Expose To The Right - on LuLa

[Scrambler]

To take that further, 1/2 the capacity to record graduations of light is in the brightest stop, 3/4 in the 2 brightest, 7/8 in the 3 brightest and so on. This is true regardless of the bit depth.

 

[Peter_Jones]

Personally, I have tried ETTR, and for me I reckon it's not really needed on my D700 (not bad for an older camera), there IS a difference, but not drastic enough for me - however I find correct white balance can have much more effective results in the final image. I found this out accidentally when meddling with ETTR - until then, I thought Auto WB was close enough. (It's not)

 

[news shooter]

Early on, this what I learned

Early on, this what I learned

Hello all.
Early on with a nikon d1, was told by an AP shooter to expose for the highlights like shooting e-6. That seemed to work in most situations. Fast forward to today. Now shooting with a Canon 1dmk2, I shoot under about -1/3-2/3 for highlight details. Also shooting with a canon 5d3. With that setup exposure is about+ 1/3 to +1.
Quite a difference but, generally works for the 2 cams
I also agree about white balancing rather than awb.
Just 1-2/3 cents.

 

[crsantin]

I've been slightly underexposing for the past year and I find it gives me better skies, clouds, tree and buildings etc, basically anything not on ground level. Shadows are not an issue for most modern digital cameras, and I tend to like shadows in my photographs. To me shadows are not something that need to be removed or lifted from a photograph.

 

[Sam N]

The problem with this method is that it seems to lower the color quality in the mid tones on modern sensors. Exposing correctly is the best way to obtain the best tonality and color quality on modern sensors.

Do you have any examples of this Edward? I can see how there could be problems in highlight colors if individual channels clip, but I don't understand how this could screw up mid-tones.

 

[peterm1]

I do not know that it is exactly correct to say that ETTR gives best dynamic range. I have always understood that it lowers noise and increases overall data captured however, which is slightly different. The problem of course as others have said is that when applied, ETTR almost invariably leaves you with blown highlights if the image you are shooting has a large range of tonal values. I have always taken the ETTR "rule" with a significant grain of salt when it comes to shooting digital because of this problem of blowing highlights. In practice its damn near impossible to avoid when shooting to the right.

Most of my shooting is done in the real world - outside in the sun where I have no control over range of tonal values in the scene being photographed. Invariably there are some very bright highlights. I rather suspect that ETTR works best in studio settings where the photographer has more or less complete control over the lighting and hence can avoid blowing highlights by avoiding having excessive highlights in the scene. In that situation you can expose to the right till your heart's content and not run the risk of having unrecoverable white blobs of blown highlight in your shot. Not so when shooting streets, or landscape scenery however, where you must take what you get and this includes large expanse of sky etc.

So I deliberately set the exposure compensation to shoot at least a third of a stop below what the camera has metered which is not a bad compromise although even this does not avoid all blown highlights. I then use highlight recovery tools in post to recover lost highlight data that still occurs.

This seems to work pretty well particularly when shooting with newer sensors and when shooting RAW images where there is a lot more information to work with. I often shoot with a Nikon D700 and find I can shoot it at up to and including ASA 1600 without excessive noise in the image. More modern sensors are even better - some of them much better. This means I can shoot in RAW without worrying too much about excessive loss of data and can treat the ETTR rule as being nothing more than a rough rule of thumb that has to be applied selectively.

 

[Ranchu]

ETTR is bunk, like the rest of that site. The purpose is not to get better pictures, it's to sell you software and books. If people just use the evaluative metering with a general compensation they like, they'll be much better off. It's shill site for adobe. Always was.

 

[tstermitz]

ETTR means expose as brightly as possible WITHOUT blowing out the highlights. So to be fair to the people who missed that definition, their comments are really about the difficulty of living close to the edge, rather than the "optimal exposure" benefits of doing so.

The specific benefits claimed by ETTR is to put as much of your image information above the noise floor, and to have sufficient bits-per-color-value to make smooth color adjustments possible.

