Metering tonality...18% or 12%?

tonylong wrote in post #12086374
I'm trying to wrap my head this stuff, but I don't have the background that you guys have.

Something that seems evident from what Wayne is saying is that we are not dealing with linear data here and so shouldn't try to "jam" linear terms into it.

So, as we know, the histogram is not a linear graph, and, for example, a gray object/card/whatever that spikes in the dead center of the histogram is not linearly half way between black and white but only shows there due to the gamma correction. Linearly, one stop lower than white is half the light, and so a linear scale would show that as "half way" to black.

Wayne, am I on the right track here?

Right.. I think that "whatever" would reflect at 22%.

(0.22 ^ 1/2.2) = 0.50, x255 = 128 50% of full scale

I hope this is perceived more as a concept than absolute numbers. The camera is busy doing other things too, shifting data, for white balance, S-Curves for contrast, etc. Which seems to affect the right end more than left end. But midscale is obviously totally out of the question.

Right, of course. Linearly, one stop down, or 50% of the light is at midscale (we just do not have opportunity to see that RAW scale). We only see gamma encoded, which is closer to 3/4 scale. And often higher in my notion. It seems pretty hard for any argument to ignore what actually happens.

An 18% card reflects 18% of the light that hits it. Ideally, it would be at 18% on a fictional linear scale. 18% is about 46 on a 0..255 linear scale.

Origin: It was said that the human eye response and the human brain sees 18% as middle gray on our perception scale. There is no calibration, just our perception. Naturally, this matters if in a bright scene or a dark scene, and all that stuff. But long ago commercial printers (without computerized color gear) used it to judge their ink output on paper. Ansel Adams picked it up in his Zone System in the 30's. Again, he printed on paper to judge it (and certainly there are several other variables too... film and paper exposure and development, etc).

Kodak absolutely does/did say in the packaging for their 18% gray card that if we expose off of it, we must open 1/2 stop (in the sun). There was a period they dropped this statement, it must have confused the customers or something, but they dumbed it down for awhile. My understanding is that they have put it back now?

It is said that Kodak (that is a link, it does not show as such here) wanted to make it be a 12% card for photography, but Ansel campaigned to keep 18%. My guess is Kodak did not care to have Ansel doing this campaign against their judgment in public.

18% cards: If we make a few assumptions...

That 100% of the full direct light (whatever that means) would be adjusted to be at histogram 100% (255).
Let's say, if a 90% white card was at 90%...

That the angle of the card gives us the maximum 18% reflection,

and I think that the moon is blue,

then it should be recorded at 18% on a linear scale, in the RAW data at the sensor. This happens to be at (0.18x255) = 46 on the linear histogram scale (but we don't have tools to see the linear RAW data).

After conversion to RGB and gamma encoding (which we do see), it should be about
(0.18 ^ 1/2.2) = 0.46, x255 = 117 46% of full scale

A little left of center is a good value. Some diehards recalibrate their light meters to put it at 50%. Even the Thom link above mentions that. But of course, there is no justification for it at all, neither 18% or 128 is the midpoint of anything. However, it is a tiny error. Just the wrong reason. The only harm is to understanding.

Heh! Getting into the linear vs non-linear is like I said seeming to be getting us lost in space! I'd love to see things boiled down to a concise "nutshell" that the layman can understand, because we here actually have to provide "stuff" to newcomers on a day-to-day basis -- stuff like this, and, well, umm, discussing depth of field and what have you!

Wilt wrote in post #12085926
So please, now apply the above concept to a system which permits 9EV of dynamic range...I see maybe 5EV in what you describe, and now want to understand what changes in the computation of a 9EV system.

It occurred to me that I may finally realize what you meant in your question about dynamic range.

In math, 1 to any exponent is 1, and 0 to any exponent is 0. Therefore, the end extents of gamma change nothing.

With 0..255 being normalized to be 0.0 to 1.0 (percentages), then

(1.0 ^ 1/2.2) = 1, x255 = 255, 100% of 255 full scale
-1 stop = (0.5 ^ 1/2.2) = 0.73, x255 = 187, 73% of 255 full scale
-2 stops = (0.25 ^ 1/2.2) = 0.54, x255 = = 137, 54% of full scale
18% Gray card = (0.18 ^ 1/2.2) = 0.46, x255 = 117, 46% of full scale
(0.0 ^ 1/2.2) = 0, x255 = = 0, 0% of full scale

The end points do not change. The dynamic range does not change.

