Jump to content

Highlight Clipping


GLC

Recommended Posts

Advertisement (gone after registration)

First let me say I love this camera.  Ive been working with M cameras for almost 20 years, I was very enthusiastic about the arrival of every digital version and very let down in one way or the other with each .. until now.  My only small gripe is that the highlights do go very quickly.  I know how to shoot for the highlights etc etc but.. but id so love a tiny bit more material to work with.  Please understand don't need 15 stops DR.

 

So my question is that is it possible that a future firmware can pull this trick?

 

 

  • Like 2
Link to post
Share on other sites

I don't have an M10 but this was one of my gripes about that change they made and why Auto ISO shouldn't include this lowest ISO. In the LFI article I read, you could clearly see the highlights had clipped in the ISO 100 photo that they were trying to pass off as "handled well".

 

Even the M Typ 240 which I have easily clips if I'm not careful. I have actually learnt to underexpose typically by about 0.7 stops to save some of those highlights, unless I am watching the histogram.

The fact is that ISO 100 is not the native sensitivity and no matter what digital processing they do, it'll never make it real, especially since all digital sensors have this "hard clipping" characteristic.

Link to post
Share on other sites

M240 handles highlights better.

 

M10 handles shadow recovery better.

 

Similar dynamic range between the two.

 

I doubt this can be changed/upgraded via firmware. My only reasoning is that if it was possible I would hope/think that Leica would have done so with the M240??

 

I find both do well enough for most applications.

  • Like 1
Link to post
Share on other sites

Explored 10 months ago: https://www.l-camera-forum.com/topic/268833-iso-100-on-m10/

 

Frankly, my Leicas since 11/2006 and the M8 have all been almost permanently set to -0.7 exposure comp. Not so much because of this sensor or that sensor, but because the regular ttl M metering covers such a narrow area, and thus is easily confused by anything dark (or light) in the center of the picture.

 

Thankfully, the M10 now has easy thumb-wheel exposure comp (and maybe the M240 did, too, but I bypassed that generation) so I can quickly spin in corrections if the meter is staring at something dark or light. but for really fast-paced journalism, -0.7 is still my norm.

  • Like 4
Link to post
Share on other sites

I do it a different way and wonder if some might no laugh at me: When shooting landscape then the sky is probably the lightest part of my picture and by no means I want the highlight being gone. So I measure the sky first and then I overexpose by max. 1½ LVs. In LR the DNG can then easily be fixed right.

  • Like 1
Link to post
Share on other sites

I do it a different way and wonder if some might no laugh at me: When shooting landscape then the sky is probably the lightest part of my picture and by no means I want the highlight being gone. So I measure the sky first and then I overexpose by max. 1½ LVs. In LR the DNG can then easily be fixed right.

 

I guess this is why a graduated ND filter is so useful for landscapes, as pulling up the shadows introduces 'noise'. When you say 'overexpose 1.5LVs' , on what base reading, the general scene as viewed? Do you mean underexpose if you want to hold highlight detail?

Edited by pedaes
Link to post
Share on other sites

Advertisement (gone after registration)

I guess this is why a graduated ND filter is so useful for landscapes, as pulling up the shadows introduces 'noise'. When you say 'overexpose 1.5LVs' , on what base reading, the general scene as viewed? Do you mean underexpose if you want to hold highlight detail?

 

Essentially the same as taking an exposure of a bobcat standing in snow.  This has been preached by Chas Glatzner exposure.https://www.shootthelight.com/Portfolio/1

Edited by richardlipow
Link to post
Share on other sites

I guess this is why a graduated ND filter is so useful for landscapes, as pulling up the shadows introduces 'noise'. When you say 'overexpose 1.5LVs' , on what base reading, the general scene as viewed? Do you mean underexpose if you want to hold highlight detail?

What I do is holding the camera up to thr sky and then I set this exposure right so that the bullet in the ragefinder shines solid without arrows. Only now I set the exposure higer by 3 to 4 clicks either at the time whell or the aperture wheel on the lens. Then I put the camera down again to my subject showing maybe ⅓ sky in my picture.

