From: Jason Porter
Date: Thu, May 1, 2008 at 12:14 AM
Subject: SD14 Problem – Unusual Saturation
To: info@sigmaphoto.com
Cc: support@sigma-photo.co.jp
Attn: @@@@@ at Sigma America

@@@@@,

As we discussed on the phone, I am writing to document my new SD14 camera’s unusual red saturation problem and provide raw X3F samples for your review.

Before I explain, here are the details of my particular unit and configuration:

Sigma SD14
serial no. [redacted]
firmware revision 1.07

Sigma 18-50mm F/2.8 EX Macro
serial no. [redacted]

Using Sigma Photo Pro 2.5 in Windows XP, using a color-corrected display (Spyder2Express calibrated).

The problem:

I have had my SD14 for one week, and like it very much! Unfortunately, I have discovered that it exhibits some unusual behavior when capturing images in bright sunlight. This became most evident when I attempted to capture a close-up shot of a red flowering plant on my front porch. Specifically, the camera seems to hugely over-saturate reds, to the point of total saturation in the red channel at 255 (per the “loupe” feature in Photo Pro 2.5) This behavior is consistent in both Program mode and Manual mode at ISO 100. In order to bring the red saturation down to the point where the red channel ceases to clip at 255, the image must be significantly underexposed in the other channels.

What is particularly disturbing is that I am experiencing significant bloom into the green channel when I capture the same exact image at ISO 50. This occurs in areas of pure red on the subject that (when captured at ISO 100) are totally saturated at 255 red, with zero information in the green channel. When captured at ISO 50 using the same Program mode exposure and metering, the red channel is saturated at 255 and the green channel jumps to around 240 in many areas! This causes the flower to appear bright neon orange with obvious two-tone banding in the image. I can’t imagine that this is normal behavior for capturing an image of a common red flower in direct sunlight, and it was not even a particularly bright day!

I understand that ISO 50 has less “headroom” than the other ISO modes (per the note on the firmware download page), but even when using center metering directly on the brightest part of the flower petals, the saturation problem in the red channel is very much present. Also, this does not explain the green-channel bleed.

The X3F samples:

I have taken a series of raw X3F images of the red flower, for your review. If you have Photo Pro 2.5 on a Windows machine and you can use that to view these, it may be helpful so that we are comparing the same output. These images are all of the same subject in the same ambient light. All images are captured using the Daylight white balance setting on the camera, and using the 18-50mm F2.8 EX lens. I have hosted these files on my own server so that you can access them and download directly at your convenience. These are organized in folders by mode, by ISO, and by exposure.

Direct your browser to: http://jasonporter.us/photography/sigmasd14/

The first files to look at are the P mode ISO 100 and ISO 50 exposures, which are in folders marked “P_ISO100_EvalMeter” and “P_ISO50_EvalMeter”.

The P mode ISO 100 Evaluative Metering exposure (no. 182) exhibits the full red oversaturation I mentioned. The same shot taken in P mode ISO 50 (no. 183) exhibits the green-channel bleed resulting in bright yellow/orange coloration. A revised version of the ISO 50 shot using Spot Metering (no. 187) and metered directly on the red area shows the same behavior, but with slightly lower overall exposure and slightly less green channel bleed.

The rest of the files are captured in M mode to demonstrate the red channel behavior at various apertures and shutter speeds. There are two sequential sets of shots at F/2.8 and F/5.6 apertures, at a series of different shutter speeds, to show both the red channel behavior and the level of general overall green/blue underexposure required to bring the red channel into control. I have also included an F/20 1/15s shot and an F/8 1/250s shot, for reference at smaller apertures… these are both “best case” shots with the red channel just beginning to fully oversaturate.

It should be noted that if the aperture and shutter speed are reduced to keep the red from blowing out completely, the green areas and the background (and the shot in general) are significantly darker than they appeared to the eye. This subject was a balanced mixture of bright red and rich green tones. In fact, the P-mode Evaluative Metering shots that were so horribly overexposed on the red highlights are the closest to being accurate on the green background out of any of the shots, and even so, they are much darker and less saturated on those areas than they should be compared to the actual rich green color of the plant.

Thoughts:

The only clue to the origin of this odd red-channel behavior that I have come across was actually with the help of an open-source Raw converter program called UFRaw. (ufraw.sourceforge.net) It is a simple Raw-converter packge similar to Adobe Camera Raw or Sigma Photo Pro, and it supports the X3F format. Using UFRaw, there is an option available to disable the “Use color matrix” option for the input color gamut (in this case sRGB). Apparently the term “color matrix” is used in that software to refer to the color-space correction factor that is applied when normalizing an sRGB (or other) image source to the workspace. When this box is unchecked in UFRaw, some of the heavy red over-saturation decreases, and quite a bit of detail is restored. Unfortunately, no similar “feature” exists in Photo Pro 2.5. It isn’t perfect, but it is MUCH better than the default “corrected” setting, which is comparable to the look of the image in Sigma Photo Pro.

In any case, it raises the question: Is Photo Pro 2.5 applying an sRGB input gamut correction when it is inappropriate to do so?

I have checked these images on a non-color-corrected display, in order to rule out any potential problem with my calibrated monitor profile. The oversaturation behavior is consistent on all 4 computers that I have checked, including three different Windows XP systems, and an Ubuntu Linux machine using UFRaw.

By the way, I would like to be clear that I’m not naive about digital capture and the necessity of image editing… I have worked with digital workflow for quite a while and I recognize that every shot will require some tweaking to produce a quality, balanced result. With a camera like the SD14 that is designed for creative photography, that is even more true. This saturation problem, however, is not something that I can easily correct in Photoshop, and it’s not something that Sigma Photo Pro 2.5 seems capable of improving. This problem is simply blowing out all of the surface detail in the surface of brightly colored objects, and I don’t think that a 2 or 3 stop (or more) overall underexposure in the camera coupled with manually correcting the levels back up in the green and blue channels in Photoshop is the best way to achieve a professional result. Maybe I’m incorrect on that, but it’s not an approach that I have ever had to take with any other camera.

Thank you very much for your time and consideration, I know this was a long read, and I salute you if you made it through the whole thing. I would greatly appreciate your input and insight on this problem. I look forward to hearing from you.

And again, thanks for making a great product!

Best regards,

Jason Porter
Roanoke, Virginia, USA

P.S. – There is one other issue I’d like to point out, and see if you have experienced it. Some of the Raw X3F files are inexplicably much larger than others. Typical shots at 4.6mp in Raw mode are usually between 9mb and 13mb, but occasionally a file will be almost double in size. Shot number 206 in the F/5.6 set folder is a good example, the file is almost 22 megabytes! Is this a known issue with the 1.07 firmware?