Pros and Cons of the Sigma SD10 Digital SLR Camera

This web site is intended to counter the misinformation, regarding Sigma Digital SLRs, that is being promulgated on the Usenet Group rec.photo.digital, by one or two individuals (posting under many different aliases). This website makes it convenient to look at the "big picture" (no pun intended). The response to this web site has been tremendous. Most of that response has been very positive.

Please contact me with any additions or corrections. It may take a while for me to update this site with any comments, since I only update it once a month or so. You can send feedback to sigmaexpert@groupnordic.us (swap group and nordic in groupnordic to make nordicgroup, I do this to prevent spam). This e-mail address should be used only for responses to this web site. I strive to be accurate. If you see something that is in error just calmly explain the problem. I'll investigate and fix it. I recently added a Change Log to the bottom of the page, in order to track the changes I've made.

The Sigma SD10 digital camera body evolved from Sigma's SA9 film camera (with the brief intervention of the Sigma SD9 digital camera). A Foveon 3.43 Megapixel sensor was substituted for the film transport, and electronics was added to write the RAW image to a compact flash card. An LCD, to preview the image, was also added. Due to the post-processing requirements of the raw files generated by a Foveon sensor, there is no in-camera conversion to JPEG images available. The SD10 corrected some of the biggest problems with the SD9, including a bizarre power system that retained the original battery of the SA9, and added more batteries for the digital functions.

The SA9, SD9 and SD10, have a proprietary Sigma lens mount, limiting the owner to Sigma lenses. While the Sigma EX line of lenses is suitable for amateurs and prosumers, in the opinion of most reviewers, the entry level Sigma lenses are not as good as the entry level lenses from Canon or Nikon.

Foveon and Sigma
It is difficult to separate the Foveon X3 Pro 10M sensor from the actual Sigma digital SLR cameras; they are inextricably linked. Sigma is the only manufacturer of a digital SLR that is using the Foveon 3.4 Megapixel X3 sensor (World Wide Licenses has a 1.5 megapixel point and shoot that uses a lower resolution (1.5 megapixels) version of the Foveon sensor). Originally, Foveon did not include "10M" in the name of the sensor, this is a marketing change with which they try to emphasize the number of photodetectors on the sensor. Nothing wrong with this, except that other manufacturers state the resolution in output pixels, not input photodetectors.

Theory versus Reality
The theory behind the X3 Pro 10M sensor is good, but unfortunately it isn't so easy to build a sensor that matches the theory! There is loss of luminance as the light passes through the layers of silicon, and the chrominance is not perfect because silicon color separation is inexact. It is amusing to read exposés, by uninformed individuals, as to how great the theory of the Foveon sensor is. The theory behind the Foveon sensor is so appealing that people just want to believe that silicon color separation is perfect, and that full color pixels provide better color than interpolated Bayer pixels. Yet even Foveon acknowledges the difficulty of silicon color separation (read http://www.x3f.info/technotes/X3SensorCharacteristics.pdf).

The Myth of Sharpness
Foveon is fond of stating that their X3 Pro 10M sensor produces a “sharper” image than a Bayer sensor. But there is another side to the story (the true side):

The Myth of the Benefit of More Chrominance Data
The human eye's ability to detect chrominance is relatively poor. Having three times as much chrominance information, per pixel, buys you almost nothing in terms of the quality of the final image. The additional chrominance data comes at the very high price of far less ISO sensitivity and far more noise. The three-layer, "full-color" sensor, is an answer to a question that nobody asked. Furthermore, because silicon color separation is so inexact, the camera must either produce only RAW images, or have tremendous processing power in the camera in order to produce JPEG images.

Foveon History

Six years ago Foveon produced the best studio camera on the market. It used three independent sensors, each with 4 million photodetectors, and a prism that separated the colors and directed the light to each photodetector. Then the results of each photodetector were combined to form an image with 4 million, full color, spatial pixels. At the time it was introduced, it was inarguably the highest resolution digital studio camera on the market. You can read about the Foveon studio camera at:
http://www.digital-photography.org/digital_photography_cameras/Foveon_digital_camera.html

Alas, the prism approach was too expensive, too complicated, too fragile, and too heavy to be usable in a non-studio camera, hence the development of the more practical, X3 Pro 10M sensor. This enabled Foveon to get out of the camera business and concentrate solely on selling sensors.

With the X3 Pro 10M sensor, the three photodetectors are layered on top of each other and the different colors of light penetrate the silicon to different depths. Foveon calls this "Silicon Color Separation." Unlike the prism method of color separation, the silicon method results in some color contamination and more noise, because silicon color separation is an inexact process, and each silicon wafer has slightly different characteristics.

With the development of low-noise, full-frame, CMOS, Bayer sensors, such as the 6, 8, 11, and 16.7 megapixel sensor used in the Canon EOS line, the Foveon approach lost its appeal. The Canon 16.7 megapixel sensor is nearly 5x the resolution, with better color and lower noise, at higher ISOs, than the Foveon X3 Pro 10M sensor, and even the 6 and 8 megapixel Canon CMOS sensors have higher resolution and a smaller crop factor than the Foveon sensor. Rumors are that Foveon is working on a higher resolution sensor (5.6 megapixel/16.8 megaphotodetector), with a smaller crop factor (1.3), but these are all unconfirmed rumors.

A pixel is one spatial element from the scene, to which a digital value is assigned. It really doesn't matter if there are 1000 photodetectors that contribute to the color of one pixel, it is still one pixel.

