Please help: M9 Exposure Problems

[diogenis]

Advertisement (gone after registration)

Hey--even though you're only using RAW, Leica is still making a JPEG to display for you as a preview It's why it takes so long to review some of the images; there's processing from the RAW file taking place.

 

So your histogram is actually measuring a JPEG made from the RAW file, and therefore not accurately telling you where your highlights have really fallen or how much shadow detail you really have. Many is the time I've seen the blinking blown out warning on the M8 but know from testing that the highlights aren't blown in the slightest; only in the way that Leica has interpreted the RAW file.

 

The problem is that a RAW file is a totally different beast than a picture you want to preview (different gamma and luminance, monochrome till it's interpreted, and linear to boot).

 

So even when you're only saving RAWs, your camera is still showing JPEGs

 

The depiction of the Raw file should be irrelevant to its histogram or the hi-lo clippings of the sensor data. Unless you record a really bright subject like facing the sun only blocked a bit, Blown highlights appear linearly. I do know that when I shoot with just a tiny level of red areas, that this does not affect the picture at all, but this doesn't mean that the blown pixels are not there. just that they are little, and the rest of the picture's gradation hides them. I refuse to believe Leica has made an inaccurate histogram function as regards to the clipping. It's such an easy task to do, and it has nothing to do with jpegs. Just to read photocell values and record there position to show in the LCD.

 

Next time you shoot, shoot some sky, over expose using the camera's meter by as little so that you only see very few red spots. Then locate them in CS or LR or something. They should be there... It should work!

[Jamie Roberts]

{snipped}

Next time you shoot, shoot some sky, over expose using the camera's meter by as little so that you only see very few red spots. Then locate them in CS or LR or something. They should be there... It should work!

 

Nope, it doesn't...what appears clipped on the LCD is not clipped in C1. The clipping indicator, from everything I've seen and read, is using the JPEG preview and not reading RAW values. If you have some proof of some other technology, I'd be happy to be wrong (the histogram is still mostly useless, though )

[diogenis]

Will check and see.

It was my feeling that it was working just fine.

In any case, even if it is the jpeg that it processes, then what's the problem? a value of 255 is 255 even when compressed...

[Jamie Roberts]

Will check and see.

It was my feeling that it was working just fine.

In any case, even if it is the jpeg that it processes, then what's the problem? a value of 255 is 255 even when compressed...

 

No, also not at all true, especially if you tweak or change the white balance...whole channels can be blown in the preview but aren't when you actually interpret the file I'd rather interpret the raw file than the camera, thanks

 

Also, there's more latitude for highlights and shadows in the raw file, obviously.

[diogenis]

Jamie, the problem is not 255 or 254. Even 255 is a highlight value of the "absolute" white which it must exist. The problem is 265, 275 and more, and their distribution, amount in the scene. Me too, I only use Raw files as a base to anything. I had as my guidance the highlights tool so far and it did the work fine, but I am not going to say no if I manage to salvage more shadows till I clip. And I don't mind that much for some clipping as soon as there is nice variation in the scene.

[lars_bergquist]

All right, here we go: Saint Ansel's Error

 

First, some historical background. Once upon a time, large low-orthochromatic or even blue sensitive glass plate negs were souped under visual observation, in the cosy red light of the darkroom lamp. The object was to achieve a negative contrast such that the neg could be printed contact more or less straight, on the only contrast grade of paper then available. The great prevalence of articles on intensification and reduction in the photo magazines of the time suggests however that this goal was seldom achieved. This was nevertheless a much beloved indoor sport. It seems that the underlying motivation behind the Zone System was to save this cherished procedure into the new era of pan cut film, which could not be processed visually. Hence, 'previsualization' and plus and minus development.

