Highlight Clipping

[Mark II]

Advertisement (gone after registration)

But is there any curve at all applied to the DNG files? With the Monochomes, the highlight ceiling is abrupt, with no roll-off until you hit the well saturation point. WIth the colour sensors, the only reason why there is anything recoverable in highlights is because the different colour channels usually saturate at different times (which is why recovery often introduces colour artefacts).

 

The advantage with (negative) film is that to some extent you can just expose for the shadows and let the highlights take care of themselves - they will still be technically overexposed, but the highlight roll-off still preserves substantial information with no hard cut-off. I do not think that any digital sensor today (not even Sony) comes close to the necessary linear range needed to completely emulate this.

 

Ironically I find my Olympus (micro-4/3 gear) better for dynamic range in quick-paced shooting than the Leicas. The Olympus sensors are substantially worse than any Leica, but they have predictable metering that is astonishingly good at making the most of what the sensor provides, no matter how difficult the lighting. I would love to have a full-frame CL equivalent optimised for use with the M series manual focus lenses for street shooting, with accurate metering available without any shutter lag...

[adan]

Yes, there is some curve applied to any raw file. The A/D converter has a "look-up" table in the firmware that can easily be factory-adjusted to delinearize the sensor output values (voltage) when converted to digital values. In theory, that can be 1:1, but in practice it is not. Because we humans have been "trained" through 180 years of seeing silver photographs, to not expect linear reproduction of light in technically-produced images (as opposed to paintings and drawings).

 

The table can easily be "curved" to produce a roll-off. Although of course there will still be a physical saturation point, beyond which a pixel blows.

 

At the core, all imaging closely approximates a linear reproduction: 1x the light input = 1x the density in the negative and 1x the brightness in the final photo. 2x input = 2x density/brightness. 10x input = 10x density/brightness. And for the most part that is why photography "works" - when we see something brighter in a print, we can count on that meaning it was brighter in the real world.

 

If silver grains were all exactly the same size in a given emulsion, film would be just as linear as digital. 5 photons = .0000001g of silver, 10 photons = .0000002g of silver, etc. Film gets its "roll-off" from the fact that film grains are not all the same size, with two effects: 1) smaller grains are a smaller target for any given photon, and thus only actually get "hit" if the incoming brightness/photon flux is high (i.e. in the highlights), and 2) a smaller grain produces less silver from a photon strike than a larger grain, modulating the "highlight signal" more delicately.

 

This shows silver-halide crystals (T-grain/Delta type) in Portra 160 - a lot of variation in the size of the "photo-sites": http://www2.optics.rochester.edu/workgroups/cml/opt307/jidong/160vc1.jpg

 

Fuji tried to approximate that with their Super-CCD sensors, where each pixel had two photo-sites of different sizes - when the large site has "blown", the smaller site is still not saturated, and "touches up" the highlight detail, just as smaller film grains can touch up the highlights when the larger, faster grains are "saturated": https://www.dcresource.com/images/news/superccd_sr.gif

 

At one point Ilford advertised a simliar approach to get even more "latitude" out of FP3/4 - dual-layered emulsions, one with larger faster silver crystals, and one with smaller, less responsive crystals.

 

And of course, even then - try to capture the natural color/tone of the sky (let expose for the shadow detail) and a defined disk of the sun with film. Film highlights can easily "blow," just not quite as soon as digital.

 

With appropriate metering, and thanks to the CA wildfires 1000 miles away, the M10 captured this.

 

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!

 

This, BTW, was a roughly similar situation with film, 40 years ago. Tri-X "exposed for the shadows" for detail in the face. I have to "recover" the sun from more than a faint lighter blur, by putting a coin on the photo paper in the darkroom, or with a photoshopped circle in a scan.

 

Which shouldn't be taken as F vs. D - just as a reminder that things are not always clear-cut.

 

Edited January 5, 2018 by adan

[Advertisement]

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

[pgk]

And FWIW on film extreme highlights were subject to what is known as high intensity reciprocity failure which meant that again somewhat spurious data was produced supplying erroneous information where none should This has conditioned us to expect this sort of result in extreme highlights in photographs.

