"What about low ISO performance?" Fri 24-Feb-12 03:00 PM by ericbowles
It seems the forums are packed with discussion of high ISO performacne of new cameras - the D7000, D4, D800, etc. What seems to be missed is the dramatic improvements in color and dynamic range at low ISO levels.
While there are situations that require high ISO levels, the vast majority of my images are at low to moderate ISO levels. A quick scan indicates I have 4 times as many images at ISO 200 and below as I have at ISO 800 and above.
Based on data from DxO, the new cameras feature significantly better color rendition and dynamic range - enough for most photographers to buy the new cameras just for those reasons. And when you are shooting at high ISO levels, you lose color quality and dynamic range. In fact, the amount of color and dynamic range you lose is pretty significant even at ISO 400-800.
The guideline for best color and dynamic range remains to stay at the lowest ISO possible or at least try to stay at base ISO.
#2. "RE: What about low ISO performance?" In response to Reply # 1
Your work as well as the work published on DxO is the context for this thread. I see a lot of reference to using very high ISO levels with the discussion focused on low noise only - and success being measured by not observing noise. But at medium and high ISO levels, the photographer is giving up something in terms of dynamic range - and often giving up performance without really being aware of the tradeoffs. You can see noise, but it is harder to observe dynamic range differences between base ISO and ISO 400 - both of which are perfectly satisfactory in terms of noise.
We all know the D3 is the low noise champ. But what we tend to forget - or at least have stopped discussing - is that there is a major benefit of using low ISO levels. The dynamic range of the D7000 is far better than the D3 or D300 - partly because it has a lower base ISO. And while the D3 has slightly better dynamic range and tonal range at a given ISO level, based on the DxO evaluation, it moves into the fair or unacceptable level at ISO 1600 - far before noise becomes an issue.
For example, moving from ISO 200 to ISO 1600 on the D3 images are still largely noiseless. But that change means a reduction of almost 2.5 EV in terms of dynamic range and 1.5 bits of tonal range.
#4. "RE: What about low ISO performance?" In response to Reply # 3
That's the kind of thing we need to think about more than just low noise. Just as you learn to know the sweetspot and the tradeoffs with a lens, camera sensors have similar tradeoffs and optimum performance ranges.
#5. "RE: What about low ISO performance?" In response to Reply # 1
I know what "Dynamic Range" is. I have never heard for "Photographic Dynamic Range" or "PDR" that you use over and over. Even googling, I only see you referring to these terms. Please provide a source that will provide a definition for this term.
#6. "RE: What about low ISO performance?" In response to Reply # 5
Vancouver (WA USA not Canada), US
You've been at Nikonians for nearly 10 years. I'm surprised you have missed this.
In any case, I got 17,900 hits on Google for "Photographic Dynamic Range". This was on the first page: Photographic Dynamic Range This idea is similar to DxOMark Landscape DR. Theirs came a year later.
I also got 1,561 hits on NikoScope. Another definition, right here at Nikonians, would be in the third paragraph here: IQ - Image Quality
#7. "RE: What about low ISO performance?" In response to Reply # 6
Bill, I think Mel's point is that the term "Dynamic Range" is quite common, but the longer term "Photographic Dynamic Range" seems to be new.
My quick search on Google and Nikonscope shows all of the references to "Photographic Dynamic Range" as opposed to "Dynamic Range" being related to you. So from my quick look, it appears that you coined the full phrase.
Since they both seem to refer to the same thing, this is really just semantics, not having any bearing on the actual subject of discussion.
However, if you think "Photographic Dynamic Range" is somehow different that "Dynamic Range" as regards photography, I would be interested in the distinction.
#8. "RE: What about low ISO performance?" In response to Reply # 2
Vancouver (WA USA not Canada), US
I don't measure Tonal Range so I'll speak solely of PDR.
We all know the D3 is the low noise champ.
Actually, the D3S is the low noise champ; soon to be dethroned.
The dynamic range of the D7000 is far better than the D3 or D300
D7000 is only better than the D3 at ISO 100, about even at ISO 200, and worse above ISO 200.
In general there's no reason to use a higher ISO setting than is necessary to achieve our exposure. Clearly how high we can go limits our photographic opportunities. So in discussions we focus on that high end performance and take the low end for granted.
Speaking solely about PDR, and not for example of tonal range or color rendition; we often shoot at a lower ISO than is necessary. If the dynamic range of the scene can be captured without apparent noise at a certain ISO setting then using a lower setting will probably be visually identical. (Not that it hurts to aim low, you never know if you'll be making an aggressive crop or a larger than usual print.)
