Exposure (and offsets)

I have found the ‘getting it right’ , exposure that is, is critical in ensuring that filmic-rgb really does a good job. Recently in this forum it was offered (aurelienpierre) that JPG (viewfinder) luminescence should be offset from the RAW data by a specific amount. It was further mentioned that Fujifilm underexpose their RAW data by 1.25 EV compared to the JPG output.
Accordingly (since I use a Fujifilm camera) I established a corrective offset preset of 1.25 EV which is automatically applied to all incoming RAW data. For me, this works 100% and I now find that rarely do I need to consider further exposure adjustments.
However, there is a fly-in-the-ointment … the exposure module (in its recent form) offers a “compensate camera exposure” capability which should be added or subtracted from any other offset but this function becomes inoperative if exposure is set through a ‘preset’.
Anybody else seeing this as a small problem?

No. The +0.5EV is supposed to be a generic starting point. If you’ve determined 1.25EV as the proper offset for your camera and the way you expose, go with your preset.

OK … but when I use the offset as a ‘preset’ it stops me from using the further ‘compensate camera exposure’ which I now have to read from the EXIF data set.

But you’ve already determined the offset for yourself.

We are talking about 2 very different things.

  1. The offset between the RAW and the JPG (that is seen in the viewfinder)
  2. The exposure compensation that is done to avoid flooding the sensor for instance.
    Both need to be accommodated in the final RAW adjustment.

Even if you account for both, both are changes made to your in-camera exposure. The exposure module in darktable takes one value. If that one value for you is +1.25EV and that is your preset and it looks good, then you’re good to go.

If this is something else, then I’m missing the point.

The camera exposure compensation compensate the exposure bias as read in EXIFs. This bias is not the systematic deviation of exposure inherent to a manufacturer, but a user-set, in-camera parameter that applies a correction on the lightmeter camera reading.

You set this bias here:

image
image

and you usually set it ≠ 0 EV when you expect your camera to clip highlights. Then, the EXIF bias is read and applied at runtime depending on the raw you open.

The + 0.5 EV or in your case, + 1.25 EV, is a systematic offset that allows to roughly match the artistic intent applied by the camera firmware when starting with a linear file. It’s set by reverse-engineering and is not reverted by the bias.

You can define a preset, and auto-apply it on new raws, that overwrites the systematic deviation and apply, or not, the exposure bias on top. The exposure bias setting is only an option, not a value, the real value is read at runtime.

All I am saying is that I am successfully applying the “systematic deviation of exposure inherent to a manufacturer” through a preset … but in doing so, the camera compensation action in the exposure module fails to operate because there is a preset already operating.
Is that more clear?

When you created your preset, was the camera compensation box ticked?

Yes, the box was ticked …
Looks like I will have to use two exposure modules if I want to auto apply both amounts.

Fresh git installed and it appears to work correctly … somebody has obviously been listening.

1 Like

From what I’ve understood, David wants +1.25ev applied to all images. In addition, if he has taken some shots using an exposure compensation of -1ev, he would like an extra +1ev added. Similarly, if the shot was taken with compensation of+2ev, then an extra -2ev should be added. This is in addition to the +1.25ev.
The issue seems to be solved now, so all’s good.

1 Like

Yesterday I fixed a bug where image history was initialized with default parameters and presets before default parameters were initialized themselves, leading to all kinds of inconsistencies. So it’s possible that the issue you have came from the same source.

1 Like

Thanks … interesting that now I rarely need to use the exposure module and as a consequence the filmic-rgb ‘auto-tune’ provides an optimum solution nearly 100% of the time. Very nice work … much appreciated

1 Like

With exposure, you target the midtones. With filmic auto-tuners, you target the white and black level. If your input images are, in general, well exposed, then of course you won’t need to change exposure much.

I think the real, underlying problem, here, was not the logic of the pixel stuff but the consistency of automatic defaults and EXIFs detection when opening a new edit.

Glad my little cleanup solved it. I hope it also fixes issues of images having both filmic and base curve applied by default.