Most of us take for granted that large sensor digital cameras have better image quality than small sensor digital cameras. Why else would someone spend $30K on one of these? Though some disagree, I feel that it is best to use equivalent images when comparing image quality from one format system to another. The term "equivalent images," as used throughout this post, refers to images with the same output (print) size, angle of view (AOV), perspective, apparent exposure, and depth of field (DOF). Joe Mama, whom I know from the DPReview forums, has written an extensive essay covering the issue of equivalence between different digital formats. There have been many discussions in the DPR forums about "equivalent" images, and the general consensus has always been that large format systems always match or exceed the image quality of small format systems when equivalent images are compared. Yet, there has been one niggling exception to this. Image quality from digital medium format systems seems to compare poorly to that of 35mm digital systems when compared at high ISO or in very long exposures in low light. This exception has been largely dismissed as evidence that medium format digital is not yet as mature a technology as 35mm digital.
Conventional wisdom suggests that digital small format systems have an advantage in deep DOF photography. However, a number of parties have debunked this apparent myth. Clarkvision has an excellent technical discussion, titled " The Depth-of-Field Myth and Digital Cameras." Quoting the conclusion of that article, Clark wrote the following:
"Given the identical photon noise, exposure time, enlargement size, and number of pixels giving the same spatial resolution (i.e. the same total image quality), digital cameras with different sized sensors will produce images with identical depths-of-field. (This assumes similar relative performance in the camera's electronics, blur filters, and lenses.) The larger format camera will use a higher f/ratio and an ISO equal to the ratio of the sensor sizes to achieve that equality. If the scene is static enough that a longer exposure time can be used, then the larger format camera will produce the same depth-of-field images as the smaller format camera, but will collect more photons and produce higher signal-to-noise images. Another way to look at the problem, is the larger format camera could use an even smaller aperture and a longer exposure to achieve a similar signal-to-noise ratio image with greater depth of field than a smaller format camera. Thus, the larger format camera has the advantage for producing equal or better images with equal or better depth-of-field as smaller format cameras."
While recently considering a candid portrait I took of my older son Oliver with the Fuji F30, it occurred to me that when a certain shutter speed is required in low light, small format systems may indeed have a significant advantage for deep DOF equivalent images. The Fuji F30 is a compact, small sensor digital camera with a 1/1.7" 6MP sensor. This photo was taken with available light at ISO 1600, 1/38s, f/5:
To get an equivalent image handheld at the same shutter speed using my Canon 5D, one would need to use the 5D at an f-number and ISO value each approximately 4.4 stops [calculated as ln(43.3/9.5) / ln(sqrt(2) using the diagonal dimensions of the sensors since the aspect ratios differ] higher than those used on the Fuji - approximately ISO 34,000, f/22. I am certain that given similar ambient lighting, my Canon 5D would not produce that nice of an image at ISO 34,000 at the same output size. Do we need to debunk the debunking of the myth of small format deep DOF advantage? In response to my questions in the DPR forums, John Sheehy gave a technical explanation for why small sensor cameras should have an advantage in high ISO, low light, deep DOF, shutter speed-limited photography.
John stated the following:
"You're overlooking the fact that the noise down in the deep shadows has nothing to do with photon counts. P&S and FF DSLR sensors vary far more greatly in photon capture than in read noise. Most of the existing Nikon DSLRs have similar or the same read noise as P&S cameras do, relative to RAW saturation (or, in ADUs).
You can't model camera noise in terms of photons collected, until you remove read noise from the equation. Most of the extra noise you'd see with a 5D or D3 pushed to ISO 70,000 is due to read noise. The fact that the most objectionable things in the image are patterned tells you right there that it is read noise that is being boosted to objectionable levels, as there is no patterning in shot noise."
I still wondered whether the much-discussed Fuji F30 6MP Super CCD was doing something special here, or whether other small sensor cameras might do as well. I therefore shot a simple test using the Ricoh GX100 and Canon 5D with 50mm lens to create equivalent images with a deep DOF in low light with shutter speed matched. The Ricoh has a 10MP 1/1.75" sensor. To match the diagonal AOV of the 5D system, I used a focal length of 10.5mm with the Ricoh. To otherwise meet the criteria for equivalent images, the Ricoh was shot at f/3.9, ISO 800, 1/10s; whereas the 5D was at f/18, 1/10s. The 5D cannot natively shoot at the necessary ISO, so I took the image underexposed at the maximum ISO of 3200 and then "pushed" the exposure during RAW processing.
Both images were processed from RAW using Adobe Lightroom 1.3. In fact, all I did was choose "auto expose" in Lightroom, and Lightroom pushed the 5D exposure by an additional 2.4 stops, almost exactly where I would have calculated it (final ISO equivalent of 16,890). The resulting histograms for the two images matched closely. All other values in Lightroom were left at default. The resulting Ricoh TIFF was upscaled by 15% (using Bicubic Smoother) to match the diagonal dimension of the 5D image. At this point, I ran both images through Noise ninja using "Auto Profile" and default settings with "Coarse Noise" checked. This somewhat helped to close the gap, but the GX100 file still has significantly less patterned noise. Click here to see the final, full-res 5D image and here to see the full-res, upscaled GX100 image.
Here are the 25% resized images, similar to what one might see in a small print. I ran both of these through Photoshop's Auto Color before resizing them using Bicubic Sharper.
First the 5D:
Now the GX100:
The 5D held up pretty well here, but the small sensor camera took the day!
If you've read through this long post, you may be wondering, "Why does this matter?" After all, most folks associate serious deep DOF photography with landscape photography, where tripods are generally used and shutter speed limitations do not apply. Low-light street photography is one application in which deep DOF often desirable, and tripods are generally not an option. In addition to obvious operational advantages (discreet appearance, light weight, etc), small sensor cameras would seem to have an image quality advantage in this setting. Another such setting would be macro photography of a moving subject, such as a flower in the setting of a slight breeze.
If anyone would like to try their hand with the RAW images used in this comparison, they will be available for 10 days or 100 downloads by clicking here (sorry for the associated pop-up ad).
Addendum: Some discussion of this post can be found at D-Spot (Hebrew) and at POTN (English).
Originally published on the old Serious Compacts blog. Older comments can be found here: http://seriouscompacts.blogspot.com/...ever-have.html
6/21/12 update: I'm not claiming that all small sensor cameras will have this advantage over all large sensor cameras. However, my own experience with many small and large sensor cameras over the years has borne out the fact that in most such comparisons of small and large sensor cameras of a given technology "generation", there is an advantage similar to the one demonstrated here for the smaller format under low light, deep DOF, shutter speed limited shooting conditions. More research and development dollars go into small sensor development than large sensor development, so it should come as no surprise that square millimeter for square millimeter, a current iPhone sensor offers better technology than a current 5D sensor.