Yours was a lengthy post, and mine a length response; probably I have missed something
#9. "RE: What about low ISO performance?" In response to Reply # 7
Vancouver (WA USA not Canada), US
if you think "Photographic Dynamic Range" is somehow different that "Dynamic Range" as regards photography, I would be interested in the distinction.
There is a very definite distinction. "Dynamic Range" is almost always used in the "engineering" sense. In terms of sensors it's determined by the smallest measureable amount of light on a per photosite (pixel) basis. Photographic Dynamic Range (PDR) is different in two ways. First, PDR doesn't go down to the smallest measureable value but rather to the point that noise becomes apparent. Second, PDR is "normalized" to a standard print size and viewing distance and not reported per photosite.
(Engineering) Dynamic Range is not useful to compare different sensors (cameras); Photographic Dynamic Range is.
#10. "RE: What about low ISO performance?" In response to Reply # 7
The terminology may be unfamiliar, but the methodology for applying circle of confusion or sensor size as a way to normalize different cameras has been in general use. You really can't compare different sized sensors without some normalization.
For the sake of this thread and to move back on topic, Dynamic Range refers to a normalized value based on the methodology of DxO which is similar to Bill's methodology.
The point of the thread is that even with low noise, you give up a lot by using high ISO levels solely on the basis of noise.
#11. "RE: What about low ISO performance?" In response to Reply # 9
Noise being apparent or acceptable? What does apparent mean?
What is the standard print size and viewing distance and how does changing that change the result
To me, both the engineering and your definition of PDR are nice theoretical points, but the real key is: compared to the human eye, how much dynamic range can the camera see and record.
The noise issue is separate and distinct in determining the usefulness of the dynamic range, and is influenced by in-camera and post processing noise reduction capabilities, as well as the the viewer's judgement as to what is acceptable.
Is PDR something that you invented as a test methodology?
#12. "RE: What about low ISO performance?" In response to Reply # 11
"What is the standard print size and viewing distance and how does changing that change the result"
The standard print used by DxO is 12x8 - or about 8 mp. It really does not matter what the size is - just that you standardize the image so you can compare different sensors. In the case of a 36mp image being downsized to 8 mp, you have 4 pixels being converted to 1. That causes a lot of noise to disappear. DxO has measured the amount. But more importantly, this approach provides the ability to compare an image as it woudl be printed and viewed.
"compared to the human eye, how much dynamic range can the camera see and record."
This is the magic question, because it varies a great deal by the ISO setting. At low ISO levels the camera records a great deal of the dynamic range, but at higher ISO levels that still have little noise, the dynamic range is much less. Take a look at the DxO measurement data to see the graphs of how it changes. Bill has similar graphs. There is a very big dropoff in dynamic range at high ISO levels.
I agree that noise is a separate issue. It seems to be all we talk about with the latest cameras is noise levels, but that is just one of the three major areas of image quality. In addition to noise, dynamic range and color range are important. The latest camera offer vast improvement in dynamic range and color range - something that gets overlooked and is probably more important for 90% of your images. But if you are shooting at high ISO levels, you are not maximizing dynamic range or color range and unknowingly might be giving up something important.
#13. "RE: What about low ISO performance?" In response to Reply # 10
>The point of the thread is that even with low noise, you give >up a lot by using high ISO levels solely on the basis of >noise.
Absolutely. And what I am finding even more interesting is what you might gain from low ISO, and is it worth it. The odd shaped curve on Bill's chart for the D4 at the low end is interesting, as it seems to indicate no point in going below ISO 200. And that above 800,a D3s is just as good. And that a D7000 beats out even the D4 at ISO100.
#17. "RE: What about low ISO performance?" In response to Reply # 9
Ann Arbor, US
>Mick, > >if you think "Photographic Dynamic >Range" is somehow different that "Dynamic >Range" as regards photography, I would be interested in >the distinction. > >There is a very definite distinction. >"Dynamic Range" is almost always used in the >"engineering" sense. >In terms of sensors it's determined by the smallest >measureable amount of light on a per photosite (pixel) basis. >Photographic Dynamic Range (PDR) is different in two ways. >First, PDR doesn't go down to the smallest measureable value >but rather to the point that noise becomes apparent. >Second, PDR is "normalized" to a standard print size >and viewing distance and not reported per photosite. > >(Engineering) Dynamic Range is not useful to compare different >sensors (cameras); Photographic Dynamic Range is. > >Regards, >Bill >
This approach seems a little flawed?
