SLR has a dynamic range of 5 stops (it's usually closer to 6 stops, but let's not quibble). When working in RAW mode, which you should be, most cameras record a 12 bit image. (Yes, we say it's in 16 bit mode, but the reality is that it's only recording 12 bits in a 16 bit space. Better than 8, but not as good as a real 16 bits would be).

A 12 bit image is capable of recording 4,096 (2^12) discrete tonal values. One would think that therefore each F/Stop of the 5 stop range would be able to record some 850 (4096 / 5) of these steps. But, alas, this is not the case. The way that it really works is that the first (brightest) stop's worth of data contains 2048 of these steps — fully half of those available.

Why? Because CCD and CMOS chips are linear devices. And, of course, each F/Stop records half of the light of the previous one, and therefore half the remaining data space available.


These three paragraphs are the crux of the matter.

Take another look at post No.17. I have had conversations with GUI, and he is extremely knowledgeable about digital photography; perhaps the most knowledgeable I've come across:

http://photography-on-the.net/forum/...d.php?t=411402


Glenn

[Scott Stephen]

The references I quoted make it pretty clear that the answer to Scott's original question is to ETTR. It DOES NOT depend on the scene/image being captured.

Quoting directly from the LL article:

Let's assume for the purposes of illustration that a digital SLR has a dynamic range of 5 stops (it's usually closer to 6 stops, but let's not quibble). When working in RAW mode, which you should be, most cameras record a 12 bit image. (Yes, we say it's in 16 bit mode, but the reality is that it's only recording 12 bits in a 16 bit space. Better than 8, but not as good as a real 16 bits would be).

A 12 bit image is capable of recording 4,096 (2^12) discrete tonal values. One would think that therefore each F/Stop of the 5 stop range would be able to record some 850 (4096 / 5) of these steps. But, alas, this is not the case. The way that it really works is that the first (brightest) stop's worth of data contains 2048 of these steps — fully half of those available.

Why? Because CCD and CMOS chips are linear devices. And, of course, each F/Stop records half of the light of the previous one, and therefore half the remaining data space available.


These three paragraphs are the crux of the matter.

Take another look at post No.17. I have had conversations with GUI, and he is extremely knowledgeable about digital photography; perhaps the most knowledgeable I've come across:

http://photography-on-the.net/forum/...d.php?t=411402


Glenn

Glenn,

That is interesting information. I had not come across the information before about the brightest stop containing so disproportionately large a share of the total tonal value data space. It is a compelling case.

To play the devil's advocate, though: What is your take on the view that that ETTL is "safer" in that ETTR has a thinner margin of error and is therefore easier to screw up? Black shadows being less aesthetically objectionable than blown highlights in many cases?

[Glenn NK]

Scott:

If one is familiar with the camera, and uses some correction on the in-camera contrast setting, then ETTR does not result in blown highlights.

What we can so easily forget is that the LCD histogram is projecting a JPEG that the camera produces for the LCD - even if we only shoot RAW (as I do). And this JPEG is not a good representation at all of the (developed) RAW file. On both my Canon bodies, I have the contrast setting at minus 3. I rarely have highlights that I cannot recover, so there is no danger at all. Admittedly, specular highlights will be blown, but show me any method of exposure that does not blow these out. Even the human eye looking at the reflections of sunlight off wave tops cannot discern any detail, and if our eyes can't, what then can?

By using ETTR, I am minimizing the black areas and maximizing the detail capture. This should apply to all bodies.

I strongly suspect that many of us are affected by the look of the LCD image (mislead?); if one is shooting only JPEG, then what you see is what you get, but this is not true with RAW images.

On occasion, I have an image onscreen in Lightroom that looks washed out, but if the histogram isn't past the right edge, the image is not washed out at all, and the rich colours can be retrieved.

Looks can be very deceiving, and a dangerous trap.

Glenn

[Carl in Louisiana]

Scott:

If one is familiar with the camera, and uses some correction on the in-camera contrast setting, then ETTR does not result in blown highlights.

What we can so easily forget is that the LCD histogram is projecting a JPEG that the camera produces for the LCD - even if we only shoot RAW (as I do). And this JPEG is not a good representation at all of the (developed) RAW file. On both my Canon bodies, I have the contrast setting at minus 3. I rarely have highlights that I cannot recover, so there is no danger at all. Admittedly, specular highlights will be blown, but show me any method of exposure that does not blow these out. Even the human eye looking at the reflections of sunlight off wave tops cannot discern any detail, and if our eyes can't, what then can?

By using ETTR, I am minimizing the black areas and maximizing the detail capture. This should apply to all bodies.

I strongly suspect that many of us are affected by the look of the LCD image (mislead?); if one is shooting only JPEG, then what you see is what you get, but this is not true with RAW images.

On occasion, I have an image onscreen in Lightroom that looks washed out, but if the histogram isn't past the right edge, the image is not washed out at all, and the rich colours can be retrieved.

Looks can be very deceiving, and a dangerous trap.

Glenn

Hey Glenn,
I am interested also, I have read the info you have posted and willing to give it a try. I have one question thus far. "Is the -3 in contrast setting good for just landscape settings or can it be used for other types of photography as well?" eg. Portraiture, sports, well you get the picture...every thing.

[Glenn NK]

Hey Glenn,
I am interested also, I have read the info you have posted and willing to give it a try. I have one question thus far. "Is the -3 in contrast setting good for just landscape settings or can it be used for other types of photography as well?" eg. Portraiture, sports, well you get the picture...every thing.

Carl:

I'm late - on the west coast.

I haven't changed my contrast setting on my 30D for four or five years and I shoot everything with it. Same for my three year old 5DII.

Three things I do watch carefully:

1) because there can be so much variation in dynamic range in one image, the type of metering set in camera can be important. For extreme closeup/macro I often use spot metering for example.

2) particularly the camera histogram,

3) and of course I check the histogram in Lightroom when processing to be sure I have the right settings.

Occasionally I get an image that "leaks out" the right side of the histogram, but most of the time I can recover it in PP.

Glenn

[Carl in Louisiana]

Carl:

I'm late - on the west coast.

I haven't changed my contrast setting on my 30D for four or five years and I shoot everything with it. Same for my three year old 5DII.

Two things I do watch carefully:

1) because there can be so much variation in dynamic range in one image, the type of metering set in camera can be important. For extreme closeup/macro I often use spot metering for example.

2) particularly the camera histogram,

3) and of course I check the histogram in Lightroom when processing to be sure I have the right settings.

Occasionally I get an image that "leaks out" the right side of the histogram, but most of the time I can recover it in PP.

Glenn

Thanks Glenn,

I have set my 60D up at -3 on contrast and will use and compare to what I already have and go from there. Thanks for sharing.