I don't understand the comment by EdwardKaraa about lowering the "color quality in the mid tones on modern sensors". I have never seen that idea before, so I'd appreciate a link. As far as I know, the color profile of the sensor is created in software and applied to the raw photon counts on the pixels. So maybe he is talking about sensor profiles he doesn't like, or the difficulty of using Lightroom to achieve the color tones he is looking for.

I also don't understand his comment about CCD cameras being "ISO-less". I've never heard that before, and it seems to contradict physics, if not mathematics. Again, a reference, would be helpful.

Modern sensors do indeed have lots of dynamic range, which makes it easy not to worry about ETTR. On the other hand, my M9 doesn't have a modern sensor, so I have to be more careful about noise.

 

[tstermitz]

I do worry about blowing the highlights when I'm shooting in cities at twilight. Bright lights when the rest of the scene is fairly dark blow out pretty quickly, and it is pretty hard to get a gentle falloff. This is one area where film may still have an edge over digital.

To help with that, I'll underexpose the main part of the scene, and add exposure in Lightroom. While I know that this leads to noise in the shadows, in this type of shooting, I often let the shadows go to black anyway.

 

[willie_901]

Many CCD cameras are isoless.
...

Name a single ISO-less CCD camera please.

 

[EdwardKaraa]

Do you have any examples of this Edward? I can see how there could be problems in highlight colors if individual channels clip, but I don't understand how this could screw up mid-tones.

I've done a lot of experimenting with the M9 and Sony A900 in the past, but unfortunately didn't keep the files. Can't tell you exactly why because I'm not familiar with the physics involved but I found that when I pull down the overexposed parts of the image (but not blown) to the middle range, there is a change in the color balance, and the tonal gradations are rougher. Some colors, especially reds may get some strange hues in some cases. The differences I'm talking about are subtle, not dramatic, but if you're shooting ETTR to maximize iq, you should realize some other aspects of iq are affected by this. As I mentioned, the best tonal gradations, at least with the A900, M9 and M(240) are obtained by exposing correctly or slightly under.

 

[EdwardKaraa]

Name a single ISO-less CCD camera please.

Leica M9

(This is added to make the text longer as requested)

 

[EdwardKaraa]

ETTR means expose as brightly as possible WITHOUT blowing out the highlights. So to be fair to the people who missed that definition, their comments are really about the difficulty of living close to the edge, rather than the "optimal exposure" benefits of doing so.

The specific benefits claimed by ETTR is to put as much of your image information above the noise floor, and to have sufficient bits-per-color-value to make smooth color adjustments possible.

I don't understand the comment by EdwardKaraa about lowering the "color quality in the mid tones on modern sensors". I have never seen that idea before, so I'd appreciate a link. As far as I know, the color profile of the sensor is created in software and applied to the raw photon counts on the pixels. So maybe he is talking about sensor profiles he doesn't like, or the difficulty of using Lightroom to achieve the color tones he is looking for.

I also don't understand his comment about CCD cameras being "ISO-less". I've never heard that before, and it seems to contradict physics, if not mathematics. Again, a reference, would be helpful.

Modern sensors do indeed have lots of dynamic range, which makes it easy not to worry about ETTR. On the other hand, my M9 doesn't have a modern sensor, so I have to be more careful about noise.

There are several ways of boosting iso. Most CCD cameras, including MFDB and the M9 boost iso in software, not by amplifying the signal like regular CMOS based cameras. The camera basically shoots always at base iso. With the M9 for instance, you can shoot at iso 160 and expose for iso 1250, then lighten the shot by 3 stops in LR. You will get same, or actually better results than shooting at iso 160 because the raw converters are always improving.

 

[willie_901]

I recently discussed this topic here on RFF.

Fact - Maximizing the analog signal-to-noise ratio at the sensor when the shutter is open provides the best technical IQ. Dynamic range depends directly on the analog SNR. There's several ways to degrade SNR and DR after the shutter closes. But, with a single exposure there is no way to increase the SNR and DR after the shutter closes. In other words the datas' maximum possible information content is determined by using the optimal shutter time and aperture parameters.

