Review – Canon S90, Part 3: Image Quality
Any engineer will tell you that all technical designs are a study in compromise, and this is palpably, and visually, true for cameras. We can cry out for tiny cameras as much as we like, but we’ll have to pay a price in image quality (IQ)—there is no way around that. So while the S90 is pocketable, is its IQ good enough that you will actually want to put it in your pocket? For those of you who don’t want to read the rest of this article, here is my answer: Yes. For those who don’t just want to take my word for it (and you shouldn’t), keep on reading.
The Trials and Tribulations of a Small Sensor
The S90’s 1/1.7″ sensor has a surface area of 43.3 mm2, by comparison an APS-C sensor has an area of about 385 mm2 and a full-frame of 864 mm2. Despite this huge difference in area, the S90 produces good images with pleasing colours. Where I’ve found the S90 to lag is in sharpness, with images having the typical telltale, mushy feel of a P&S camera. I can recover some detail with judicious sharpening, but I’ve found I need to pay a lot more attention and work harder with the sharpening than I need to with my DSLR. Happily, Canon decided to scale back and only crammed 10MP into this little sensor, but the pixels are still much smaller, and thus noisier, than those in a DSLR. Another issue with small sensors is achieving the wide angle FoV, especially with zoom lenses; in most cases wide angles will exhibit noticeable barrel distortion. This can be corrected in postprocessing, and I’ll be taking a look at how well the S90 files cope with this.
No lab tests here, just some good old-fashioned real-world photography. Exhibit A: Bright, sunlit buildings (which are actually misshaped, that’s not lens distortion) with shade in front. The sky is nice and bright, as are the lit parts of the buildings, yet there’s plenty of detail in the shadowed area. Honestly, this is all I need to know.
The S90 at Low ISO
I will not be publishing 100% crops at various ISOs, as those can be found elsewhere on the internet (for example, at the Imaging Resource S90 samples page). Mine will be more of a qualitative comment on low ISO performance. OK, here it goes…
It’s very good. At the very least, it’s good enough. Good enough for what? Good enough for me.
Seriously, I don’t know what else I can say. Search on Flickr and you’ll find almost 4,000 photos tagged with S90, take a look and see what you think. Colours are nice, JPEGs are sharp, RAWs are a bit soft (so you can apply your own sharpening to taste) and life is generally good. Granted, it’s a P&S, so you shouldn’t compare it to a DSLR (not that that’s stopped some people), but I don’t see a difference in web-sized images. There are differences at the pixel level, but I suspect these won’t translate to differences visible in 8×10 prints. In fact, I plan to test this in the near future to see if it’s true&mdasg;I’ll keep you posted.
Here’s a snapshot at ISO 80, 40mm-equiv., f/5.6, and pretty much auto everything else. I shot in RAW and converted with ACR, but I only applied my standard S90 settings to the image, meaning it looks quite close to how the camera-made JPEG would have looked. I love the sky, which really was that blue.
For all the pixel-peepers out there, I’m also including a 100% crop for your viewing pleasure. I didn’t perform any special sharpening, which I normally would if I were cropping an image this much (like my white squirrel). Some may have noticed that the tones of blue in the sky don’t transition smoothly. Welcome to the world of small sensors (although DSLRs can also render tonal gradients in this way). I don’t think this would be a problem in a print.
Here’s another pic at ISO 80 showing pleasing colours (shot at the long end, wide open):
The S90 at High ISO
Now things begin to get interesting! A P&S claiming 3200 as its maximum ISO has some big proving to do. Again, if scene comparisons is what you want, please go to the Imaging Resource S90 samples page. What I’m going to do here is show you a scene at ISO 1600 in 3 different versions: as it came out of camera via the JPEG engine, as a RAW processed with DPP, and as a RAW processed with ACR. Instead of showing 100% crops, you can just take a look at the full-res images by clicking on the thumbnails below.
There are clear differences: The out-of-camera JPEG shows the least amount of chroma noise, yet there is little detail left and the colours are dull; the DPP-processed image (with automatically chosen noise reduction settings) shows slightly more detail and is less washed out, but the colours are still dull; finally, the ACR-processed image (with no real noise reduction) is clearly noisier, yet the colours are truer to life and there is a bit more detail. There’s no such thing as a free lunch, so pick your poison: dull, featureless, noiseless images, or vibrant, slightly-more-detailed, noisy ones. It’s your choice! If you plan to use external noise reduction software, shooting RAW is definitely the best option; if you process with DPP, don’t forget to turn off noise reduction before exporting to your preferred noise reduction program.
