APD / ADX documentation

Home / APD / ADX documentation

I’m reading about APD and ADX specifications and came across the following sentence in TB-2014-005 :

The universal ADX-to-ACES transform and universal ACES-to-ADX transform are described in other documents detailing the Image Interchange Framework.

Can someone point me to the document(s) referred ? I didn’t find anything in the others Technical Bulletin ?

Related question : in the CTL code base I see the ADXtoACES transform but not the other way around, it’s pretty easy to create an inverse CTL manually but I’m wondering if there is a reason for this ?

Thanks !

PS : don’t know if I’m in the right forum section, feel free to move the thread


The documents that talk about APD and ADX are the SMPTE standards 2065-2 and 2065-3.
At the bottom of the ACES doc page.

The ADX to ACES transform was in much demand for a bit, and so was put into the transforms.
The inverse was not a straight-forward transform, because in general you would go from
ACES to ADX with the intent of filming it out, and because of the reduction of film use and the
particulars of film recording, it was thought it would be better for customers to approach the
recording facilities about the best transform from ACES to ADX given their calibration, recording stock,
and Lab. So the transform is possibly not an exact inverse of the math. Other uses of ADX might be possible, but it important to remember that the film side chain has a lot of particulars that make back and forth between ACES and ADX and then ACES have some color conversion issues.


Thanks for the details Jim.

I understand your point on the ADX to ACES vs. ACES to ADX transforms, would it be correct to say that the former is close to an IDT an the latter would be a special case of ODT (not involving the RRT) ?

Trying to get a high level understanding of the ADX to ACES transform, several points are not obvious to me :

  • What are exactly Channel Dependent vs. Channel Independent densities and how the matrix was derived ?
  • Why is an 1D LUT necessary to model the conversion from density to log-exposure for the low end part of the density range ?
  • Finally, was the EXP_TO_ACES matrix derived in a way similar to what is described in the procedure for IDT design (P-2013-001) ?

Other uses of ADX might be possible, but it important to remember that the film side chain has a lot of particulars that make back and forth between ACES and ADX and then ACES have some color conversion issues.

I have seen it used as a grading colorspace pretty often, if you have any comment or more details on the possible color conversions issues involved I would be interested.

I hope these are not inappropriate or wrong questions.


Yes, your first paragraph is good.

-on channel dependence,
in the film system, layers of the color response are coupled to each other. As you change green exposure, the red also changes. These are generally coupled together in descriptions of the “interlayer effect” which includes both different channels responding to a single color and the inhibitors present in film to reduce this effect.

Independent channels can change just Red and only the Red changes. The matrix was supplied by Kodak and Fuji as an average that would work across all films. If you know how a film was constructed, you can make a better matrix, but that is not for mere mortals. 

– In general, 1D Luts are needed to change the behavior and conversion of the toe in converting into linear. (or vice versa) It can also be though of as ‘unbuilding’ the tone.

  • on the EXP_TO_ACES matrix, no this was a different process for the film system, and we had a lot of help
    from Kodak and Fuji color scientists.

Yes, it can be used as a grading space. But it works better when you have a film emulation LUT as the viewing LUT. In essence, ADX becomes the ‘negative’. You can always grade direct to a final output, but you may feel that the controls do not get you to where you want to go. The color conversion issues happen because you are trying to represent what was on a film that was scanned, and that does not easily go back to scene referred. Our conversion is a best effort at an average, but coming from film, certain colors are limited, we can’t reproduce the subtractive colors on film, and there is coupling between the colors that remains in a film scan that we have not removed. There can all be overridden by a colorist, but sometimes it is a chore.

There is never an inappropriate question 


Thanks Jim for all the details and precisions !

As for using ADX as a grading space, it’s even used for born digital images sometimes, but I think this make your point even more relevant. If I understand correctly, used naïvely the conversion from ACES to ADX then ADX to ACES (simple inverse of the former, not the output/film transform discussed earlier) will not preserve all the scene dynamic range and maybe constrain the colors to what is expected of an average print film.