Friendly competitor Phil Askey of DPreview.com managed not only some sunny shots, but much better ones than I'd likely have come up with. (He's a much better photographer than I. :-(

I thus thought that it would be interesting to do some night-time photography with the SD9, since nobody had as yet published any night shots taken with the camera. In the interests of providing some comparison between the SD9 and another camera, I duplicated my SD9 shots with the Sony DSC-F717 I was carrying as my main photographic tool while covering the show.

The photos below are the result of this exercise, although I need to insert a big disclaimer/caution right up front, to the effect that they were all shot with very poor camera support. The only camera support I had along on the trip was a tiny SLIK pocket "tripod" (I use the term advisedly) with legs formed of flexible metal sleeving material. It was quite a bit better than hand-holding the cameras, but far from a rigid support. It also only worked for landscape-format shots, meaning that any vertical-format images below were shot by bracing the cameras against convenient light or sign posts, stonework, etc. The bottom line is that none of these photos can be considered to show much of anything about the respective cameras' sharpness or resolution. I think they do serve as useful bases of comparison in terms of image noise, dynamic range, and color rendition.

The table below shows a sampling of the photos I shot, with 1:1 crops to give you some idea of image noise without having to download the full-sized images.

The SD9 model used here was a late-model prototype, one rev before the version that was equivalent to the final production models. (A "500-series" prototype unit.) As such, it's noise levels are apparently slightly higher than those expected from production models, although Foveon staff felt the differences would be relatively minor. (Note too, that I refrained from applying any sharpening to the images here, to avoid the increase in noise levels that sharpening would have produced.)

Since I posted this article, a number of readers have written to inquire about it, and as is usually the case, a couple of questions came up more than once. I thus thought it would be good to address these before moving on to the images:

Q: Why did you compare the SD9 with the F717? They're aimed at different markets and price points.

A: The simple answer is "because that's what I had access to." I had the 717 along to shoot photos of Photokina with, and didn't have access to a D100 or D60, which are admittedly more nearly comparable camera models. I did feel though, that the 717 was similar enough to the SD9's performance class to at least serve as some kind of reference point, something I was interested in, as there've been no direct comparisons between any other camera and the SD9 to date. A too-direct comparison would be unfair to the 717, since it sells for a bit over half the price, with the lens included, but the resolution should be somewhat equivalent, allowing for the 2x improvement in resolution-per-pixel afforded by the SD9's X3 sensor technology from Foveon. The chip and therefore the pixels in the 717 are a good bit smaller than those on the SD9, so if anything, noise factors would favor the SD9.

Q: Why do the SD9 shots all have their EV adjusted in the software, post-exposure? Doesn't this accentuate the noise?

A: Definitely yes, although perhaps not as much as one might think. While more signal in the sensor would help with noise levels, to some extent increasing exposure times would also increase the amount of thermal noise in the sensor elements. - So while doubling the exposure time (running the camera at half its rated ISO) would definitely help noise levels, there'd also likely be some additional noise fromthe longer exposure as well.

It's apparent from looking at the exposures, that the Foveon sensor isn't really operating at ISO 100 when it's run at that level (nor ISO 400 when run there). - Look at the first shots of the plaza with the hot dog stand. Compare the shots taken at ISO 100 with both cameras. The combination of (uncompensated) exposure times and f-stops of the two cameras amount to very similar exposure values for the two cameras, but the SD9's shot had to be tweaked in their software to produce anything close to the same brightness in the midtones of the final image. (Note that I'm talking about the exposure as delivered by the camera, not after the software-derived EV boost. - Compare the SD9's 1/8 sec at f/4.0 to the Sony's 1/13 at f/2.0, and you're awfully close to the same amount of light being admitted to each sensor. This suggests that both cameras exposed based on an ISO of 100, but the sensor in the SD9 produced a much darker image with that amount of light. - And also note the importance here of computing EV based on the actual camera exposure, not based on the boost applied in software as some readers apparently have done. What we're interested in is how much light is falling on the sensor, not what we had to do to the image to make it come out looking as bright as the one from the 717.) I really think these two shots are pretty equivalent. It's possible that the results might have come out a bit better with the SD9 if I'd run it at an effective ISO rating of 50, but as of the time of my doing that shooting, I didn't know that the sensor was underperforming in terms of ISO. (It was only the second time I'd had my hands on the camera, and I didn't have access to a computer with a proper CRT simultaneous with my use of the camera.) Given this behavior, it seems that the fairest comparison would be to run the SD9 at ISO 200, but with an exposure compensation of +1.0EV dialed into the camera's exposure system, and then compare those results to those of other cameras operating at ISO 100. Bottom line, had the camera been exposing properly, the noise levels would doubtless have been somewhat lower, but it's hard to say by just how much. (I'd make a wild guess at the noise level dropping by roughly half, which would still leave it rather high, particularly at ISO 400.)

