What’s the origin of correctly exposed middle gray?

[silvan schnelli]

I feel quite confident in lighting with a meter and placing my exposure within a scene. I also often feel like I grasp the concept of middle gray, as it is one of the most basic/essential ones. However, every-time I’m asked to explain it to someone, numerous questions arise within myself.

After a lot of pondering I have came to the conclusion of what my understanding would be: 

Middle gray when talking about cinematography is a sort of response setting by the camera. It doesn’t refer to a  certain luminance from an object or illuminance hitting the sensor. 

Light meters use the EV formula to ensure that that similar exposure values can be generated at different scene illuminance values and/or camera settings. 

Furthermore different manufactures can place middle gray at different code values, e.g. the recent example of logc3 and logc4. 

Also important to note is that a middle gray patch, visually reflects half of the light it hits, this is 18% of the light as human vision is logarithmic.

Question

But the most fundamental question I would have is who or what exactly determines this correctly exposed middle gray. Yes middle gray is the middle tone, but what exactly is middle tone? I’m guessing this is where the calibration constant comes into play, which seems to practically determine what this is. At least this would be my guess?

Correctly exposed Middle gray doesn’t refer to a specific luminance reflecting off an object, or a specific illuminance hitting the sensor. It isn’t based on being in the middle in a log encoded image (EI can change this). 

In conclusion my understanding is: Scientists/Photographers noticed that when an object reflected 18% of the light projected on it appeared to be half of the light incident.

However, correctly exposed middle gray must have been another decision made. Where it was agreed upon that that specific rendering of this 18% gray by the camera at specific settings (governed by the EV equation), is considered correct. 

What was the goal of middle gray?

I’m guessing it was to help achieve a balanced image, as in this doesn’t appear too dark too bright.
 

Monitors and Cameras:

So I am assuming that middle gray became a standard for the way monitors and cameras were manufactured and calibrated. 

 

I apologies if the question is a bit pedantic and verbose. It just never felt right for me to simply say middle gray is when 18% gray appears the way it should be or correct. I think also the term correct exposure can often become dangerous as it makes students (coming from a current student) believe that if the shot isn’t correctly exposed (the Caucasian skin-tones aren’t 1 stop over middle gray) it must be wrong. I know this must partly also come from the film days and the fear of a thin negative, with films superior highlight handling and lower latitude in the shadows.

I have grown accustomed to trying to light for the way I want it to look, as is always advocated for by cinematographers such as Sir Roger Deakins ASC BSC. Of course also coming into play is having enough information to make it through the whole post pipeline, but EI and over/under exposure can help with this. 

Edited October 31 by silvan schnelli

[Karim D. Ghantous]

There is such a thing as correct exposure based off 18% gray - that's the definition. 😉 It doesn't mean you must be correct, though.

I cannot give you the proper explanation. But 18% makes sense, because if you double it twice, you get 36 and 72. And if you halve it twice, you get 9 and 4.5. If gray were 50%, you could only double it once.

To make things more interesting, the WhiBal card is not 18% gray - more like 12% or something like that:

https://whibal.com/

[Nicolas POISSON]

I struggled to get it as well. Here is the way I understand it now. Short answer :

- when exposing to the right (ETTR), you match the brightest part of your scene to the clipping threshold of your sensor.

- when using a grey card, you match the median brightness of your scene to the middle of the DR of your sensor. Statistically, the grey card happens to fall at this median brightness for most scenes.

 

Long answer :

You have a scene with a certain contrast (think of its histogram), you want to shoot it with a camera whose sensor has a certain dynamic range. Exposure is about how you match both.

There are basically two techniques.

The first one is well-know from digital photographers : Expose To The Right (ETTR). You look at the histogram, and open or close the aperture so that the brightest spots of the scene fall just below the clipping point of the sensor. This way, you maximize the light (the incoming signal) that reaches the sensor, without clipping. You get the best signal/noise ratio possible since the sensor is the main culprit for adding noise.

But this technique has drawbacks :

First, when shooting film (whether stills or movie), there is no obvious threshold of clipping. There is a highlight roll-off. You cannot match "to the right" since this is a fuzzy "right".

Second, even in the digital age, there is another problem for film-makers. Imagine you shoot several angles of two actors talking, that will be later inter-cut. You have a hot spot in the frame of the shot (a lamp shade). This angle has a high contrast. Exposing to the right, the actor’s face will fall relatively low with respect to the sensor’s dynamic range. Now you switch to the counter-shot where the lamp shade does not appear. The contrast is much lower. Exposing to the right, the other actor’s face will fall way higher.

You have the best possible signal/noise ratio for each angle. This may sound like a good thing, but noise will be inconsistent. The angle without lamp shade will be cleaner. The problem is that the human visual perception is hard-wired to detect changes, because change is potential danger (whether a tiger in the jungle or a bus coming while you are to cross a street). This is the same for human hearing, and that’s why the ambulance has flashing lights and varying tone.

When angles are inter-cut, the varying noise level is more noticeable. So you do not want the best signal/noise ratio for each angle, you want a constant level of noise. For the counter-shot with lower contrast, you will not ETTR. You will expose like the shot, and loose some quality on purpose. A simple way to do this is to have a common reference in both frames that you will set at the same level on your sensor. Here comes the middle-grey card.