 

Now to the zones themselves. Asked "what is Zone V?" most aficionados would answer "It is a mid-tone with 18% reflectance". But that is a point value. A zone is a range, in this case with a reflectance ratio of 1:2 between the limiting values of the zone, and abutting other similar zones. In order to find out what Zone V realy is, we must find these limit values by multiplying and dividing, respectively, the 18% mid-zone value by the root of 2 (1,414 ...). Then we can do the same with the other zones. Here are the values for Zone V and the higher zones:

 

Zone V: 12.7 -- 25.3% (mid-value 18%)

Zone VI: 25.3 -- 51% (mid-value 36%)

Zone VII: 51 -- 102% (mid-value 72%)

 

Hey, we're only at Zone VII yet, and we are already finding reflectances above 100%! How can any diffuse highlight reflect more light than falls on it? So Zone VIII in its entirety should print white, or be allowed to blow out, because only specular highlights inhabit it. Still, the Zones march on:

 

Zone VIII: 102 -- 204% (mid-value 144%)

Zone IX: 204 -- 407% (mid-value 288%)

Zone X: 407 -- 815% (mid value 576%) ...

 

Now, Ansel clearly did not know whereof he was talking. His excuse is that even simple reflected-light meters were just coming in, were of low sensitivity, patently non-linear and un-standardised. It was not easy to grasp this field. He may also have been misled by the fact that with the long shoulder of film, in a badly overexposed neg, or with a subject with very large and deep shadows and a deliberately strong exposure ('Zone V' less than nominal reflectance), it was still possible to tickle out highlight detail well into 'Zone VIII'. But we can't do that.

 

The only thing that remains is that the maximum diffuse highlight cannot reflect more than 2.5 f-stops light than an average mid-tone. The only thing we have to do is to nab that highlight, and the rest of the image follows (unless we deliberately blow the highlights in order to save shadow detail). And this nabbing of the highlights is what the incident meter does. No in-camera histogram can do that, because it cannot know the difference between different types of highlight.

 

Below is a Kodak Gray Scale correctly exposed. Three steps = one 'zone', i.e. one f-stop. Step 7, 'M', is the 18% midtone. Steps 0--1 are 'Zone VIII', and you can clearly differentiate between the steps here. Step 16, 'B', marks the midpoint of 'Zone II' and is the last step that can be differentiated from its lower neighbour. Conclusion: Omnis lux est divisa in partes VII ... Or, Trust your incident meter.

 

Lars Bergquist -- The old man from the Age of Film

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[lars_bergquist]

Advertisement (gone after registration)

Correction:

Below is a Kodak Gray Scale correctly exposed. Three steps = one 'zone', i.e. one f-stop. Step 7, 'M', is the 18% midtone. Steps 0--1 are 'Zone VIII', and you can clearly differentiate between the steps here. Step 16, 'B', marks the midpoint of 'Zone II' and is the last step that can be differentiated from its lower neighbour. Conclusion: Omnis lux est divisa in partes VII ... Or, Trust your incident meter.

 

Lars Bergquist -- The old man from the Age of Film

 

I'm very sorry -- steps 0--2 are of course Zone VII (three steps make one 'Zone'). My excuse is that I am a bit dense just now, due to a bad head cold. But step 0 is about the highest reflectance that can be technically achieved in print, and is about 90%, if my befuddled memory serves me right. Zone VII remains the highest where detail can be distinguished. It seems that I was unable to follow my own argument! And now for a nice cuppa.

 

Lars Bergquist -- The old man from the Age of Film

[jaapv]

Jamie is right that the clipping warning is read from the jpg in the end. you can see it happening on the LCD -or at least you could on the M8; I havent checked the M9- The warning starts at RAW, and as the Jpg is generated you see the alledgedly clipped area expanding. Quite a large safety margin.

[Jamie Roberts]

{snipped}Conclusion: Omnis lux est divisa in partes VII ... Or, Trust your incident meter.

 

Thanks Lars! Of course it didn't occur to me, even after reading Adam's own texts, that multi-contrast papers and filters would help

 

And I pretty much (pretty much) always trust my incident meter; my shenanigans with the M (and other cameras) meter tries replicate incidence whenever possible and save those highlights or push those shadows (unless they're not worth being saved or pushed )

[lars_bergquist]

Thanks Lars! Of course it didn't occur to me, even after reading Adam's own texts, that multi-contrast papers and filters would help

 

Not only that: In the beginning there were not even graded papers. So you simply had to get negative contrast right. With contact printing, opportunities for burning in and shading were also very restricted. So already graded papers and enlargement printing made the Zone System largely irrelevant. And the changeover to roll film made individual development impossible. (Though I have heard of people who tray-developed ortho film until the separate frames were visible, then cut the strip up with scissors to develop the frames individually -- but this did not catch on, understandably. And pan film put an end to that game too.