Q: Doesn't the fact that you hand-held the exposures increase the noise?

A: Huh? - I have to say this question surprised me, as I can't imagine how anyone could think that the camera support would affect image noise. Handholding a long exposure will certainly affect image sharpness, but noise is purely a function of the sensor and camera electronics.

Q: What did you do (as the photographer) to make the camera underexpose so badly

A: Nothing. This isn't rocket science, you basically point the camera and take the picture. While some of the lighting was rather contrasty here, other subjects (the hot dog stand plaza and the cathedral spire) were actually fairly evenly illuminated.) My guess is that the underexposure was the result of an incorrect calibration between the camera electronics and the ISO that the sensor was actually operating at. Per my note above about the uncompensated exposure values for the photos of the hot dog stand, it's pretty clear that the SD9 was actually exposing correctly based on an ISO of 100 in that shot, as the overall EV matched that of the 717 pretty closely. The issue was simply that the sensor wasn't as sensitive as the camera thought it was.

Here are the sample photos, scroll down below the table to see my comments on the two cameras.

		Sigma SD9 Lets start with a food stand, outside the Cologne Haptbanhoff (main train station). They offered the obligatory panoply of varieties of bratwurst, but the main attraction was fried potato pancakes. Afficionados raved about them, but I found them glutinous and greasy. (The bratwurst was indeed excellent though.) This shot snapped at ISO 400 with the SD9, f/20 sec, f/4.0. The camera was handheld, braced against a light pole, so don't use this to evaluate the sharpness of the sensor. Adjustments in the Sigma software were: Exposure: +1.0 Contrast: -0.9 Shadow: +0 Highlight: -0.5 Saturation: 0 Sharpness: 0 Color: 6M+1Y
		Sigma SD9 The same scene, again shot with the SD9, only this time at ISO 100, 1/8 sec, f/4.0. Adjustments in the Sigma software were: Exposure: +1.7 Contrast: -1.2 Shadow: 0 Highlight: -1.3 Saturation: 0 Sharpness: 0 Color: 6M+1Y
		Sony DSC-F717 Here's the same scene shot with the Sony F717, at ISO 100, 1/13 sec, f/2.0. (That fast Zeiss lens on the 717 is really great for night shots. Image is just as it came from the camera, no adjustments. (The scale is a bit different from the SD9's too, due to the different aspect ratios. This shot takes in roughly the same left/right angle as the shots above, but the 4:3 aspect ratio results in much more vertical area being included.) (Apologies for the slight tilt in these images - I used an upright on the food stand on the right to align the vertical, forgetting that there'd be both barrel and perspective distortion affecting that part of the image. The result is the whole image is rotated counterclockwise a bit - I didn't want to correct this in Photoshop though, as the result wouldn't have been an original camera image then.)
		Sigma SD9 Next up are some shots of the Cologne cathedral. The one at left was shot with the Sigma SD9 at ISO 400, 1/6 sec, f/3.5. Adjustments in the Sigma software were: Exposure: +2.0 Contrast: +0.2 Shadow: +0.5 Highlight: +0.7 Saturation: -0.1 Sharpness: 0 Color: 6C+1M
		Sigma SD9 "By popular request" (thanks, Ulysses), here's the same shot as above, but only pushed +1.0 EV. ISO 400, 1/6 sec, f/3.5. Adjustments in the Sigma software were: Exposure: +1.0 Contrast: +0.2 Shadow: +0.5 Highlight: +0.7 Saturation: -0.1 Sharpness: 0 Color: 6C+1M
		Sigma SD9 This shot unfortunately shows the difficulty of capturing a sharp image at slow shutter speeds, even with the camera braced against a light pole. - Still useful for evaluating the noise level though. ISO 200, 1/2 sec, f/3.5. Adjustments in the Sigma software were: Exposure: +1.9 Contrast: +0.2 Shadow: +0.3 Highlight: +0.3 Saturation: 0 Sharpness: 0 Color: 1M+1Y
		Sony DSC-F717 A shot of the same subject, with the Sony F717. No adjustment, exactly as it came from the camera. ISO 400, 1/2 sec, f/3.2. (Note the slower shutter speed, even though shooting at ISO400, as with the first SD9 shot above. Note too though, that no post-exposure adjustment was needed.)
		Sigma SD9 A much brighter subject, exposed for the backlit map. Shot with the SD9 at ISO100, 1/60 sec, F/3.5 Adjustments in the Sigma software were: Exposure: +0.8 Contrast: -0.3 Shadow: +0.2 Highlight: -1.9 Saturation: +0.7 Sharpness: 0 Color: (none)
		Sony DSC-F717 The same shot with the F717, only at ISO 400, 1/80 sec, f/3.2. (There's really no particularly valid comparison to be made here, the SD9 clearly has lower noise at ISO 100 than does the F717 at ISO 400, which shouldn't be surprising. The framing of the two photos are also somewhat different...)
		Sigma SD9 Cologne had a very active street scene at night, with dozens (hundreds?) of little pubs and restaurants, and broad expanses of pedestrian-only plazas. This was one such. Shot at ISO 100, 1.5 seconds, f/4.5. Adjustments in the Sigma software were: Exposure: +0 Contrast: -1.0 Shadow: -0.2 Highlight: -0.3 Saturation: +0.1 Sharpness: 0 Color: 1C +1M
		Sony DSC-F717 The same shot with the F717, ISO 100, 0.8 seconds, f/4.5. This is the shot exactly as it came from the camera.
		Sony DSC-F717 In this case, the 717's shot was a bit darker in the midtones and shadows than the one from the SD9 after being hand-tweaked in the Sigma software, so I tweaked the Sony's shot in Photoshop, boosting the gamma somewhat to lighten the midtones.
		Sigma SD9 Here's another "plaza" shot, captured with the Sigma. Shot at ISO 100, 1.2 sec, f/4.5. This shot shows off the capability of the Foveon/Sigma software, which let me preserve both highlight and shadow detail on this very contrasty shot. Adjustments in the Sigma software were: Exposure: +1.1 Contrast: -0.9 Shadow: +0.1 Highlight: -2.0 Saturation: -0.1 Sharpness: 0 Color: 2M +1Y
		Sony DSC-F717 This is the shot straight from the Sony F717. ISO 100, 1/4 sec, f/2.1. A super-contrasty subject like this one is an example of where the Sigma/Foveon software and native "raw" file format of the SD9 comes into its own. Notice how much more highlight detail is visible in the SD9's shot above, after processing through the software.
		Sony DSC-F717 A slight gamma boost on the above shot in Photoshop, to brighten the midtones and shadows somewhat. It loses a bit more detail in the highlights though.