Fact - Increasing ISO above base ISO decreases SNR and DR if you do what the camera's light meter tells you to do. Increasing ISO decreases exposure (total light energy recorded by the sensor when the shutter is open) because you decrease shutter time and, or apeture. Less exposure means a lower SNR. Under exposing and then pushing during post production is analogous to reducing sensor size with optimum exposure. In both cases less total light is recorded for a given combination of shutter time and aperture.

Fact - Different brands use different data stream technologies. The technique required to obtain the best possible shadow-region detail depends on what camera you own. Often a compromise has to be made between DR and shadow-region detail.

Fact - For raw files, there is no disadvantage to letting highlights clip when those highlights are not essential to the purpose of the photograph. Highlights can be lost two ways. You can loose them at base ISO by using an exposure that exceeds the sensor's analog, full-well capacity. Or you can clip them above base ISO even though the sensor is typically used exposed far below its full-well capacity. The analog signal is amplified by according to the ISO setting. With excess amplification the signal level exceeds the maximum DC input voltage specification of the ADC. When this happens you have to under expose the sensor even more (or lower ISO). In both cases (exceeding sensor or ADC input specs) highlight region information is lost. But the SNR is always lower in the second case.

Fact - Years ago the original concept of ETTR was based on a misunderstanding about modern analog to digital conversion technologies. Back then ETTR was the right advice for the wrong reason. Maximizing exposure happens to produce a histogram that resembles the histogram achieved by following ETTR. It was pure coincidence that ETTR produced superior IQ.

Fact - Maximizing exposure only considers the effects of read noise. The uncertainty in the light amplitude itself (shot or photon noise) can not be altered. A perfectly exposed sky at base ISO will appear noisy in an extreme crop due to the photon noise.

Fact - Content trumps IQ. Spontaneity demands that maximizing exposure is often impractical. At the same time naive gratuitous use of high ISO degrades IQ because SNR and DR are thrown away.

 

[willie_901]

Leica M9

(This is added to make the text longer as requested)

Your statement is not supported by data published by Jim Kasson in a LuLa Forum post. Please see the last graph at the end of Kasson's post. The M9/M8 data stream is ISO-less below ISO 640. Above ISO 640 (not an extravagant ISO level these days) the signal-to-noise ratio decreases significantly. The M8/M9 data stream design clearly uses two different amplification stages.

By contrast Jim Kasson's data from Nikon CMOS sensors is ISOless over a much larger ISO range.

 

[Ranchu]

I don't understand the comment by EdwardKaraa about lowering the "color quality in the mid tones on modern sensors". I have never seen that idea before, so I'd appreciate a link. As far as I know, the color profile of the sensor is created in software and applied to the raw photon counts on the pixels. So maybe he is talking about sensor profiles he doesn't like, or the difficulty of using Lightroom to achieve the color tones he is looking for.

You should try it yourself and compare, to see which you like better. You're asking someone to give you an example that proves something based on how it looks. If you need a reason that makes sense in order to try it, here - when you ettr, you are by definition underexposing the midtones which will degrade the color quality. If you expose properly at the time of exposure and blow the highlights you're going to blow anyway, you aren't pulling the colors up you're exposing for them.

edit: I've just read EdwardKaraa's response, that you are both talking about OVER exposing scenes that are WITHIN the range of a digital sensor. I can't remember ever experiencing such a scene, but I guess it must have happened. Which makes ettr (or HAMSTER, which I prefer!) even more pointless. Still, you should check for yourself.

🙂

 

[EdwardKaraa]

You should try it yourself and compare, to see which you like better. You're asking someone to give you an example that proves something based on how it looks. If you need a reason that makes sense in order to try it, here - when you ettr, you are by definition underexposing the midtones which will degrade the color quality. If you expose properly at the time of exposure and blow the highlights you're going to blow anyway, you aren't pulling the colors up you're exposing for them.

edit: I've just read EdwardKaraa's response, that you are both talking about OVER exposing scenes that are WITHIN the range of a digital sensor. I can't remember ever experiencing such a scene, but I guess it must have happened. Which makes ettr (or HAMSTER, which I prefer!) even more pointless. Still, you should check for yourself.