The Sigma SD9 and SD10 have 3.4 million sensor pixels. This is lower than any other current D-SLR. Each pixel has 3 photodetectors, for a total of 10.2 million photodetectors, but there are only 3.4 million, spatially-separate, sensing locations. The number of sensor locations is what determines the ability to resolve fine detail. Competing DSLRs have 5-16.17 million measuring locations. The Sigma RAW processing software, used to convert the RAW data into an image, can produce 14 megapixel images using interpolation, but the interpolation process adds no detail.

Cameras that use Bayer sensors, measure one color per pixel, not three colors per pixel. However despite this, they yield the same resolution as a Foveon sensor with the same pixel count.

"Pixel-Speak" Campaign

http://www.x3f.info/technotes/x3pixel/pixelpage.html

Foveon wants to market photodetectors, not pixels. This is because they have a sensor with 10.2 million photodetectors, but only 3.4 million pixels. Unfortunately, everyone else quotes resolution in output pixels, which in the Bayer sensor, is equivalent to the number of input photodetectors.

Foveon has taken a two-pronged approach in their marketing campaign.

First, they created a new term for a photodetector, calling it a "pixel sensor," hoping that consumers will equate "pixel sensor" and "pixel." They are very careful to never leave out the qualifying word of "sensor," so they can't be accused of claiming that their sensor is 10.2 million pixels; they simply never mention the fact that they have 3.4 million pixels (unless you go back to some of their old press announcements, where they accurately describe the characteristics of their sensor).

Second, they are trying to claim that each photodetector on a Bayer sensor is only 1/3 of a pixel, by redefining a pixel as "an RGB triple in a sampled color image." Quite amusing, since if they want to follow this flawed line of reasoning, they would claim that each photodetector on a Bayer sensor is only 1/4 of a pixel, since a Bayer sensor is 50% green, 25% red, 25% blue.

The former is marketing semantics, and while not technically lying, it is misleading. It's called "implicature," in the trade. The latter is simply incorrect. By all accepted definitions of a pixel, each spatially distinct photodetector on a Bayer sensor translates to a spatially distinct output pixel, with the chrominance determined by interpolating from adjoining pixels (usually at least eight adjoining pixels).

Let's compare Foveon-Speak to ROW (Rest of World), using the Foveon X3 10M sensor, and a Bayer 6 megapixel sensor.

Foveon-Speak
X3 10M: 10.2 million pixel sensors, 3.4 million pixels
Bayer 6 Megapixel Sensor: 6 million pixel sensors, 1.5 million pixels

ROW
X3 10M: 10.2 million photodetectors, 3.4 million pixels
Bayer 6 Megapixel Sensor: 6 million photodetectors 6 million pixels

You can understand why Foveon wants to count input photodetectors (renaming them "pixel sensors"), while the ROW wants to count output pixels. Automotive analogies spring to mind, i.e. displacement versus horsepower, horsepower versus 0-60 times, driving range versus miles per gallon, etc..

The new Sigma SD-14 continues the "Pixel-Speak" campaign. It's a 4.65 megapixel sensor with 13.9 million photodetectors, but Sigma is already promoting it as a 13.9 megapixel camera.

Pixels≠Photodetectors
Foveon and Sigma attempt to re-define a pixel as a photodetector . Their position is that since they have three stacked photodetectors, that contribute to the color of each spatial pixel, that each photodetector should count as a pixel (not a spatial pixel, but some kind of a pixel!). I believe that Foveon's rationalization is that if they took one layer, and used it as a senor in a Bayer camera, it would be considered a 3.43 megapixel camera; they have three layers, so they have a 10.2 megapixel camera. The flaw in their logic is that these three layers are in one camera, in the same focal plane, and not spatially distinct.

A good layman's explanation of the Foveon X3 Pro 10M sensor was posted on Usenet, which I took the liberty of paraphrasing:

'In the Foveon X3 sensor, there are 3.43 million individual spots on the focal plane, where data is taken from. Each of those 3.43 million spots has 3 photodetectors, which use the inherent filtration of the silicon to pull off data that corresponds to three discrete colors. The three data streams are incorporated into one pixel, giving (in theory) excellent color for that one pixel. Since there are 3.43 million instances of this going on, the sensor outputs an image of 3.43 million pixels.'

There is nothing wrong with Foveon and Sigma explaining about the X3 Pro 10M sensor, and how it has as many photodetectors as a 10 megapixel Bayer, but they should not be calling the X3 Pro 10M a 10.2 megapixel sensor. And in fact, both Sigma and Foveon used to explain how the X3 Pro 10M sensor layered three photodetectors at each pixel location. It was only recently that they tried to redefine a photodetector as a pixel. In my view, and the view of many others, the attempted redefinition of a photodetector as a pixel is extremely misleading, and greatly reduces the credibility of both companies.

One source of confusion is that pixels really refer to the picture that is output, not the input from the sensor's photodetectors. On a Bayer sensor, there is basically a one to one correspondence between input photodetectors and output pixels. On the Foveon X3 Pro 10M sensor, there is a three to one correspondence between input photodetectors and output pixels. As Sigma states on their web site, "...for every pixel on a FOVEON^® X3^TM image sensor, there's actually a stack of three photo detectors, forming the first and only full-color capture system." Actually this statement is slightly inaccurate, because the pixels aren't actually on the sensor, but it's close enough. Foveon states: ""The Foveon X3 image sensor in the Sigma SD9 contains over 10.2 million color photodetectors, which are organized in 3 layers within the sensor to form 3.54 million full-color pixels."

The problem for Foveon is that it's become industry practice to quote pixels, not photodetectors. But it's not unreasonable to quote pixels, because the number of spatially distinct photodetectors, that measure luminance, are the key to resolution.

