Pages: 1 [2]   Go Down

Author Topic: ISO 640 has less noise than ISO 100....WT@#?  (Read 14986 times)

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: ISO 640 has less noise than ISO 100....WT@#?
« Reply #20 on: August 15, 2012, 05:22:01 pm »

True although in this example it wasn’t a light starved situation. The same shutter and aperture was used for both captures.

I only adjusted the ISO setting. But yes, had the 100 ISO image had those two stops more photons and had been equality normalized, it would look better than the ISO 800 image.

And this is the point.  ISO 800 has less read noise in photon equivalents than ISO 100 on many cameras.  The only disadvantage of ISO 800, then, is that it clips 3 stops earlier than ISO 100.  In your example, you had the headroom for the particular exposure that the image didn't clip at the higher ISO.
Logged
emil

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: ISO 640 has less noise than ISO 100....WT@#?
« Reply #21 on: August 15, 2012, 05:36:58 pm »

i just saw a noise/iso comparison chart for the 5D MarkIi that shows ISO 640 having less noise than ISO 100.  Is this a joke?

Be careful of what units are being used in the measurement.  The relevant thing is how much noise there is relative to a given light signal being captured (ie the amount of photometric exposure).  An image exposed for middle gray at ISO 640 captures 6.4 times less light than one exposed for middle gray at ISO 100.  So one has to do a careful calculation of the noise (both the photon noise and the electronic read noise, and I suspect your comparison chart included only the latter) to determine which ISO has less noise.  As a rule of thumb, when exposing according to the meter (which takes into account the ISO setting) generally the lower ISO wins on noise, because the factor of overriding importance is the amount of light captured, and the lower ISO setting allows more light to be captured without clipping in the raw data (assuming that opening up the exposure doesn't cause motion blur or too shallow DOF).
Logged
emil

texshooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: ISO 640 has less noise than ISO 100....WT@#?
« Reply #22 on: August 15, 2012, 05:56:30 pm »

What I hear all the time is that ISO 160, 320, 640 is better than ISO 200, 400, 800.  It turns out that is only sometimes true. Pulling the ISO from 800 to 640 will reduce noise, but could clip your highlights due to loss of dynamic range. It depends on what's more important to you: noise or clipping.
Logged

Wayne Fox

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4237
    • waynefox.com
Re: ISO 640 has less noise than ISO 100....WT@#?
« Reply #23 on: August 17, 2012, 02:39:00 pm »

The example of Andrews isn't surprising nor should it be.  For the ISO 100 image to improve noise, exposure needed to increase so the histograms/data collected would have been equivalent. But his test does show that on a dSLR, increasing ISO to get more information/expose further to the right can be a benefit if you can't increase exposure with aperture/shutter speed, as long as the reduced dynamic range for the higher ISO isn't a problem.
Logged

Ellis Vener

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2151
    • http://www.ellisvener.com
Re: ISO 640 has less noise than ISO 100....WT@#?
« Reply #24 on: August 18, 2012, 09:16:47 am »

This seems to be one those things one can test to one's own satisfaction to see the capabilities of the specific camera model you use, any in camera settings you prefer (like ADL for Nikons) that have an effect on a raw file, and your choice of raw processing program and settings.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: ISO 640 has less noise than ISO 100....WT@#?
« Reply #25 on: August 18, 2012, 10:29:57 am »

This seems to be one those things one can test to one's own satisfaction to see the capabilities of the specific camera model you use, any in camera settings you prefer (like ADL for Nikons) that have an effect on a raw file, and your choice of raw processing program and settings.

Exactly so! I remember testing this issue years ago with my Canon 20D. A full exposure at ISO 800 produces noticeably less noise in the shadows, lower midtones and upper midtones than the same exposure at ISO 100 (which is 3 stops underexposed). This is old hat.

However, this situation does not apply to all sensors. Recent Nikon sensors have such superb dynamic range at base ISO, the noise in absolute terms is about the same at all ISOs. Not exactly the same. There may be some marginal benefit in using a high ISO in preference to underexposing at a lower ISO, but the advantages of avoiding any risk of overexposure, even of specral highlights, can be interesting. Sometimes one can even see the filament in a switched-on light bulb.

As Emil Martinec has pointed out, we should distinguish between absolute noise and relative noise in relation to the signal srength. With Canon DSLRs the absolute noise is less at higher ISOs, at a given exposure.
Logged
Pages: 1 [2]   Go Up