Digitizing a B&W Negative While Preserving Its Subtlest Textures

Results in images

My intention was to include in this document comparative examples of images obtained in diffuse light versus collimated light. However, on-screen comparison is highly problematic. It is impossible to show through this medium what appears to me as a genuine improvement in the "prints" I have obtained. While it is not razor-sharp, I achieve what I hoped for: a correct rendering of the silver grain of TriX 135 at 400 ISO for an A3+ enlargement, while allowing the application of the best algorithms in Darktable without resulting in a mush of pixels caused by halos.

However, the result on the grain is not an end in itself. It is primarily an alert for textures (skin, fabrics, graininess, feathers, sands). With experience, I have gotten into the habit of post-producing my digital images in order to adapt the processing to the finesse of details and textures according to the subject, print format, and intended use: for optimal sharpness in some cases, for more softness in others. This is a new magical power offered to the digital printer This new power must be tamed, and as a little fox of my acquaintance said, for this "you must be very patient." I remember Tonnie. She had agreed to be my model for portraits as part of Jean-Pierre Sudre's workshop. I eagerly developed the films and, coming out of the darkroom to wash and glaze the contact sheets, I saw Jean-Pierre talking to a very old man. I salute and, out of discretion, continue my work. When the last contact sheet came out of the machine, Jean-Pierre, who was keeping an eye on me, said to his interlocutor, "Since you are here, you will correct Alain's work. You will see, he makes good portraits." And I blushed with pleasure... Then he continued, "Alain, meet Brassaï." And I blushed even more, intimidated and impressed. Brassaï examined the dozen contact sheets intensely. He grabbed the grease pencil and, while complimenting me, drew a crop and added, "This is the one." This was followed by a good hour of exchanges with this delightful man about photography and societal evolutions. A few days later, I found Tonnie at La Chope, Place de la Contrescarpe in Paris. Anxious to see her reaction, I gently opened the 30x40 box, revealing the print to which I had devoted all my care. Immediately, she stood up without a word and left. I never saw her again. I still have this print, which I cherish more than any other. Brassaï liked this portrait. Sudre and I did too. But it hurt Tonnie, to whom I wanted to offer it... I was an ass. Today, I see clearly that the unique sharpness of the Rolleiflex Tessar 3.5 on a FP4 125 ASA and a deliberately "slightly contrasty" print did not do justice to the velvety grain of her skin. I should have made two prints: one for the workshop and one for Tonnie..

Little by little, I learned to marvel at the almost magical ability of the silver grain, when well-defined and visible, to give way to the details of a fabric, of skin. And these details are only visible to the extent that this grain is well-defined and chiseled.

This reminds me of the remark by an expert from the Christie's salesroom, examining a large painting https://www.christies.com/features/Constable-Painting-View-on-the-Stour-near-Dedham-at-Auction-7416-1.aspx (130 cm × 185 cm) by the 19th-century English painter John Constable, depicting a landscape: "If you step back a bit, you can really see the expression on the face [of the little figure on the boat raising the anchor], but if you get closer, it's just a brushstroke, nothing more." What is this phenomenon? I cannot explain it, but it is one of the important, yet little-known, advantages of the silver process.

Initial Results with a White LED

Example 1 - Calque vs. Bare LED, Comparison on a Portion of a 60 × 80 cm

I chose the image below for its dark values, which are precisely those whose rendering in point light is the subject of the main criticisms. There are also light flat areas where the grain is most visible. These three examples present:

1. The raw result with diffuse light by adding a calque 25 mm from the LED,
2. The raw result with point light,
3. The result with point light plus post-processing in Darktable or RawTherapee with the "diffusion or sharpness" module with dematrix sharpness preset (which I've since abandoned).

In the third image:

1. The effect obtained on this 60 × 80 would be very different, and sometimes inappropriate, on different formats.
2. Without the slightest halo effect appearing, the sharpness and contrast are greatly enhanced here. The rendering of the skin texture is much improved, although the grain becomes unpleasant in the dark areas (yes, yes, on a positive the grain is white ;-)

I insist that the images in this document give only a very imperfect idea of the printed result. At this level of demand for rendering, it would be illusory to think that one could evaluate it on-screen in order to print with certainty. Only test prints can decide.