History of Sigma and Foveon Statements on the Resolution of the X3 Pro 10M Sensor

Sigma originally wrote (http://www.sigmaphoto.com/Html/news/sd9press.htm): "The breakthrough Sigma SD9 camera features the FOVEON^® X3^TM image sensor which is the world's first image-sensor that features three layers of photo detectors. The layers are embedded in silicon to take advantage of the fact that red, green and blue light penetrate silicon to different depths allowing full-color to be measured at every pixel. The layers are positioned to take advantage of the fact that silicon absorbs different colors of light at different depths, so one layer records red, another layer records green and the other layer records blue, This means that for every pixel on a FOVEON^® X3^TM image sensor, there's actually a stack of three photo detectors, forming the first and only full-color capture system."

How much clearer can you get? Again "...for every pixel on a FOVEON^® X3^TM image sensor, there's actually a stack of three photo detectors..."

Or shall we look at a Foveon press release (http://foveon.com/press_sigma.html): "The Foveon X3 image sensor in the Sigma SD9 contains over 10.2 million color photodetectors, which are organized in 3 layers within the sensor to form 3.54 million full-color pixels."

No doubt Foveon feels that since their sensor is comprises of three layers of photodetectors, while a Bayer sensor has only one layer, that is unfair for them NOT to be able to redefine a photodetector as a pixel. But the Foveon advantage was not supposed to be increased resolution, it was supposed to be that their sensor produced more accurate color (since each pixel had its own set of three photodetectors (RGB), rather than having to calculate the color of each pixel from adjoining pixels as is done on a Bayer sensor). But as it turned out, the Bayer method of calculating the color of each pixel is just fine, especially on high megapixel sensors, while the Foveon layered sensor has its own set of problems. The bottom line is that the digital SLRs with the Bayer sensors produce more accurate color that the digital SLRs with the Foveon sensors.

References on the Definition of a Pixel
Here's a list of references to look at. These explain what a pixel is. To save space, I have just assigned numbers to each URL: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19. Please let me know if any of these links are broken.

Bayer sensors do color interpolation. A 6 megapixel Bayer sensor has 3 million green photodetectors, 1.5 million red photodetectors and 1.5 million blue photodetectors.

A 3.4 megapixel Foveon sensor has 3.4 million green photodetectors, 3.4 million red photodetectors and 3.4 million blue photodetectors, but these are overlaid, so there are only 3.4 megapixels.

There are trade-offs. On a Bayer sensor, the color of each pixel in the final image is determined by the colors of adjoining pixels. On a Sigma camera with a Foveon sensor, the software does spatial up-scaling.

The advantage of the Foveon approach was supposed to be more accurate color, and if silicon color separation was perfect then this would in fact be the case.

Spatial Resolution
It's important to understand that no matter how many photodetectors you have for each pixel, it does not increase the spatial resolution. More accurate color is perceived by the brain as higher resolution, but the Foveon X3 Pro 10M sensor does not produce more accurate color than a Bayer sensor (the original Foveon camera, which consisted of three separate sensors and a prism, actually did produce very accurate color).

Color Accuracy
Foveon claim to fame was supposed to be more accurate color, because the color of each spatial pixel would be calculated from the three layered photodetectors, rather than calculated from the colors of adjoining spatial pixels. Unfortunately for Foveon, their sensor has its own issues with color accuracy.

Sensor Size
Sensor size is also important. With a smaller sensor, you are using only the center portion of the lens. While this results in less chromatic aberration, you lose a lot of the wide angle capability of the lens. The tiny sensor in the Sigma SD9 and SD10 results in a very large crop factor. In a way, the Foveon sensor is ideal for Sigma, because shooting through the center of the lens helps to mask optical issues with the lower end lenses (the center of the lens is optically better).

A professional photographer may limit themselves to 80% of the full frame when using wide angle lenses*. This is essentially a 1.25 crop factor. This reduces chromatic aberration from the edges of the lens. So if you're shooting mainly wide angle, the less expensive Canon EOS-1D Mark II, with the 1.3 crop factor, may be sufficient. But the 1.7 Sigma crop factor is just too large, and results in the need for very wide angle lenses to compensate. The next Foveon sensor will probably be larger, and will have a smaller crop factor (rumors are 5.6 megapixel with a 1.3 crop factor). The Canon 1.6 crop factor, and Nikon 1.5 crop factor are also a pain. It means that for a wide angle lens, you must go down to something like the Canon 10-22 EF-S lens, which puts the rear element further into the camera body (the EF-S lenses cannot be used on Canon's film SLRs or on their full frame or 1.3 crop factor digital SLRs).

What you really want is larger individual photodetectors since noise is lower on the larger photodetectors, and larger overall sensor size (as close to full frame as possible). It's ironic that larger geometry semiconductor fabrication processes are usually better for the production of larger sensors with larger photodetectors, though there are other devices that also benefit from larger geometry processes. Canon is the only producer of high-quality, very high-resolution, full frame sensors. Canon also produces a 1.3 crop factor 8.2 megapixel sensor that is very good.

*Thanks to a poster on the dpreview Sigma forum for pointing out the fact that professionals often limit themselves to 80% of full frame.

