In my article about re-evaluating my position on crop cameras, I mentioned that a 20MP crop sensor was diffraction limited pretty much right out of the gate with the f/5.6 lenses that make crop camera setup appealing to me. When I said that, I knew full well that one of the biggest advantages of digital photography is that software can overcome many optical problems, including diffraction. Not only that, but I also knew full well, that Canon has been advertising that their Digital Photo Professional (DPP) software — the stuff that comes with tier cameras — since version 3.11 can in fact do that.
The big question: How good is diffraction correction?
Canon first introduced diffraction correction in DPP version 3.11, as part of the Digital Lens Optimizer (DLO) module. With it came support for 29 lenses and 25 cameras going back as far as the 30D. The current version of DPP, 4.3.31, the supported lenses have expanded to over 100 EF, EF-S, and EF-M models.
Smartly, Canon’s early efforts were focused on L and EF-S lenses, as those are the two places where diffraction correction makes the most sense. L lenses because professionals demand the highest levels of image quality, and EF-S because the high pixel densities in the crop cameras and generally slow max apertures of EF-S lenses, mean that diffraction is a bigger problem sooner.
The trick, or trouble, is how to effectively test it. The diffraction corrected is part of the DLO process, and that can either be enabled, or not. The DPP UI doesn’t provide a mechanic to enable just diffraction correction, and leave the image otherwise unprocessed.
I should also point out, that my objective here wasn’t to produce an end all set of data, consider this as a preliminary test. I wanted a quick way to see if there was any real utility to using DPP to do at least lens correction processing on images before working on them in Lightroom as I normally would.
The methodology I settled on was to compare across two images. This way I could make an image at a wide aperture where diffraction spot would be much smaller than a pixel, and a second image at a narrow aperture where diffraction would be visible. Of course, this is imperfect, as stopping down also tends to increase sharpness until the diffraction limit is hit too.
For this test I used my EOS 5D mark III, and my EF 100–400mm f/4.5–5.6L IS II USM at 100 mm. Everything was shot from a tripod at ISO 100, 1/125, with a flash providing the illumination. I made a series of images at whole stops from f/4.5 to f/32 then processes them in DPP with only DLO enabled. All the rest of the settings were either unchecked or had their values set to 0.
The resulting images were exported as 16-bit TIFFs in the WideGamut RGB color space and imported into Lightroom where I matched brightnesses and applied my normal sharpening and noise reduction.
As far as matching brightnesses go, it was necessary to push the f/22 image up by a stop [?] to match the brightness of the f/4.5 image. This does mean that there will be more noise in the f/22 image. However, much like it’s necessary to match volumes when testing audio gear as louder will often be perceived as better. I felt it was necessary to match the brightness in these images to insure that the difference in brightness wouldn’t be perceived as a difference in image quality.
As a secondary test I also processed the raw images in Lightroom alone, using my standard processing settings. (click the images to enlarge to 100% magnification)
|Canon DPP||Lightroom 6|
As far as diffraction reduction goes — or really you could say lens correction as a whole — you want to compare between the two images processed with the same software.
As far as diffraction correction goes, I’m actually really impressed. Yea, there’s a little poorer detail at f/22 than there is at f/8 or f/4.5, but the difference is really negligible. There are certainly slight differences, even in alignment from image to image, but by and large fine details that should have started to become softened at f/22 are rendered quite sharp.
Actually, I’m quite impressed with Canon’s DLO correction as a whole. Remember, all 6 of these images were taken from the same 3 raw files. The only difference was what software was used to convert and process them.
Typically, and especially for a lens like this which while high quality isn’t a prime that can be super optimized, you’d expect to see an improvement in sharpness as you stop down until you reach the diffraction limited aperture. Beyond the diffraction limited aperture, you would expect to see the image become less sharp.
With Canon’s DLO, the image at f/4.5 and f/8 aren’t all the different as far as fine details go. The f/8 image is a tad better, but not so much that it’s going to be noticeable in a big print. But at f/22, the fine details look only slightly worse than they did at f/4.5 when the lens was wide open, and you can still perceive the separate lines — where they exist in my crappy print — in the test patterns.
On the other hand, Lightroom’s engine doesn’t have nearly the fine detail response across the board. But it also has a much more obvious curve to performance. At f/22 the white intermediate areas in the test stripes are fairly dark gray, while they’re still quite light in the DPP processing. Moreover, there’s very clearly a reduction in fine detail that doesn’t get removed in DPP. Notice, for example, the scratches in the large black diagonal stripe in the top center.
Long story short. Lightroom’s rendering engine is, no surprises, worse than Canon’s DPP engine for Canon cameras and lenses, at least for fine details. Of course, this is not unexpected. Moreover, to the point, the diffraction correction in DPP seems to work pretty good. It certainly narrows the gap between what I think should be the optimally sharp image for my 5D mark III at f/8 and a the well past the diffraction limited aperture at f/22.
The only real downside to this is that Lightroom doesn’t have a native way to use an external RAW engine to do conversions, and 22.3MP 16-bit TIFFs are about 128MB and DPP doesn’t support compression them. That said, while it’s a pain in the butt, I can certainly see dragging some images through DPP and back to get the best possible quality.