M246 weak in the highlight range?

[wolfloid]

Advertisement (gone after registration)

I'm on the verge of buying a Leica M246. Almost everything I shoot ends up in B&W and anyway, I have other cameras for colour. I have a decent set of M lenses and shot M6s for years. However, I am held back by what I see and what I have read about poor handling of highlights.

I have viewed hundreds of images online, and what I see (when it is relevant) is very abrupt, and to my eyes, ugly highlight falloff. I do know that it is possible to avoid blown highlights by under exposing, relying on the ability to pull up detail in the shadows, but that is not really the issue, since even when the highlights are 'saved' the falloff seems very abrupt. By far the bast work I've seen is in the dull light of Iceland, or north Norway, where the highlights are snow, and exposure is obviously tweaked to keep the detail there. However, because of the duller light, the shadows are a) either less important - serving as a contrast to the white snow, or b) easier to pull up since they are not that dark given the lighting range. Images I've seen from brighter, sunnier places often have dramatic highlight fall off, which is very unattractive, if not blown highlights, where keeping some detail would add to the image.

When I look back at my B&W film images taken over many years in the tropics, I see very few problems with highlights, unless contrast is extreme and I need to preserve mid-tone values. Colour filters may change the tonal relationships, but they do not change the overall contrast range. So I don't see colour filters as a solution to the seemingly inherent limitations of the monochrom 246 sensor. Can anyone explain  how they have overcome this, and preferably demonstrate long highlight falloff in bright regions with the M246?

A second question: is it not possible for Leica to obtain a better sensor (broader dynamic range) and build in a film like S curve?

[a.noctilux]

Hello,

What you had observed is what I observe in M246 or Monochrom files.

Some ideas as Monochrom user, not to try to convince you to buy one Monochrom of course...

- to "preserve highlight", I use my Monochroms mostly manual settings, as I know which part of the scene "would blow out"

- experiments are the only way to "learn to take the most out of the gear", and Monochrom need a long learning curve

- I learned to adapt lenses I choose to the "contrast rendering" of their sensors = less contrastting lenses like older Leitz/Leica in M or in LTM mount

( some are 70 years old, non coated 😇 )

- I learn to use colored filters (still learning) and I appreciate, happy to be able to view the "direct changing in contrast with each filter" with LV of M246 (not on MM1 that I always use)

- post processing can be long to "preserve highlight" (and I think that bother less some people using Monochrom), so I see and understand why most pictures showed on line "always has some part blowing out"

- I remember that Monochom is a kind of monochrom slides, so as I was fond of Kodachrome (time to re- view those nice slides, for me), I had already the basic to master "mono-slides"

(at some point I dreamed of Agfa Scala that I've never use), and the behaviors of Scala transfered to Leica Monochrom nice files, happy ending for the dream 😉.

...

- and in the end, when I want the nice S curve of film, I just use film that is as fun anyway

 

[Advertisement]

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

[Overgaard]

I have the same observation and wrote about it some time back when the M 246 was new:

M246: https://www.overgaard.dk/leica-M-Monochrom-Type-246-Digital-Rangefinder-Camera-black-and-white-sensor-page-29A-The-Leica-M246-Dynamic-Range.html

I would look at using a M10-P for black and white (convert the DNG), go back to the excellent Leica M Monochrom (Leica M9 edition) or wait patiently for the M10-P Monochrom (with or without screen).

If you look at the images from the Leica M Monochrom here, I feel you can see an overall difference in tonality and soul of the camera:

MM: https://www.overgaard.dk/leica-M-Monochrom-Henri-Digital-Rangefinder-Camera-black-and-white-sensor-page-21.html

I would expect a future Monochrom to be an actual monochrome camera, not a conversion of a color sensor to black and white. 

The M246 seems to be at the end of its product life (not online at the USA Leica Online Store anymore), and it's time for a monochrome version based on the M10-P (as Leica 'usually' release a Monochrom version around this time (give or take 6 months) of the product cycle. Though many parts not ready yet of a camera can delay the release). 