The center section is boosted, with extreme emphasis on the low end.

FWIW, this is about the same as the middle slider in the Adobe Levels dialog box. This center slider is sometimes also called gamma. It is the same exponential gamma curve, but with the variable exponent. It changes brightness, but does not change contrast (the Levels end points change contrast). (in great contrast, the control named Brightness merely shifts the entire histogram, which definitely affects contrast too).

The Levels center slider 1.0 unadjusted position represents the original sRGB gamma 2.2... the 1.0 is relative to that. It boosts center values in the same way, with strong emphasis on the dark end, but without changing the extent of the 0 or 255 points. Same thing, same curve. Video gamma has just already done this in degree of gamma 2.2 for all RGB data, including in all histograms that we will ever see.

Then ultimately, for viewing by the human eye, in every case, at the last possible time, the file gamma data is decoded back to linear, for viewing by the analog human eye... which should still be a faithful reproduction of the original linear RAW data. This decoding (from gamma to original analog) is done by merely showing it on a CRT monitor, or a LCD monitor specifically must overtly decode it first, within the monitor electronics. The eye has a logarithmic response curve in the brain, making us think we see the original 18% as "middle gray". There is no other connection with the concept "middle", and certainly it has nothing to do with the center of any histogram (histograms are gamma encoded).

Printers too must deal with the same RGB gamma data, but they also need most of it for themselves. Until relatively recent times, Apple computers used gamma 1.8. They did use the same monitors as Windows, and these monitors obviously required gamma 2.2, but Apple split this up. The image file used gamma 1.8, which was intended for their early laser printers, and then the video hardware board added another 0.4 for the CRT monitor, to be an unspoken 2.2. The last few versions have now accepted the world standards of 2.2.

Bottom line, gamma 2.2 is just how digital life is. The RGB data has been encoded, and in any histogram we can see, 18% shifts to 46%, and 50% shifts to around 73%. If you want to ponder any numerical histogram numbers, gamma is the very first thing you should learn. Just being aware is sufficient.

Again, these numbers are only theoretical, and not precise in practice. The camera is busy doing other shifting, for White Balance, and S-curves for contrast, and final result will not be numerically precise. But 18% will never be 50%. It happens to be coincidentally close at 46%, which does confuse the troops who falsely imagine 18% should be 50%, but its nearness is simply for an entirely different concept (called gamma).

Thanks for the continued thought on the matter, Wayne, appreciate it. I still have some difficulty with this fundamental gap in understanding...
When the exposure system is 5EV wide, your-2.5EV example of 18% (or thereabouts) works. But what if you have an exposure system (like we truly do with today's cameras, with 8.5EV of width in an exposure range) ?!...it takes just over 4EV simply to hit the midpoint of its range, as exemplified by this test report http://www.dpreview.co​m …ws/canoneos60D/​page13.asp, so what happens in the computations?

It would seem that from 0-255 in a 5EV wide exposure width vs. 0-255 in a 9EV exposure width still has everything squished into the 0-255 range of tonality. If we widen to 16 bit JPG, 0-65535 in a 5EV wide exposure width vs. 0-65535 in a 9EV exposure width still has everything squished into the 0-65535 range of tonality. So the dynamic range of the exposure system has no direct correctly to the width of the tonality, it would seem.

Wilt wrote in post #12097037
When the exposure system is 5EV wide, your-2.5EV example of 18% (or thereabouts) works. But what if you have an exposure system (like we truly do with today's cameras, with 8.5EV of width in an exposure range) ?!...

Gamma encoding has nothing to do with the dynamic range. Assuming we really do have actual data content to fill the bits fully, we could say we compute dynamic range by simply continuing to divide by half, each time being one stop. We do not run out of range until we hit zero. Which for 8 bits, that happens after 8 divides, or 8 stops. This is about bits (the storage container, and ASSUMES we have actual data available to fill those bits) - But it is not about gamma.

Gamma is just way to represent the data while in storage, so to speak. Before any human eye sees it, it must be equally decoded back to the original data. No change. Decoding it one way, and decoding it back in opposite way, should be an equal original result. Digital is simply math numbers, there is nothing physical which could erode away.

