Canon S90 vs Ricoh GR Digital III: Lens Performance


Hall of Famer
Many people who are interested in a Ricoh GR Digital III are also interested in the Canon Powershot S90 IS. Like the Ricoh, the Canon is small (smaller than the Ricoh, in fact), light, has excellent manual controls, a proven heritage, a high end small-sensor, and a fast, wide lens.

Beyond that, Canon managed to pack in a nearly 4x zoom and image stabilization, two features the GRD III lacks. So what gives? For one thing, barrel distortion. Severe with the S90, modest with the GRD.

Other than barrel distortion, how does the Canon lens hold up? We'll see in this comparison.

Technical details:
- Both cameras shot in RAW mode with self timer on a sturdy tripod
- Each image taken in triplicate, refocused for each shot, with the best result used for comparison
- IS was disabled on the Canon
- Each camera was used in aperture priority mode at the lowest ISO (ISO 64 for the GRD, ISO 80 for the S90)

I had a difficult time deciding whether to fix the barrel distortion or not in the Canon CR2 files. On the one hand, correcting the barrel distortion puts the Canon files at a disadvantage since resolution is lost in the process of correction. On the other hand, the correction chops out the edges and corners of the Canon files, eliminating the least sharp portions of the frame. In the end, I decided that few photographers will choose to leave pronounced barrel distortion in their images, and therefore the more useful comparison is that using the corrected Canon files.

The next decision was whether to use Canon's Digital Photo Professional (DPP) or something else to process the S90 files. In this case, I felt that using DPP for the S90 files and a different RAW processor for the GRD files would introduce an unnecessary variable with regards to sharpening and noise reduction. Instead, I converted the Canon files to DNG in Lightroom 2.6 Release Candidate. In the process of conversion, Lightroom addresses the S90 barrel distortion.

The native GRD and corrected S90 DNG files were then converted in Lightroom 3 Beta with identical sharpening for the sake of this comparison. I used version 3 Beta for the final processing because it lacks the detail-destroying luminance noise reduction of Lightroom 2.

Here is an overview of how the resized images looked with regards to barrel distortion:


Click image to view larger version

The yellow boxes below depict the regions used for 100% crop comparisons:


The GRD III cannot be set to f/2, so we'll start by looking at how the two lenses perform wide open, with the GRD lens set to f/1.9 and the S90 lens to f/2... One last disclosure: Both cameras were overexposed in the wide open comparison. There was a lot of light, and camera testing ain't easy.

Okay, here we go, center:


Neither of these lenses has a problem capturing center detail at any f-number.

Bottom edge:


More detail and contrast for the GRD.

Left edge:


Better detail from the S90 file here. Looking further, there is a bit of purple color fringing in high contrast areas of the S90 file.

Right edge:


GRD is looking better here on the right.

I think it's worth pausing here to try and explain the discrepancy between the left and right crop comparisons. While I cannot rule out contributions from field curvature and a touch of decentering affecting the GRD lens, it has become obvious to me that my S90 lens is affected by a small amount of decentering. As a result, the right edges of my S90 frames are more blurry (and likely the left edges are less blurry) than they would be if the lens were perfectly centered.

I've owned a ton of compacts, and my S90 lens decentering is no worse than the decentering on any number of others. This is a very common occurrence if one looks for it*.

*Several individuals describe a similar issue in this DPReview forum thread, including one fellow who exchanged his camera and received another one with the same issue. A soft right edge/corner is also notable in DC Watch's review unit as seen in this picture.

Moving on, here is the left upper corner:


Purple fringing again evident in the S90 crop.

Right upper corner:


If you think that is bad (S90), you ought to see the extreme corner before barrel correction!

Little has changed in the center at f/2.8:


Bottom edge at f/2.8:


Left edge at f/2.8:


To my eye, the GRD has nearly caught up.

Right edge at f/2.8:


The S90 has improved but still struggles. Again, purple fringing is present.

Left upper corner at f/2.8:


Right upper corner at f/2.8


Center at f/4:


Bottom edge at f/4:


Left edge at f/4:


Right edge at f/4:


Left upper corner at f/4:


Right upper corner at f/4:


Center at f/5.6:


Bottom edge at f/5.6:


Left edge at f/5.6:


Right edge at f/5.6:


Left upper corner at f/5.6:


Right upper corner at f/5.6:


1) The S90 lens has pronounced barrel distortion relative to the GRD III lens.
2) The GRD III lens shows better overall sharpness when considering the entire frame at a variety of f-numbers. That said, the S90 lens puts in a very respectable performance.
3) The S90 suffers from color fringing, whereas the GRD does not.

11/25/09 Addendum:

I purchased a second S90 to see whether it too would be affected by decentering, and I can report that it is largely free of this issue. It isn't perfectly symmetrical, but it's lens is more centered than the one tested above. As a result, the left edge/corner performance is a bit worse than shown above, and the right edge/coner performance is a bit better than shown above.

The net difference is essentially zero. In other words, the second tested S90 remains a notch below the GRD with regards to edge/corner performance, and the difference persists at all f-stops.

Originally published on the old Serious Compacts blog. Older comments can be found here: Canon S90 vs Ricoh GR Digital III: Lens Performance