Example 2 - Light Table vs. White LED, Comparison on a Portion of an A3+

The image above on the left was digitized using diffuse light on a small opal plastic light table. On the right is the same image digitized with point light, without local contrast/clarity. The grain is chiseled, regular (too much?), and present. Once hung on the wall, this is the rendering I prefer.

On the Obtained Images

Initially, I was delighted with the progress obtained compared to everything I had seen before, whether with flatbed scanners or diffuse light on a DSLR. But I must admit that little by little, I was bothered by small defects that persisted and whose cause I could not understand or eliminate. I could not accept them:

• There was still this crumbly rendering of the dark areas. It really spoiled the images in which the dark areas played a major role. Image after image, it appeared to me that another difficulty might be attributable to the same elusive cause: the grain became even more unpleasant precisely in areas where the contrast had to be greatly increased.
• In post-processing, I could not obtain all the nuances given by the more or less diffuse light sources used in the past. For certain images, I wanted this "veloute" that is so docile in diffuse light. I was frustrated because I saw clearly that there should be no insurmountable technical impossibility and that only my insufficient mastery of the tools was to blame.

Results with a Green LED

From the First Test with the Green LED, the Result Amazed Me!

These three images below are each a portion of a TriX 135 in 60 × 40.

I have never before obtained a grain with such a precise appearance. Everywhere it is chiseled, sharp, and perfectly dispersed. It even becomes possible to discern differences according to the development process used: whether by myself, by a professional laboratory, or by François Huchet! If only I had found this LED sooner!

Post-Processing is Easier and the Dark Areas are Finally "Normal"

The various modules in Darktable/RawTherapee seem easier to use. The artifacts that appeared very early when increasing the contrast now manifest later.

Finally, the dark areas are no longer "lumpy." They now appear as they should be, without discontinuity in the texture (except if carelessly clipped, of course), and without requiring specific treatment.

And What About the Elusive "veloute"?

I have not forgotten my objective of being able to obtain a rendering of the "diffuse light" type with just a touch of diffusion finely dosed in post-processing, while altering the precision of the textures as little as possible. Whether to attenuate the presence of the grain or to obtain this delicious "veloute" that photographs in medium or large format sometimes offer, I admit that I still did not achieve the expected results...

It was Aurélien Pierre who came to the rescue, giving me three decisive tips for a very specific configuration of Ansel/Darktable and its use in such a context. Thank you, Aurélien!

• In Preferences/Processing/Pixel interpolation select: bicubic.

• Optimize use of green light only Aurélien Pierre : "[Configured this way] if you scan under quasi-monochromatic green light, the sharpness of your scan will be maximum because only the green photosites of the camera sensor will be used. In practice, this is equivalent to completely removing trichromy from the graphic pipeline” :

  • In the Dematrixing module, change the PPG method (default) to VGN4.
  • Download the IdentityRGB-elle-V2-g10.icc profile https://github.com/ellelstone/elles_icc_profiles/blob/master/profiles/IdentityRGB-elle-V2-g10.icc and place it in
    Windows - C:\Users\[YourName]\AppDataLocal\Ansel\colorin\ and
    Linux: ~./config/ansel/color/in/ and /out/

  • In the Input Color Profile module, select this new profile as input color profile AND working profile.

  • In the Color Calibration module, CAT tab, select Adaptation None and in the Gray tab, set the Green channel to 1, and the other two (R and B) to 0.

• In the Diffusion or Sharpness module, create a “Grain attenuation” preset:
  - Itérations : 10
  - Center radius: 0 px
  -Radius range: 186 px
  - 1st order speed: + 1 %
  - 4th order speed: + 10 %
  - SEdge sensitivity: 2,00 ou 1,20
• Compensate for the loss of contrast by a new Diffusion or Sharpness instance with the "Add local contrast" preset, or with the rvb Color Balance module.

Thus, it becomes easy to adjust the number of iterations in the two instances according to the result obtained on paper. I insist again: on-screen, it is impossible to evaluate such subtleties that vary according to the enlargement ratio.