[pgh]

I'm less interested in why the Leica highlights clip - whether or not it's technically defined as dynamic range or not.

 

All I know is that the Leica sensor contains less image data to work with in total than other normal sensors today, and the ability to either recover slightly blown highlights or to tone down near highlights yields a crappier rendering than other sensors, and this is a practical (not theoretical!) loss for many in many different genres of photography. Whatever you want to call it, a raw file from an M10 has less data to work with for the final print of the image, and is a compromise if you care about being able to work with a broader tonal range within your Raw file. This isn't just an issue with sky/ground landscape work, it comes in to play in many real life situations that have dynamic light. A common one, that is very documentary in nature - is an indoor environmental portrait exposed for the indoor ambient window light, where a window is still in the frame. In other modern sensors you can pull back the blown out windows to a degree where detail outdoors may be relevant, and the transitions to the blown out areas look much better. Shoot that image with the M10 and the blown out backgrounds look...well, like I said, more like renderings from digital cameras of the last generation. There's just less to work with, period. It's not the end of the world, but it's solidly disappointing for such an expensive camera. 

[jaapv]

Could you please quantify "sensor data", whatever that may be, and compare to the values of other sensors?

[pgk]

I'm less interested in why the Leica highlights clip - whether or not it's technically defined as dynamic range or not.

 

If you aren't interested in why the highlights clip then you cannot appreciate the compromises which may have been made in order to make Leica dMs work with their lens line up. If it was as simple as just sotting in a different sensor then that would be great, but as we know using wide M lenses on 'better' sensors is problematic. So there are technical problems which probably have consequences. If you want such problems to be overcome then it means bigger lenses, etc., etc.. For the scenarios you describe you may need a different camera and different lenses. The expectation that everything is possible is unrealistic.

[pico]

Advertisement (gone after registration)

Or maybe with a software update, Leica can give us a metering method that protects highlights. My D750 has such a metering method, where the camera calculates the exposure in a way that it won’t clip the highlights. the standard center weighted exposure calculation can be arranged in this way with a software fix easily.

 

The camera cannot fix what the meter cannot see.

[Mark II]

I am fairly certain that the raw output is almost entirely linear. The only are only a few reasons why this might not be the case.

 

One is to compress the raw file data (which is unlikely to be an issue for a slow frame-rate 24MP camera), and the other is to try to make the raw file look better (for example, to improve DXO scores...). The M240 and 262 series do some of the latter, with some NR at higher ISOs (which should not affect linearity) and some clipping to hide fixed pattern noise (which is supposedly what causes the green shadow tint in extreme edits).

 

I suspect that the curves that you are thinking of are applied only in JPEG output and when processing the raw files in Lightroom etc.

 

FWIW, I have been experimenting with An ISO-less shooting style for night images, to maximise dynamic range. I set the sensor to base ISO and a fixed shutter speed based on the brightest highlight that I care about preserving. Everything else is then lifted in post-processing (3 to 4 stop push). I do not see any non-linearity when doing this, other than the (annoying) green shadow cast (this is on an M262). The resulting images are very slightly noisier than if I had shot at higher iso - but they are also sharper and can take more NR - and of course there is much better highlight retention than a non ISO invariant camera can provide.

[pgk]

FWIW, I have been experimenting with An ISO-less shooting style for night images, to maximise dynamic range. I set the sensor to base ISO and a fixed shutter speed based on the brightest highlight that I care about preserving. Everything else is then lifted in post-processing (3 to 4 stop push). I do not see any non-linearity when doing this, other than the (annoying) green shadow cast (this is on an M262). The resulting images are very slightly noisier than if I had shot at higher iso - but they are also sharper and can take more NR - and of course there is much better highlight retention than a non ISO invariant camera can provide.

 

This is how I have shot all images for several years using my M9s. FWIW I do the same with my Sony A7II and its files appear to respond less well to such treatment than do the M9 files, as I've observed shadow noise increasing more. You can of course make of that what you will .... .