Camera (in order of Megapixel Resolution)	Megapixels	Crop Factor	Photo- Detectors per Pixel	Sensor Size	Pixel Pitch	Frame Speed (frames/sec)	Technology
	More is Better	Smaller is Better		Larger is Better	Larger is Better	Higher is Better
Sigma SD9	3.4	1.7	3	21x14	9.1	1.9	CMOS
Sigma SD10	3.4	1.7	3	21x14	9.1	1.9	CMOS
Sigma SD14	4.65	1.7	3	21x14	?	?	CMOS
Olympus E1	5	2	1	18x14	6.8	3	CCD
Canon EOS-300D	6	1.6	1	23x15	7.4	2.5	CCD
Canon EOS-10D	6	1.6	1	23x15	7.4	3	CCD
Canon EOS-20D	8.2	1.6	1	23x15	6.4	5	CCD
Nikon D70	6	1.5	1	24x16	7.8	3	CCD
Nikon D100	6	1.5	1	24x16	7.8	3	CCD
Fuji S3 Pro	6	1.5	2	23x16	Not Specified	2.5	CCD
Pentax *istD	6.1	1.5	1	24x16	7.8	2.6	CCD
Pentax *istDs	6.1	1.5	1	24x16	7.8	2.8	CCD
Canon EOS-1D Mark II	8.2	1.3	1	29x19	8.2	8.5	CMOS
Canon EOS-1Ds Mark II	16.7	1	1	36x24	7.2	4	CMOS
Canon EOS-1Ds	11	1	1	36x24	8.8	3	CMOS
Canon EOS-1D	4.1	1.3	1	29x19	10.8	8	CCD
Kodak DCS SLR/n	14	1	1	36x24	7.9	1.7	CMOS
Kodak DCS SLR/c	14	1	1	36x24	7.9	1.7	CMOS
Konica-Minolta Maxxum 7D	6.1	1.5	1	24x16	Not Specified	3	CCD
Olympus E-300	8	2	1	18x14	Not Specified	2.5	CCD

I have also prepared some charts that graphically show the difference in resolution and sensor size of most of the digital SLR cameras. As you can see, Sigma has the lowest resolution, and close to the smallest sensor size. The colors go from red (worst) to green (best).

Film Equivalency
With a film camera, you can change the type of film to suit your particular situation, i.e. low ISO film for the best resolution in bright light, higher ISO for lower light. Obviously you can't change your digital camera's sensor in the same way.

Often you see inquiries about digital SLRs couched in terms of how close the camera is to a certain film type. Only in 2004 did we see a digital camera that could be considered close to medium format, the Canon EOS-1Ds Mark II. But we also saw the Canon 20D which is a lower priced D-SLR that rivals 35mm professional film.

The "Film Equivalency" column shows how the camera would be rated in terms of film cameras. Since film equivalency is more "informed opinion" than absolute fact, I have separated it from the camera specification table.

A 6 megapixel Bayer sensor has a mosaic filter placed over the sensor that result in 1.5 million red pixels, 1.5 million blue pixels, and 3 million green pixels. Each pixel has one photodetector. The colors in the final image are calculated, based on the values of the specific pixel, and the eight surrounding pixels. A camera with a 6 megapixel Bayer sensor is marketed as a 6 megapixel camera because the sensor has 6 megapixels.

A 3.4 megapixel Foveon has 3.4 million pixels, each with three photodetectors. There is no mosaic filter. The colors are separated based on how deep each color of light penetrates the silicon. The colors in the final image are calculated from the three photodetectors that comprise each pixel. In theory, the Foveon sensor should have more accurate color for each pixel, though this is not actually the case, because the Bayer mosaic filter is much better than the Foveon silicon filtering. The silicon filtering is imprecise and results in a lot of color contamination between layers.

The reason why the Bayer sensors are so much better, is because manufacturing a very precise mosaic filter is not difficult, while fine tuning the absorptive properties of silicon is virtually impossible (it will vary from sensor to sensor). The extremely powerful post-processing software that is used with Foveon sensor based cameras is able to help correct the color accuracy, but it still isn't as good as a high end, high resolution, Bayer sensor.

An image with more accurate color would be perceived as having higher spatial resolution than an image with equivalent spatial resolution but less accurate color. So three photodetectors per pixel, if in fact they provided more accurate color, would be an advantage.

Foveon clouds the issue of pixels by claiming that each of the three photodetectors that comprise each spatial pixel, should be counted as pixels. In reality, the Foveon sensor has 10.2 million photodetectors, but only 3.4 million pixels.

3CCD Camcorders

Some higher end digital camcorders use three CCD sensors, and use a prism to separate the light into red, green, and blue. This is similar to what Foveon used to do on their older studio camera. However the camcorder manufacturers don't overlay the sensor information, they offset (shift) it, to create kind of a pseudo-Bayer mosaic, but without the need for a mosaic filter.

The down side of the 3CCD approach is that, given the same total sensor area (with each CCD of a 3CCD camcorder being 1/3 the size of the CCD in a single CCD camcorder), the low light performance isn't as good as what is obtained from a single CCD (with a primary color filter as opposed to an Bayer type mosaic filter). The higher end consumer 3CCD camcorders use bigger sensors which solve the luminance issue. For around $1200 you can buy a three CCD camcorder with good low-light performance. The prosumer camcorders also use 3CCDs with larger sensors.

In short, when camcorder shopping, avoid the low-end 3CCD camcorders such as the Panasonic PV-GS120. If you're spending only $600, buy something with a large single CCD, and a primary color filter (such as the Canon Optura 30, 40, or Xi). For more advice on camcorders, visit http://nordicgroup.us/camcorder .

These are not in any particular order. As I become aware of more reasons I simply add the new reasons onto the bottom of the list.

Some people mentioned some very weak reasons to buy the SD10, such as 'it comes with an AC adapter,' and 'it's readily available on-line, while the Canon 10D is very hard to find.' I left the desperately bad reasons out. Anyone who purchases a $1200-1300 device and worries about $50 for an AC adapter, clearly isn't thinking straight, unless the AC adapter is the ONLY difference! Similarly, when a desirable product is difficult to buy, due to supply issues, the prudent thing to do is to wait for the product, not buy something that isn't as good.

