Exposure adjustment layer discrepancies between 32 bit and 8 bit

[GBoucherMontreal]

Hi,

I noticed I get different result if I use the exposure adjustment in 32 bit or 8 bit. I've compared by converting the 32 bit to 8 bit with exposure and gamma at 0 and 1 respectively. Here's a comparison to illustrate what I mean. I've used the ligthness on the lab scale. I'm curious to know if I'm making a mistake or it's only due to rounding of numbers during conversion from linear to gamma.

[rob day]

Try setting your Info Eyedropper to 32-bit then you’ll get an real number readout for both 32- and 8-bit:

[GBoucherMontreal]

Thanks a lot, I didn't know we could set the eyedroper to linear values. It's great to have lab values with decimals. But my question was more about: why photoshop gives a different results between 32 bit and 8 bit, if you take the same value and your remove one stop with exposure in both environment. The layer adjustment doesn't give the same result. As you switch to 32bit, you see the dark values all getting lighter but not the light values they stay the same.

 

 

[rob day]

I’m not seeing that here with CC2020:

[GBoucherMontreal]

Thanks Rob,

I think it's created by the rounding of variable between 8bit and 32bit. I'm on CC2020 too. My scales were created in 8 bit. Unlike your exemple, my 20 was not at 20,00, it was at 20,10 or 19,90. I recreated your exemple with a 20,00. Now I get the exact results you had. 

[rob day]

Even though the UI scale is 0-100, the 8-bit L channel would still have 256 gray levels, so the conversion from a higher bit depth down to 256 levels should be around 0.39453125 increments (101/256).

Same problem with the Color Picker’s CMYK 0-100 values—even offset printing halftone screens are capable of more than 100 gray levels.

[GBoucherMontreal]

Hi Rob,

I retried and got the descrepancy again. I think, I found where it's coming from. It happens if my workspace and the document are in sRGB IEC61966-2.1

  

[GBoucherMontreal]

@rob day 

To be more precise the modification happens when I convert switching between 32 and 8bit when my working profile is sRGB IEC61966-2.1.

But it happens even if we start with only 8bit documents. Exposure result is different between Adobe1998 and sRGB IEC61966-2.1 color profiles

  

[GBoucherMontreal]

I use PRTSCR for screengrabs and it changes the gray value. Color picking from them would be inacurate.