In this way, the possibilities are endless. I finally have the magical tool that allows me, as I hoped, from a single digitization, to emulate all kinds of light sources, adapting to all enlargement ratios, to all subjects... I love it!

After printing a few images, I can finally draw some conclusions:

• In general, I prefer prints without grain attenuation.
• In small format, 15 × 20 for example, the grain is barely perceptible, but the details are more precise.
• In A3+, the grain becomes obvious if the image is held in hand. On a wall, it no longer exists. However, I encountered a case where the grain interfered unpleasantly with the wrinkles of a small face in the image (30 mm). Grain attenuation fulfilled its role perfectly.
• But for those who prefer more softness, I finally have the tool to satisfy them in all formats.

Yet, a Problem Remains

A few days later, I was no longer obtaining the same results. The grain appeared crumbly, pasty, and ugly again, and I could not explain why. Many times, I redid all the settings, without success. By checking the dates of my last RAW files, I ended up realizing that this brutal degradation had occurred after a change in the negative format. This seemed to indicate a setting problem. Yet, I had just redone them several times, with more care than ever! I then questioned the focusing process again, despite the great ease obtained thanks to the use of Live View directly on the DSLR screen. I dreamed of the assurance provided by the Scoponet https://objectif-argentique.com/le-scoponet/, unfortunately unusable in our case. Unable to find a solution, I was disheartened. My prototype was no longer of any use to me...

⊕ A ruler and a toothpick Then, while walking in the woods, a simple idea finally came to me:

• Add a millimeter graduation on the column of the enlarger and stick a toothpick in front of it,
• Make the focusing as precise as possible on the DSLR screen and note the position on the graduation.
• Do a first test. Then, then taking the toothpick as a reference, do additional tests by shifting the focus, millimeter by millimeter, up and down.

Big surprise! That's where the problem lies.

How could such a situation have occurred? It is in total contradiction with what I have often read online, stating that in point light, focusing is "not critical"...

The image below makes me smile again because I find in it all the subtleties I was hoping for. To obtain it, I had to bring the enlarger head 3 mm closer (which seems enormous to me) compared to the focusing done in Live View. The table shows all the results obtained by shifting the focus by 3 mm increments. Yes, it's ugly: display at 100%, with the most basic post-processing possible, without contrast enhancement, to allow a fair comparison. I chose an image without interfering details and with dark areas, as well as light grays, which are the most critical zones.

This "chameleon" aspect of the grain, capable of appearing so differently depending on the amplitude of the focusing error, is surprising, to say the least! The most disturbing thing is that, in the light areas of the image, I find this strange effect of an "interstitial network" in the grain that often bothered me during my tests, without understanding the cause. Or especially these big white dots, which become so intrusive in the dark areas of the image. It gave me so much to do in post-processing, without ever obtaining a completely convincing result. I admit that I was content with it, considering that it was still better than with all other techniques. But there was something flawed that I did not question early enough!

And all these problems miraculously vanish with a green LED!

One might think that comparisons made on a 100% display of a 36 MP image do not make much sense for smaller print formats. But for the formats that concern us here, A3 or A4, I affirm that it allows for a correct evaluation of the results.

Final Assessment in a Few Images

Example: Diffuse vs. Collimated White LED

The three images below, processed as specified in the captions, are a portion of a TriX 135 which, in its entirety, would be displayed at 13 × 18. Even in such a small format, noticeable differences can already be seen. A click on the red icon allows you to get an idea of all the details that are blurred by the diffuse light, without our knowledge, in the most common enlargers, but also in ALL scanners!
Cuba 2000 - La Havana - Photo by François Huchet

Example: Collimated Green LED with Varied Post-Processing Attenuation

And finally, an example of what I mean by the extended possibilities offered by this process to the printer to "interpret" the image as they see fit. Here, it is a portion of TriX 120 dating from 1975 ;-)

In Conclusion

I dream that someone competent in optics could take up the "Rube Goldberg machine" to which I have arrived and make it a coherent, compact, and convenient tool. And of course, that their work be published and freely usable.