Sigma SD1 Image Quality
Saturation & Hue Accuracy
Muted color with slightly below average hue accuracy. Hue error increases and saturation falls rapidly at ISOs above 800.
|ISO 100||ISO 200||ISO 400||ISO 800|
|ISO 1600||ISO 3200||ISO 6400|
|In the diagram above, the squares show the original color, and the circles show the color that the camera captured. More saturated colors are located toward the periphery of the graph. Hue changes as you travel around the center. Thus, hue-accurate, highly saturated colors appear as lines radiating from the center.|
Skin tones. The Sigma SD1's Caucasian skin tones looked fairly realistic in outdoor lighting using auto white balance, with a healthy-looking pink tint, though darker skin tones can be a little too red. Manual white balance produced flatter, warmer and more yellow skin tones, though reddish cast still exists in darker areas. Where oversaturation is most problematic is on Caucasian skin tones, as it's very easy for these "memory colors" to be seen as too bright, too pink, too yellow, etc.
Hue. The Sigma SD1 showed a lot of small color shifts relative to the mathematically precise translation of colors in its subjects, resulting in a Delta-C color error after correction for saturation of 5.51 at base ISO, which is a little higher (worse) than average these days. Like saturation, hue accuracy dropped significantly at higher ISOs. Hue is "what color" the color is.
The Sigma SD1 offers seven preset Color Mode options. Contrast, sharpness and saturation can be adjusted in eleven steps. (Note that B&W and Sepia are not supported when RAW or RAW+JPEG mode is selected.)
Mouse over the links above to see the effect of the presets on our Still Life target. Click on a link to load the full resolution image.
The Sigma SD1 lets you adjust the image saturation and contrast in 11 steps, from -5 to +5. As can be seen below, the saturation adjustment provides fine-grained adjustment over a useful range of control, however some colors are affected more than others. The saturation adjustment has almost no impact on contrast, though. That's how a saturation control should work, but we've often found interactions between saturation adjustments and image contrast (and vice versa) on the cameras we test.
|Saturation Adjustment Examples|
The series of shots above shows results with several different saturation adjustment settings including the two extremes. See the Thumbnails index page for more (look for the files named SD1OUTBSATxx.JPG). Click on any thumbnail above to see the full-sized image.
| See full set of test images
See thumbnails of all test and gallery images
Exposure and White Balance
Indoors, incandescent lighting
Very warm results with Auto white balance and the Incandescent setting was too red. Good color balance with Manual setting, though colors were muted. Negative exposure compensation required.
|Auto White Balance
|Incandescent White Balance
|Manual White Balance
Indoors, under normal incandescent lighting, color balance was very warm and orange with the Auto white balance setting, while the Incandescent setting was too reddish. The Manual white balance setting by far produced the most accurate results, if just a touch cool. Colors were quite muted, though. The Sigma SD1 required negative exposure compensation for this shot (-0.3 EV) which is unusual since the average needed among cameras we've tested for this shot is about +0.3 EV. Our test lighting for this shot is a mixture of 60 and 100 watt household incandescent bulbs, a pretty yellow light source, but a very common one in typical home settings here in the U.S.
Good exposure but muted, slightly warm color outdoors.
|Auto White Balance,
|Auto White Balance,
The Sigma SD1 struggled a bit under harsh sunlight, yielding muted colors as well as a lot of clipped highlights, though exposure accuracy was better than average. We needed to use +0.3 EV exposure compensation for a bright face in our "Sunlit" Portrait shot. That's less than the average of +0.7 EV normally require for this shot, but it did lead to a lot of clipped highlights in the model's shirt and flowers. We found skin tones just a touch yellow and flat in our "Sunlit" Portrait shot with Manual white balance, preferring the more pinkish skin tones Auto white balance provided here. Colors were a bit warm and muted in our Far-field outdoor shot with Auto white balance. Default exposure was a little hot, clipping some highlights in the white building, though shadow detail was quite good if a touch noisy.