By normalizing with respect to print size, you unfairly negate some of the advantages of different sensors. Depending on the print size selected, you can favor either large photosite sensors or small-photosite sensors respective strengths. From what I gather, the results of this test would change based on the (arbitrary) selection of print size.
That the small-photosite sensors seem to have equal "PDR" to the large photosite sensors indicates that the chosen "standard print size" is too small to fully utilize the small-photosite sensor, since small photosite sensors are getting the advantage of downsizing.
Also, you seem to unfairly disadvantage DX/CX sensors (lowered ideal limit?). I still don't really understand the basis for this.
Per-pixel "engineering" DR is more useful to compare different sensors' relative performance. "PDR" may be useful as a practical check ("Even though the D800 has worse per-pixel noise than a D4, I can still print at X size and ISO Y and it will look fine"), but not a truly neutral metric.
#18. "RE: What about low ISO performance?" In response to Reply # 17
Certainly there is value in looking at how pixels perform, and the relative pixel pitch. But it does not make sense to look at the "trees" without looking at the "forest".
To look at pixels would mean looking at every camera with a different print size. Can you imagine looking at a 8x12 image on one camera because it is the equivalent of a 2400x3600 sensor (8 MP) at 300 DPI, and then look at an image that is 16x24 because the camera is 33 mp? You can't compare images that are vastly different in size and expect to make sense of the evaluation. Normalization puts the image in the context of the final print or presentation.
Said another way, if your goal is to look at 20x30 prints, why would you judge image quality by looking at 4x6 prints or web sized images from any camera?
The disadvantage for DX format on FX cameras is that once you go to crop mode, you have a smaller image. If you are producing a standard sized print, that shows up in terms of image quality. The image size of an DX crop on an D700 camera is about 5 megapixels. That's smaller than a D70 image file. The small file could be a problem producing 20x30 prints as you will have to convert each pixel - including the noise - to 6 pixels after resampling.
You don't have to downsize the files of large megapixel cameras - you could upsize the file from the D700 to match. But the comparison would be the same either way.
Using a standardized size for comparison is the generally accepted approach for image quality analysis. It's used by DxO in their testing, and used by all the photo magazines in reviewing print quality.
#20. "RE: What about low ISO performance?" In response to Reply # 3
Vancouver (WA USA not Canada), US
the D4 actually as slightly less dynamic range at the lowest ISO than at 200
The drop won't make any difference in real photography but it is interesting and has been noted by other researchers.
Since I display actual data this could simply be within the accuracy of the data. In other words, maybe it's flat rather than sloping.
There are some hints in the data that the camera may actually operate the sensor in a different mode at low ISOs. I participated in a lengthy discussion on the topic at dpreview. Participants included Marianne Oelund and Eric Fossum. No conclusion was reached, we need more data than is currently available.
#21. "RE: What about low ISO performance?" In response to Reply # 20
I read elsewhere on the site that the D800 is still being tweaked by the engineers and the camera isn't the finished deal yet. Does this mean that images already available won't be the same as ones produced by the camera that is launched onto the market? There could be differences? If so then will the work already done on the image quality have to be done again?
#22. "RE: What about low ISO performance?" In response to Reply # 21
Vancouver (WA USA not Canada), US
will the work already done on the image quality have to be done again?
When the D800 (and D4) become available they will get additional testing.
The chief reason is that the "gold standard" for my measurements requires taking photos of a specific test target or is a lens/body cap shot. We don't have these for the D800 or D4. Often with a proper test target results are better, it's seldom that they are worse.
Because I'm using NEF files for my measurements it's highly unlikely that any firmware changes will have an effect. I think there's a low probability that last minute firmware changes would affect JPG output but not the NEFs.
#23. "RE: What about low ISO performance?" In response to Reply # 5 Sun 26-Feb-12 01:23 PM by agitater
Mel, a number of people, apparently independent of Bill, have jumped on the PDR term and definition as something distinctly different than engineering DR. Here's the link to Luminous Landscape's article (written by someone other than Bill):
Whether or not the method of measurement has more relevance than engineering DR as tool for helping to estimate a camera's capability in various lighting conditions and subjects is not widely understood yet IMO.
I searched the quoted term using Google and Bing and found enough non-Bill hits to indicate some interest out there in this approach to measurement. If you search for the term without quote marks in the Google or Bing search fields, you'll get nothing but conventional DR references.