I (and Canon) consider high ISO to start at 800, so let’s see what the S90 can do at that sensitivity. Enter the surprised squirrel. Shot at ISO 800, 105mm-equiv. f/4.9, 1/100s (and slightly cropped, I confess):
Let’s satisfy our innermost pixel-peeping urges and look at a 100% crop (feel free to click to enlarge):
I added a touch of sharpening to improve the squirrelness. This is not a great image, but it’s not that bad either, bearing in mind it’s from a P&S with the lens wide open and at ISO 800.
Final verdict: The S90’s high ISO capabilities are impressive for a P&S. No, it’s not as good as a DSLR, but we shouldn’t be comparing it to one (although some people do). That said, it was only a few years ago that DSLRs had similar high ISO performance, so I’m happy to see this good behaviour in a P&S. If you’re wondering why I didn’t show an ISO 3200 shot, it’s because it’s one stop worse than ISO 1600, and if you use it, to so at your own risk and don’t come crying to me if your pictures have no detail.
Oh what the hell, here’s a sample at ISO 3200. I’m only posting it at websize, though. Processed via ACR with Chroma and Luminance noise reduction added to taste.
I only had one real opportunity to test this out, so bear with me. Here are 3 examples; click on them for the full size.
They look fine to me, which makes me wonder why Canon were afraid to add a bulb shooting mode (or exposures longer than 15s). I mean, that’s why they didn’t give the S90 a bulb mode, cos they were afraid. Right…? Add this to the list of things that baffle me about this camera. And yes, I understand that to add a bulb mode would have meant adding a connector for a cable release…but this wouldn’t have been a bad thing to begin with.
Let’s not beat around the bush: The lens on the S90 provides a healthy dose of barrel distortion at all focal lengths, although it is worse at 28mm-equiv. It seems the camera’s firmware corrects some of it in-camera, and I say “it seems” because I have not been able to find any official confirmation of this fact. Tom Niemann, creator of the PTLens software tells me the RAW files don’t have quite as much correction as the JPEGs out of camera, which means the camera must be performing some type of correction. And speaking of PTLens, I contacted Tom about adding the S90 to his list of supported cameras and he was most gracious. He sent me instructions on what photographs I should take with the S90, and after I sent him the files, he calculated the correction parameters from them and created a profile for the S90. I have been using PTLens for over 2 years and am very happy with it; and at $25 it is one of the cheapest programs (or Photoshop plug-ins) available. You can download a trial version here.
|Uncorrected||Corrected with DPP||Corrected with PTLens|
Canon’s RAW converter, DPP, can also completely correct distortion. When using it early on (before Adobe released a profile for the S90’s RAW files) I did all distortion correction with DPP. Having compared it to PTLens, I see only a slight difference, with DPP correcting a bit more than PTLens. Forgive me, but I haven’t photographed graph paper to measure which correction is truer, but PTLens’s looks good enough to me, so that’s what I’ll be using. On the plus side, I have found no loss of detail in in-focus areas associated with the correction. I have, on some occasions, seen a slight loss of detail in out-of-focus areas in some images. I find this strange and cannot readily explain it, but because the difference occurs only occasionally, in a part of the image that is not as important, and the loss of detail is not great, I have not lost any sleep over it.
Because correcting distortion adds an extra step to my post-processing, I only apply it when absolutely necessary. I’ve found that in most cases the barrel distortion is either not noticeable, or not objectionable. Oh, and DPP can automatically correct vignetting too (although you have to tell it to do it), but PTLens requires manually selecting the parameters. If this statement seems like an afterthought, it’s because it is—I just haven’t found vignetting in the S90 to be a big deal.
The last issue associated with distortion correction via software is the reduction in the FoV. Because the image is “squeezed in” from the sides, the edges have to be cut off so as not to leave the photo with the shape of a bow-tie. From those of you expecting me to tell you how many degrees of FoV you lose at each focal length, I beg forgiveness—I had neither the time nor inclination to test this. To the eye, it doesn’t seem like much is lost, and if instead of 28mm-equiv. the wide end is 29mm-equiv., then so be it; I will add that to the list of compromises I am happy to make in order to have this camera in my pocket.
The S90 does macro at the wide end, which I find annoying. Macro reproduction is about 1:10; to obtain the same magnification with an APS-C DSLR you’d need a lens capable of 1:3 macro. I haven’t done any serious macro shooting, but here’s a flower at close to maximum magnification:
And (you guessed it), here’s a 100% crop:
Make your own mind up.
Again, I was lazy. I just took 3 pics at different apertures shooting into the Sun. Click thumbnails for larger version and judge. I think flare control is very good, and I especially like those Sun spikes at f/8.