 

When checking now the result on the small screen and look at the histogram then the exposure is correct or to the right.

 

Why do I do that. I do not want that the dark areas of the picture are too dark. And the sky is still not overexposed.

  • Like 2
Link to post
Share on other sites

I don't have an M10 but this was one of my gripes about that change they made and why Auto ISO shouldn't include this lowest ISO. In the LFI article I read, you could clearly see the highlights had clipped in the ISO 100 photo that they were trying to pass off as "handled well".

 

Even the M Typ 240 which I have easily clips if I'm not careful. I have actually learnt to underexpose typically by about 0.7 stops to save some of those highlights, unless I am watching the histogram.

 

The fact is that ISO 100 is not the native sensitivity and no matter what digital processing they do, it'll never make it real, especially since all digital sensors have this "hard clipping" characteristic.

The point is that highlight clipping is a photographerschoice. Some (specular) highlights should be clipped, some not. Highly subjective.

  • Like 1
Link to post
Share on other sites

I agree that just about all centre-weighted metering on Leicas (from my experience of M9, M240, SL, TL2, CL) causes at least 2/3 stop over-exposure in a landscape scene with sky. Where there is no sky in the frame (and no other bright lights), then I find the metering much more reliable without EC.

 

On the SL and CL it is much easier to adjust exposure on the fly with the histogram and exposure compensation. I find my exposures on the M240 are getting better simply because of the useful feedback I'm getting from the CL (I rarely use the SL for landscapes). 

 

This isn't an equipment problem; I'm just gradually getting better at using my brain as a photographer should.

Edited by LocalHero1953
  • Like 1
Link to post
Share on other sites

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.

Edited by fatihayoglu
Link to post
Share on other sites

Effectively protecting highlights by metering (sort of auto ETTR) requires metering off the sensor, which the M10 only does in Live View. So possible in theory only in that mode. Whether it’s possible in practice is unknown in the M10 system design.

Edited by mdemeyer
Link to post
Share on other sites

So my question is that is it possible that a future firmware can pull this trick?

 

I doubt it. I think that the problem is that whilst there may be some recoverable data in the clipped highlights, it is actually 'spurious' or unpredictable information. So even if you recover it, there may be posterisation and/or unwanted colours creeping into it. If as I believe to be the case, it is spurious, then there is no way of determining what it should actually be, so it cannot be accurately recovered. Certainly recovering clipped highlights is possible in software but banding and false colours creep in, which reinforces the above idea, and they are usually unacceptable.

Link to post
Share on other sites

The Nikon D750 has a separate "sensor" devoted to metering, with 91000 pixels all to itself. So it can take 91000 separate readings of separate parts of a scene, and analyze them to find the highlights, etc.

 

http://imgsv.imaging.nikon.com/lineup/dslr/d4/img/features01/img_33.png

 

(Nikon has been doing segmented (what they call "Matrix") metering going all the way back to the FA, which had just five (as compared to 91000) metering areas, that could be compared and contrasted by a metering CPU. http://imgsv.imaging.nikon.com/history/chronicle/cousins14-e/img/nikon_fa_03.gif - They have come a long way in 30 years. ;) )

 

The Leica digitals with LV/EVF (as mdemeyer says) can do something similar using the imaging CMOS itself. But the normal Leica M metering used with the window viewfinder is still essentially the metering system of a 1986 film M6 - one big "pixel" or photodiode that sees only a circle in the middle of the image, via a lens and the gray reflective stripes on the shutter.

 

It cannot see the whole image, and it cannot differentiate different brightnesses within the metering area it does see. It can only measure one average brightness in the central ~10% of the image.

 

Therefore it has a hardware limitation that no software update can improve. Not without gutting the cameras for a newer, larger meter sensor and wiring, and repainting the entire shutter light gray (which will likely also result in a lot of flare.)

Link to post
Share on other sites

OP, I'm with you. I find all of this talk about metering a bit beside the point. Any sensor will blow highlights if you blow the exposure enough. 

Yes, learning to use the meter on an M is a different experience, but that doesn't change the fact that a properly exposed image on the M in a scene with dynamic light will yield less highlight data than on another camera.

