Metering, some myths explored

D-Noc wrote in post #18105547
Exactly. But I guess it all depends on how true-to-life you want your image to be. If you use ETTR right up to the point where the clipping would occur, you should be able to pull it down by -x in post to get a natural looking image. But x might not be exact same amount as you overexposed with when you ETTR but in practise, does it matter much if the image is still LOOKING natural?

As stated earlier I don't use ETTR as such, but my 6D does seem to leave most of the histogram on the dark half if I set the exposure as suggested by the built in metering. So I overexpose by ~1/3 to 1/2 stop to get a better centered histogram. In a sense this is ETTR, but not to overexpose at all, more to get exposure right. The right amount of additional exposure will of course also be dependent on the subject.

I still prefer to get exposure as close to what I really want in camera. I think in regards to my way of handling a camera, ETTR is just an unnecessary step, as I haven't yet had an experience where I thought "Phew.. Im glad I used ETTR".

Not the histogram!!!

LOL, allow me to explain a bit.

First, regarding how true-to-life, I was gonna add a footnote to one of my posts above stating that this discussion does not really apply to 'creative/artistic licence' or experimental photography, although I'm willing to make an exception for silhouettes. I'm basing everything on the premise that a photograph is as accurate a representation of a scene as it's possible to obtain with the equipment we have available at the time. As a matter of fact, in the kind of work that I specialise in (architecture, business portraiture, products) true-to-life is paramount. But even when shooting for personal purposes, I want my photos to represent what my eyes saw.

Now, regarding your camera and its seeminly tendency to underexpose: what metering mode are you using?

Also, what kind of scenes are you shooting?

As stated above, the tonal values and their distribution will have a major impact in the exposure determination (both the one done by the camera and the one done conscientiously by you). If you're shooting a black cat in a dark street, you don't want the histogram in the centre, but pushed towards the left; conversely, if you're imaging a high-key scene you'll not want the histogram in the centre either, but pushed towards the right. Histograms are mere representations of the distribution of tonal values, but they are not to be used for exposure determinations. There's no right or wrong histogram— actually, as a matter of fact, since the histogram simply reflects the tones in the image there's no wrong histogram, but there certainly are wrong exposures.

Wilt wrote in post #18105925
...And the sugar is only to stave off hunger.

Innovative idea, though! ; D

Alveric wrote in post #18105939
Now, regarding your camera and its seeminly tendency to underexpose: what metering mode are you using?

Also, what kind of scenes are you shooting?

As stated above, the tonal values and their distribution will have a major impact in the exposure determination (both the one done by the camera and the one done conscientiously by you). If you're shooting a black cat in a dark street, you don't want the histogram in the centre, but pushed towards the left; conversely, if you're imaging a high-key scene you'll not want the histogram in the centre either, but pushed towards the right. Histograms are mere representations of the distribution of tonal values, but they are not to be used for exposure determinations. There's no right or wrong histogram— actually, as a matter of fact, since the histogram simply reflects the tones in the image there's no wrong histogram, but there certainly are wrong exposures.

I know that a centered-biased histogram might not always be desired, but for the sake of argument, let's say the image in question is a portrait taken indoor. Outside is an overcast. Light is good as you have some big windows with a lot of ambient light coming in. The subject is well lit and the background is a light grey backdrop.
The camera is set to Spot metering, metering of a Caucasian face. We'll just use a 85mm at f/2.8 and a decent shutter speed of 1/400s. Let's say that this gives a "perfect exposure" (according to the built in metering).

In some cases, This would give me a shot were the subjects face is clearly underexposed, even though the spot metering is done on the subjects well lit face. In such a case, the histogram should be center biased from my understanding, as the face is mostly mid tones to high tones and the background would be low tones to mid tones.
But, the histograms tends to be stacked in the dark half, and after a + 0.5 to +0.66 exposure adjustment everything looks well exposed and the histogram seems more spread out.

I am not saying something is wrong as such, I am merely stating that I know this happens so I usually let the camera do its metering and then, if the scene "normal" or "average", I dial in a +0.5 exposure compensation and get the result I expect. That's all

In that case, with an average caucasian face, did you up the exposure by one stop? (Average caucasian skin = ~+1 above mid grey.)

