SL2-S FW 2.0 Highlight weighted exposure metering challenges in backlit scenes

[BernardC]

Advertisement (gone after registration)

11 hours ago, SrMi said:

I am not following.

If I expose at ISO 100 using at 1/100 sec, then at ISO 400 I would need to expose at 1/400 sec instead of 1/100 sec, thus reducing the exposure and increasing noise. When using automatic metering, and I need -2 EC at ISO 100, then I will need -2 EC at ISO 400 as well.

What I was trying to say is that if your exposure at ISO 100 is 1/250 at F:5.6, and you apply -2 stops of EC (thus 1/250 F:11), it's the same thing as-if you had metered at ISO 400 and applied no compensation (1/250 F:11). The reason why you wouldn't apply compensation is because you have two extra stops of headroom already.

It's a bit more complicated than that, but not by much. People who use incident metering typically will not only check the main light, but also the lighting ratio. If the ratio gets too high (say the highlights are getting 4 stops more exposure than the main light), you would put-up a few reflectors, add some fill light, use a scrim to tone-down the key, or recompose your shot to minimize the lighting ratio.

11 hours ago, SrMi said:

In SL2, and in most other cameras, ISO is not implemented using DNG metadata but in-camera gain amplification.

Yes and no. Some manufacturers, like Sony, apply very aggressive pre-processing at higher ISOs. Others, especially cine manufacturers, are very transparent about the fact that changing the ISO doesn't change the dynamic range, it just changes the distribution of DR above and below the mid-point. (I'll ignore dual-ISO sensors for now).

Here's an Arri chart that's worth 1,000 words:

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!

Context is provided here.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[SrMi]

2 hours ago, BernardC said:

What I was trying to say is that if your exposure at ISO 100 is 1/250 at F:5.6, and you apply -2 stops of EC (thus 1/250 F:11), it's the same thing as-if you had metered at ISO 400 and applied no compensation (1/250 F:11). The reason why you wouldn't apply compensation is because you have two extra stops of headroom already.

I will rephrase your example, adding some details that I assume are part of the discussion.
The metering says that default exposure at IS100 is 1/250@f/5.6, and you apply -2EC to preserve highlights (no highlight headroom left). The exposure at ISO 100 is now 1/250@f/11. You will clip the highlights if you raise the ISO to 400 and use exposure 1/250@f/11.

Are you thinking of a different situation?

[BernardC]

19 hours ago, SrMi said:

The metering says that default exposure at IS100 is 1/250@f/5.6, and you apply -2EC to preserve highlights (no highlight headroom left). The exposure at ISO 100 is now 1/250@f/11. You will clip the highlights if you raise the ISO to 400 and use exposure 1/250@f/11.

The Arri diagram answers this much more clearly than I can. It shows 6.6 stops of headroom (above 18% gray) at EI 400, and 5.6 stops at 200. Presumably it would show 4.6 at EI 100, if the Amira went that low.

I am unsure which part isn't clear. If you meter at 100, you get four and a half stops of headroom. In our example, that's not enough, so we compensate the exposure by two stops to get six and a half stops of headroom (and two stops less of "legroom"). That's the same thing as starting at EI 400, and not applying any compensation because we haven't exceeded the available 6.6 stops of headroom!

My original point is this: there isn't that much real-world difference between a "base" EI of 100 and 400 (or 800). Using the lowest EI implies that you will apply compensation to keep the highlights from blowing-out. Using a higher "base" EI implies that you won't need compensation, because you already have lots of space above and below middle gray.

• One method provides the most theoretical dynamic range, because your files are always at the threshold of clipping. The downside is the extra effort expended on each shot (on-set and in post).
• The other method provides more consistent files, but it leaves some DR on the table.

[Tinbird]

Of course I realize that my comments here are coming in more than 3-years after the last comment was posted, however, since all of these forum threads are archived and reviewed by many as they study a topic, I thought I'd chime in for future readers studying this subject.

