Go Back   Steve's Digicams Forums > Digital SLR and Interchangeable Lens Cameras > Olympus dSLR

Reply
 
Thread Tools Search this Thread
Old Mar 17, 2011, 8:46 PM   #1
Senior Member
 
Join Date: Dec 2006
Location: Davenport, IA
Posts: 2,093
Default 1/3 ISO and Noise, Examples

I just ran a test of my e620 shot every available ISO setting from 100 to 3200. The tests were shot with noise filter off, f/3.2, 1/200, lens cap on. I'll post five photos per section and some thoughts in the last section. This section has 100, 125, 160, 200, and 250.

A. C.
Attached Images
     

Last edited by ac.smith; Mar 17, 2011 at 9:49 PM.
ac.smith is offline   Reply With Quote
Sponsored Links
Old Mar 17, 2011, 8:53 PM   #2
Senior Member
 
Join Date: Dec 2006
Location: Davenport, IA
Posts: 2,093
Default

This section contains 320, 400, 500, 640, and 800.

A. C.
Attached Images
     
ac.smith is offline   Reply With Quote
Old Mar 17, 2011, 9:03 PM   #3
Senior Member
 
Join Date: Dec 2006
Location: Davenport, IA
Posts: 2,093
Default

This section has 1000, 1250, 1600, 2000 and 2500.

A. C.
Attached Images
     
ac.smith is offline   Reply With Quote
Old Mar 17, 2011, 9:45 PM   #4
Senior Member
 
Join Date: Dec 2006
Location: Davenport, IA
Posts: 2,093
Default

Final section with just 3200. These photos where shot in jpg but I believe the results would would follow the same pattern in RAW without application of advanced NR techniques.

Observations:

1) Confirmed that the noise increase as ISO is raised is non-linear with the intermediate 1/3 stops noisier than predicted by the whole stop increases.

2) The non-linear noise increase appears to be principally choma noise. The luminace noise appears to increase in a more linear fashion.

3) The 1/3 stop just below a whole stop is generally noisier than that whole stop while the second 1/3 stop below a whole stop is less noisy than that whole stop.

4) This phenomenon can be tracked in the file size of these photos since I saved the crops with no compression.

Thoughts:

1) In the studio one should probably use full stop ISOs only and I think that would be practical.

2) This pretty much confirms Gregg's experience that 2000 is the best ISO for his hockey work as it is less noisy than 3200 and results in enough faster shutter speed over 1600 to be worthwhile.

A. C.
Attached Images
 
ac.smith is offline   Reply With Quote
Old Mar 19, 2011, 6:50 AM   #5
Senior Member
 
bobbert's Avatar
 
Join Date: Dec 2006
Location: Toronto Canada
Posts: 600
Default

Very interesting... thanks for this careful analysis. I'll pay attention to these intermediate settings form now on.
bobbert is offline   Reply With Quote
Old Mar 19, 2011, 9:28 AM   #6
Senior Member
 
Join Date: Dec 2006
Location: Davenport, IA
Posts: 2,093
Default

Thanks for your comments. I had read reports of this but the information was always presented as graphs of a numerical analysis. I wanted to see the visual effect. By the way, anyone that has software that do the a numeric analysis feel free t download the files. The ISO is in the file name.

Another observation, the results are exactly the same with noise filter on or off. I think this is because the image is all noise so there's nothing for the software to distinguish between noise and image so it does nothing.

A. C.
ac.smith is offline   Reply With Quote
Old Mar 19, 2011, 7:35 PM   #7
Senior Member
 
Join Date: Dec 2006
Location: Davenport, IA
Posts: 2,093
Default

After removing the effect of the original jpg compression and re-saving the files. I have treated the noise at ISO 100 as the "0" or reference noise and plotted the percentage increase in noise (based on file size) against ISO. I have also used the ratio of the files size at ISO above 100 and the file size at 100 converting the increase to decibels, plotting those in the second graph. That graph uses a logarithmic horizontal scale.

I have also attached the spreadsheet containing the data. An additional look at just the base ISO and multiples suggests that even with the noise filter switch off there is some noise filtering going on.

A. C.
Attached Images
  
Attached Files
File Type: xls ISO Increase.xls (27.5 KB, 93 views)
ac.smith is offline   Reply With Quote
Old Mar 24, 2011, 6:19 AM   #8
Senior Member
 
tkurkowski's Avatar
 
Join Date: Apr 2007
Location: Maryland, USA
Posts: 3,625
Default

AC, I forgot to also say thanks for this careful analysis. It also emphasized the importance of calibrating a monitor. I've never done that to my monitor at work, which is where I first opened your OP. On that monitor I could barely detect any noise in the low-ISO photos. It wasn't until I got home that I could clearly see the different noise levels on my calibrated monitor.

Ted
tkurkowski is offline   Reply With Quote
Old Mar 24, 2011, 9:46 AM   #9
Senior Member
 
Join Date: Dec 2006
Location: Davenport, IA
Posts: 2,093
Default

Thanks for your comments Ted, that's worthwhile info. All my monitors are uncalibrated and while I can see differences they don't really pop out until I'm past 200.

I have a two monitor setup at work with two identical model Dell LCDs, both set to factory default. Yesterday while examining some test photos over at DPR that attempt to compare the "sharpness" of the e520 and the e620 I found that the 2 monitors differed enough to make doing such a close comparison very hard. The photos themselves had size and jpeg compression differences and even after attempting to compensate for those vagarities the monitor differences made it hard to draw a clear cut conclusion. Your comments and my experience yesterday seem to make a pretty good case for calibrating monitors if one is doing any kind of crtical work.

A couple of other notes about my tests. I've also run the tests saving in raw and while I don't have the numbers into the spreadsheet yet multiples of the base ISO have smaller file sizes than intemediate ISOs just as in jpg.

Finally I plan to do the test with noise reduction (dark exposure subtraction) turned on. Theoretically this leaves a completely empty image, will this drive the jpeg compression algorithm bonkers? Will it warp the space-time continuum? |:-)

Last edited by ac.smith; Mar 24, 2011 at 9:49 AM.
ac.smith is offline   Reply With Quote
Old Mar 24, 2011, 9:52 AM   #10
Senior Member
 
tkurkowski's Avatar
 
Join Date: Apr 2007
Location: Maryland, USA
Posts: 3,625
Default

Quote:
Originally Posted by ac.smith View Post
I have a two monitor setup at work with two identical model Dell LCDs, both set to factory default. Yesterday while examining some test photos over at DPR that attempt to compare the "sharpness" of the e520 and the e620 I found that the 2 monitors differed enough to make doing such a close comparison very hard. The photos themselves had size and jpeg compression differences and even after attempting to compensate for those vagarities the monitor differences made it hard to draw a clear cut conclusion. Your comments and my experience yesterday seem to make a pretty good case for calibrating monitors if one is doing any kind of crtical work.
A number of years ago Pantone released a very affordable calibrator called a ColorPlus - around $100 IIRC (of course, at that price it just does monitors, not printers and it's Windows only, not Mac). I've used it ever since. I don't know what's available today at a similarly affordable price.

Ted
tkurkowski is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 8:45 AM.