I have been testing the Canon 7D and 5D Mark 2 sensor chips for resolution and in the process came across an interesting relationship between ISO setting and sensor noise nad resolution.
Here is how it works. As you increase the ISO in the camera the residual signal noise on the chip increases. It is like turning up the volume gain on a radio that is not tuned to any station. As the volume goes higher you begin to hear the amplifier “hiss” in the speaker. This is the same with sensor chips, increase the ISO gain an you get background noise in the image. This also affects the lens/chip relationship for resolution.
What I did not expect to find was that certain ISO settings show less noise than the others. I call these “Prime ISO” settings. They are in these particular cameras the ISO setting of: ISO 160, 320, 640, 1200 and 2500.
Lower ISO settings will be smaller files and higher ISO are larger files. As the sensor noise increases the “Hot” pixels add “1′s” to the binary file making the file total larger. In the case of these cameras the ISO 100 file is about 5MB smaller that the ISO 6400 setting do to the noise. The ISO 160 file is smaller than the ISO 100 which is unique and goes against our common logic where we would expect the lower ISO to have less noise but it is not true.
What is interesting is that the “Sweet” ISO setting are the ones you should be using most of the time to get the images with the best Lens/Chip resolution and the least noise in your image file.
I will besting other camera models in the near future to see how other sensors respond to ISO changes.
The Red Dot on the chart show the best ISO settings.










Hello, I got this from Stephen Burns. I belong to his group.
This is interesting.
I have a Canon G9. Do you have any info on that camera?
Lee,
I find it interesting that the ‘Prime’ ISOs land on a multiple of the Powers of 2. Which leads me to think that sensors are manufactured to a pixel count constraint of the Powers of 2. Being that is true, there are no leftover pixels that need to be rounded up or down which would cause digital noise if rounding took place. So ISO’s that fall on a Powers of 2 multiple would inherently be less noisy.
Joe
Very interesting. I have a Canon Mark II DS.I will presume these tests will be relevant to my camera.
Thanks for giving me a math problem I can solve. My kids are in 5th and 8th.
Michael