While I don't presume to speak for him/her, I think the argument that @SrMi was advancing was that for still cameras, increasing ISO's lead directly to increasing noise levels, and decreasing dynamic range, and hence, decreasing "image quality." While it's certainly true that "image quality" can be a subjective phrase, the fact that increasing ISO's axiomatically impart increasing noise levels and consequent decreasing fine image detail preservation, is empirically evidenced by scientific laboratory measurements (see charts, linked below). Ditto (but generally perceived much less consciously and much less objectionably) that increasing ISO's axiomatically impart decreasing dynamic range. Note that while the "objectionableness" of noise levels is, indeed, largely a matter of personal perception and personal preference, both lines of resolution (detail preservation) and dynamic range can be (and routinely are) measured empirically, and precisely, in lab environments.

This doesn't impugn or criticize the "organic feel" that film grain (or any visible texture) imparts to, and that many of us enjoy seeing in, a photograph (and yes, even digital image noise -- which, admittedly, looks a whole lot different than silver halide granularity, and is generally less tolerable to many viewers than the "organic feel" of film grain -- still imparts a certain texture to a photograph, which some find pleasing). A photograph is not a painting or a line drawing, and the visible presence of film grain or digital noise in an image can help reinforce a photograph as art...or at least as something other than a pure, "sterile," literal, unimaginative or "voiceless" precisely accurate copy of something seen. There's nothing wrong with any of that, I think most of us will agree.

Notwithstanding, many genre's of photography -- and "modern appetites in photography" more broadly -- clearly reveal a strong (albeit not unanimous) demand for maximum resolution, sharpness, and clarity in photographic images. Some find this trend "soul-less," artificial, "plastic," and uninteresting, and personally, I think that's a valid concern, or complaint (much the same criticism has been leveled at the sterile precision of digital audio). Thankfully, for those who like to see and feel a little texture and some distinct "voicing" in their still photographs, there are many options to satisfy this need, from digital noise, to film grain simulations, paper and canvass simulations, and more.

In summary, and to @SrMi's original point about ISO's (digital noise) harming "Image quality," the charts linked below would seem to empirically confirm that increasing ISO's lead directly and invariably to increasing noise levels and decreasing dynamic range, both of which, as perceived by many today, adversely impact "image quality" (mostly noticeably by obliterating fine image details). Yes, there does exist variation in personal preferences and in personal perceptions as regards the "objectionableness" of any given degree of image sharpness reduction or any given degree of reduction of very fine image detail preservation, but, given the legitimate empirical data, there really can't be much denying that increasing ISO above the manufacturer's "base ISO" WILL ALWAYS increase noise and decrease dynamic range, in digital still images, given the current state of the art in 2024. The "base ISO" for all modern digital cameras is that which the sensor manufacturer determined through laboratory technical testing produces the lowest noise response (technically, the highest S/N ratio). This value also, almost always, produces the highest dynamic range (ISO/Noise Response and DR are inversely proportional, and mostly linear, throughout their range, as the charts linked below confirm). This is how and why manufacturers specify "base ISO" values in the first place. I'm pretty sure this is the point @SrMi was driving home in his/her original post. Again, (1) these comments (and the charts linked below) are directly exclusively at still photography. Video is an entirely different (and perhaps even more complex) subject for a different thread on a different day; and (2) noise response, dynamic range, and other technical variables -- each of which play an important role in a viewers perception and enjoyment of a photograph -- are valued or "weighted" differently by different people. Data is data and feelings are feelings. Photography seems to have one foot firmly planted in each quagmire, and we ignore either at our own peril.

~ Douglas

https://www.photonstophotos.net/Charts/PDR.htm

https://www.dxomark.com/Cameras/Compare/Side-by-side/Leica-SL--Typ-601--versus-Leica-Q2-versus-Leica-SL2-S___1058_1299_1360

[SrMi]

8 hours ago, Tinbird said:

Of course I realize that my comments here are coming in more than 3-years after the last comment was posted, however, since all of these forum threads are archived and reviewed by many as they study a topic, I thought I'd chime in for future readers studying this subject.

While I don't presume to speak for him/her, I think the argument that @SrMi was advancing was that for still cameras, increasing ISO's lead directly to increasing noise levels, and decreasing dynamic range, and hence, decreasing "image quality." While it's certainly true that "image quality" can be a subjective phrase, the fact that increasing ISO's axiomatically impart increasing noise levels and consequent decreasing fine image detail preservation, is empirically evidenced by scientific laboratory measurements (see charts, linked below). Ditto (but generally perceived much less consciously and much less objectionably) that increasing ISO's axiomatically impart decreasing dynamic range. Note that while the "objectionableness" of noise levels is, indeed, largely a matter of personal perception and personal preference, both lines of resolution (detail preservation) and dynamic range can be (and routinely are) measured empirically, and precisely, in lab environments.

