Hello, new poster here!
when I load a dslr-scanned negative into raw therapee on a neutral setting, apply desaturation and enable “film negative”, I get artifacts in highlights. Users on reddit helped me determine my scans are not the issue since when loaded into darktable or digiKam, the scans convert fine and display smooth fields of brightness instead of “sand” in highlight areas. Please see the uploaded photos, I’ve also attached negatives to show what rawtherapee gets to interpret. I tried consulting the settings but only found profiles that I was told to keep on “neutral”… No idea what else to change to make rawtherapie change it’s mind. Results from rawtherapee are comparatively darker to digiKam results (again, rawtherapee profile set to neutral) and show the afforementioned highlight artifacts.
I discovered that when I apply the autogenerated film-like curve, it seems to alleviate the problem somewhat (still not as good as digiKam result) but at the expense of skewing the look of the images. All positive images come from rawtherapee.
Welcome!
In the last two examples, I can see no “obvious artifacts in highlights”, even at 600%. What am I missing?
When I simply invert the posted negative in the GIMP it becomes obvious that your RT is doing much, much more than just that.
Did you shoot this in colour on your camera or did you set your camera to BW? In the shot of the paddock it looks like there is a lot of film grin being recorded.
Three things come to mind.
-
Film grain - your png examples look like they were shot using film processed in a fairly strong developer
-
Dark/Flat renditions accentuate film grain
-
“Auto-Match” tone curve requires a jpg reference - RT is not giving what you want/expect
Bonus Point: When converting film negatives directly into a digital image, many/most people fail to realize/understand that old film prints raised the mid-tones. It was in the nature of the process. Film negatives are, actually, quite flat and look horrible when converted without knowledge of the entire silver halide BW process.
@Terry Hey, Terry!
The camera used for scanning is shooting in colour, monochromatization is on the part of the software. I’m sorry sorry sorry I didn’t mention it, but the pictures are crops of a 120 film, equivalent to 1/4 of the image. That’s why you’re seeing so much grain! I zoomed it in so that it would be easier to spot the problems in the highlights. I’ll add it to the post so no more confusion arises.
EDIT: I wanted to edit but for the life of me can’t find the little crayion icon to edit the original post.
@xpatUSA hey, xpat!
the artifacts in the last image are located in between the horizontal poles of the fence. The meadow was lit up and was basically a solid chunk of light, but with the inverted negative, there are so many little grains that aren’t evident from the negative (and the other programs I used indeed don’t interpret these areas as having tiny little holes in them). I mentioned this in another reply, the reason why those scans look so grainy is because they’re 1/4 crops of actual images with only the problematic areas isolated.
I’m glad we’re on the same page about “it being obvious that my RT is doing much more than just inverting”. Could you (or anyone) please help me identify WHAT IT IS?!? 
I’m reiterating, the only thing I’m asking RT to do is: a) monochromatize, b) invert the image (through the “film negative” plugin). All of this on a “neutral” setting. Yet there seems to be the “much, much more” part that RT is doing. No idea what it might be or how to turn it off (or indeed even influence it), there isn’t anything in the settings that I found as pertaining even broadly to this issue.
@ChristopherPerez Hey, Christopher!
ad 1) the grain is due to it being a drastic crop. It’s a 120 shot of foma 100 processed in rodinal 1+100 and the grain is pretty much invisible on a normal computer screen without zooming or squinting really hard. Sorry for not mentioning this in the original post.
ad 2) that’s a thing, yeah, Rawtherapee just makes the images dark. When I simply invert them in another software, the darker the scan, the lighter the resulting positive. However, when I use RT, the darker the scan, the darker the positive as well! To reiterate, I’m using the “film negative” plugin in rawtherapee. This can of course be remedied with adding gain, but that creates other problems and doesn’t help to explain the unintuitive behaviour. I even tried scanning three times on different exposure times to see if it had an effect. What startled me was that the result (on neutral, after applying “monochrome” and “film negative”) were identical!
ad 3 (and bonus)): Could you elaborate on the jpg reference? You’re right in that I don’t know the history of the silver halide process. It’s certainly the first time that someone is suggesting midboosting to me as anything else than a matter of preference. I’ll give it some thought as that probably is why the midboost is a default for film-like auto-curves in RT. For now, though, I’d very much like the “quite flat and bland” positives that don’t feature startling artifacts and can be edited into a more likeable image. Or would you suggest that the midboost curve is necessary to securing a problem free positive? That would go against everyone who told me to keep the profile on neutral, but i’d give it a thought since the problem with artifacts only arises on the neutral. It would mean that the “neutral” setting is the reinterpretation and not the film-like auto-curve, which should be the baseline. Am I reading you right?
That’s interesting. Normally Rodinol at 100:1 can be fairly smooth. Though your comment about your sample images being an extreme crops is well noted.
In RT, try this. Load film negative image and before applying any other tools → Exposure tab → Tone Curve 1 → take the right corner of the curve and slide it to the base of the graph + take the left corner of the curve and slide it up to the top of graph. Snug the ends of the inverted curve against the graph showing the histogram.
Regarding RT “Auto Matched Tone Curve” I realize I may have confused the problem. What Auto-Matched does is take the thumbnail jpg of an image RAW original and matches the curves of the thumbnail.
For instance, if you shoot “Standard” in-camera, the jpg thumbnail curve will be applied to the RAW file. It does not seem to take the thumbnail’s tones, only the curve. Someone can correct me if I’m wrong.
In your case I’m not sure since I’m not really certain what the nature of your problem is.
I added that comment to suggest a much longer commentary on how digital BW is very different from film silver-halide based BW. Taking your hybrid approach, when we digitize a film negative and simply invert it digitally we’re missing out on the true “magic” of a BW print.
Further, if the goal is to make an all digital BW image that retains the “luminosity” and tonal range of chemical process prints there is also a “problem.” Current digital BW is all too often simply a desaturation exercise (even when applying color filters). Compared to an old chemical silver print the mid-tones in digital are much darker and the highlights tend to have too much tonal separation.
I’m writing something that I’ll post on my blog in the near future on this very topic.
Sorry, I don’t know what “the ‘film negative’ plugin” is.
Converting color to grayscale is an enormous subject, abounding with various curves, algorithms, opinions and no doubt AI these days.
Some light reading for you: Martin Cadik - Perceptual Evaluation of Color-to-Grayscale Image Conversions
Personally, I prefer the straight luminance (Y) method and do any required editing after conversion.
This? Film Negative - RawPedia
Thanks, most helpful. I do have that in RT 5.8, but he said “plugin” which I took to mean an addition, not “comes with”.
@afre yes, thank you for covering for me
@anon8280290 yeah, it’s a complex thing no doubt, but not the source of the problem since the problem appeared with the monochrome setting turned off.
@everyone I think I’ve got a bit of a breakthrough! While messing around with digiKam (on which I had settled until I would fix rawtherapee), I found the “Gamma” setting to be included with brightness and contrast. After fiddling with it, I realised I haven’t seen this particular sort of slider (that decides what gets to be black and what not, in my understanding) in rawtherapee and googled to find out if there was one. Not only there is, but it’s tucked away under color management (as you all most certainly know). I clicked around and found out that raising the slope and gamma to higher values made the contrast between the sandy highlights and the surrounding area much smoother. It is also why RT gave me a comparatively darker look than other apps - I’ve never even touched any of these settings so my gamma and slope wer set low the entire time. The results aren’t ideal yet but I at least now know how to compensate for the highlight problem.
I would be interested in seeing the out of camera rgb image. The blue channel is typically very noisy, the green channel is the least noisy. The process of converting the rgb to greyscale may increase the noise depending on how the software handles it. Also it could be kept greyscale in look but maintain rgb channels and some of the tools in GIMP need rgb to function. Could you share one of the out of camera rgb images so I can see if that makes any difference?
Be that as it may, you originally said:
Therefore, the link I posted remains relevant for education or for interest …
… or not. Your choice, of course.
Meanwhile, Y = about 0.3R+0.6G+0.1B works well enough in the highlights for me. 
You may find it funny but I have no idea what you mean by
Could you please elaborate? The problem for me is resolved with the gamma/slant slider discovery - I’m able to get satisfying results where I wasn’t able to before. Thanks to all of you! I’d appreciate some further directing on this topic because despite reading the rawpedia documentation, i have no idea what either gamma or slope mean or do I just discovered that both higher => more uniform highlight areas.
@Terry If you don’t mind it not being the exact same negative, I can supply (due to space issues, I delete stuff way too quickly sometimes).
_JSM0299.NEF (42.1 MB)
Certainly!
In apps that can convert color to grayscale there is usually a basic option called “luminance” and ‘Y’ is the CIE symbol for luminance, formula as above approx. but with many more digits and some small variations based on color space such as BT.709.
So, an 8-bit color pixel of R,G,B=50,100,200 will be about 95 in 8-bit grayscale or 95,95,95 in RGB.
Glad to hear it.
One can select a defined conversion method like ‘Luminance’ and know immediately what one will get, or one can indeed waggle sliders until it looks satisfying on the screen.
I looked at the supplied file. I converted it to a B/W looking RGB image in darktable. I used negadoctor to do the inversion and I am sure programs like RT have similar tools. The advantage of retaining it as an RGB image is that you can easily tint the image to sepia if you like. The only real advantage of a single channel greyscale image is the smaller file size, but then you lose some editing options and tools available for processing the image.
_JSM0299.NEF.xmp (14.8 KB)
@Terry yeah, I’ve also started using darktable more, and after overcoming the insane first hour or two when you don’t know how anything works, I¨ve fot some results quite close to what you’re showing me now! I reckon I might switch to darktable, but first I’ll do a few side by side comparisons of what I can make from the same scan in both.
If you post your image as a PlayRaw, you’re likely to get many responses, with sidecars, so you can learn from the steps others took.
Be sure to read the first, pinned post, and add a licence (there should be one added automatically when you create your post).