Understanding gain settings

[Steve Mullen]

Originally Posted by Rick L. Allen

Craig, Simple answer is that negative gain settings (-3 db) etc. are a simple way to add the equivalent of a neutral density filter.

From the point of view of the shooter, you are correct that what happens is one stop less sensitivity

BUT, putting an ND filter in front of the sensor actually cuts the light falling on the sensor. Any sensor (including a mic.) outputs voltage as the input increases. At some point, this relationship ends. Typically, "bad" things happen at this point.

Actually, things may go bad long before the peak is reached because the output is very likely to be linear over only the lower end of the range.

Then there are the Sample&Hold, OP amplifiers, etc. that are in the signal path. Are they able to handle the signal when the peak levels are reached?

To really cover the range 16- to 24-bits may be needed, but typically only 12- to 14-bit paths are used to keep cost in bounds. (That's why audio has moved to 24-bits.)

The word to cover this issue is over-saturation. Putting ND filters in front cuts the light which lowers the chance of saturation.

David is correct -- there is an amplifier in the signal path. There is some point where it's gain value begins to add "unacceptable to the designers" noise. The point before this occurs is labled ZERO -- even though it may not literally be 0dB. But, it's likely very close to 0dB.

Above this point amplifier induced noise increases.

Below this point amplifier induced noise decreases.

So, when you have enough light, switching gain to negative values results in less recorded noise.

[David Heath]

Originally Posted by Steve Mullen

Below this point amplifier induced noise decreases.

So, when you have enough light, switching gain to negative values results in less recorded noise.

It's hard to generalise, but coming back to my earlier point there must come a point where quantisation levels in the recording codec will become the limiting factor. No matter how much the camera front end noise is reduced, how "fine" the noise input into the codec becomes, values have to be rounded to the nearest of 256 values (in an 8 bit system). Hence RECORDED noise doesn't decrease proportinally.

This then means that a move to a 10 bit recording system will enable lower levels of camera noise to be registered, so an incentive for working at lower gain settings as the "nominal" 0dB, and - if camera sensitivity is defined as ISO at 0dB - an apparently less sensitive camera!!

Which to me is another good reason for not defining sensitivity as the ISO rating at 0dB!

[Pietro Impagliazzo]

Originally Posted by Eric Stemen

from my understanding when the rating goes above 0db is when noise is becoming noticeable.

I think manufactures should really go to ISO ratings.

That's what RED does.

Anyway, no other company shoots RAW yet so when Sony/Canon/Pana decide start developing RAW cameras this may be a better choice.