This doesn't impugn or criticize the "organic feel" that film grain (or any visible texture) imparts to, and that many of us enjoy seeing in, a photograph (and yes, even digital image noise -- which, admittedly, looks a whole lot different than silver halide granularity, and is generally less tolerable to many viewers than the "organic feel" of film grain -- still imparts a certain texture to a photograph, which some find pleasing). A photograph is not a painting or a line drawing, and the visible presence of film grain or digital noise in an image can help reinforce a photograph as art...or at least as something other than a pure, "sterile," literal, unimaginative or "voiceless" precisely accurate copy of something seen. There's nothing wrong with any of that, I think most of us will agree.

Notwithstanding, many genre's of photography -- and "modern appetites in photography" more broadly -- clearly reveal a strong (albeit not unanimous) demand for maximum resolution, sharpness, and clarity in photographic images. Some find this trend "soul-less," artificial, "plastic," and uninteresting, and personally, I think that's a valid concern, or complaint (much the same criticism has been leveled at the sterile precision of digital audio). Thankfully, for those who like to see and feel a little texture and some distinct "voicing" in their still photographs, there are many options to satisfy this need, from digital noise, to film grain simulations, paper and canvass simulations, and more.

In summary, and to @SrMi's original point about ISO's (digital noise) harming "Image quality," the charts linked below would seem to empirically confirm that increasing ISO's lead directly and invariably to increasing noise levels and decreasing dynamic range, both of which, as perceived by many today, adversely impact "image quality" (mostly noticeably by obliterating fine image details). Yes, there does exist variation in personal preferences and in personal perceptions as regards the "objectionableness" of any given degree of image sharpness reduction or any given degree of reduction of very fine image detail preservation, but, given the legitimate empirical data, there really can't be much denying that increasing ISO above the manufacturer's "base ISO" WILL ALWAYS increase noise and decrease dynamic range, in digital still images, given the current state of the art in 2024. The "base ISO" for all modern digital cameras is that which the sensor manufacturer determined through laboratory technical testing produces the lowest noise response (technically, the highest S/N ratio). This value also, almost always, produces the highest dynamic range (ISO/Noise Response and DR are inversely proportional, and mostly linear, throughout their range, as the charts linked below confirm). This is how and why manufacturers specify "base ISO" values in the first place. I'm pretty sure this is the point @SrMi was driving home in his/her original post. Again, (1) these comments (and the charts linked below) are directly exclusively at still photography. Video is an entirely different (and perhaps even more complex) subject for a different thread on a different day; and (2) noise response, dynamic range, and other technical variables -- each of which play an important role in a viewers perception and enjoyment of a photograph -- are valued or "weighted" differently by different people. Data is data and feelings are feelings. Photography seems to have one foot firmly planted in each quagmire, and we ignore either at our own peril.

~ Douglas

https://www.photonstophotos.net/Charts/PDR.htm

https://www.dxomark.com/Cameras/Compare/Side-by-side/Leica-SL--Typ-601--versus-Leica-Q2-versus-Leica-SL2-S___1058_1299_1360

To clarify: increasing ISO alone decreases noise unless the sensor is ISO invariant, in which case it has no influence on noise. Noise is determined by exposure, not by ISO.

However ...

Increasing ISO limits the maximum possible exposure and therefore reduces the max possible DR and least possible noise. Also, when coupled with automatic metering, increasing ISO will decrease exposure and thus increase noise.

[Tinbird]

2 hours ago, SrMi said:

To clarify: increasing ISO alone decreases noise unless the sensor is ISO invariant, in which case it has no influence on noise. Noise is determined by exposure, not by ISO.

However ...

Increasing ISO limits the maximum possible exposure and therefore reduces the max possible DR and least possible noise. Also, when coupled with automatic metering, increasing ISO will decrease exposure and thus increase noise.

Hi, SrMi. Yes, I did read the excellent FastRaw Viewer/Raw Digger article you linked previously, and it was quite interesting (I also use both of the aforementioned software titles from them). Thanks for the comments and for your valued insights into these fascinating technical topics. Peace and blessings wherever you are, and of course, happy image-making!