🙂

Indeed, I am talking about some loss in the tonal range when darkening the shot in ACR. I do not notice this loss when lightening the shot. In fact the colors look better when the shot is slightly underexposed and lifted in post.

 

[Pherdinand]

The analog signal is amplified by according to the ISO setting. With excess amplification the signal level exceeds the maximum DC input voltage specification of the ADC. When this happens you have to under expose the sensor even more (or lower ISO).

?? you surely mean something else than what you wrote?

 

[moreorless]

I would agree with him that "shooting to avoid blinkes" often isn't the best method but I wouldn't say most of the examples in that article are really very challenging lighting.

I think it helps as well if your not trying to lift your shadows too much, I see this a lot in landscape shooting, rather than brightening shadows just enough to get some detail into the print people take it way too far and try and make all the detail clearly visable which also shows up more noise(or lack of detail after its been removed).

 

[willie_901]

?? you surely mean something else than what you wrote?

I meant exactly what I wrote.

Exposure and image brightness are not equivalent unless ISO is set to base ISO. Gratuitously increasing brightness by using a high ISO setting will reduce the signal-to-noise ratio and might reduce dynamic range. It can also cause clipping. If clipping occurs above base ISO, the first thing to do is reduce ISO.

Details follow.

The sensor operates at a single amplification level. For still cameras the sensor's maximum analog DC voltage output is fixed. The sensor always operates at this base ISO (typically 100 or 200).

There are only three possible exposure states for each pixel (sensor well) the instant after the shutter closes.

1/ maximum exposure: The DC voltage is just below the maximum possible voltage without exceeding the pixel’s full-well capacity. When this happens the signal-to-noise ratio and dynamic range are for that pixel are optimal.

2/ under exposure: The DC voltage is significantly below the maximum voltage threshold.

3/ overexposure: The pixel’s full-well capacity is exceeded. This is often called clipping, but technically it is not clipping. It is analog signal overload.
The shutter speed and aperture are the only parameters that determine which exposure state each pixel was in when the shutter closed. Of course all three states can exist simultaneously in different pixels.

After the shutter closes the analog DC voltages may, or may not, be increased using analog amplification. When the ISO parameter is set above base ISO, the signals are amplified, ISO parameter determines the amplification level. The amplification does not increase exposure. Instead it increases brightness after the shutter closes.

Suppose I am in a pub where the average light level is about 4 EV. I set ISO to my camera’s base ISO The required the lens aperture is 2.8. The camera is on a tripod. The camera light meter estimates the shutter time should be ¼ sec. I press the shutter button and the DC voltages from the sensor range from just below the maximum of 5 volts to ½ volt. Not a single pixel is overexposed. Unfortunately people are moving around and the result is unacceptable. The shutter time must be decreased.

Now I increase ISO to 1600. The meter changes its shutter time to 1/30. This photograph reduces motion blurring to an acceptable level.
But the sensor was underexposed by 4 stops because the shutter time decreased. Less light reached the sensor. Every pixel is underexposed (state 2 above). Since ISO was increased by a a favor of 4, the signals’ DC voltages were amplified 4 fold. Note the voltage level inputs to the analog-to-digital converter are identical to the ¼ sec. base ISO exposure. However the signal-to-noise ratio and dynamic range are well-below the sensor’s potential. This means the shadow regions will have less detail compared to the ¼ sec. exposure.

Shutter speed and aperture determine exposure of the sensor. The ISO amplification determines the brightness after the shutter closes.
Suppose I wonder how well auto-ISO works on my camera. The camera decides ISO should be 3200. I press the shutter. Again, the sensor is underexposed by 4 stops. But now the pixels’ DC output is amplified by a factor of 8. This means the DC signals from the brightest regions exceeds the analog-to-digital converters maximum input voltage threshold. Now those regions will clip. They don’t clip because the sensor was over exposed. They clip because the ADC’s maximum voltage threshold was exceeded due excess signal amplification.