clipping indicator lower threshold ?

In the clipping indicator in v3.3 what lower threshold should I be using if I only view images on a monitor rather than print them please ?

It’s currently the default -12.69 however I’ve never seen any clipping of black in my images. Even if I deliberately drop the exposure and confirm via the colour picker there is definitely black in the image (0, 0, 0 RGB), no clipping shows. This happens on all my images.

The clipping preview mode is set to luminance only.

If I set the lower threshold to -9.00 clipping does show so it is working.

It may be working as expected if a monitor can always produce pure black, however I’m following the recommended workflow for using filmic where the manual says “However, it is important to avoid negative pixels in black areas”. If I use the default -12.69 then I’m never going to see negative pixels.

I’m obviously not understanding something.

I raised a similar query when this was introduced - apparently this is filmic doing its job: over-exposure warning : add meaningfull modes by aurelienpierre · Pull Request #6253 · darktable-org/darktable · GitHub.

In terms of negative pixels, the important thing is to make sure that the input to filmic doesn’t contain negative pixels. The clipping warnings are at the output of the pipeline which, presumably, filmic has constrained to the appropriate bounds.

1 Like

Chris - thank you very much, this has been driving me daft.

Yes the clipping does show if I disable filmic and set the threshold to the default -12.69.

I’m going to have a proper read of the issue to understand it more.

Important part is :

But that also tells me that users are so used to false positives in that alert, that they find it strange if a piece of soft in the pipe actually prevents clipping efficiently.

Don’t overthink clipping. What you saw in darktable previous 3.4 was mostly false positives.

If I use the default -12.69 then I’m never going to see negative pixels.

-12.69 EV = 0.00015133150634020836 > 0 so you will most definitely see negatives. But filmic will have them fixed anyway. Just look at the picture, ensure you don’t have black blotches, and if everything looks ok, it’s going to be ok.

However, it is important to avoid negative pixels in black areas”.

Yes, at the output of the exposure module. Because they will get clipped later. Because they are invalid.

2 Likes

This is because filmic crushes the blacks at 0.0152% by default (the “target black luminance” value in the display tab). If you want to have a clue which areas will be pure black, change this value to “0”.

Filmic doesn’t crush blacks, it specifically avoids black crushing by raising them. Pure black doesn’t exist and is related to each medium, meaning variable. And willingfully make filmic fail at that for the sake of… seing false positives makes zero sense.

1 Like

Hi Aurélien
I didn’t understand this, do we need to turn off filmic, check the blacks and turn it back on?
Thanks

You are right, my mistake. However, in my opinion, crushing or lifting does not make any noticeable difference in this particular case, as it prevents the clipping indicator to highlight the blacks. As the filmic is not necessarily the last operation in the pipeline (at least based on the default module order), other modules (local contrast, tone curve, color zones, etc.) will break this lower threshold of black levels. Moreover, clipping indicator could help significantly by highlighting the blacks and whites, particularly when you work with a low-end, not calibrated display.