[pgh]

Could you please quantify "sensor data", whatever that may be, and compare to the values of other sensors?

An equivalent exposure on a Nikon or Sony camera will provide more information across the spectrum of total light to total darkness in the image, and the file is more robust in post processing. The scenarios I listed regarding portraiture are where I encounter it most, but it is also applicable in landscapes. I, and many others, prefer files that are lower in contrast with a more neutral rendering of tones across the spectrum. In post, this means often brining up shadows and bringing down highlights a little bit, to achieve a look that may recall color negative film to a certain extent.

 

With Sony and Nikon cameras you can do this more consistently across a wider range of lighting conditions before encountering light that simply clips. With the M10 it is most noticeable the highlights, although the shadow recoverability isn't that great either. It just feels more like an inflexible file, which is how files felt 2 generations of cameras ago.

 

I think dXo compared it to the value of other sensors accurately. I'm just a photographer without a science background. All I know is that with equivalent exposures, highlight data is lost consistently. If you don't know what you're missing, I can see how this doesn't matter - but after the D800 and Sony cameras came out - and after working with these cameras for 5 years, it feels like a step back in technology. If you haven't shot with other cameras I'd encourage you to do so, in dynamic light, to get a sense of what I mean in practice. 

[pgh]

If you aren't interested in why the highlights clip then you cannot appreciate the compromises which may have been made in order to make Leica dMs work with their lens line up. If it was as simple as just sotting in a different sensor then that would be great, but as we know using wide M lenses on 'better' sensors is problematic. So there are technical problems which probably have consequences. If you want such problems to be overcome then it means bigger lenses, etc., etc.. For the scenarios you describe you may need a different camera and different lenses. The expectation that everything is possible is unrealistic.

I am not interested in why in terms of the technology, because it doesn't change the fact that it's an inferior result, as others here have also noted. That's not how a good company operates - making excuses for middling performance. They should fix it, or at the very least be very clear why they will never be able to keep pace with current sensor technology. If Leica cannot solve this problem somehow they should make it known. Right now it looks like they're putting out mediocre sensors - both in terms of resolution and in terms of the range of light they are capable of capturing with detail - in their extremely expensive cameras and we are left to wonder why this is the case. If it is all a matter of compromise we have to wonder why they are compromising much of anything to make such an expensive 35mm format camera. What would it cost to implement better sensor technology? It's not like the Leica fan club is generally a price sensitive lot...

 

Of course I am interested in maximizing technique to reduce this issue. 

[pgk]

An equivalent exposure on a Nikon or Sony camera will provide more information across the spectrum of total light to total darkness in the image, and the file is more robust in post processing.

 

Well this is not my experience. In practice I find that the M9 files are as robust as Sony A7II files for how I shoot. The problem is trying to extrapolate technical information into practical viability. The RAW files from the cameras are already subject to processing. Their 'robustness' to any additional processing depends on you and your processing requirements/workflow (experience and skill). In some cases this will mean that the advantage is with one camera system, in others it will favour a different one. Translating data into practicality is fraught with difficulties because the variance in how we do things is substantial and cannot be taken into account easily.

[pgk]

What would it cost to implement better sensor technology?

Almost certainly it would require a new generation of 'electronic' M lenses which would be able to transfer information to the camera and allow for better integration of optical performance and software adjustment, alongside redesigned wide-angle lenses which would have to be larger. Its not about cost, its about a fundamental change to the system. There are other solutions offered by Leica which do this; they are just not rangefinder based. The easier alternative might be to add an electronic rangefinder to an SL/CL/TL type camera .....

[M11 for me]

I think these experiences with different cameras over many years as a photographer are very interesting to take into consideration.

[M11 for me]

Almost certainly it would require a new generation of 'electronic' M lenses which would be able to transfer information to the camera and allow for better integration of optical performance and software adjustment, alongside redesigned wide-angle lenses which would have to be larger. Its not about cost, its about a fundamental change to the system. There are other solutions offered by Leica which do this; they are just not rangefinder based. The easier alternative might be to add an electronic rangefinder to an SL/CL/TL type camera .....