Very high resolution, ~2,200 lines of strong detail from JPEGs, about the same from converted RAW files.
|Strong detail to
~2,200 lines horizontal
|Strong detail to
~2,200 lines vertical
|Strong detail to
~2,200 lines horizontal
SPP processed RAW
|Strong detail to
~2,200 lines vertical
SPP processed RAW
Our laboratory resolution chart revealed sharp, distinct line patterns down to about 2,200 lines per picture height in both the horizontal and vertical directions in JPEGs. Complete extinction of the pattern didn't occur before the limits of our chart (4,000 lines) in both directions, and as expected, no color moiré was present (though some chroma noise can be seen if you look closely at some of the resolution patterns). There are some pixel defects that were not well corrected, but that's not unusual.
We weren't really able to extract any more resolution here using Sigma's Photo Pro RAW converter; in fact, the software's slightly aggressive default sharpening produced moiré patterns earlier than camera JPEGs. Note that Adobe Camera Raw does not yet support the SD1's RAW files at time of writing (mid-January 2012). Use these numbers to compare with other cameras of similar resolution, or use them to see just what higher resolution can mean in terms of potential detail.
As expected, the Sigma SD1's spatial resolution isn't much higher than a traditional Bayer sensor producing images with similar dimensions, though the lack of an anti-alias filter does indeed result in higher per-pixel sharpness. To see how color resolution compares to a conventional sensor of similar pixel count (not photosite count), and even significantly higher, see below.
These color starburst patterns (variations on the Siemens Star target) found in our Multi target are useful for evaluating color-dependent resolution as well as demosaicing issues. The spokes in the starburst don't touch except at the very center, so where they touch in these images is an indication of the effective resolution limit. The green/black target provides the best resolution for standard Bayer filters used on most sensors, as they have twice as many green photosites as blue or red, while the blue/red target is worst-case, as there is only one of each in the repeating four-color Bayer pattern. The Foveon sensor on the other hand has a photosite for each primary color (RGB), stacked on top of each other similar to how film emulsion works. Color is not interpolated as it is with a Bayer filter, so color resolution is much better for the same number of pixels. Mouse over the links and compare how far into the starburst the spoke pattern remains intact. As you can see, the 14.8-megapixel SD1 easily outresolves the 16-megapixel K-5, especially in the blue/red target. Results are closer in the green/black target, but the SD1 still resolves noticeably more. (Note that the crops in these comparisons are from converted raw files.)
Here, we've taken a higher resolution Bayer camera (the 25-megapixel Sony NEX-7) and resized the image to roughly match the SD1's size for easier comparison. Again, mouse over the links and compare how far into the spoke pattern remains intact. As you can see, the Foveon sensor still outresolves the Bayer sensor, though black/green resolution is closer. The difference is especially evident in the blur/red target, where a lot of demosaicing artifacts and aliasing limit the NEX-7's resolution compared to the SD1. Unfortunately, we don't have a calibrated target with resolution scales to determine actual color resolution figures, but it's clear that Sigma SD1 outresolves significantly higher resolution Bayer sensors when it comes to color.
Sharpness & Detail
Excellent sharpness straight from the camera, with very minor edge-enhancement on high-contrast subjects. Moderate noise suppression artifacts at base ISO.
Sharpness. The Sigma SD1 produced images with excellent sharpness and very good detail, doing much better than a typical 16-megapixel Bayer sensor would do at base ISO. Very few edge enhancement artifacts are visible on high-contrast subjects in the crop above left, which is also excellent. Edge enhancement creates the illusion of sharpness by enhancing color and tonal differences right at the edge of a rapid transition in color or tone.
Detail. The crop above right shows some detail loss due to noise suppression, as darker areas and areas of low contrast in the model's hair show some smudging and loss of definition. Still, a pretty good performance at base ISO. (A sharper lens probably would have helped here, since we shoot the indoor portrait shot at f/4 and the 17-50mm lens isn't that sharp at wide apertures.) Noise-suppression systems in digital cameras tend to flatten-out detail in areas of subtle contrast. The effects can often be seen in shots of human hair, where the individual strands are lost and an almost "watercolor" look appears.
RAW vs In-Camera JPEGs
As noted above, the Sigma SD1 does a an excellent job capturing loads of detail in its JPEGs. As is usually the case, more detail can be obtained from carefully processing RAW files, though, without introducing additional artifacts. Take a look below, to see what we mean:
In the table above, mousing over a link at the bottom will load the corresponding crop in the area above, and clicking the link will load the full resolution image. Examples are all shot at ISO 100, and include in-camera Fine JPEG, the matching RAW file processed through Sigma's Photo Pro 5.2 software using default settings, and another conversion with Photo Pro with ots sharpening and noise reduction set to the lowest settings, then sharpened in Photoshop with an Unsharp Mask of 200% and radius 0.3.