Better than what's available for the Canon 20D or Nikon D70. This software evolved from the Foveon software that was developed for their studio camera. However this software still doesn't have the functionality needed by professionals, so Adobe Photoshop would still be needed (in the event a professional ever used a Sigma D-SLR). Some of the functionality that PhotoPro doesn't provide includes:

Adobe Photoshop now supports the Sigma RAW format and is what would be used by a professional.

The SD10 has an IEEE 1394 interface. The Canon 20D has USB 2.0, which is as good as IEEE 1394. The Nikon D70 has a USB 2.0 interface, but it operates at USB 1.1 speeds.

For Nikon D70 users, the best option is to remove the Compact Flash card, and place it into a USB 2.0 card reader. This is what most people do anyway, no matter what the camera.

With Canon and Nikon, you have amateur lenses and professional lenses, with a huge price difference; there is nothing in the middle. Sigma's EX line of lenses is a good compromise for the prosumer, since they are of higher optical quality than the entry level Canon and Nikon lenses, but much less expensive than the professional Canon and Nikon lenses. Of course these same Sigma EX lenses are available for the Canon and Nikon bodies, so it really isn't an advantage, except for the fact that the Sigma lenses for non-Sigma mounts, often have limitations.

While the Sigma D-SLRs don't have a good lithium-ion battery solution as other cameras, some users like the ability to substitute AA non-rechargeable batteries for their Ni-MH AA batteries, if they get caught out with flat batteries.

These are not in any particular order. As I become aware of more issues I simply add the new reasons onto the bottom of the list.

Image Quality

While most reviews agree that the Sigma SD10 image quality is not as good as the Canon 300D, 20D or Nikon D70, though there are specific situations where the SD10 comes close, or exceeds the Canon and Nikon cameras. Image quality, especially at higher ISOs and in lower light, is the primary reason to opt for the Canon or Nikon offerings.

Since "image quality" is rather vague, I have not included it as a stand-alone item. I have broken it down into the reasons for the image quality issues. These include the ISO limitations of the sensor, sensor noise, the low megapixel resolution of the sensor, color accuracy issues with the sensor, the auto-focus system, the lens quality, etc. Some of these items are very inter-related to each other (i.e. noise and ISO limitations), but are often singled out as problems so I have listed them separately. I realize that there is some repetition here.

Bottom line, unless you're a studio photographer, shooting at low ISOs with excellent lighting, the Sigma SD10 is not the best choice.

At lower ISO speeds the noise level is okay, but as you go up in ISO, the noise increases greatly. The SD10 does improve somewhat over the SD9, mainly due to the addition of microlenses on the sensor.

The higher ISO speeds need to be avoided because of the noise issue. Canon and Nikon D-SLRs don't suffer as much from this.

The slower flash sync limits your ability to use fill flash. It's not terrible, but it's not as good as what's available from Canon and Nikon.

Sigma probably has the most limited AF system on the market today. It has only one Auto-Focus Sensor. This is worse than many point and shoot digital cameras. There are no user-selectable auto-focus points. Also, the auto-focus doesn't work well in low contrast situations (Canon is especially good at AF in low contrast).

Part of the auto-focus problem can be attributed to the fact that many people are using the low-end Sigma lenses. With the Sigma EX lenses, this is less of an issue. But of course the Sigma EX lenses are more expensive than the very capable non-L lenses from Canon, and the amateur lenses from Nikon.

The auto-focus speed on the SD10 is actually pretty good, on a par with the Canon 10D, though not as good as the Canon 20D or the Nikon D70.

The Foveon sensor lacks an anti-aliasing filter. This results in a high degree of aliasing. Amusingly, one reviewer stated that the aliasing was less pronounced when using the low-end Sigma lenses, because the blurriness masks the aliasing. This is actually true; the sharper the lens, the more pronounced the aliasing, but it's not a good reason to use poor lenses.

Some people claim that the images from a Foveon sensor camera look sharper than images from other 3.4 MP cameras. But this imagined sharpness comes at the expense of inaccurate rendering of diagonal edges and thin lines, and sometimes moiré effects. So don't be misled by people who claim that the Sigma is "ultra-sharp" while the Canon is "soft."

One of the biggest criticisms of both the SD9 and the SD10 is the color accuracy, especially at speeds greater than ISO 400. The Canon 20D and Nikon D70 do far better, even at ISO 1600.

Ironically, the claim to fame of the Foveon sensor is supposed to be better color accuracy, since it has three photodetectors at each pixel. Theoretically, the Foveon type of sensor should provide better color accuracy. In fact, the original Foveon camera had three separate sensors, not one sensor with three layers. This did help with color accuracy. But Foveon couldn't go down-market, into consumer-level cameras, with such a complicated three-sensor system (it used a prism to separate the light beam). At lower ISOs, the Sigma cameras do well.

This is related to the lenses, but since you can only use Sigma lenses on the SD10, it's fair game. If using the non-EX, lower-end, Sigma lenses, there is significant chromatic aberration in high contrast images. The Canon, non-L, EOS lenses do not have nearly the CA as the Sigma low end. Generally, the Canon non-L lenses are a bit lower than the Sigma EX lenses in quality of optics.

Sigma is Primarily a Lens Manufacturer
Sigma is the largest independent lens manufacturer (some would claim that they are not independent, since they do manufacture the SA mount lenses for their own cameras, but this is a tiny portion of their business). Sigma manufactures low to mid-range lenses for SLR cameras made by the top tier of SLR manufacturers, in addition to making lenses for their own line of film and digital SLRs. Their higher end line, the EX line isn't bad, it's often optically better, than the consumer grade lenses from Canon and Nikon.