Also, remember to point your spot to the area sitting between the shadow and the highlight: that's where the mid-tone/real colour lives.

D-Noc, your histogram, in the shot metered as you described,

• puts the pixels for the face AT THE CENTER of the histogram (because you metered on the face), when in fact they belong a bit to the RIGHT of center in the 'correctly exposed' shot.
  
• All the pixels in the photo (and in the histogram) are shifted to the left from where they should be (if metered by an incident meter).
  
• If the face was +0.66EV brighter than 18% tonality, all the pixels are shifted -0.66EV (to the left) on the histogram. Because the 'perfect exposure' was based on a target (the face) which is 0.66EV too bright (+0.66EV brighther than 18% tonality).
  

Alveric wrote in post #18106022
In that case, with an average caucasian face, did you up the exposure by one stop? (Average caucasian skin = ~+1 above mid grey.)

Also, remember to point your spot to the area sitting between the shadow and the highlight: that's where the mid-tone lives.

No, good point. I always assumed that the highlights/shadow areas of a face would approximately even out. Silly me. Never assume anything

Wilt wrote in post #18106028
D-Noc, your histogram, in the shot metered as you described,

• puts the pixels for the face AT THE CENTER of the histogram (because you metered on the face), when in fact they belong a bit to the RIGHT of center.
  
• All the pixels in the photo (and in the histogram) are shifted to the left from where they should be (if metered by an incident meter).
  
• If the face was +0.66EV brighther than 18% tonality, all the pixels are shifted -0.66EV (to the left) on the histogram. Because the 'perfect exposure' was based on a target (the face) which is 0.66EV too bright (+0.66EV brighther than 18% tonality).
  

As you describe, +0.66EV brighther than 18% tonality, but it does NOT 'spread out' pixels on the histogram (i.e. no increase in spread between 'black' and 'white'.

You would dial in +0.66EV of Exposure Compensation because the 'target is 0.66EV brighter than 18% tonality'. (but your eye would have to estimate how Usain Bolt's face compares to 18%, and how Gwynyth Paltrow's face compares differently to 18%, for you to dial in the correct amount of EC.

True. Alveric already hit the nail on that one
But.. I didn't really mean spread out as such... more like the massive buildup in the totally dark area that are spread out as the exposure is raised.

I think there are some misunderstandings in this thread. I'm still trying to get a grip on metering and exposure. Maybe someone can confirm my ideas about all this.

The 18% Gray Card
The 18% means that the surface reflects 18% of the incident light. In terms of stops, this translates to 2.5 stops of light loss. My guess is that this value was selected because 2.5 stops was the midpoint of the print medium used back when this became a standard value. Perhaps the paper had a 6-stop dynamic range. In that case, it would make sense to use it as an indicator for exposure. Also, since film was being used, and because film does not record light linearly, the dynamic range of the scene being photographed could be compressed.

Metering at 12.5%
What the midpoint seems to be, through testing of Canon's auto exposure algorithm, is 3 stops. This translates into a reflectivity percentage of 12.5%.

In today's world, and correct me if I'm wrong, prints and computer monitors have about an 8-stop dynamic range. It seems to me that the midpoint should be about 3.5 stops now. I remember testing this with the 5D3 and I could've sworn that I found the midpoint to be 3.5.

How does this affect auto exposure?
There are two predominant considerations among us from what I've gathered:

1. How do we avoid clipping highlights in our image?
   
2. How do we align the image's tones with how we perceived the scene in reality?
   

Frugivore,

To further confuse you,

• consider the fact that in the nine Zone System (1 thru 9), the middle is Zone 5 and each zone to the left/right is one full f/stop difference.
  That also corresponds to 'midtone' which happens to match the 18% gray tone, between 0% and 100%. Suddenly your speculation about 'middle of 5 f/stop range' further muddles in contradiction to the 10 f/stop Zone System. How can Zone 5 have 4 f/stops lower and higher, flanking it, according to what you posit?
  