You will have to forgive me here for not presenting some evidence for what I am about to say, but a) I simply do not have time right now as I’m travelling in 3 days, and b) The file I was going to use as an example I converted the CR2 to DNG, and then deleted it. You’ll just have to take my word on this. The S90 does quite well in dealing with CA, which I have managed to induce only in areas of very high contrast. It seems to favour blue or green fringing, which is much less unsightly than the purple variety. DPP does a wonderful job of dealing with it, although it’s not quite automatic and you have to play around with sliders to get the best result. ACR does not handle it quite as well, but is still acceptable. In general, I have not found CAs to be a problem, but do be careful of tree branches against cloudy skies.
RAW vs JPEG
I don’t want to get into a general debate about one shooting style vs the other, just give my opinions as they pertain to the S90. Those photographers who shoot straight JPEGs have nothing to worry about with this camera, whose JPEG engine works very well up to ISO 800 (in part aided by its superb auto white balance). From ISO 800 onwards the camera does a great job of killing noise, but at the expense of lost detail—for some this is not an issue, and at web viewing sizes there is little difference. If you can dial in the JPEG settings so that images come out with the colour, saturation and contrast you like, you’re good to go out of camera.
I shoot RAW, not because there’s better image quality to be obtained, but because there is more latitude when postprocessing images, which I regularly do. Most of my photography is in B&W, which involves a conversion with selective channel processing, not to mention highlight and/or shadow recovery—a RAW file fares much better under such assaults. If all I did in postprocessing was bump up the contrast and maybe add a bit of sharpening, I would shoot JPEG without a doubt.
No camera is perfect, and the S90 is as far from it as the best of them, yet it is still eminently useful. I will not be returning it nor selling it off, and I foresee it giving me a few good years of use as a pocket cam. I Despite its drawbacks, none of them make me feel like I’m fighting the camera every time I take a photograph. As I’ve grown used to it during these past weeks I have figured out a workflow to photographing with it that I’m happy with and seems to give me the results I want. I’m still learning the ropes when it comes to postprocessing (it was only admitted into the Adobe stable of supported cameras some 3 weeks ago) but so far I am happy with the IQ I can extract from the RAW files.
As far as the user experience goes, I find the camera user-friendly, and what’s more important, photographer friendly. I have larger than average hands yet can handle the camera just fine. Having the shutter button so far into the top plate took some time getting used to, but now it’s a non-issue, as is the rear control wheel that is annoying so many other people—it just depends on how you hold the camera. As for another requirement of the forum brigade, autofocus speed, I don’t feel like I’m qualified to grade it, especially after reading the multitude of opinions on the autofocus speed of the Olympus E-P1. The S90’s autofocus speed is faster than I can focus manually, yet slower than a speeding bullet.
Will it make my DSLR obsolete? Absolutely not, but that’s not why I bought it. I wanted the best IQ I could carry everywhere in my trouser pocket, and the S90 is just that: The smallest compromise I could make in the smallest P&S I could find.
To end this 3-part review I want to list the pros and cons I find for the S90. Many of the cons could be easily fixed via firmware, and while I would love Canon to address them themselves (as I’m not the only one with similar complaints), I suspect the folks at CHDK will do it sooner (some hope here).
- Small, light and pocketable
- Good JPEGs and RAW files (superb for a P&S)
- Impressive high ISO performance for a P&S
- Fast f/2 aperture at the wide end
- Superb auto WB which can be manually fine-tuned
- PASM shooting modes plus very useful Custom mode
- Highly customisable buttons and dials (including lens ring) allowing direct or quick access to most important features (i.e., superb user interface)
- Very fast start up time
- Powerful, adjustable onboard flash
- Accurate AF (with assist lamp in low light)
- Can review images with camera turned off
- Very nice, bright, crisp LCD screen that can be used in strong sunlight light
- Manual mode disables auto ISO (Canon, give us auto ISO in M mode!)
- Manual focus offers a poor distance scale and no indication of depth of field
- Manual focus mode “disappears” when camera goes into sleep mode
- No hyperfocal focus mode
- No way to turn off sleep mode
- No RAW shooting in modes other than P, Av, Sv or M
- The lens is very slow at the long end (f/4.9); if Canon could make it f/3.5, or even f/4, it would be so much better
- One cannot customise the auto ISO range and the manual doesn’t even tell what the range is for each shooting mode/scene
- No bulb mode (with 15s being the longest exposure available)
- No cable nor IR remote release
- No hotshoe (though not much of a problem for me)