Are you then saying that the M system is at the end today? There is no doubt that the sensor technology will further progress.

[pgk]

Are you then saying that the M system is at the end today? There is no doubt that the sensor technology will further progress.

 

Why should it be at an end? Its a highly competent camera system which can take stunning images. In the past choices were simpler  - 35mm, 2 1/4" square, large format - and depending on what was required a different choice could be made. Now we all too readily assume that all camera systems are capable of all things, but the M never has been and still isn't. Just because it can't compete in certain aspects of photographic requirements has never stopped its use in the past, why should it now?

 

The real question we should be asking ourselves is whether a camera system we choose is fit for our purposes, not whether it contains the 'ultimate' sensor. I will continue to use the M system because I enjoy using it even if Leica never exceed 24 MPixels and don't substantially change its capabilities. It does what I want and for what it doesn't I have other options which will no doubt bounce along the roller coaster ride of competing for more pixels and everything else - but I don't and probably won't enjoy the experience of using them as much.

[M11 for me]

Why should it be at an end? Its a highly competent camera system which can take stunning images. In the past choices were simpler  - 35mm, 2 1/4" square, large format - and depending on what was required a different choice could be made. Now we all too readily assume that all camera systems are capable of all things, but the M never has been and still isn't. Just because it can't compete in certain aspects of photographic requirements has never stopped its use in the past, why should it now?

 

The real question we should be asking ourselves is whether a camera system we choose is fit for our purposes, not whether it contains the 'ultimate' sensor. I will continue to use the M system because I enjoy using it even if Leica never exceed 24 MPixels and don't substantially change its capabilities. It does what I want and for what it doesn't I have other options which will no doubt bounce along the roller coaster ride of competing for more pixels and everything else - but I don't and probably won't enjoy the experience of using them as much.

I forgot the Smiley in my post.

[robert blu]

Ti each his own of course but I prefer “less data in the sensor” and the sensor to be in an M body than more data in a sensor if this is in a canikon body :-)

It depends on what we need/desire and what we use/buy, we cannot have everything!

robert

Edited January 6, 2018 by robert blu

[adan]

I do think there is a fundamental - misunderstanding - of the purpose and "design brief" for the Leica M digitals (and the M10 especially). Which is, as closely as possible, to replicate a Leica M7, that just happens to have a 24x36 digital sensor inside, instead of film. In size, compatability with 40-year-old lenses built for film use, viewing/focusing system, etc. etc.

 

Anything else is secondary and expendable, if it compromises that core goal. Battery capacity? What will fit in an M7-sized camera. Shutter speeds? Only what will produce no more noise and shake than an M7. Video or time exposures longer than 120 sec.? Not if they require a heat sink and/or a more spacious body to dissipate sensor heat. A sensor with more this or that? Electrical lens/camera connections? Only if it works well with those older lenses.

 

What a Nikon 750/850 can do is irrelevant - unless it will work in an all-metal body the size of an M, also containing an optical RF/VF, and function well with digital-unfriendly compact M lenses, going back 40+ years.

 

When Nikon can put the functionality of a D750/850 into an all-metal SLR body the size and weight of a Nikon FM2, or even a plain-prism Nikon F (their Df failed miserably at doing that) and work with, perhaps, a 1959 21mm f/4 Nikkor lens....

 

http://www.mir.com.my/rb/photography/companies/nikon/nikkoresources/ultrawides/21cm.jpg

 

....then we (and Leica) can worry about whether the Leica sensor holds up in comparison. I strongly suspect Nikon will have to throw out a great many features and capabilities - including perhaps some aspects of image quality - to achieve that.

 

Now, if someone with some serious engineering cred - an advanced degree in EE plus some experience in the industry - can show Leica how to achieve better highlights or some other technical goal without compromising their core design goal, I am sure Leica would love to see a presentation about how to do that.

[Advertisement]

Hallo,

Look here: Leica M