Avoid the Entry Level Sigma Lenses
The entry level Sigma lenses should, in most cases, be avoided. While there are a couple of lenses with acceptable optics, they are clunky, and lack the build quality of Nikon and Canon entry level lenses.

Sigma EX Lenses
With Canon and Nikon charging premium prices for their professional lenses, Sigma has carved out a profitable niche with their EX line, especially for amateurs that want something a little better than the entry level lenses from Canon or Nikon, but less expensive than the Canon and Nikon professional lenses.

No Professional Lenses
Sigma does not have a professional line of lenses, nor do they have any known plans to offer one. Though for most users, the EX line is fine.

While the SD9 was virtually unusable for night shots due to noise, ISO limitations, and the poor AF system, the SD10 has improved the situation somewhat. However it still is below the level of the Nikon D70 or the Canon 20D and 300D.

As one reviewer wrote: "the camera is quite usable for after-dark photography, but you'll have to accept higher noise levels than you'd find with cameras such as the Nikon D100 or Canon's EOS-10D and Digital Rebel," and, "the d-SLR competition still smokes it when it comes to image noise with long exposures," (see http://www.imaging-resource.com/PRODS/SSD10/SD10A12.HTM)

The SD10 has a 3.4Mp sensor. This is the lowest resolution of any currently available digital SLR. The competition uses 5 Mp sensors at a minimum. Sigma claims that the type of sensor they use has an effective resolution of 10.2 Mp, but this is untrue. The SD10 has 10.2 megaphotodetectors, but not 10.2 megapixels. The fact is that the Sigma, with the Foveon sensor, is lower resolution than many point and shoot digital cameras.

Here is the most concise and accurate explanation of Foveon versus Bayer sensors that I've seen:

On both types of sensor, the color information for each pixel is calculated based on the color information derived from the photodetectors. In theory, since each pixel on a Foveon sensor has three color photodetectors, the color of each pixel should be more accurate, but this is actually not the case because of issues with the layered photodetectors.

This problem is due to the Foveon sensor. The way the Foveon sensor works, leads to poor color discrimination, especially in the blue channel. The blue layer is "contaminated" by the green layer and the red layer underneath. The Green layer is "contaminated" by the red layer. If you look at sample images that are used to promote the Sigma SLRs, you'll notice that they tend to emphasize images with a lot of red.

Ironically, the early Foveon based cameras did not have this problem. These cameras used three separate sensors, not one layered sensor. So there was no contamination from other layers.

Okay, I'm comparing the Sigma EF500 flash for the 20D with the Canon 580EX flash for the 20D, because there is no other flash for the SD10 other than the Sigma EF500. The Sigma flash is less powerful (more comparable to the Canon 420EX). As with lenses, the build quality of the Sigma flash is not up to the level of the Canon and Nikon flash units, and lacks some useful features.

The tendency to over-expose or under-expose some shots is more related to the limitations of the camera body and the poor auto-mode. You may want to do flash photography using manual mode.

Reviewers have complained about dust on the sensor, despite the presence of a protective cover. Note that the sensor area is not hermetically sealed, just covered.

A Nikon or Canon owner can use their own camera maker's lenses, as well as lower cost, after-market, lenses from Sigma, Quantaray, Tamron, Tokina, Vivitar, etc. Sigma SA mount owners are locked into a limited selection of Sigma lenses.

With Sigma, you MUST use a flash attachment for low light conditions. On the Canon 20D, Nikon D70, and most other consumer and prosumer D-SLRs, you have a small pop-up flash that is quite useful at times. The pop-up flash certainly has limited range, but their are many instances when it is sufficient, especially as a fill flash. With the Sigma, you'd better ALWAYS carry along that EF500 flash attachment.

One reason for Sigma's low frame rate (6 frames at 1.9fps) is because the buffer is so small.

The SD10 shoots only in RAW mode. There is no option for JPEG. This means that a) you MUST post process every photo on a computer, and b) storage requirements are very high. The Foveon sensor is a 3.4 Megapixel sensor, but the stored images are an average of 8MB. On other D-SLRs you have the options of RAW or JPEG mode. Clearly Sigma didn't want to bother with putting in a JPEG compression engine. Remember that Foveon based cameras have always been RAW mode only, because the post-processing requirements are such that it would be expensive to build in the post-processing engine. This was not an issue when Foveon based cameras were only used in studios (the original Foveon based camera had no in-camera storage at all). The upcoming, 1.5 Mpixel, Foveon-based, Polaroid point and shoot camera, does have a JPEG mode.

On the plus side, RAW only mode forces you to become skilled at post-processing.

While the SD10 eliminated the bizarre dual battery system of the SD9, Sigma still doesn't offer a good Li-Ion rechargeable battery solution.

Sigma announced, that sometime in late June 2004, they will have a Li-Ion battery solution. Click here for details. The price is around US$90. The Sigma Li-Ion solution requires two RCR-V3 type Lithium-Ion rechargeable batteries, but the charger charges only one battery at a time. This is a royal pain. Most people charge their batteries overnight, or have two sets, so one is charging while the other is being used. You'll need two chargers, and four batteries for two sets of batteries, so that one set can be charged while you're using one set. The Sigma batteries and charger can be purchased at Adorama, click here for a direct link.

There are other RCR-V3 solutions as well, but they all also require two batteries and two chargers. Furthermore, a single, after-market, proprietary Li-Ion battery for a Canon or Nikon is less costly than a single, lower capacity, RCR-V3 battery, and you need two RCR-V3 batteries for the Sigma SD10.