As you can see, Sigma's Photo Pro software produced an image very similar to the camera JPEG, though with slightly better detail. Reducing sharpening to a minimum in Photo Pro, then sharpening in Photoshop showed an improvement in fine detail, though turning noise reduction down did result very fine noise in the form of individual red pixels. We suspect processing the SD1's RAW files in Adobe Camera Raw would yield even better results, however ACR does not yet support the SD1's X3F RAW files at the time of writing (late-January, 2012). We'll try to come back and update this section when Adobe adds support for the SD1's RAW files.
Here's another example of how Photo Pro can do better than the in-camera JPEG engine when sharpened in Photoshop. Notice how threads can be seen in the cloth in the RAW converted example. This is better performance than some ~25-megapixel DSLRs. The lack of an image-softening anti alias filter coupled with a sensor that requires no color interpolation allows the SD1 to resolve fine chromatic detail at low ISOs well beyond what most Bayer sensors with similar or even significantly higher resolutions can do.
ISO & Noise Performance
Very good handling of noise vs detail at low to moderate ISOs, but image quality quickly deteriorates at higher ISOs.
|ISO 100||ISO 200||ISO 400|
|ISO 800||ISO 1,600||ISO 3,200|
Images were reasonably clean at ISO 100 through 400, with just a touch of luminance and chrominance noise visible in the shadows. As mentioned, though, colors are already quite muted at base ISO. Chrominance noise becomes noticeably stronger at ISO 800 in the form of large pink and green blotches, though luminance noise was still pretty good. At ISO 1,600, luminance noise became quite a bit stronger, and chrominance noise also increased, though detail was still pretty good. Images were pretty ugly at ISO 3,200, though, with strong luminance and there were large areas of purple blotching. ISO 6,400 was very poor, with stronger noise, a strong shift of midtones and shadows toward green, and a drastic drop in saturation. ISOs 3,200 and 6,400 also showed noticeable horizontal banding in darker areas. Overall, below average high ISO noise performance for an APS-C sensor. Note that the Sigma SD1 offers no control over noise reduction strength, at least not with the latest firmware available at time of writing (version 1.04.0.6360).
Note that these shots were taken at f/4 with the bundled Sigma 17-50mm f/2.8 lens. This lens isn't as sharp as our reference Sigma 70mm lens at f/4 we normally use for these shots, however the 17-50mm was (ironically) the sharpest lens we could locate in the Sigma SA-mount. Below is an ISO crop table from our Still Life shots which are taken in brighter light at f/8 where the lens is quite sharp.
|ISO 100||ISO 200||ISO 400|
|ISO 800||ISO 1,600||ISO 3,200|
Here, we can see fine detail looks better than in the Indoor Portrait shots, however noise follows a similar trend with below average performance at ISOs above 1,600 along with a drastic reduction in saturation and a shift towards green.
Of course, the impact of noise and detail loss are highly dependent on the size the photos are printed at, and pixel-peeping on-screen has surprisingly little relationship to how the images look when printed: See the Print Quality section below for recommended maximum print sizes at each ISO.
Extremes: Sunlit, dynamic range and low light tests
Very high resolution but with poor highlight preservation. Fair low-light performance, capable of capturing bright images in near darkness at low ISOs, though autofocus struggled and higher ISOs were very noisy.
|-0.3 EV||0 EV||+0.3 EV|
The Sigma SD1 struggled with the deliberately harsh lighting in the above test. Though the camera only required +0.3 EV compensation to render a bright face in our "Sunlit" Portrait shot (the average for this shot is about +0.7 EV), the SD1 clipped quite a few highlights in the mannequin's shirt and flowers, though shadow detail was quite good if a bit noisy. We suspect much better dynamic range is available from the SD1's RAW files. It's also interesting to note that the 0 EV image is practically identical to the -0.3 EV image, using the same shutter speed (1/50s) in Aperture Priority (f/8) auto exposure mode, so the camera seemed to ignore the exposure compensation adjustment between -0.3 and 0 EV here.
