What's ISO on a digital camera?

[Heika Werner]

I wonder whether anyone can answer these questions:

• How closely do the "ISO" settings on digital cameras approximate ASA or ISO film speeds?
• What am I actually doing electronically when I increase ISO on my DSLR?

[ohenry]

They're pretty darn close, depending on your camera.

You're increasing the gain when you increase your ISO (I know one of our electronic wizards will expound upon the soon)

[JKSeidel]

ISO on a digital camera can vary quite a bit from the film 'equivalent'. It depends completely on make and model of camera. As an example, on one Canon model ISO 100 performs to the equivalent of 65 ISO film. ISO 200 is like film ISO 180. ISO 400 is close to ISO 400 film. Finally ISO 800 behaves like ISO 875 film (if such a beast existed). These numbers were taken from a similar discussion on rec.photo.digital.slr-systems (a USENET newsgroup). A different model may behave completely different.

The ISO adjustment on a digital camera can be thought of more like a gain control for the sensor. Just like other electronic gain controls, the higher you set it, the lower the signal-to-noise ratio gets. In other words, the more sensitive you make the sensor, the more electronic noise is introduced from various sources. Newer cameras have better ways of dealing with this problem, so in most of the newer DSLRs noise isn't a huge concern until the ISO 800 (or greater) setting. Some DSLRs now have excellent performance right up to ISO 800. ISO 1600 usually results in noticable noise in even the best DSLRs. I'm not sure about Nikon, but Canon's ISO 3200 setting isn't a true extension of gain. It is an interpolated version of the camera's 1600 setting. That is why Canon DSLRs have ISO 3200 disabled by default and not even labeled 3200 (most models list it as 'H').

Kinda complicated and I probably messed up explaining it somewhere. But that's pretty close to the way it works.

[Jon]

Because ISO doesn't always very closely match ASA (as said above), think of ISO on YOUR camera (no matter which it is) as an arbitrary scale used to tell you the relative sensitivity to light of the sensor.

[JDArt]

Originally Posted by Jon

Because ISO doesn't always very closely match ASA (as said above), think of ISO on YOUR camera (no matter which it is) as an arbitrary scale used to tell you the relative sensitivity to light of the sensor.

I'd say it's a little better than arbitrary, and I say that because the exposure readings I get from my light meter are pretty darn accurate when I set both the meter and my camera to the same ISO value. Now, this meter was made to be used with film cameras, but my digital cameras speak the same language, apparently.

Yes, there are some discrepancies. I note that my Nikon cameras tend to calibrate better than my Canon cameras with the meter, but there are variations from camera to camera.

[ohenry]

Originally Posted by JDArt

I'd say it's a little better than arbitrary, and I say that because the exposure readings I get from my light meter are pretty darn accurate when I set both the meter and my camera to the same ISO value. Now, this meter was made to be used with film cameras, but my digital cameras speak the same language, apparently.

Yes, there are some discrepancies. I note that my Nikon cameras tend to calibrate better than my Canon cameras with the meter, but there are variations from camera to camera.

My Canon 1Ds agrees closely with my light meter also. Close enough that I set the ISO the same on both camera and light meter. But, I think I'll do another test across the range of 50-400, just to see again.

[scoundrel1728]

Originally Posted by JKSeidel

As an example, on one Canon model ISO 100 performs to the equivalent of 65 ISO film. ISO 200 is like film ISO 180. ISO 400 is close to ISO 400 film. Finally ISO 800 behaves like ISO 875 film (if such a beast existed). These numbers were taken from a similar discussion on rec.photo.digital.slr-systems (a USENET newsgroup). A different model may behave completely different.

If this is correct - and I don't mean to imply that it isn't - then a nominal x2 increase in the gain of the imager actually results in something like a real increase of x2.2, so increasing the gain of the imager by 3 stops results in an actual increase of 3-1/2 stops. *I wonder if anyone has found this to be true in practice that images exposed at a nominal ISO 100 are about 1/3 stop darker than those exposed at ISO 400 for this particular model.

Originally Posted by JKSeidel

I'm not sure about Nikon, but Canon's ISO 3200 setting isn't a true extension of gain. It is an interpolated version of the camera's 1600 setting. That is why Canon DSLRs have ISO 3200 disabled by default and not even labeled 3200 (most models list it as 'H').

"Interpolated" is somewhat of an odd term. *I interpret this to mean that the extra boost occurs in the software processing rather than an actual hardware gain increase, something like an in-camera version of lightening the image as you would in your image editor.

[Jon]

Originally Posted by JDArt

I'd say it's a little better than arbitrary, and I say that because the exposure readings I get from my light meter are pretty darn accurate when I set both the meter and my camera to the same ISO value. Now, this meter was made to be used with film cameras, but my digital cameras speak the same language, apparently.

Yes, there are some discrepancies. I note that my Nikon cameras tend to calibrate better than my Canon cameras with the meter, but there are variations from camera to camera.

Yep...True...That being stipulated though, my D2H is dead-on-balls accurate (that's a technical term) with my L358; S2 #1, however, was 1/3 of a stop under, while S2 #2 is 1/2 stop over.

[JKSeidel]

Originally Posted by JDArt

Now, this meter was made to be used with film cameras, but my digital cameras speak the same language, apparently.

That is where some of the issues lie when comparing film ISO to digital 'ISO' settings. I've read several discussions on the issue and they get technical beyond my willingness to sit down and decipher the jargon and theories. One salient thing I do understand is that the accuracy in comparing the two mediums varies because of film's inherent design to approximate the human eye's way of seeing thing with a non-linear gamma. Film response graphs have a 'heel' and 'toe' while digital's is completely linear. The main point of contention amongst the techies is that the meter is responding to the middle part of the response curve where film and digital are approximately equal. The variance in determining how digital vs. film ISO compare should take into consideration the total curve (or in the case of digital lack of a curve) and how they differ. This is usually a color management discussion that inevitably winds up spilling over into other areas (in this case ISOs). Also the difference between measuring incident light vs. reflected.

I'm pretty sure you can Google this under Google Groups (USENET) under rec.photo.digital and rec.photo.digital.slr-systems and find several discussions about which cameras vary widely in practice from their ISO setting.

Roger Clark has some interesting articles (http://clarkvision.com/) about this and a heap of other techie subjects pertaining to digital imaging. Fascinating stuff if you are tech minded.