Upload & Sell: On
| p.1 #2 · Understanding 'Native' ISO and resulting non-native ISO's |
The native ISO of a sensor is that which has the highest light saturation capacity when metered at middle gray, which also yields the highest dynamic range offered by that sensor. The fake "low" ISO setting on cameras is produced by exposing 1 stop above metered gray and then digitally reducing the exposure by 1 stop to produce the same output brightness, albeit with less dynamic range. ISOs above native ISO are handled either via analog amplification, digital multiplication, or a combination of the two.
For traditional CMOS sensors which have high sensor electronic read-out noise (all Canon sensors and Nikon D3/D700/D3s sensors), analog amplification yields lower effective read noise at higher ISOs than it does at native ISOs because the amplification increases the relative signal to the constant read-out noise. This doesn't mean the resulting image at High ISO will be less noisy, since most of the noise in the image is the result of photon shot noise, which is a function of only exposure, and typically when you double the ISO you halve the exposure, which means 1/2 the amount of light reaching the sensor. On most entry level/prosumer cameras there is only a single amp which only handles full ISO increments (200, 400, 800, etc..), and so the intermediate levels are achieved by over/under exposing by the fractional stop difference and then adjusting digitally, thus the intermediate ISO levels will not be optimal. Pro-level cameras typically have two amps, one for the full ISO increments and another for the 1/3 stop increments.
For newer CMOS sensors which have low sensor electronic read-out noise (Sony EXMOR sensors), there is not much to be gained via analog amplification vs digital since the relative read-out noise is so low to begin with. Still, it has been measured that the D800 uses analog gain up to around ISO 1000 and digital gain thereafter.
Edited on Jan 11, 2013 at 08:39 PM · View previous versions