Please file an issue https://github.com/darktable-org/darktable/issues/new?assignees=LebedevRI&labels=scope%3A+camera+support&template=new_camera_support.md&title=<Enter+your+camera+name+here> for that.
Might the problem be that the raw samples visible on raw.pxels.us aren’t from the same scene, as requested
in “Raw Samples Wanted”?
Opened #7623.
Please take a look, I’ve not changed anything except the title. If you need any info I will be glad to provide it for you.
Many thanks
What I meant is that there was not supposed to be any change in the engine of Filmic. However, when I align the Filmic settings (after importing a file) in dt 3.4.0 to 1,5 contrast and 33% latitude, I get different results. That is why I asked the question.
Could somebody explain why the Filmic default values have been changed?
it a process of optimizing - and since the complementary modules for scene referred workflows also evolved defaults are tuned …
Does it mean that when loading an image in dt 3.4.0 with automatically applied scene-referred mode (i.e. Filmic contrast 1,35 and latitiude 25%) I will get very similar results if I would do it in dt 3.2.1 with default Filmic settings (i.e. contrast 1,5 and latitude 33%)?
Sorry for bothering you with these questions, but I had a fine working process in dt 3.2.1 which does no longer give the same results. Before modifying the workflow I want to be sure I know reasons.
no, different values - different results … just save your dt 3.2.1 settings as a preset and autoapply this if these setting are better for you.
That is what I have done. I clicked filmic’s reset, changed the contrast to 1.5, set the latitude at 33% and checked the box to use custom middle-grey values, then saved this as a preset to auto apply with all raw files.
A small exercise that I did.
The first picture shows Filmic graphs after loading an image in dt 3.4, but which had been developed in dt 3.2.1.
The second one presents Filmic graphs after for the same photo after developing it with the same values in d t3.4.
All the Filmic, Local Contrast, Haze, etc settings are the same (maybe except demosaic). No difference in jpg between the two (at least at a fist glance) but a bit different handling of the blacks on the graphs. Are there really no changes in Filmic maths?
The added 0.01 so you don’t hit 0 as its a log function so the graph just ends up looking a bit different at the lower end…I think you can still se the target black to 0 if you want…I think this used to be zero 
1 Like
Also in 3.4 you can set the color space of your histogram.
1 Like
That’s it!!! It’s been driving me daft trying to understand why my filmic curves/graphs have nearly all had orange tails since upgrading to 3.4. Setting the target black luminance at 0% makes them look like they used to (but I’ve no idea whether this is good or bad thing).
1 Like
Just use your eyes…I do notice the indicators but I used to live by them and I think I had worse results…I use the new ev mapping view which resonates with me and I know what functions make what changes so I just go that route…if you have any lingering filmic issue I suggest that you try that view and move some sliders around and watch the mapping that happens…it will perhaps clarify some things…for example I rarely used the shadows highlights setting in filmic to much use other than to satisfy the warnings if needed now I am much more aware it shifts the latitude up and down and I move it left often way left to help resaturate the highlights when needed
I have been using my eyes, and do get get very good results with filmic, but I can’t help being driven daft by warnings that suggest that I’m doing something wrong. It’s like trying to concentrate on driving a car when a warning light is nagging you 
I’ll follow your advice and try the alternative EV mapping view.
2 Likes
I am totally the same…have to fight the urges to be too technical…try the ev view even just to experiment with the sliders and to compare the mapping to the image…it sort of reveals the inner workings of filmic…its like you get thinking that filmic is an s curve however if you set up everything to make it essentially a strait line with a slope of 1 it still does log tone mapping so I found it helped me to look at it more broadly…
Ther are some issue wirh darktable 3.4.0, not sure it’s only on my end
Photoflow
Darktable artifacts in the highlights
Darktable lch reconstruction shifted to the left
Here the raw for testing only
IMG_20181014_164524.dng (24.8 MB) IMG_20181014_164524.dng.xmp (6.0 KB)
There’s issues in the shadows too
Photoflow
Darktable green noise
Rawtherapee and Photoflow are pretty much equivalent and don’t show these issues
Thank you very much. Now, when I think about it, I am almost ashamed for having asked.
The fact is that I have never even played with the display sheet.
There are definitely artifacts in the (clipped?) highlights introduced from darktable and not from the demosaicing algorithm itself.
From this PlayRaw I get weird artifacts using the match greens option in demosaicing .
170322_Waterloo_Bridge_37.CR2.xmp (9.9 KB)
Here is possible to observe the darktable’s green bias noise (just demosaicing with amaze and addind 2EV in exposure)
Darktable
Photoflow
Rawtherapee has blue-ish shadow but the noise level is the the same of Photflow
Also looks like Darktable doesn’t round to nearest when exporting to 16/8 bit tiff.
Which church is that? Is it Wren?
Edit oops followed the trail back to the playraw and indeed it is a Christopher Wren church. (Same architect as St Pauls)