All that changes is the value of the numbers we see on the histogram scale.

it takes just over 4EV simply to hit the midpoint of its range, as exemplified by this test report http://www.dpreview.co​m …ws/canoneos60D/​page13.asp, so what happens in the computations?

It would seem that from 0-255 in a 5EV wide exposure width vs. 0-255 in a 9EV exposure width still has everything squished into the 0-255 range of tonality. If we widen to 16 bit JPG, 0-65535 in a 5EV wide exposure width vs. 0-65535 in a 9EV exposure width still has everything squished into the 0-65535 range of tonality. So the dynamic range of the exposure system has no direct correctly to the width of the tonality, it would seem.

The curve you reference is the farthest thing from a histogram. It is simply an analog graph, with arbitrary 1 stop divisions. It has absolutely nothing to do with digital data, histograms, or gamma.

But I don't want to go there. Not my fight.

However, FYI, your 5 EV is simply a number you made up somehow, with which I strongly disagree. 8 bits can store 8 stops of data... 2 to power of 8 is 256. This data ranges 255, 128, 64, 32, 16, 8, 4, 2, 1, 0. Some call it 9 stops, between steps. My monitor does not show 1 however. I don't care to nitpick that either. And of course, this is only the range of the storage container... story is not yet about the actual data captured there.

My guess about the origin of your 5EV range is that 18% is roughly 2.5 EV (at least, 46 is somewhere between 32 and 64 - assuming linear, which we cannot see). No argument. But you insist on calling 18% 50% (blatantly wrong), therefore you double it and call it 5EV range. Really, that is laughable.

Stops down from 46 are 23, 11, 5, 2, 1, 0. These most conservative numbers are a good 6 more stops below it. Whatever the numbers, it is extremely greater than 2.5 stops below it, and more than 5 stops overall.

Yes, we could use 16 bit data, and could represent greater dynamic range because yes, the numbers are larger, and it takes longer to hit zero. You might be able to bypass a JPG file, and retain 16 bits, but you still have to deal with a 8 bit video card and a 8 bit printer. And anything you print on paper and view in reflected light will suffer drastically.. Pitifully low dynamic range.

Look, Wayne, I am trying for understanding but you continue to ridicule my inability to grasp the math used in your own description...

(1.0 ^ 1/2.2) = 1, x255 = 255, 100% of 255 full scale
-1 stop = (0.5 ^ 1/2.2) = 0.73, x255 = 187, 73% of 255 full scale
-2 stops = (0.25 ^ 1/2.2) = 0.54, x255 = = 137, 54% of full scale
18% Gray card = (0.18 ^ 1/2.2) = 0.46, x255 = 117, 46% of full scale(0.0 ^ 1/2.2) = 0, x255 = = 0, 0% of full scale

...with comments like " therefore you double it and call it 5EV range. Really, that is laughable. "

I am sorry that my (lack of) genius in this falls short of your inability to help anyone comprehend.

We do need to boil this all down to what a layperson can comprehend and apply to photography!

From Wilt's original post (if I can recall) if you meter off a 12% card to center it in the meter then an 18% card will show up just to the left on a histogram. Wayne has layed it out mathematically to land at a 46% position on the histogram, if I understand correctly.

So, isn't that simple enough? If you have an 18% gray card, in good light meter with it set up by 1/2 or so of a stop?

Tony,
I believe that summary statement is what we arrived at, before the thread got sidetracked with Gamma mathematics and how it might (or might not) have bearing on centering the 18% gray card in the middle of the histogram. Pragmatic application of available tools and the technique of using them, vs. theory and possible application.

Wilt wrote in post #12097465
Look, Wayne, I am trying for understanding but you continue to ridicule my inability to grasp the math used in your own description...

(1.0 ^ 1/2.2) = 1, x255 = 255, 100% of 255 full scale
-1 stop = (0.5 ^ 1/2.2) = 0.73, x255 = 187, 73% of 255 full scale
-2 stops = (0.25 ^ 1/2.2) = 0.54, x255 = = 137, 54% of full scale
18% Gray card = (0.18 ^ 1/2.2) = 0.46, x255 = 117, 46% of full scale
(0.0 ^ 1/2.2) = 0, x255 = = 0, 0% of full scale

...with comments like " therefore you double it and call it 5EV range. Really, that is laughable. "

I am sorry that my (lack of) genius in this falls short of your inability to help anyone comprehend.

