linear dodge(add) mode different in 32bit than 8bit mode

[willanie]

We are rendering raster passes in cg which need to be added together to reassemble a beauty render.
The layer blending mode being used is linear dodge.
There is a dramatic difference in the result in 32bit mode compared to 8bit mode.
We normally work with adobergb1988 color workspace.

Does anyone know what is happening?
What is the math being used in 8bit mode with the adobergb colorspace?
The problem is similar with sRGB profile.
If set up a linear color workspace icc the 8bit mode behaves just like 32bit mode.
Help.

[Mike Hale]

This question doesn't really seems to be about scripting and you might have better luck in one of the Adobe user forums such as the Photoshop Windows or Photoshop Mac forums.

Having said that, I wouldn't expect any of the blend modes to give the same results in 32bit compared to 8 or 16bit.

Even in a 32bit only workflow, the '32bit preview options' ( found under the view menu ) can effect how the blend-mode looks. With AdobeRGB and sRGB being gamma mapped color spaces there should be a difference between those and a linear color space.

So it seems to me that you have answered your question. Use a linear color space for the 8bit images.

[willanie]

Thanks Mike,

I just posted the same issue on the photoshop forum
The reason I posted it here is that I assumed more technical photoshop users are on the scripts forum.

The problem is no one in our facility nor our clients wants to work with a custom linear icc profile.

My intent is to write some javascript to process the layers while in 32bit mode to pre-compensate for the gamma space
arithmetic being done in sRGB or adobeRGB color profiles in 8bit mode.