Proof is Better Than Theory
Last week I was theorizing that my current cameras are all that I need. Of course, the flaw in this statement is that one doesn't know what one doesn't know. Rather than rest comfortably on my untested theory, I decided to put it to the test — at least as best I could.
I had two questions that I set out to answer based on two real-world scenarios:
- Will I see any improvement in prints of the size I've been making (for folios and chapbooks) from a camera with more megapixels? (I'm basically making prints roughly 8x10".) That is, will a down-sampled image from a higher megapixel camera produce a visibly better print with even more detail?
- Conversely, will I seen any compromise in quality if I do want to make a larger print (say, 16" on the long side) with my existing camera? Would an up-sampled image from my 12 megapixel camera produce a visibly inferior print? This is a much more challenging test and one I was really curious about in this age of ever-increasing megapixel hype.
Note that I am limiting this test to something I might actually someday do, not some fantasized 30x40" print that I would never likely do. I'm not sure I've ever made even a 16x20, so there is not much likelihood that I will need greater than that capability in my near future. It's just not me. So, if you are an ultra-large print fan, you can bail out now. Nothing to see here.
My current camera is a 12 megapixel Panasonic G2. I could be tempted into its 16 megapixel bigger brother, the GH2 (or newer GH3 or even the G5) so I could use all my existing lenses. I chose to test using the GH2 because control images are available on that model. The other camera I've been looking at is the Sony a77 — a serious upgrade to my system that could suck up a lot of cash for both camera and lenses, but also a serious step up in capabilities and specs.
Bottom line for the purposes of this test
First, apples and apples. Rather than look at a bunch of random images, it's always best to eliminate as many variable as possible. I downloaded the "Photographic Test" RAW files from dpreview.com, all shot at the lowest (and presumably highest quality) ISO. Different lenses, but there would be in the real world, too, so I guess that's a variable that will, ahem, vary.
Second, in my real world I always fuss with an image in post-processing — sharpening, noise reduction, clarity expansion, etc. Since I wanted a real-world test, I processed each image as I normally would in the real world — sharpening, noise reduction, clarity.
Using these RAW files as my starting point, I up-sampled or down-sampled as needed and then layered the images in a single registered Photoshop file. There were two resulting PSD files — a 6000 pixel file (with the G2 and GH2 up-sampled using Photoshop's recommended "Bicubic smoother"), and a 4000 pixel file (with the GH2 and the a77 down-sampled using "Bicubic sharper"). I know there are various alternatives to these simple Photoshop resampling methods that have adamant adherents, but for this test I just wanted to keep it simple. If you are interested, the two layered PSD files can be downloaded from here in a single 378mb zip file.
Here is a screen video with my comments about the on-screen pixel-peeping phase of the process. Caution here: Pixel peeping is not real-world, and trying to do a video of pixel peeping is just asking for trouble. Nonetheless, I think you will find this an interesting milepost on the way to the printed conclusions below.
In the video there were discernable differences that were clearly visible, especially at 100% on-screen magnification, I'll grant you that. But, visible differences when viewed on-screen at 100% magnification — something no one ever does (except for us pixel-peeping photographers) — is decidedly not real-world. Even then, the differences were modest and certainly not persuasive, at least not $5,000 worth of persuasion. Besides, that wasn't the criteria of the test: I was interested in printed visual differences. You'll just have to print out my test PDF for yourself and look with your own eyes.
As the final step, I placed all these files into two InDesign documents at 100% scale and printed them on glossy baryta paper for maximum sharpness using my Epson 4880. If you want to do the test on your printer, here are the PDF files.
Drum roll please. And the winner is . . . wait for it . . . ain't no winner.
No, the prints are not identical. By staring hard and examining minutely at close distance (roughly 8-9"), there are minute differences between prints from the three cameras. Barely. In fact, what's less than barely? Microbarely? The differences are so subtle and so insignificant that I had to work hard to see them — really hard. I was looking to see the differences and searching with diligence to find them. Who, in the real world, does that?
From 3-feet, there is no discernable difference at all. Period. Double period with an exclamation point.
Yes, but my eyes are 58 years old and peer through bifocals. Just to verify my impressions, I had my young assistant — whose eagle eyes catch everything — take a look. She agreed with my conclusions in spite of the fact that she knows I would have preferred her to find enough difference to help me justify the new camera purchase.
I'll go even further: Without the text labels on the prints to identify which camera they were from, after shuffling I'm sure I could not have told you which image was from which camera — that's how indistinguishable they were from one another. And I guess that's the real-world point I was trying to determine in the first place.
And here is the most interesting part of this little test: all the above applies equally well to the up-sample (G2 and GH2) prints at 16" and to the down-sampled (GH2 and a77) prints at 10". No visibly discernable differences at these print sizes that anyone would ever notice, let alone care about if they did. None. Q.E.D.
Shooting Holes in the Test
There are lots of holes in this informal test that must be acknowledged.
- If you need or want to make larger prints — larger than 16" on the long side — all bets are off. You might very well find you need to go with the camera with more megapixels, I presume. I've not tested larger print sizes.
- If you need or want to crop your images, go with the camera with the most megapixels.
- Because the up-sampling of the G2's 12 megapixels up to 24 megapixels produced acceptable results (1.5X linear increase in size), then it might be reasonable to think the a77 could be up-sampled to 48 megapixels and make 25" prints. Testing would be needed to confirm this, but it's be a pretty safe bet that 1.5X linear increase in print size would work fine here, too. It would not, however, be an intelligent bet that the G2 could be up-sampled to 48 megapixels without serious problems. Unless you want to view prints from across the stadium.
- Lots of the cameras' features do not compare, but that wasn't the objective of the test. I was only looking at print quality. The Sony a77 is a dream camera in terms of functions and capabilities, but I cannot say it produces better prints at the sizes I make. Just for grins, I stopped by my local Sony store last weekend and took a look. It's a tank and an amazing machine — powerful, flexible, fast, responsive, intuitive, and that great Carl Zeiss lens line to back it up. I wish I could justify it with more than mere lust. The only downside was that is it uncomfortably heavy compared to my G2 system. Oh, yes, and then there's the $5 grand for camera and accoutrements.
- I started with RAW files and tweaked from there. I can offer no opinions about in-camera JPGs because I almost never shoot them.
- Also, there were visible differences in the straight-out-of-the-camera RAW images. But who in their right mind uses them? Post-exposure processing is part of the process and I would never dream of leaving an image unsharpened or otherwise untweaked to pull out everything there is to be pulled out. Again, real-world results using real-world processing was the objective of this experiment.
My initial assertions turn out to be correct — I would not see any improvement in my prints — at the sizes I typically print. Damn. Clearly, for PDF and other digital publications, this conclusion holds true even more so. Oh well, I guess I'll stick with what I've got for a while. Or, I'll have to start cooking up a reason for making wall size prints . . .