 

Edited May 31, 2019 by Overgaard

[Jager]

What you observe with the rapid (and ugly) rolloff of highlights certainly isn't limited to the M246... it is a characteristic of all digital sensors.  (And I'm going to gently disagree with Thorsten's inference that the M246 is qualitatively worse in that regard than the original Monochrom).  Digital sensors clip suddenly.  They don't exhibit the forgiving toe that most negative film emulsions do.

They, in fact, behave very much like slide film... which likewise have the highlights blowing up suddenly, losing all detail.

It's not an issue of having insufficient dynamic range, a characteristic which speaks to the range of tonality that can be expressed.  Current generation digital sensors generally have greater dynamic range than did their predecessors, but they clip highlights in exactly the same way.

There's an oft-repeated notion that a conventional color digital camera, a la an M10, somehow handles highlights better.  They don't, of course.  But with their three color channels there's sometimes some remaining data in one or two of the channels after one channel has clipped... which leads to the myth.

Manufacturers will often bias digital camera exposure systems to underexpose by some amount, to prevent users from inadvertently walking over the highlight cliff.  The real answer, of course, is to meter knowingly and with precision.

Just like you had to do with slide film, back in the day.

 

 

 

[wolfloid]

Thanks for the replies. I'm not convinced it has nothing to do with dynamic range. Surely, as you say, extending the 'range' of tones would also, at least theoretically, extend the range of highlight tones, and it is that which seems to be missing. Can no one devise a highlight filter or algorithm to control the rate that extra photons (above a certain quantity/rate) hit the sensor?

[a.noctilux]

While waiting for "algorithm highlight control" (which can never come), I learn to use colored filter to "boost" some contrast in different colored scene.

In Monochrom, as in film day, red (or yellow/orange/green, etc.) filter can do (well er...sometimes ! ) magic if used correctly.

[Overgaard]

Advertisement (gone after registration)

Orange filter may be a good idea. I sent y MM to Leica for adjustment recently as I looked at some older pictures and decided I wanted to use it again. Back when the M246 came out I didn't realize the difference between the two, but in retrospect, looking at the pictures they produced made me want to utilize the MM again.There's a crispness and detail to the MM pictures - when you hit it just right -  that the M9, M240 and M10 doesn't have.

https://www.overgaard.dk/leica-M-Monochrom-Henri-Digital-Rangefinder-Camera-black-and-white-sensor-page-21.html

Edited June 2, 2019 by Overgaard

[EoinC]

Generally the M246 is very good at recovering shadow detail. Underexposing sufficiently as to bring the highlights below ‘blocky’ clipping, I find works very well.

[jankap]

My Sony has the following exposure mode:

Highlight:

Measures the brightness while emphasizing the highlighted area on the screen. This mode is suitable for shooting the subject while avoiding overexposure.

Edited June 5, 2019 by jankap

[adan]

On 6/1/2019 at 3:15 PM, wolfloid said:

Thanks for the replies. I'm not convinced it has nothing to do with dynamic range. Surely, as you say, extending the 'range' of tones would also, at least theoretically, extend the range of highlight tones, and it is that which seems to be missing.

You are assuming that the DR is equally distributed (or "should be equally distributed") around "middle gray"- which is not the case. The DR extension can be quite asymmetrical, with proportionally more range at the shadow end.

Not totally useful, yet, but I've been exploring the M10 tone curves, and happen to have had a chance to compare the M240 curves (upon which the M246 sensor is based). Would love to get an MM or M9 (or even M8) CCD for comparison, but have not run across one to borrow, yet.