Be forewarned that not all RCR-V3 batteries will work with the SD9 and SD10. Problems have been reported with Moby Power and Delkin (eFilm) RCR-V3 batteries, while Nexcell and Amondo appear to be fine. The problem relates to the initial current draw at power-up (some RCR-V3 batteries cannot handle the initial surge of current at power-up). See http://www.thomasdistributing.com/rcr-v3-index.htm for the Nexcell RCR-V3 batteries.

Be especially wary of the no-name RCR-V3 batteries sold on eBay. Stick with Nexcell or with the Batteries Sold by Sigma!

Search the Sigma dpreview forums for the posts about this issue. The bottom line is that you're probably going to be better off with the highest rated AA cells (2500mAH) then the very expensive RCR-V3 solution.

This is one area where Sigma has work to do, or maybe not. Canon is ahead of Sigma in ergonomics. Sigma is strictly old-school, with multiple dials to select functions, rather than one multi-function knob. However, some traditionalists may prefer the Sigma approach, because the Canon approach has a learning curve.

The separate dials for aperture and shutter priority are a pain, though again, some people prefer this approach.

Bottom line, most experts agree that the SD9 and SD10 lag the Canon 20D and Nikon D70 in terms of ergonomics, but there are personal preferences involved here as well.

Always carry a small flashlight if you're going to be using the SD10 in the dark, because the LCD information panel isn't backlit.

Sigma has a glass cover over the sensor. But dust still manages to get onto the sensor. To blow the dust out requires the removal of the internal glass cover. This is not required on other D-SLRs. If the glass cover actually hermetically sealed the sensor, then it would have been a good feature. One persons stated that they were worried about the accumulation of moisture behind the glass cover, in high humidity environments.

The Sigma SD10 lacks a USB 2.0 interface. The Canon 20D has a USB 2.0 interface. USB 2.0 is much more common on computers than 1394, though on a desktop computer it's easy to add a 1394 PCI card. This isn't a huge issue, as you can always remove the Compact Flash card and stick it into a USB 2.0 card reader.

According to Steve's Digicams, it took one hour to transfer 20 images from the SD10 to the computer via USB 1.1. The lesson here is to either use Firewire, or a card reader hooked to a USB 2.0 interface.

Sigma cameras produce only RAW images. This wouldn't be so terrible if the conversion to usable images was simple. But it's not. Expect to spend hours with Photo Pro enhancing each individual image. As one reviewer wrote (of the SD9): "there is extensive post-processing to get rid of the tendency to go yellow. Initially this didn't perturb me...but it does get tiresome considering point and shoots can do a better job."

This is a major problem with the Sigma SD10, at least for users that have the need for continuous shooting at a fast rate. The small buffer, and the slow processing results in an abnormally low frame speed. Only the Kodak digital SLRs are slower than the Sigma SD10. But Kodak at least has somewhat of an excuse, since they have a 14 megapixel, full frame sensor, which requires more processing and storage time.

Now to be fair, in a studio environment, or for shooting landscapes, the frame rate is not all that critical. But for action shots, wildlife, etc., it's a big issue. Canon did a good job on the 20D, with 5 frames per second, as opposed to 3 frames per second on the older 10D.

You can purchase an optional adapter for this functionality (PC Synchro Terminal Adapter [ST-11]). This same issue occurs on the Nikon D70 where you have to purchase the AS-15 Sync Terminal Adapter. The Canon 10D and 20D have this capability built in; there is no need to purchase an accessory.

A full frame digital camera has a focal length multiplier of 1. But since full-frame digital camera sensors are very expensive, only the most expensive digital SLRs use them. The smaller the focal length multiplier the better. For example, a 20-35mm wide angle lens on a full frame camera will give you a range of 20-35mm. On a camera with a 1.7 focal length multiplier it will give you a range of 34-60, which is not wide-angle. You need to get extremely wide angle lenses to still have wide angle capability on a digital SLR with a large focal length multiplier. A professional probably would not go below the 1.3 focal length multiplier of the Canon EOS-1D Mark II, unless they had no need for wide angle capability. Sigma is only a little worse than the mainstream consumer and prosumer digital SLRs, and is better than the Olympus digital SLRs.

Sigma carried the full frame viewfinder over from their SA9 film camera. This causes a problem because the Foveon sensor is much smaller than full frame; a lot of what you see in the viewfinder doesn't end up in the photograph. Sigma came up with a half-baked solution, which is really no solution at all. They made the edges of the viewfinder opaque, so you can still see what's outside the actual area that will be in the photograph. They call this a "sports finder," which is supposed to help you prepare for action photos since you can see the subject approaching the actual area of the viewfinder that represents what will be captured.

The problem with this approach is that the area inside the viewfinder, that represents the captured area, results in a small viewfinder image. Remember, one design issue with D-SLRs is that you cannot use the LCD as an electronic viewfinder, you only have the optical viewfinder, so it had better be good.

The proper way to have done the viewfinder would have been to redesign it to match the sensor size. This would have resulted in a larger view of the area that is actually captured.

Don't you just love it when a manufacturer of a product tries to turn a negative into a positive? You mostly see this on automobiles, where manufacturers decontent from one year to the next, and the salespeople have to rationalize the decontenting to the customer.

Shutter lag is normally much better on digital SLRs than on point and shoot digital cameras. But the SD9 and SD10 manage to have a relatively long shutter lag. Steve's Digicams has a very good paragraph on this problem, see the seventh paragraph over at http://www.steves-digicams.com/2003_reviews/sigma_sd10_pg7.html for details, as the reviewer concludes, "While this performance might seem robust compared to consumer digicams, it lags its competition in the dSLR market."