Because digital cameras are more like slide film than negative film (in that they tend to have a more limited tonal range), we test them in the harshest situations to see how they handle scenes with bright highlights and dark shadows, as well as what kind of sensitivity they have in low light. The shot above is designed to mimic the very harsh, contrasty effect of direct noonday sunlight, a very tough challenge for most digital cameras. (You can read details of this test here. In actual shooting conditions, be sure to use fill flash in situations like the one shown here; it's better to shoot in open shade whenever possible.)
Just like saturation, the Sigma SD1's contrast setting offers 11 settings, between -5 and +5.
|Contrast set to lowest,
|Contrast set to lowest,
At its lowest contrast setting, the SD1 did a pretty good job of preserving highlight detail while bringing nice detail out of the shadows.
|Contrast Adjustment Examples|
The series of shots above shows results with several contrast adjustment settings, including both extremes. While you can see the extremes, it's hard to really evaluate contrast on small thumbnails like these, click on any thumbnail to go to the full-size image.
One very nice feature of Sigma's contrast adjustment is that it has very little effect on color saturation. Contrast and saturation are actually fairly closely coupled, it's a good trick to be able to vary one with out the other changing as well. Sigma did a good job here.
A key parameter in a digital camera is its Dynamic Range, the range of brightness that can be faithfully recorded. At the upper end of the tonal scale, dynamic range is dictated by the point at which the RGB data "saturates" at values of 255, 255, 255. At the lower end of the tonal scale, dynamic range is determined by the point at which there ceases to be any useful difference between adjacent tonal steps. Note the use of the qualifier "useful" in there: While it's tempting to evaluate dynamic range as the maximum number of tonal steps that can be discerned at all, that measure of dynamic range has very little relevance to real-world photography. What we care about as photographers is how much detail we can pull out of the shadows before image noise becomes too objectionable. This, of course, is a very subjective matter, and will vary with the application and even the subject matter in question. (Noise will be much more visible in subjects with large areas of flat tints and subtle shading than it would in subjects with strong, highly contrasting surface texture.)
What makes most sense then, is to specify useful dynamic range in terms of the point at which image noise reaches some agreed-upon threshold. To this end, Imatest computes a number of different dynamic range measurements, based on a variety of image noise thresholds. The noise thresholds are specified in terms of f-stops of equivalent luminance variation in the final image file, and dynamic range is computed for noise thresholds of 1.0 (low image quality), 0.5 (medium image quality), 0.25 (medium-high image quality) and 0.1 (high image quality). For most photographers and most applications, the noise thresholds of 0.5 and 0.25 f-stops are probably the most relevant to the production of acceptable-quality finished images, but many noise-sensitive shooters will insist on the 0.1 f-stop limit for their most critical work. A full discussion of all the data Imatest produces is really beyond the scope of this review: Visit the Imatest website for details of what the program measures, how it performs its computations, and how to interpret its output.
JPEG. The graph at right (click for a larger version) was generated using Imatest's dynamic range analysis for an in-camera SD1 JPEG file with a nominally-exposed density step target (Stouffer 4110). At default settings and base ISO, the graph shows 9.76 f-stops of total dynamic range, with 6.1 f-stops at the "High" quality level. These are below average results for an APS-C sensor. Compared to the Nikon D7000, the SD1 scored much lower at the "High" quality level (6.1 vs 7.97 f-stops), while total dynamic range was slightly lower (9.76 vs 10.0 f-stops). The SD1's higher than average noise levels is mainly to blame for low high quality score. Note though that this measurement has a margin of error of about 1/3 f-stop, so differences of less than 0.33 can be ignored.
RAW. It'll probably be some time before Adobe Camera Raw supports the SD1, so we'll make an exception and use Sigma's Photo Pro RAW conversion software to see if we can improve upon its JPEG dynamic range scores. The graph at right is from the same Stouffer 4110 stepchart image captured as a RAW (.X3F) file, processed with Sigma's software. The Sigma SD1's RAW file scored essentially the same in total dynamic range compared to the JPEG (9.78 vs 9.76 f-stops) however the score at the highest quality level increased about 1.8 f-stops from 6.1 to 7.92 f-stops. This is still a below average score, though, and it's worth noting here is that Photo Pro's default noise reduction settings reduced overall noise relative to the levels in the in-camera JPEG (compare the noise plots in the graphs), which tends to boost the dynamic range numbers for the High Quality threshold.