 

But this point in time was of course exactly the one when the System became the Zone Cult. Today, just as you can recognise the Salafite Moslem male by the shortness of his trousers legs, you can spot the Zone Cult male by his spot meter.

 

The old man

[mjh]

Conclusion: Omnis lux est divisa in partes VII ...

Yes, assuming that lighting is uniform. In that case, the relative brightness of a subject only depends on its reflectance. But that’s the trivial case and not a particular interesting one in that it doesn’t present any problems. That’s why photographing a gray scale doesn’t tell you anything about your camera’s dynamic range – any compact digicam could get that right. But if lighting is not uniform, a subject may appear brighter, not because it has a higher reflectance, but because it receives more light to begin with. With non-uniform lighting the scale of tonal values grows considerably, and that’s when exposure gets really tricky and dynamic range becomes an issue.

 

In real-life scenes we have non-uniform lighting simply because there is light and shadow – subjects in the shadow will appear darker, regardless of their reflectance. And because of the ubiquity of non-uniform lighting one cannot simply equate Zone V with a reflectance of 18 percent (plus/minus 1/2 EV). Put your Kodak card in the light and in the shadow, and a spot meter would give you quite different readings. Now which one should represent Zone V? That’s for the photographer to decide. Using an incident meter implies that you don’t care about the rendition of tonal values of subjects where the light source you’ve measured doesn’t reach (or only as much as you might be able to salvage some shadow detail in the raw converter), and that, too, represents a decision on part of the photographer.

[diogenis]

Ok as expected and promised, I took a couple of pictures on the outside, Raw. No lightmeter used, directly manual, using the sunny16, both at f/4. 1st one spot on, @1/2000, 2nd one over by 1/2 stop @ 1/1500.

 

The first one is almost like I wanted it to be, or how I would chose expose. There are very little blown areas, EXACTLY where shown by the review screen when you chose info.

the other, a bit over with many more blown highlights, still shown EXACTLY where they should be. That means that the tool works as it should.

 

Recorded difference per channel is some 20-25 levels from the blown areas and for this lighting condition and half stop.

 

Now, how come one persuade me that an external -certainly accurate, VERY accurate I would add- incident light meter, but with no connection with the end result (what the sensor actually records) is a better option than using camera's meter and validating with the histogram? Especially, when you have the thing inside your cam, for free!

 

I can send the Raws if you want em

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[diogenis]

Also, the first pic, shows the full DR of the sensor: from near 0 for blacks to 255 for blown whites. Both these regions are clearly shown in the histogram, with very low distribution. Actually one has to hunt black areas in C1 to be able to see channels at 3-5 brightness level (railings on the right down hand corner)

In the other photo, due to the half stop overexposure, blacks begin higher.

Forgive me about WB, it's a bit off, but this has nothing anyway with clips

[Jamie Roberts]

Don't need the RAWs; the highlight clipping indicators are still often very wrong (and they are based on JPEG interpretation) in my workflow.

 

That they're sometimes (or even often) right doesn't mean I shouldn't ignore them In a JPEG workflow they're great; in a raw workflow I do things differently.

 

As I said before, too, the histo doesn't tell you anything really useful either. If I'm maintaining or pushing shadows highlights will blow (and that's ok) and if I'm maintaining highlights at all costs then shadows will look "bunched up"... in my experience with the Kodak sensors, beyond their ISO limit there's lots of shadow detail (and more in the m9 than the M8 by a stop or two).

 

@ MIchael & Lars--I still think an incidence meter (or something that approximates an incident measurement) is still a better starting point for many kinds of shots than a mostly unpredictable reflective meter.

 

Though for me, the Leica M, with it's heavy center weight, is almost ideal for the way I meter and shoot--it's one of the reasons I can work relatively quickly with an M. Matrix meters especially drive me round the bend, and almost always do the wrong thing for what I want to do. Maybe it's just practice, but every time I rely on them I end up doing more in post than I want to

 

So both Michael and Lars are correct; but I'm more with Lars on the incident side (again, as a place to start). Do you have to adjust exposure from an incident reading? Yes, but placing "zones" is not entirely equivalent to the Zone System (and I'm one of those guys, that, yes, cut up 35mm negs in a distant darkroom a long time ago ). And had a professor who made us work with negative density and development, despite multi contrast RC papers. I remember him doling out fixed contrast fibre for our final assignments

 

Conveniently, my incident meter also has a spot meter, for those rare times when I can't measure the light falling on a subject... my M has neither; hence the workarounds.

