10 megapixels vs. 6 megapixels - the sequel

[jlw]

A number of people seemed disturbed by my recent thread in which I subjectively observed that in one particular picture-taking situation, the pictures I shot with my 6-megapixel R-D 1 seemed better in overall quality than the ones I shot the next night with my 10-megapixel Nikon D80.

To subject this to something more closely resembling a "controlled test," I did some more careful test photos today. Rather than let the cat out of the bag right away, I think I'll show the results first and then disclose which are which. (No fair peeking at the EXIF data!)

Test conditions: both cameras set to EI 1600; Nikon's "High ISO Noise Reduction" at default setting; both with 50mm lenses (C-V Nokton for the Epson, 50/1.8 AF Nikkor for the D80) set to f/8. I set exposures by making test shots at different exposures until I got histograms as far to the right as possible on both cameras' displays without entering the clipping region; as it happened, the shutter speeds were very close to each other.

One important difference from my earlier non-test is that in this case I was using diffused daylight; the first time around, the lighting was tungsten. Various people, such as Sean Reid, have noted that tungsten lighting is an especially tough noise-control challenge since the gain of the blue channel has to be turned up so drastically to get neutral color rendition. So, eventually I'll have to repeat this process with tungsten light to see whether the results differ significantly from daylight.

Now, here's what I got. The first set of attachments show my test scene overall; I've downsized the photos to be RFF-friendly and cropped them to look as identical as possible, so this is just a reference:

 

[jlw]

10 megapixels vs 6 megapixels - part deux

10 megapixels vs 6 megapixels - part deux

Now here are two fairly anonymous sections cropped from the original, full-size raw files. They are shown pixel-for-pixel at actual size. The purpose of this part is to judge the "smoothness" of the actual image. (Of course, since JPEG compression adds its own texture -- even at the 100% quality setting I used -- your judgments lose a bit of accuracy here.)

 

[jlw]

10 megapixels vs 6 megapixels -- part III

10 megapixels vs 6 megapixels -- part III

As Sean Reid pointed out in his Leica M8 review saga, images of different pixel counts often have to be output at constant sizes. He used this position to rationalize the M8's subjectively higher noise level compared to the R-D 1, noting that downsampling the M8's 10-megapixel file yielded a subjective smoothness similar to the R-D 1's 6-megapixel file.

So, I've taken the same approach here: I downsized the Nikon's file to the same width as the Epson's, using settings for both that would yield an 8-inch-wide print at 240 printer pixels per inch (the figure I usually use for my Epson R800 printer.) Then I cropped sections out of both files so I could show them pixel-for-pixel at an RFf-friendly file size.

Here they are:

 

[jlw]

10 megapixels vs 6 megapixels - conclusion

10 megapixels vs 6 megapixels - conclusion

So, have you decided which versions you preferred? Then here comes the "reveal," as they say on Trading Spaces and Overhauliin'.

The "a" pictures came from the D80. The "b" pictures came from the R-D 1.

Even before knowing the answer, you probably had decided on which rendering you preferred. (And if you were honest, you probably also had decided the differences aren't huge either way.)

Here's my personal takeaway:

-- The Nikon pictures look somewhat sharper and more detailed, even when downsampled to the same final print size.

-- On the other hand, the R-D 1 pictures look smoother in tone (see the purple slices) and still look somewhat smoother even when the Nikon's images have been downsampled to the same final print size.

Of course, information doesn't become knowledge until it acquires utility. Here's how I'm going to apply what I learned:

-- I'll lean toward the R-D 1 for low-light shots in which smooth tones are more important, such as people pictures.

-- I'll lean toward the Nikon for low-light shots in which fine details are important, such as interiors and architectural details, or when I know I'm going to need to make big prints.

-- In most situations, though, the image-quality differences are small enough to be trumped by other considerations. In other words, in situations in which an RF is preferable to an SLR, I'll continue to use the R-D 1 without much worry about how much fine detail I'm giving up; in situations in which an SLR is preferable, I'll use the D80 without much worry about the extra noise.

Of course, that's still pending what I find out about tungsten light; the smoothness differences in my first set of pictures (and in my other experience using both cameras under tungsten light) were striking enough that I still think that may be a significant difference.

I'll also need to try the Nikon's other settings for high-ISO noise reduction, to see what effect it has.

That's enough sorta-science for now, though -- I've got pictures to take!

 

[rogue_designer]

Nice breakdown, and I think your summary of usage is about what I would have expected, and leaned towards regardless. Thanks for taking the time to come back to this.

 

[iml]

This sort of confirms my subjective impressions from playing with a few 10MP DSLRs.

Ian

 

[Gabriel M.A.]

Very interesting. Thank you for the demonstration.

 

[Joe Mondello]

That was interesting.

I had decided I liked A better than B but as you say, not by much.

I am a D200 shooter with a ton of Nikon lenses and an R-D1 re-furb arriving tomorrow, so this is timely for me!

 

[pvdhaar]

I've had a 6MP picture that I took with a DSLR printed at 1.5x1meter. Nothing wrong with it. Clean as a whistle. My pixelpeeper-colleagues made no remark whatsoever on the technical side of things..

This leads me to firmly believe that if you can't make a good looking picture with 6MP, there's nothing wrong with the megapixel count, but with the picture itself..

 

[IGMeanwell]

This an interesting comparison

I continue to look over to the RD1 with green eyes as it seems as though it would be refreshing to use.

I'll be honest... one reason I have continued using my 6 mp DSLR is I haven't encountered a situation where I have had to enlarge to the point of making the 6mp sensor an issue. Even at ISO1600

The only major selling point of the D80 for me is the bundle of SD cards I own and the viewfinder (which is excellent)

The other issue is file size, 10 mp files are huge!!! I just managed to upgrade my computer which now has 250GB (and a DVD RW drive) ... however before that it was a 40GB with a 200gb external drive and I was suprised on how fast I have been filling up that hard drive (with RAW files).

but honestly whenever I do purchase a second digital body (DSLR or Rangefinder)... I will be looking for the best dynamic range, shadow detail, and whatever utilizes the best image size with its given sensor whether its 6mp or 12mp

 

[HansDerHase]

Another evidence that 6 MP is enough for all practical purpose.

Thanx for posting.

 

[freeranger]

I opened the first 2 images side by side. The one I found most appealing was picture b. Very interesting. - thanks for fuelling my GAS.😀