I am sorry too. And I was quite incorrect (wrong) to attempt to compute stops on exponential encoded gamma data. I blame Charles Poynton.

We never see pictures using gamma encoded data, everything is first decoded back to linear for human eye consumption.
CRTs, LCDs, and printers all decode it before output - in different ways, but they all do it.
But also, we cannot see RAW histograms, we only see gamma encoded histogram data.

As I have tried to explain repeatedly, your notion that histogram midpoint is the middle of anything is wrong too. After gamma encoding, you can only get nonsense from that. And while 18% might artificially be called middle gray (for entirely other purposes, from long before digital histograms existed), but 18% is a low value, which is NOT the middle of the histogram either (not linear, not gamma). The term middle gray absolutely does not refer to the center of any digital histogram. Both may use the word middle, but with very different meanings, not related. Middle gray only refers to the conceptual human eye/brain response (which leaves out histograms).

One stop intentional underexposure of the 255 right end is 50% of the light, by definition, and certainly half can be called middle of the available range. Yes, this 50% is at 128 on RAW linear scale at the sensor (255/2), which is the conventional viewpoint we always see everywhere - it is an easy explanation. And it is correct, HOWEVER WE CANNOT SEE RAW DATA. We have no tools, and all we care about seeing is our RGB conversion.

This one stop down at half value is around 73% of the gamma encoded histogram which we can see. Why is that difficult? (the precise number 73% is difficult, cameras are busy manipulating it - but they, and their RAW processors, certainly do output gamma encoded data).

18% ideally is 18% on the RAW sensor, but ends up about 46% in the RGB gamma histogram we can see. Why is that difficult? This simply happens, every time.

Gamma does have a reputation for being difficult, and it can be made difficult, but really, it is pretty simple - and the overwhelming big thing is that IT EXISTS. The world convention is that we just simply do it. All cameras, all scanners, all photo editors, they output gamma encoded image RGB image data. All RGB data in the world is gamma encoded.

We all know this number 2.2. We all know about sRGB and Adobe RGB being gamma 2.2. What we don't seem to understand is that this absolutely means all RGB data values are in fact gamma encoded. We don't even care why anymore, it just happens. This is always undone before the human eye sees it, but all RGB data is gamma encoded. It must be really difficult to try to ignore this fact.

And it obviously explains why 18% graycards come out about 46%, and why the midpoint (one stop down of underexposure) moves up to about 73%. Really, it is quite good stuff to know. If you ever do choose to think about gamma, then just look around, and you will see.

Heh! The one "good" thing I see is why it is so important in digital shooting to not underexpose!

Wayne, help us out -- we need to have a statement that we can give to the layperson (someone like me) about how to approach exposure using a meter and a histogram in practical terms!

tonylong wrote in post #12098028
Heh! The one "good" thing I see is why it is so important in digital shooting to not underexpose!

Wayne, help us out -- we need to have a statement that we can give to the layperson (someone like me) about how to approach exposure using a meter and a histogram in practical terms!

Wow! You caused me to stop and think, what is my method? And I'm not sure I have anything I can put in words. I'm not really conscious of an actual method, it is just something that happens. I doubt I can help, but I will try a description. Don't expect much.

I don't use a gray card. Not saying it's bad, if it helps, but I just don't see the need. Kodak's gray card advice was always to tell us to open 1/2 stop in sun. There are always ifs and buts however. Kodak said that many years ago, when good advice for negative film was "Don't Underexpose". It's latitude could handle a pretty good degree of overexposure, but there had to be something there. Digital however, is more like reversal slides - cannot handle overexposure. Slight underexposure is not such a bad thing.

I do check the histogram, but I know that I am never conscious of its center. There is nothing there of any interest. Maybe if I were photographing gray cards?

I am conscious of the right end, a large unfilled gap is rarely good. General scenes ought to fill most of it, however you have to also be guided by the subject, a few special dark/black subjects OUGHT to have a large unfilled gap at right. Typically bright white is near 255, and lesser bright is a bit more distant. For my amateur portraits, under studio light, I try to be sure to back off just a bit, to not overdo it. Nothing there that should be 255. It just looks better, and I go by look.