[lars_bergquist]

All right. So suppose that you have an extra dynamic margin and can hold detail up in Zone VIII. Then it is Zone VIII and not Zone VII that you have to determine. But you do still have to determine it, no?

 

And how are you going to do that? Incantations to your raw developer won't do the job.

 

A Kodak Gray Scale can in fact help you determine how large your dynamic range is, with your no doubt very superior raw developer. Just set the card up, determine the correct 18%-based exposure, then overexpose by one f-stop. If you can still discern detail in steps 0--2 (now become Zone VIII instead of Zone VII) then you can indeed overexpose by one f-stop in order to save shadow detail. Repeat the experiment with 2 stops extra exposure ... but remember, there is in fact a limit to this sort of exercise. By upping exposure, you also up stray light in the shadows, and with film, more than one stop extra did not bring any improvement in shadow detail -- while, as I wrote before, three extra stops did not destroy highlight detail. But that was film. With digital, this long shoulder of the characteristic curve simply does not exist.

 

Also note how in Diogenis' second picture the wires seen against the bright sky are already visibly being eaten away -- by one paltry half-stop of exposure increase! This is the phenomenon called irradiation (Überstrahlung) and it is a great destroyer of detail.

 

The old man from the Age when Developers Came In Cans

[diogenis]

Don't need the RAWs; the highlight clipping indicators are still often very wrong (and they are based on JPEG interpretation) in my workflow.

 

That they're sometimes (or even often) right doesn't mean I shouldn't ignore them In a JPEG workflow they're great; in a raw workflow I do things differently.

 

As I said before, too, the histo doesn't tell you anything really useful either. If I'm maintaining or pushing shadows highlights will blow (and that's ok) and if I'm maintaining highlights at all costs then shadows will look "bunched up"... in my experience with the Kodak sensors, beyond their ISO limit there's lots of shadow detail (and more in the m9 than the M8 by a stop or two).

 

Jamie,

My workflow is also entirely sushi (Raw ). I know I sound like a broken record, but the fact that the camera with its light meter has yet to fool me, it means that it works. It also means that there might be a situation which has yet to appear that it will not.

In this particular photo the histo tells me the following:

1. that exposure is perfect: with blacks and whites that span through the entire dynamic range of the M8's sensor. Only the WB is off by 1500 K, because I had left it to manual.

2. that there is indeed a small very small portion of clipping occurred, I don't know: 2,3,5%?

3. This kind of info you obtain real time and it's correct, regardless of compressions and/or jpegs

4. Even if you have your trusty best of the best of incidence meter, you can only get real life results by checking the info button and histograms, as this is actually what the sensor has recorded.

 

I'm not judging in particular your workflow using an light meter. Absolutely not, besides you still need something to evaluate EV.

For me, and to simplify things, I want to stick and learn the sunny16 rule or the light meter of a camera worth $3500 when I bought it.

 

Edit: I also forgot to mention that the histogram also tells me that clipped blacks -if exist- are indeed very little.

Edit2: Furthermore, the histograms on BOTH C1 and camera info are the SAME! And in all 3 channels

[diogenis]

Lars, nothing eats my cables lol.

I can zoom to 100% and all the detail is there. Which sector of the frame do you mean?

map the sector using the numeric keyboard, and cut it in 9 sectors then tell me the key, so I can magnify it

[Jamie Roberts]

Also, the first pic, shows the full DR of the sensor: from near 0 for blacks to 255 for blown whites. Both these regions are clearly shown in the histogram, with very low distribution.{snipped}

 

No, it shows the DR as mapped to an exposure level, gamma space (non-linear) and colour space. Consequently, it's simply not always right. It's often right, but not always.

 

That it matches here is a happy accident. But I bet you could pull at least another two stops out of the shadows, and another stop out of the highlights (maybe not the sky, since it's likely on the edge anyway).