But why middle grey ? The reference could be anything, as long as it is the same for both shot and counter-shot. It could be a white board as well. In the printing world, the darkest black possible reflects about 3 % of light, while a white sheet of paper reflects about 90 %. In a logarithmic scale (the human perception is logarithmic), 18 % reflection is halfway between those two values. So if the print uses all the levels from black to white, middle grey will statistically happens to fall at the middle of the histogram.

Now a print only has 5 stops of DR, while modern camera easily has 10 stops or more. We can shoot scenes with greater contrast. But if the contrast expands toward both sides, which means the scene has simultaneously spots brighter and darker than what we could record with only 5 stops of DR, the 18 % middle grey remains at the middle. Take any scene you want, add a middle grey card to it, it will often fall at the centre of the histogram, whatever the contrast of the scene.

Of course this is not always the case. Sometimes the histogram is unbalanced, or you do not pay the same attention to every objects in the scene. This is where you would use exposure compensation.

Also, with high DR cameras, you may want to set the middle grey a bit shifted toward higher values. This means you expect that most scenes will not have spots that much brighter than the middle grey. Shifting the middle grey up means you will expose higher, and get cleaner shadows.

Now, there is no defined IRE level for middle grey, as this completely depends on the curve used (whether log or rec709). On an A7sIII, The same scene, lit the same way, with the same exposure parameters (aperture, shutter speed) will have different histogram in S-log and S-log3.

Edited November 1 by Nicolas POISSON

[silvan schnelli]

@Nicolas POISSON

thank you for the lengthy reply, although when it comes to exposure I do usually just expose correctly (as in how I want it to look) and use EI accordingly to shift the latitude, so technically over and underexposing. 

I do see the appeal for ETTR, especially since pulling down brightness is much more forgiving than pulling it up. But for the reasons you have mentioned, as well as digital cameras having very good shadow latitude and with the fact that ETTR you technically are losing control of the images latitude distribution, I wouldn’t expose this way.

My question though, is a bit more related to how did they decided how a correctly exposed 18% gray appears like with respect to the appropriate camera settings. Which includes things like how did they view these images. 

[Simon Wyss]

Screen luminance is standardised.

From the projection flux of light necessary for standard screen luminance the diffuse density of prints is derived as a middle grey. From the prints’ middle grey a negative middle grey or density is calculated.

After the negative exposure is determined by the amount of light reaching it you have a given diaphragm opening for a given exposure time (shutter angle width, frame rate). The diaphragm opening determines the flux of light for taking pictures where a middle grey is deducted from snow white. Since the exposure index of a cinematographic material is given, say, ISO 100, the last variable is the luminance of a snow white or grey surface. We can choose between white as reference or grey. Sometimes checkerboard or printing dot greys are used.

Development of negative and positive is standardised, too.

[silvan schnelli]

@Simon Wyss thank your for the reply. I’ll have to look more into it and read some SMPTE papers.

[Jeff Bernstein]

Reading SMPTE papers rewarded this cinematography.com user :

 

"When at Technicolor helping with the dailies on Raging Bull, I solved this problem for Michael Chapman, ASC, when I found the solution in an old SMPE pamphlet from the 1930s . . . It earned me my very first screen credit."

 

[Eric Eader]

Hello,

Mascelli's Cine Workbook from late 60's and early 70's had an explanation that went like this: 4% (Black) x 88% (White), then take the square root of that answer to get 18%.

4x88=352

Square root of 352=18.766166     I don't remember the reason for rounding down.  (Latitude?)

If I still had the workbook I would give his explanation.  4%= still some detail in black; 88%= some detail in white, maybe?  Based on seven stops of exposure.  Three above and three below middle.

As for middle gray...  there are some gorgeous face/skin tones to be found around middle gray.  Sure some women were filmed a little hotter than mid-gray but others were stunning at or just below.

Check out the B&W work of Ernest Haller,  Joe MacDonald,  Oswald Morris,  James Wong Howe,  Lucian Ballard,  Phil Lathrop,  Loyal Griggs,  Burnett Guffey,  Ernest Lazlo,  Robert Burks,  Charles Clarke,  Joseph Ruttenberg, Robert Surtees,  Russel Metty,  Stanley Cortez,  Joseph August,  Lee Garmes,  Gaetano Gaudio,  George Barnes,  Conrad Hall,  Russell Harlan,  Robert Krasker,  Sven Nykvist,  Greg Toland,  and the only man to win an Academy Award for cinematography without even being nominated: Hal Mohr.

Looking back over your post I've realized you may be more interested in shooting color not B&W. Sorry.

With 18% gray you help to establish color balance, and a good relationship between light and dark tones all within the working limits of your equipment, set colors, costumes, and make-up etc.

With the greater latitude of digital, 15-17 stops, 18% presents a challenge because the above formula doesn't work.  And yet 18% persists.

For those with an interest of the earliest days of Lighting, with photography as medium for testing, check out "Light and Shade"  by M. Luckiesh, a research engineer for the newly created General Electric Corp.  The book is more than 100yrs. old and I don't know about the quality of the reprints (expired copyrights). Since there was no one to ask for help, he had to experiment...  test, test, test.  He also wrote "Color and its Applications" with lots of color chips.  Experiments and thoughts from the pioneering days of the 'modern era.'