[jaapv]

Advertisement (gone after registration)

Quite interesting technical discussion, but the end question is of course how this works out in practice. My pragmatical solution is to display the histogram in the EVF and use EV compensation to get as full a curve as possible, keeping an eye on the righthand side for clipped highlights and the zebra pattern for the location of those patterns - some highlights may be blown without impacting the image.
As a side remark: The fact that many sensors are at least partly invariant and the advent of dual base ISO makes the theoretical approach somewhat more convoluted. 

[jrp]

I tried highlight weighted for a while, until I discovered that it underexposed more than it needed to to protect the highlights. Now I just use the zebras 🦓 and multi meter. 

[jaapv]

I quite agree - I think Leica aimed this feature at the less experienced photographer - after decades of using slide film and digital photography since 2004 it would be a sad thing if I were unable to handle such a photographic basic.

[hdmesa]

3 hours ago, jrp said:

I tried highlight weighted for a while, until I discovered that it underexposed more than it needed to to protect the highlights. Now I just use the zebras 🦓 and multi meter. 

 

2 hours ago, jaapv said:

I quite agree - I think Leica aimed this feature at the less experienced photographer - after decades of using slide film and digital photography since 2004 it would be a sad thing if I were unable to handle such a photographic basic.

It varies greatly depending on the scene contrast and the ratio of highlights to non-highlights. For example, shooting a landscape with a thunderstorm taken right after sunset — but the sun is still hitting a small part of the anvil of the storm up at the top. That often requires -2 or more stops of EC to keep that portion of the image from blowing. So Leica’s Highlight Weighted is still an evaluative mode. It reads the scene and determines how important a given highlight area is based on proprietary logic. Nikon is the opposite: their version of highlight weighted is only looking to save the highlights and at all costs to darkening the rest of the image, yet it’s smart enough to still ignore specular highlights. While more aggressive, this is much more predictable for knowing what EC to apply and having it fluctuate less as the camera composition changes slightly.

Edited December 26, 2024 by hdmesa

[jaapv]

The method using zebra stripes and histogram or spot metering and histogram is far more precise. I should hope that a photographer can outthink the peanut brain of the camera. 

[hdmesa]

16 hours ago, jaapv said:

The method using zebra stripes and histogram or spot metering and histogram is far more precise. I should hope that a photographer can outthink the peanut brain of the camera. 

But what if I'm the one with the peanut brain and not the camera? 😂

• Zebras are video, right? For stills it's just that crazy flashing highlight warning. It's either so large in the image and oppressive that it hurts my eyeballs, or it's so small in the image as to not be noticeable. I have to hunt around the image to make sure I got all the flashing areas to stop.
• Leica's histogram is of no use to me since it's an average. It can't be trusted. To be useful for me, I need a live view RGB histogram (Canon does it). 
• Spot metering makes sense if I have 8x10 view camera time on my hands I guess.

All these things are either imprecise, impractical, or both (for me). I get acceptable results by knowing what the exposure preview in the EVF looks like compared to how the RAW histogram will look in RawDigger. So yeah, I've gotta use my peanut brain. But I wouldn't mind having an RGB histogram as a backstop to experience and intuition.

[jaapv]

Blinkies if you will. A rose by any other name…

 

Hmm. I think my pea brain plus histogram wins by the results. 
The exposure preview in the LCD is quite unreliable as your eye adapts to the brightness of the ambient light and needs time to adjust to the brightness of the EVF. Whatever else, I think that regardless of method experience makes correct exposure instinctive. 

[SrMi]

Blinkies are key elements for me when setting exposure. Yes, they are intruding, but exposure is typically not good if large blinky areas are visible in the frame. Also, the histogram is unavailable with LPC on, but blinkies are.

I cannot judge an exposure by the brightness of the EVF.

The key is to avoid clipped highlights, which give digital images an ugly look. The beauty of negative film is that it does not have clipped highlights. Clipped blacks are OK.

[hdmesa]

4 minutes ago, SrMi said:

Blinkies are key elements for me when setting exposure. Yes, they are intruding, but exposure is typically not good if large blinky areas are visible in the frame. Also, the histogram is unavailable with LPC on, but blinkies are.