Anyway, here is how the tone curves map out for M240, M10 and (for reference) Canon 6D - the DR is distributed about 4 stops on the highlight side of middle gray, and about 7.5 stops on the shadow side of middle gray (based on metering for a gray card with each camera - heavy blue line). Not symmetrical at all, but about a 2:1 ratio - and that is "normal" for all cameras I've checked so far. Specific to blown highlights, the M240 gets brighter faster up to a point, but rolls off a bit more slowly than the M10 (doesn't crash through 255 like a rocket as much). Both actually "blow" medium gray identically when it is 4 stops overexposed. Both have the same total DR of 8+4=12 stops±.

The M240 renders medium gray about 0.5 stop lighter, however

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!

 

[a.noctilux]

Thanks Andy, it is usefull for futur reference.

I see here that same solution is adopted by Canon and Leica to "resolve" same problem.

I see that is very clever approch indeed with asymetrical "curve".

I thought that the "curve" was straight, but in real (my) use not "so straight", so this is explaination.

Sort of "film curve" that the OP asked for in "next" software.

[wolfloid]

10 hours ago, EoinC said:

Generally the M246 is very good at recovering shadow detail. Underexposing sufficiently as to bring the highlights below ‘blocky’ clipping, I find works very well.

Yes, yours are some of the better results I've seen, and give me some hope.

[wolfloid]

Thank you Andy! Very educational.

[wolfloid]

 

6 hours ago, a.noctilux said:

Sort of "film curve" that the OP asked for in "next" software.

 

Yes, it certainly looks like it, or are we deceiving ourselves? I too thought the digital sensor response to light was a straight line. Is there any way to compare this to shooting 400 ISO film?

[adan]

On 6/5/2019 at 10:36 AM, wolfloid said:

I too thought the digital sensor response to light was a straight line.

Generally it is, straight off the sensor.

1 photon in ≈ 1 electron, or pico/nano/whatevercoulomb of charge, or pico/nano/whatevervolt of electrical potential out. But that would really look screwy to us humans used to 190 years of seeing non-linear silver-based photographs. So in the initial digitizing and processing of the image between sensor and SD card, an S-shaped tone curve is mapped onto the linear output, to make the result look more normal to our eyes.

It does attempt to replicate the logarithmic S-shape ("toe," "straight-line" and "shoulder") of the "Hurter-Driffield" curves of processed film, in a generic way (not replicating any particular film - although Fuji's "Film Type" settings may do that): https://en.wikipedia.org/wiki/Sensitometry

Those "built-in" firmware curves (and differences between them from various cameras, especially other digital Ms) are exactly what I am attempting to visualize in my own test. To get a handle on how different (if at all) the "dark, contrasty" M10 curve is, compared to others. A camera-maker has "artistic license" to play with the tone curves it programs into the camera firmware (thus Leica changed the curve mathematics, going from the M(240) to the M10).

When you make curve or other exposure adjustments in PhotoShop/LR/ACR/whatever - your curve-adjustments are on top of the baked-in "looks like a photograph" curve from the camera's firmware.

[pico]

28 minutes ago, adan said:

When you make curve or other exposure adjustments in PhotoShop/LR/ACR/whatever - your curve-adjustments are on top of the baked-in "looks like a photograph" curve from the camera's firmware.

Adan, is there any way to over-ride the baked-in parameters?

(I'd drive the 1,000 miles if you gave a workshop.)

[Magnus_L_Andersson]

adan,

very interesting information, but a few questions:

1) How do you find out the presented values? Do you have access to the actual firmware code, or by decoding the dng files for a well defined shot?

2) According to Erwin Putz, the MM ACR import looks very much like Ilford Delta 100 if I’m not wrong. Then it would mean that “Adobe Monochrome” + “your curve in case of MM” = “Ilford Delta 100”, or “your curve in case of MM” = “Ilford Delta 100” - “Adobe Monochrome” assuming monotone curves and that none or in combination blow highlights or turns black. Any take on this?

/magnus 

Edited June 6, 2019 by Magnus_L_Andersson

[Magnus_L_Andersson]

pico, assuming you have the baked-in curve it’s no problem to neutralise in PS or LR given a few prerequisites on the actual shot like assuming the curve doesn’t blow highlights or turn things black that aren’t suppose to be so.