So mostly I am just guided by the previous image on the rear LCD, what it looks like there. If it is not quite right, I simply fix it so it is. It is my picture, and I know what it should look like. TTL flash needs more watching than sunlight.

LCDs are typically too bright, and I've experimented slightly to know to reduce my cameras LCD brightness one stop, to be more trustworthy, more what I think it should look like. My computer LCD uses a Spyder2Express calibration to bring that down a little too. Older now, but requirements are simple, and it is sill well done. Be aware that you can copy the picture to computer, but leave it in camera too, and view both at the same time, to compare LCDs. Color depends on White Balance, but brightness depends on exposure.

And, I shoot RAW (for its versatility and control), so I pay little attention to white balance until later, and I have no issue with slight exposure tweaks later. WB does affect the rear LCD picture however, so it has to be ballpark, but flash and daylight are all the same (I use a white card for critical studio color).

I am of a certain age now, but I was doing extensive dark room work over 50 years ago, in the last half of the 1950s. So I've seen a few things since, but frankly, I'm not good enough to shoot JPG. Seems impossible to always get it just right. In contrast, RAW is so easy, solves so much. This was true of film too, so much can be fixed in the dark room, or by the processor (but film loses out to digital's immediate result preview).

If I had a system, it would have to be named "Just watch, and do what you see you need to do". After awhile, you just sort of already know. We pick it up pretty fast.

The best advice for histograms is "be aware what they are, and what they show" (and they do show gamma). A histogram is NOT a light meter. The histogram has no clue what it is, or how it OUGHT to be, it only shows you what is, for this scene and this exposure, right or not. Specifically, it is a bar chart of how the pixel colors are distributed. If the subject has some bright white, it probably ought to be at the right end. If not, then maybe not. We also have eyes and a brain, which are very useful tools.

The best advice for a camera reflected light meter is to "be aware of what they are , and what they show". Beginners usually assume the camera light meter sees all, knows all, and is never wrong. This is NOT the case. The meter is a very dumb computer that knows nothing.

Incident meters are one thing (much better, but cannot be in a camera), but reflected meters are something entirely different. The meter has no clue what the subject is, or how it ought to look (no human brain recognition of anything). The reflected meter is greatly fooled by the fact that reflective colors reflect brightly, and less reflective colors reflect less. All the meter knows is this result, but it has no clue what it is or how it ought to be. It does not know if it is white that ought to be brighter, or black that ought to be darker. It simply does not know nothin'.

So the only possible course of action for the reflective meter is to try to make all scenes come out "middle gray" (whatever that means). I don't even mean gray (B&W days), it might be a middle green or middle blue tone. If it is a spot meter, the spot area is averaged to one color, which is adjusted to be middle gray (the rest of the scene comes out however it does with that exposure). If it is a center meter, that larger center area is averaged to come out middle gray. Matrix... same idea, but harder to say, some other dumb computer is in there meddling, and usually, neither we nor it has any clue about what it is going to do. This averaged result (of the metered area all averaged to one tone, and set to be middle gray) is often about right for an average, or typical, or wide ranging scene, with many things in it. But not always. It depends on the scene.

So be a little leery of your reflective meter. Understand what it shows. Use it, but be aware of your special subjects, and also check the results. Do what you see you need to do.

If you are unaware that ALL reflective meters strive to make ALL scenes average out to be middle gray, then you might be interested in http://www.scantips.co​m/lights/metering.html

On that last post, Wayne, 100% agreement about value of histograms. In fact, yesterday I wrote this reply to someone, about judging 'correct exposure' via histogram, and about the (lack of) 'correctness' of blind adherence to ETTR.

https://photography-on-the.net …hp?p=12092698&p​ostcount=7

Wayne, what you have described sounds vaguely familiar -- oh, I know -- it sounds like the kind of "seat of the pants" approach I (and many others) take!

I do have a gray card, I know I do, It's buried somewhere amongst my belongings, but the fact is that I never use it because I shoot in so many conditions I just take things as they come. I've never done studio work -- the only controlled lighting I shoot in is when I'm using a speedlight (or very rarely two) and then it's finding the balance between ambient and flash lighting on the fly and away I go.