• 12.5% gray tone is often said to be the true reflectivity of the target's brightness which causes many meters to register an exposure which captures an 18% tone 'in the middle between white and black'. So midtone gray in the range of white-to-black is not necessarily the 'tone for metering proper exposure' We can see proof of this here, where I metered the 18% area, and it is about 1/3EV to the left of 'middle' on the histogram

1) which film emulsion you use,
2) what developer you use for processing B&W,
3) which FILM TYPE (i.e. B&W/transparency/color​neg)
4) which digital camera you use

Of course, on POTN we are mostly considering digital (so ignoring points 1-3), where so many think 'Canon vs. Sony' and the 11-12EV DR in Canon and the 13DR of Sony. Look at this comparative illustration of the high end of the range of brightnesses ...

Wilt wrote in post #18106245
Frugivore,

To further confuse you,

• consider the fact that in the nine Zone System (1 thru 9), the middle is Zone 5 and each zone to the left/right is one full f/stop difference.
  That also corresponds to 'midtone' which happens to match the 18% gray tone, between 0% and 100%. Suddenly your speculation about 'middle of 5 f/stop range' further muddles in contradiction to the 10 f/stop Zone System. How can Zone 5 have 4 f/stops lower and higher, flanking it, according to what you posit?

An 18% card translates into a 6 stop range, not 5:

1. 100% reflectivity
   
2. 50%
   
3. 25%
   
4. 12.5%
   
5. 6.25%
   
6. 3.125% (anything below this registered as black)
   

12.5% gray tone is often said to be the true reflectivity of the target's brightness which causes many meters to register an exposure which captures an 18% tone 'in the middle between white and black'. So midtone gray in the range of white-to-black is not necessarily the 'tone for metering proper exposure' We can see proof of this here, where I metered the 18% area, and it is about 1/3EV to the left of 'middle' on the histogram

I think you're thinking about this the wrong way.

If you shoot a frame filled with a single tone of gray, regardless of how much light this card or surface reflects, what is the RGB value on the histogram?

Take this value and plug it into a calculator to determine how many stops below the maximum brightness level of a pixel. Multiple that by 2 and add 1. That is the dynamic range that the camera is programmed to expose for. If the RGB value is about 100, that would mean that the auto exposure system marks -3 stops as the middle tonality. That's a 7 stop range.

Whether you clip your highlights, when using the auto exposure system, depends only upon whether your scene is over the DR programmed into Canon's system. I think it might be 7 stops. Maybe 8. Outdoors on a sunny day, you'll blow highlights. Overcast or very even lighting? Your highlights will be well below clipping.

frugivore wrote in post #18106280
An 18% card translates into a 6 stop range, not 5:

1. 100% reflectivity
   
2. 50%
   
3. 25%
   
4. 12.5%
   
5. 6.25%
   
6. 3.125% (anything below this registered as black)
   

And 18% is smack dab in the middle.

I think you're thinking about this the wrong way.

If you shoot a frame filled with a single tone of gray, regardless of how much light this card or surface reflects, what is the RGB value on the histogram?

Take this value and plug it into a calculator to determine how many stops below the maximum brightness level of a pixel. Multiple that by 2 and add 1. That is the dynamic range that the camera is programmed to expose for. If the RGB value is about 100, that would mean that the auto exposure system marks -3 stops as the middle tonality. That's a 7 stop range.

Whether you clip your highlights, when using the auto exposure system, depends only upon whether your scene is over the DR programmed into Canon's system. I think it might be 7 stops. Maybe 8. Outdoors on a sunny day, you'll blow highlights. Overcast or very even lighting? Your highlights will be well below clipping.

But you have not attempted to resolve the conflict with the 9 zone 9 f/stop range between 'white' and 'black'. Nor have you reconciled it with 11 or 13 f/stops of dynamic range in the sensor. Open ended puzzles.

Wilt wrote in post #18106284
But you have not attempted to resolve the conflict with the 9 zone 9 f/stop range between 'white' and 'black'. Nor have you reconciled it with 11 or 13 f/stops of dynamic range in the sensor. Open ended puzzles.

Sorry, I meant to address this and posted too quickly. I've updated the post.