Since the SD9 and SD10 are not popular, there are less accessories available for them. For example, there is no waterproof case available for the Sigma digital SLRs, while you can get one for the Canon 10D, 20D, and Nikon D70 (see http://ikelite.com/web_pages/1digital.html).

Sigma has no plans to move up-market to the prosumer or professional level, and could not do so without a massive investment in product development. This is unlikely to occur. Not only would they require a new body with a new sensor, they'd have to produce a line of professional grade lenses that rival the Canon L series and the Nikon professional lenses.

On the other hand, Sigma is manufacturing a professional grade SLR camera for Kodak, with a Canon lens mount (though it has received very poor reviews). So if you ever want to move up from the SD10 to the DCS-14/c, you're going to have to buy lenses all over again. With Canon, you can start with the EOS-300D or 20D, and later graduate up to the professional level digital body while continuing to use your existing lenses (except for the Canon EF-S lenses which will work only on the 20D and 300D). You could even move to the Sigma/Kodak body, provided the genuine Canon lenses work properly with this body. However, the Fill-Factory 14 megapixel sensor used in the DCS-14/c has not been well received.

Very, very few retail camera stores carry the Sigma D-SLRs, or SA mount lenses. You will probably have to purchase everything, sight unseen, from a web or mail-order retailer unless you live close to New York City. You will often see Sigma lenses for other mounts (Nikon or Canon) at larger camera retailers, but rarely any SA mount stuff (unless you go to B&H or Adorama in New York City).

The SD10 body sells for about $1350 ( http://www.adorama.com/SGSD10.html). You can buy the far superior Canon 20D body for around $1250 (on sale), or the Nikon D70 body for about $900. Plus, with the 20D and D70 you'll have a wider variety of lenses available and you can use the consumer level Canon or Nikon lenses and still obtain fine results. With Sigma, you have to move the more expensive EX series of lenses in order to get good results.

Note that the price comparison of the SD10 against the other models, are made by using the lowest price from a reputable dealer (i.e. Dell, Buydig, Amazon) and prices vary over time, with sales, rebates, and coupon codes. The problem for the SD10 is that it's only carried by a few dealers, and hence there is no competition on price. One person on the Sigma forum stated that 20D was more expensive than the SD10 at Adorama, which is true (the 20D is $1370 versus $1350 for the SD10). But the 20D can usually be found for around $1300-1325, and sometimes for as little as $1250 when Dell combines 10% off with a $ off coupon.

While I have strived to be as accurate as possible, you may want to also look at other information sources about the SD10.

Sigma Forum on dpreview.com: http://forums.dpreview.com/forums/forum.asp?forum=1027 Warning, a lot of mis-information is posted on the dpreview forum, by a very few people. Any posts to that forum that say anything negative, no matter how accurate, are met with a barrage of fact-free attacks. However in the past few months, there has been a big change in that group, with a lot more people admitting that they are stuck, and talking about switching to other systems.

If you want to get more information about the Sigma D-SLRs there are a couple of forums that are available. rec.photo.digital is the Usenet forum, but I suggest you enter it via a regular newsreader, not via Google Groups. For some reason, rec.photo.digital has been attacked by a bunch of low-lifers, so considerable subject filtering, and blocking of senders is necessary in order to get anything of value.

There is also the dpreview Sigma forum. You can get some good information on this forum, but unfortunately there are some very insecure people over there, who get extremely upset if anything negative is posted about the SD9 or SD10. Anytime anyone says anything negative about Sigma or Foveon, you'll see a flurry of posts along the lines of:

The Canon EOS-1Ds Mark II is the only professional, low-noise, full-frame, high megapixel, digital SLR camera available. Canon knows this, and charges a premium price for the EOS-1Ds Mark II, around US$8000.

I hope that this web site has cleared up some of the questions regarding the Sigma Digital SLR cameras.

The bottom line is that the Sigma SD10 is simply not a good value. It uses a low resolution, 3.43 Mp sensor, while the competition use at least 6 Mp sensors. It doesn't have as wide a variety, or as high a quality, lenses available. It lags the competition in auto-focusing, noise, and night-shot capability. Some of these flaws could be overlooked if it were exceptionally low priced, but it isn't--it actually costs more than the Canon 20D (when the 20D is on sale, which is always). However, I would not buy one no matter how inexpensive because the low resolution is just too big of an issue. There is simply no reason to buy a low-resolution D-SLR when Canon and Nikon have such fine six and eight megapixel models available at such attractive prices.

If you'll only ever shoot at low ISOs, in good light, and, most importantly, already have a collection of SA mount EX lenses, then the SD10 could be a good choice. The worst thing you could do would be to start spending money on a collection of lenses.

Foveon is reportedly working on a higher resolution sensor, with a 1.3 crop factor. If Sigma puts a higher megapixel, lower crop factor, Foveon sensor into an SLR, and the other issues are resolved, then the SD11, or whatever the next model is called, may be a good choice.

Model	Film Equivalency
Sigma SD10	APS
Olympus E1	APS
Kodak DCS SLR/n	35mm Consumer
Kodak DCS SLR/c	35mm Consumer
Canon EOS-300D	35mm Consumer
Nikon D70	35mm Consumer
Fuji S3 Pro	35mm Consumer
Pentax *ist D	35mm Consumer
Canon EOS-20D	35mm Professional
Nikon D100	35mm Professional
Canon EOS-1D Mark II	35mm Professional
Canon EOS-1Ds Mark II	Medium Format