Low Light. The Sigma SD1 was able to capture usable images down to the 1/16 foot-candle light level (about 1/16 as bright as average city street lighting at night) at all ISO settings, though noise was an issue at higher ISOs and lower light levels. As is often the case, the SD1's metering system struggled a bit with getting the exposure right at the lowest levels though, so we used manual exposure for these shots. Color balance with auto white balance was fairly neutral, just slightly cool, but shifted toward magenta at higher ISOs. Noise was low to moderate up to ISO 800, though a hint of horizontal banding can already be seen in the red channel at this ISO. ISO 1,600 and above were quite grainy with obvious chroma noise, color shifts, desaturation and banding.
The camera's phase-detection autofocus system was only able to focus on the subject down to just below the 1/2 foot-candle level with the 17-50mm f/2.8 lens unassisted. That's quite poor for an SLR, especially one costing as much as the SD1, though the lens' softness at f/2.8 may have impeded low-light AF performance. The camera was however able to autofocus in complete darkness with the AF assist enabled.
(Keep in mind that the longer shutter speeds here demand the use of a tripod to prevent any blurring from camera movement. A useful trick is to just prop the camera on a convenient surface, and use its self-timer to release the shutter. This avoids any jiggling from your finger pressing the shutter button, and can work quite well when you don't have a tripod handy.)
How bright is this? The one foot-candle light level that this test begins at roughly corresponds to the brightness of typical city street-lighting at night. Cameras performing well at that level should be able to snap good-looking photos of street-lit scenes.
NOTE: This low light test is conducted with a stationary subject, and the camera mounted on a sturdy tripod. Most digital cameras will fail miserably when faced with a moving subject in dim lighting. (For example, a child's ballet recital or a holiday pageant in a gymnasium.) Digital SLRs like the Sigma SD1 do much better than point & shoots, but you still shouldn't expect a quick autofocus lock with moving subjects.
Very nice 24 x 36 inch prints at ISOs 100/200; ISO 800 shots are better at 11 x 14; ISO 6,400 shots are unusable at 4 x 6.
ISO 200 shots are also quite good at 24 x 36, with nice color detail and contrast; and fine for wall display at 30 x 40.
ISO 400 prints are very good at 20 x 30, with the only exception being low contrast in our target's red swatch.
ISO 800 prints are usable at 13 x 19, but color darkens in the mids, making for an odd antiqued look. There were also odd, desaturated splotches in green areas. We will call 11 x 14 inch prints good here.
ISO 1,600 shots look a little scorched for lack of a better term; colors are dimmed to almost brown, and purple and green blotches appear in the shadows. The blotchy desaturation we saw at ISO 800 was much worse here. Detail is dulled at 8 x 10 inches, and we preferred the 5 x 7 inch print.
ISO 3,200 prints are severely faded at any size. Blobs of purple and green chroma noise are more prominent in the shadows. The 4 x 6 inch print does not pass as good but could be usable in certain situations.
ISO 6,400 prints have color so badly faded that we don't recommend shooting at this setting at all. Prints might look OK if reduced to black and white for an aged, grainy film-noir look.
Overall, the Sigma SD1 performs well from ISO 100 to 400, with amazing detail for its 14.8-megapixel output resolution. Also know that images from a Foveon sensor generally respond to significant upsizing better than images generated from a Bayer sensor due to the lack of color interpolation artifacts, so with some careful manipulation during post-processing, you should be able to make great-looking low ISO prints much larger than those straight from the camera. From ISO 800 and up, though, color fades and darkens badly, with an antiqued look that's nostalgic at first, fading to disappointing. There's nothing wrong with keeping ISOs low to enjoy the astonishing color resolution, but it's important to know this limitation when considering the Sigma SD1. The rapid degradation of image quality from ISO 1,600 up is a perfect illustration of this point, as it goes from 11 x 14 at ISO 800 down to 5 x 7 at ISO 1,600.
Testing hundreds of digital cameras, we've found that you can only tell just so much about a camera's image quality by viewing its images on-screen. Ultimately, there's no substitute for printing a lot of images and examining them closely. For this reason, we now routinely print sample images from the cameras we test on our Canon Pro9000 Mark II studio printer, and on the Pixma MP610 here in the office. (See the Canon Pixma Pro9000 Mark II review for details on that model.)