What I was looking for, though, was a way of putting Wilt's findings to work together with your understanding of the linear/nonlinear stuff. Wilt presented the fact that using a 12% gray card in "good lighting" and centering that in the reflective meter would result in an 18% gray card showing up as a notch above center in the meter and then slightly to the left of center in a histogram. So, for those who want or need to use an 18% gray card to set exposure, they would want to set it to where the gray card meters that notch up, 1/2 stop, 2/3 stop, depending on the camera controls.

It seems like what you reacted to is the "idea" of that 18% gray card being called "medium" whereas (as we know and now are quite aware) 18% gray is much farther down the "linear staircase".

But, do you disagree to the practical approach I just stated, as to someone who wanted to get a "medium" tone using the gray card with a bit of positive "compensation", given good light?

tonylong wrote in post #12099341
Wayne, what you have described sounds vaguely familiar -- oh, I know -- it sounds like the kind of "seat of the pants" approach I (and many others) take!

Yes, that must be the right description. The point I was trying to make though is that reflective meters vary so much, depending of the reflectivity of the scene. So I don't want to say untrusted, I do think they are very precise, in a "calibrated" and repeatable sense, but just not always very knowledgeable of the situation in front of my camera. That is where the photographer comes in. Metering a gray card could be a solution, but I just pay attention to results. We kinda already know, from experience. It seems of advantage to try to learn to know what the meter is going to do.

Incident meters turn their back to the subject, and meter the light directly without ever seeing the subject. Then even if the subject is pure black, or pure white, they will come out black or white, because the light itself is metered accurately.

Reflective meters must meter the reflection from the subject, which really distorts things, and both the black subject and the white subject will be made to come out a middle gray. This needs watching.

When I say middle gray, I don't mean 12% or 18% or 50%, I just mean whatever the meter does in that case. But specifically, it will not be black or white. I'm not sure it needs more evaluation.

With studio lights, I just use an incident meter, and never even think about it... real pleasure, it's going to be right. Not to say I don't sometimes tweak it slightly later, which is exposure, but its more for looks, not for exposure as such.

But for ambient light and hot shoe flash with a reflective meter, yes, I do pay rather close attention to results.

I disagreed with the reasoning, mostly concerning what should be at the center of the histogram, and specifically why it should be there. I'm sorry to pass, but the details seem in the way of salvage attempts.

Here's the deal:

1. Reflection is analog (linear), and 12% and 18% are a half stop apart (standard ISO values, if you will). So if we meter on an 18% card, and open Kodak's 1/2 stop, that is the same as metering on a 12% card, all else equal.

2. In a perfect world, metering with an incident meter, photographing a 12% card should come out at 12% and 18% should come out at 18% on a linear histogram, which are at 30 and 46 on a 0..255 RAW scale. However, we can never see this linear scale - we cannot see RAW data. At best, we would see 38% and 46% of full scale (255) on a gamma scale after RGB conversion.
EDIT for clarification: Metering the reflection from such a card is basically this same result, if the card is appropriate for the meter.

3. With a reflected camera meter, photographing any of the individual 12%, 18%, and 90% cards are going to come about the same middle gray. This is just what reflective light meters do. All things come out the same (assuming our field of view only includes the card ... nothing else to average in).

So... if we imagined it ought to be histogram midpoint, what would be unique about which card we used in that case?

The other cards should vary with the same settings though, and 12% and 18% are a half stop apart.

http://sekonic.com …20FLASH%20MASTE​R_spec.asp

says L-358 uses K=12.5
Not much to question there.

This article
http://en.wikipedia.or​g …xposure_meter_c​alibration

explains a bit, and says "ISO 2720:1974 recommends a range for K of 10.6 to 13.4 with luminance in cd/m². Two values for K are in common use: 12.5 (Canon, Nikon, and Sekonic[1]) and 14 (Kenko[2] and Pentax); the difference between the two values is approximately 1/6 EV."

My own notion is I bought the meter, from Sekonic or the camera maker, and paid for their expertise. I'm not sure my qualifications are greater than theirs. I just try to learn to use it.

Well, so in practice I'd say we are all on the same page, we just got tangled up in the linear/non-linear thing so we just need to bear in mind that "things are not necessarily as they appear"...