I shot most of these samples at ISO 100, feeling that this would give the best view of the best-case noise performance of both cameras under fairly dark shooting conditions. In hindsight, I wish I'd shot more samples at ISO 400, as that probably would have been more representative of what most people would do under these conditions. - It was also at ISO 400 that the differences in noise performance between the SD9 and the F717 became the most apparent,with the SD9 images becoming quite noisy. Noise is an area of concern for the Foveon technology used in the SD9, and the SD9's images were certainly more noisy at ISO 400 than those from the F717. At ISO 100, the difference wasn't as great, but the F717 still showed considerably lower noise. We'll have to wait until I can test a production-level SD9 before forming any formal opinions in this area though.

As I worked with them, the contrast between the two cameras was pretty dramatic, both at the time of exposure, and afterward, viewing the images on a computer.

The DSC-F717 was much easier to get good-looking photos from, requiring very little attention to the details of the exposure, and almost no tweaking of its images after the fact. Its exposure system delivered correct exposures most of the time, apart from obvious situations where spot metering or manual exposure compensation were required. (The map-sign shot being one example of this.) The images it produced also showed very little noise, and excellent color. (The 717's white balance system is really remarkable for its ability to deliver correct-looking color under an amazing range of shooting conditions.) Overall, the DSC-F717 made even tricky night shooting almost a point & shoot experience.

By contrast, the SD9 needed a good bit more post-exposure tweaking to achieve what I considered acceptable results, and as such is much more suited to people interested in tweaking their photos extensively post-exposure. Perhaps as a result of my test unit's prototype status, the camera tended to underexpose most of its shots by a good f-stop or so. I also felt that its color was somewhat undersaturated. These tendencies were balanced by the extraordinary control provided by the Foveon/Sigma software that accompanies the SD9. Derived from Foveon's studio-camera product, this is clearly some of the very best software I've seen to date for manipulating color and tone in photos. It's easy and intuitive to use, but very powerful in its capabilities. -- If you're the type of photographer who wants to hand-adjust each exposure to achieve perfect tonal balance, the SD9 is the camera for you.

Overall, (IMHO) the Sony DSC-F717 delivered better images directly from the camera (without adjustment), with lower noise and better color. On the other hand, the SD9's native raw format and excellent software gave me more control over the images I eventually ended up with. The two cameras are obviously very different, intended for quite different audiences, but I suspect that many people will be making a choice between the SD9 and an "all in one" prosumer camera like the F717. At this point, it's hard to arrive at any final conclusion about the SD9, given the prototype nature of the camera I was working with. Stay tuned though, I'm hoping to get a sample of a final-revision model fairly soon, to put through the full battery of my tests.

What can be said at this point is that the SD9 and its Foveon-derived sensor technology do appear to offer intriguing possibilities at the upper end of the "prosumer" realm. Stay tuned for a full test report in the (hopefully) near future!