Edit: I did a search and found this article which supports my hypothesis:

http://ralphnordstromp​hotography.com …otograph/ansel-adams-hdr/

frugivore wrote in post #18106280
If you shoot a frame filled with a single tone of gray, regardless of how much light this card or surface reflects, what is the RGB value on the histogram?

I just set my 40D on JPG, shot a white sheet of paper and a real 18% gray card. and the LR readings were 54%(out of 100%) and 57% (out of 100%) in LR, and with Paintshop Pro it read about 130/255 and 137/255.

frugivore wrote in post #18106280
Take this value and plug it into a calculator to determine how many stops below the maximum brightness level of a pixel. Multiple that by 2 and add 1. That is the dynamic range that the camera is programmed to expose for. If the RGB value is about 100, that would mean that the auto exposure system marks -3 stops as the middle tonality. That's a 7 stop range.

Whether you clip your highlights, when using the auto exposure system, depends only upon whether your scene is over the DR programmed into Canon's system. I think it might be 7 stops. Maybe 8. Outdoors on a sunny day, you'll blow highlights. Overcast or very even lighting? Your highlights will be well below clipping.

Yes I have seen ability in the 40D image to underexpose by -3.5EV difference, yet I could still discern the crosshair mark seen in the Photovision target pictured earlier.

But wait, if my 40D and the most recent 5DIV will BOTH put that same photo at 128, just where does your calculation account for improvement/difference in the brightness (if any) and the considerable difference in the noise floor and the major difference that exists at the low end?! Post 39 illustrations #2 and #3 show different number of segments (14-17 segments) to the right, where the sensor finally quits differentiating, yet on a scale of 0-255, 255 simply is 'max' and fails to distinguish 14 steps vs. 17 steps. The same comments apply at the bottom side. The location of the Sony 0 is lower than the location of the Canon 0 value on the 0-255 scale, but both are 'max black'

Wilt wrote in post #18106326
I just set my 40D on JPG, shot a white sheet of paper and a real 18% gray card. and the LR readings were 54%(out of 100%) and 57% (out of 100%) in LR, and with Paintshop Pro it read about 130/255 and 137/255.

Yes I have seen ability in the 40D image to underexpose by -3.5EV difference, yet I could still discern the crosshair mark seen in the Photovision target pictured earlier.

But wait, if my 40D and the most recent 5DIV will BOTH put that same photo at 128, just where does your calculation account for improvement/difference in the brightness (if any) and the considerable difference in the noise floor and the major difference that exists at the low end?! Post 39 illustrations #2 and #3 show different number of segments (14-17 segments) to the right, where the sensor finally quits differentiating, yet on a scale of 0-255, 255 simply is 'max' and fails to distinguish 14 steps vs. 17 steps. The same comments apply at the bottom side. The location of the Sony 0 is lower than the location of the Canon 0 value on the 0-255 scale, but both are 'max black'

I'll have a deeper look into DPR's chart. I'm not that familiar with their testing methodology. Thanks!

Wilt, I'm not sure how you got those values. I set up a white seamless paper on the floor and turned off the lights. I then took a picture with a 200mm lens shot at f/5.6 and with a flash to illuminate it set at 20mm zoom. All this to get an evenly lit image, as much as possible. I opened my result in Photoshop which shows values of about 102, which is about 13% maximum brightness. If I increase the brightness by 3 stops, I will be exactly at the clipping point. This was with the 5D3 and I got the same result with my 70D:

frugivore wrote in post #18106388
Wilt, I'm not sure how you got those values.

I simply put a white piece of paper under uniform illumination and shot it in Av mode at ISO 100. Then I shot an 18% gray card under the same uniform illumination and shot it in Av mode at ISO 100. Both stored as JPG.

I took both the above JPGs, and read them into LR and also into Paintshop Pro. I used the respective white balance eyedroppers to neutralize the image (since it was not shot in daylight). Then I used each eyedropper to display numeric values under the sample point.

• LR sampler measures on a scale of 0% -100% in tenths of percent, a separate value for R, G, B; 'middle' is 50%
  
• PSP sampler measures on a scale of 0 - 255, a separate value for R, G, B; 'middle' is 127 or 128