I cannot judge an exposure by the brightness of the EVF.

The key is to avoid clipped highlights, which give digital images an ugly look. The beauty of negative film is that it does not have clipped highlights. Clipped blacks are OK.

I guess I've gotten really good at judging the EVF image. I don't find it difficult to visually judge highlights in the EVF at the default brightness setting. The SL EVFs are so good. I just use negative EC until I see the detail fill in on the highlights. Once I see the full detail, I can bounce back and forth between that exposure and 1/3 stop higher and see the faintest highlight detail disappear and reappear, which confirms where the clipping is happening. Well, at least it tells me where the clipping is for the JPEG. I usually bump up the exposure 1/3 to 2/3 stops over that to get a true ETTR histogram in RawDigger. So basically I get one exposure that's true to the JPEG clipping, then I bracket up higher to get a slightly more efficient and saturated exposure for use with the DNG in post.

Regarding blinkies, can that be mapped to on/off via a function button? That would be ideal for me versus having to hop through the info screens to turn it on when I need it. 

[SrMi]

1 hour ago, hdmesa said:

Regarding blinkies, can that be mapped to on/off via a function button? That would be ideal for me versus having to hop through the info screens to turn it on when I need it. 

Yes, the "Clipping" function can be assigned to a button (SL-3).

 

Edited December 27, 2024 by SrMi

[hansvons]

On 12/24/2024 at 8:45 PM, Tinbird said:

Video is an entirely different (and perhaps even more complex) subject for a different thread on a different day; and (2) noise response, dynamic range, and other technical variables -- each of which play an important role in a viewers perception and enjoyment of a photograph -- are valued or "weighted" differently by different people. Data is data and feelings are feelings.

Well thought-out post, Doug. There‘s a reason why the terms ISO, base ISO, invariance, texture, highlight-weighted exposure, etc. pop up and stir controversy. It’s not an issue of education or the lack thereof. It’s a matter of how you were "raised". I grew up in news gathering and later cinematography, and now enjoy shooting stills photos. Others had long careers in stills photography and stepped later their toes into video (which is different from cinematography, even if digitally acquired). 
The biggest difference is probably how you look at exposure and what your goals are (that‘s where data vs feelings kick in). In cinematography, the human skin is the measure for everything as filmmaking is mostly about shooting environmental portraits. Add to that the desire to provide leeway as a shot is always a sequence within a changing environment, qand you‘ll end up at the Arri chart in @BernardC post above because as @SrMi rightfully brought up that cl,ipped nightlight must be avoided at all costs, especially on faces and clouds (that‘s why the race for the high dynamic range in cine cameras tops the race for sensor resolution). 
 

Then the issue of noise vs clean vs texture. Again feelings vs data. Many still photographers crave for the perfect picture which can be seen objectively from a technical standpoint. No clipping highlights, no noise in the shadows, perfect colour separation, sharpness from corner to corner or perfectly rounded bokeh ball, the highest possible resolution,  you name it. Other couldn‘t care less and want character (whatever that is), imperfection, mood, you get the idea. The latter have a similar view on image making as cinematographers because their job is precisely this: supporting with their images  a director‘s vision of a film‘s story. It rarely about photographic perfection. It’s mostly about mood. And that entails texture.

Knowing that cinematographers shooting digitally crave for organic texture (it gives the eye something to hold on regardless of sharpness) and thus tend to expose at the edge of what their equipment can deliver (and there are other reason for that, such as moody lighting), Arri incorporated the option in their latest model to select specific textures that are baked into the camera files (at least the Log-C ProRes444 files that 90% of productions use). 

I‘m sure the controversy about the "right" exposure will never be settled as the goals can be highly different. However, in technical terms, at a given scene and environment, we all agree that the lowest possible ISO setting that is not clipping the  whites will deliver the cleanest picture. Acknowledging that for many clipped highlights are understood as a failure, Leica added the option to measure exposure highlight-weighted. In real-world scenarios  the same can be achieved by adding 2 stops of leeway and chose a higher ISO setting or dial down EV compensation by 2 stops. This is the exact same thing if a sensor is in its invariant range. That‘s why some people say the base ISO is the lowest number that doesn’t cripple DR, others look at that from their daily work and have it 2 stops higher. Here, the controversy may start again 😉. Cheers!!