 

IOW, where the OE is extravagant, the camera's indicators are great. But when we're talking about the difference of a stop, say, on a bride's dress, or on the bride's face, many is the time the camera indicators are downright wrong

 

It just saves me time in post to ignore the histo and the blinkies.

 

And Lars is right. How do you determine, precisely, where you want values in a scene like this. Come to that, what is important here? The stripes in the overhang? The trees? The cars below?

 

Or is it just the average illumination? Ok. I don't know why in terms of meaningful exposure, but as an even distribution (sort of) of levels, it's OK

[Chroma project]

Yeah, but...I honestly couldn't care less what a histogram says... because:

 

1) a histogram only shows you highlights, and not whether they're significant or not... which means you might be holding speculars and other junk you don't want or need to hold. Same goes for shadows.

 

2) the histogram measures JPEG response not raw. So in fact the M9 has a ton of latitude; I've pushed ISO 1600 to a veritable ISO 5000 (by metering) and printed the results and they're great. The histogram is a lie, in other words.

 

3) Expodisc is kind of overrated for exposure. A true incident meter is much easier to use in tricky light because the light falling on your subject is more important generally (95% of the time) than the light reflected by your subject. The histogram is not nearly as important as understanding that, IMO.

 

4)you can calibrate your camera to the meter with studio lights so you're never guessing again You need to understand what illumination will give a measures RGB highlight of about 242/242/242 from the raw file and then compensate for the difference.

 

But all that is pretty complex for the weekend shooter.

 

So my advice? ignore the histogram because it lies; get an incident meter and use it till you get the light OR just point your M9 meter at the most significant highlight you want to hold and overexpose +-1-2 clicks of the shutter or aperture. Sometimes that thing you're pointing to is not "white," remember, but it often is .

 

Don't change the exposure again till 1) the light changes or 2) you want a different interpretation of the subject or have a different subject

 

Your shadows will fall where they fall, but they can usually be retrieved in post (of course when you're shooting ISO 2500 in very low light that's not going to be the case )

 

Very good ,thank you James.

[Advertisement]

Hallo,

Look here: Leica M

[diogenis]

No, it shows the DR as mapped to an exposure level, gamma space (non-linear) and colour space. Consequently, it's simply not always right. It's often right, but not always.

 

That it matches here is a happy accident. But I bet you could pull at least another two stops out of the shadows, and another stop out of the highlights (maybe not the sky, since it's likely on the edge anyway).

 

IOW, where the OE is extravagant, the camera's indicators are great. But when we're talking about the difference of a stop, say, on a bride's dress, or on the bride's face, many is the time the camera indicators are downright wrong

 

It just saves me time in post to ignore the histo and the blinkies.

 

And Lars is right. How do you determine, precisely, where you want values in a scene like this. Come to that, what is important here? The stripes in the overhang? The trees? The cars below?

 

Or is it just the average illumination? Ok. I don't know why in terms of meaningful exposure, but as an even distribution (sort of) of levels, it's OK

DR is DR. Fundamental to your sensor/camera system.

On this particular shot, I am only telling that it displays the whole range of possibilities one has with an M8. This is what this camera can do and nothing more.

What YOU can do is decide where you want to focus in terms of detail/luminosity.

If you want the skies for whatever reason, you underexpose by little, if you want the trees you overexpose. By doing so you shift left or right the luminosity levels your camera can capture, but on the other hand you cannot do this without consequences. Besides your adjustments lie within a very small +- 1 stop a bit more?

 

On my particular photo Jamie, you can't pull anything out of it. If you change exposure you will lose either shadows or light. If you want however, more trees or more skies, you won't avoid post besides this is why it is there. On a bribe's face, dress or whatever there are tools that can independently enhance (viveza2 i.e which is on this month's LFI btw). If you don't want that, then use camera's lightmeter with it's strong center metering or your trusty incident meter and no need for pp. Besides, within church you won't find contrast like that where you need to play with exposure. Light is controlled and camera's DR is enough and can forgive even mistakes.

 

Problems begin when one encounters conditions like Michael is describing. Oh and you can indeed chose to ignore histo, but you should not do this when both blinkies and distribution of them is high in your histos. And it doesn't mean that they are not there either.

 

But with a nice M9, the new noctilux or even the summilux 50 or 35mm, an expodisc to meter WB, and good available lighting I cannot imagine someone not be able to focus composing.