See comment 2) in previous post that in a way describes the way to address it. You create the inverse of the baked-in curve and add it to the inverse of the default profile used by ACR for the camera and save it as a LUT that is used as a profile when importing. However, I’m just familiar with MM and don’t know about any other camera.

I have started to write about my way of achieving this kind of thing at my new and far from ready site under the topic of image processing at photography.anderssoneklund.se 

Edited June 6, 2019 by Magnus_L_Andersson

[adan]

6 hours ago, Magnus_L_Andersson said:

1) How do you find out the presented values? Do you have access to the actual firmware code, or by decoding the dng files for a well defined shot?

I photographed the target setup below, metering off only the standard 18%-reflectance gray card in the center. 50mm Summicron for the M cameras, 50mm f/1.8 for the Canon 6D. I included color patches "just in case" I want to explore color clipping or blown channels in specific colors - haven't done that analysis yet.

Metering was with the internal camera meters, cross-checked against a hand-held Sekonic reflected meter also reading off the gray card. (The camera meters all matched my Sekonic, BTW)

I then bracketed from that metered exposure from -9 stops underexposed to +5 stops overexposed in half-stops, by changing shutter speed. That gave me 30 pictures ranging from extremely underexposed to extremely overexposed. I discarded those that were "beyond clipped," leaving me with 25 sample pictures for each camera in which the gray card value was between 0 and 255.

I opened all the pictures from each camera sequentially in Adobe Camera Raw (target output space Adobe RGB 1998) with the default ACR exposure settings (that is, exposure, contrast, highlights, shadows, black, whites, clarity all set to "0,") and the default linear curve. I desaturated the pictures temporarily to get just the "gray" value (R=G=B).

I then just slid my cursor around over the center of the gray card to read the card's value for each image, noting small texture or noise variations of 2-4 points on the 0-255 digital 8-bit scale, and finding the "average brightness value" for the center of the card, which number I recorded.

E.G. from the straight metered exposure, I got 120 from the M10 and 122 from the 6D and 151 in the M(240) series. At 2 stops under, the gray card, uncorrected in processing, came out as a dark gray with 8-bit values of 46 for the M10 and 49 from the 6D and 62 for the M(240) - and so on for all 25 exposures from each camera.

I then just graphed those values on the vertical axis, and the exposure ± on the X axis (across the bottom), at 1/2-stop per unit.

Pretty much an analog of how film curves are measured - bracket exposures of a gray card to get a series of negs from "blank" film to DMax, and then measure the density of the series of negatives with a densitometer, and graph those densities against the exposures (or more typically, the logarithms (Log E) of the exposures).

https://www.kodak.com/uploadedfiles/motion/US_plugins_acrobat_en_motion_education_sensitometry_workbook.pdf

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!

[Advertisement]

Hallo,

Look here: Leica M

[adan]

7 hours ago, pico said:

Adan, is there any way to over-ride the baked-in parameters?

would connect

It probably can be neutralized, more or less as Magnus suggests, and also "more or less" in the general sense, by applying a tone curve in LR/ACR/PhotoShop/etc. that is the inverse of the curve I showed above.

As an extremely rough approximation of this, using the M10 curve I graphed previously:

1) flip the S-curve not vertically or horizontally, but across the "linear graph" curve of very approximately 45° slope that would connect black and white with a straight line. Such as (purple is the inverse curve, black the "linear output desired"):

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!

2) redraw that purple inverse curve in the Tone Curves section of your processing program, which might produce a result like this (original, and with added "inverse" tone curve). Basically, lightening the lower 80% of the tones as much as the M10's native curve darkens them, and darkening the upper 20% of the tones as much as the M10 native curve lightens them. Thus at least removing the strong mid-tome contrast.

BTW - I'm out on the edge here, so happy to accept critique of either the concept or the execution.