 

The comparisons to film are also a bit limited. Why should we compare a modern camera to 20+ year old technology? The point of comparison is other modern sensors of the same size. In that sense, the sensor on the M falls short when it comes to highlights. I've taken to underexposing images a bit more than I'd like, personally, because at least I can pull up the shadows some. It feels like shooting digital back in 2009 though. For the past 5 years or more I've generally been able to expose properly or even slightly overexpose, where pulling back highlights resulted in less image degradation than pulling up shadows. I, also, am not someone who needs crazy DR most of the time, I rarely shoot landscape work etc, but this stuff applies to highlights in images too, whereas with other cameras the highlights can brought back to have a more subdued, subtle appearance with a more natural looking fall off towards the highlight. This is very applicable for environmental portraiture as well. The modern sensors let one shoot portraiture in dynamic natural lighting situations - and have it look good - that weren't really possible with the older sensors. This comes in to play on a recent magazine shoot where I photograph someone in shade, with sunlit areas in the scene. On another camera, I have the data to bring these two parts of the scene more closely in line with eachother, while keeping that interesting light rendered in a detailed, natural way in all parts of the scene. Situations like this are where the shortcomings are noticeable. You simply can't make as many of these situations look good as much of the time. On the M10 that sunlit area is more quickly blown, and it's hard to pull much back and have it look decent - at least compared with other current sensor options - or you shoot to underexpose your subject and pull it up in post, but then the more important part of the photograph isn't rendered with as high quality as you'd want. 

 

People who say it doesn't matter or one doesn't need it I think miss the point. You pay $6.8k or more USD for this thing, nothing about it should be average, and for the few images it matters, well, it matters. And yes, this is an issue that translates directly in to making better prints, even at smaller sizes.  

Edited by pgh
  • Like 1
Link to post
Share on other sites

It is not the Dynamic Range, but the shoulder falloff of the exposure curve. The more steep it is, and it should be to maximize exposure latitude within the dynamic range, the sooner highlights can be clipped irrecoverably as there will be fewer data left for recovery. 

 

Highlight clipping is not an indicator for dynamic range. It can be abrupt or smooth regardless of the sensor design. Smooth will cost exposure latitude, but it will be less demanding of exposure metering.

A sensor with a small dynamic range may have easily recovered  highlights if the exposure curve flattens out excessively at the end in one or two of the colour channels, a sensor with a huge dynamic range may well have abruptly clipping highlights if the curve remains near-linear throughout.

 

So easily clipped highlights are the price you pay for exposure latitude, commonly referred to as the misnomer dynamic range.

  • Like 3
Link to post
Share on other sites

It is not the Dynamic Range, but the shoulder falloff of the exposure curve.

Maybe I misunderstood what you mean by “exposure curve”, but isn’t the DNG output completely linear? (Ie the raw sensor ADD output - give or take oddities around clipping near the black point - the cause of the green shadows on the M 240 etc).

 

For me, the limited issue with the dynamic range aid not about shadow recovery, but about the highlights when a scene is sub-optimally metered. With film I have immense highlight recovery potential, and the negatives are much more tolerant of over exposure than when shooting digital. This matters most when shooting fast moving street scenes, where there is no time to use careful metering or chimp-and-tweak-exposure techniques.

Link to post
Share on other sites

The DNG output has nothing to do with this; it is about the dynamic range. That can be defined as the curve that represents the sensor response to the number of photons. It runs from the noise floor where no signal but noise can be detected to the fully activated pixel that cannot output more despite getting more light. Mainly linear, the shoulders of the curve are not.

Highlight recovery is a digital trick. A pixel cannot produce more signal than 100%. The exact light/output ratio varies per colour channel. The Sofware reconstructs detail that cannot be rendered in one colour channel from data that can still be gathered in one or two of the remaining channels. Obviously, this reconstruction cannot be perfect for lack of data. The more data, the better the recovery. A monochrome sensor cannot recover a highlight at all, through lack of colour channels.

Negative film is the opposite of digital, there the detail in the highlights will peter out gradually, but the shadows will block abruptly. A sensor behaves like slide film.

  • Like 1
Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...