Robin Balas
Well-known member
It's normal to use log20 when describing DR in digital photography. Then 120dB corresponds to 20 stops and 13 stops is 78dB.
But this is all theory, maybe they "feel" they need 10dB for roll-off and get out of the shadows, so they claim 13 usable stops out of a measured 88dB signal?
If the RAW file is in a native gamma (Linear state) then you need at least 1 bit for each stop, some people including myself feel that last shadow stop beeing described by only 1 level isn't very usefull. So I would say based on what I see that you need 2 bits more than the number of usefull stops you have.
If they do some math trickery getting out of linear and into log you could achieve more stops than bits. BUT the sensor characteristics is Linear for all silicon, and either some funky analog amplification or using a high bit A/D converter and digitally getting out of linear prior to writing the RAW file is needed.
It seems like the latter is in use today as many now claim more stops than the bits available.
MHO.
But this is all theory, maybe they "feel" they need 10dB for roll-off and get out of the shadows, so they claim 13 usable stops out of a measured 88dB signal?
If the RAW file is in a native gamma (Linear state) then you need at least 1 bit for each stop, some people including myself feel that last shadow stop beeing described by only 1 level isn't very usefull. So I would say based on what I see that you need 2 bits more than the number of usefull stops you have.
If they do some math trickery getting out of linear and into log you could achieve more stops than bits. BUT the sensor characteristics is Linear for all silicon, and either some funky analog amplification or using a high bit A/D converter and digitally getting out of linear prior to writing the RAW file is needed.
It seems like the latter is in use today as many now claim more stops than the bits available.
MHO.
Last edited: