I have been having a bit of a dilemma with my Sigma SD Quattro H (SD Q H), thinking I have a bad sensor or a firmware problem. The noise I have been getting out of my new SD Q H has been quite out of hand and unacceptable for the product photography I bought this system for. The SD Q H is an “upgrade” from my SD15 that I have gotten several years of outstanding results from.
As with most any project I take on, I start out with suggestions from those going before me and “tweak” things as I get used to my gear. Unfortunately I had to trash pretty much all of the images from one photo session and many of the images from several other sessions due to noise problems. The noise was so bad I would have had to do extensive patch work to make these pics marginally acceptable.
First off, at the suggestions from my predecessors, I set things up as follows: (1)Use the lower ISO’s (100-400) (2)Skip processing in Sigma Photo Pro (SPP). (3)Write to a 12 bit DNG file and use a commercial RAW processor.
I have never liked SPP so taking these suggestions has been no problem. The only problem is, at risk of repeating myself again, the results were terrible
Not knowing what I might find, I took both a color test image and a resolution test image out into the back yard; as it was cloudy the light was pretty even. I took a series of images using my 105 mm fixed length portrait lens.
I took images in: (1)Super fine detail (SFD) mode. (2)Images of each target at ISO 100 writing both DNG and X3F files. (3)Images of each target at ISO 400 writing both DNG and X3F files
I didn’t bother with the higher ISO’s since most everyone agrees that noise from the Foveon sensor above 400 rapidly becomes unacceptable.
I went through each sequence and target without moving the camera to facilitate comparing images.
The DNG images were processed using Adobe Camera Raw (ACR) with only lens correction. The X3I files and the X3F files were converted into TIF files using SPP with no processing.
The X3I files were useless. I was unable to align the 7 images SFD mode takes. I think I will have have to use my old workflow of auto bracketing 5 shots for my HDR work.
With the other files there were some glaring issues. I increased the image size to 300% to illustrate the point. Yes, I know…but you can still see the problems at 100% magnification too. And I do enlarge up to 24 x 30 on occasion.
There are huge differences between ISO 100 and ISO 400 images. In the DNG images the noise in the ISO 400 image was much more apparent. In the X3F files the line definition in the ISO 400 file was much degraded. Though you can’t see it at first glance, the line definition is as bad or worse in the ISO 400 DNG image. You can see it better once you treat the noise. (click on the images to enlarge them)
Comparing the DNG to the X3F files there is no way I can say the DNG images are remotely close to the quality of the X3F images. Only in certain situations can I call the DNG images acceptable to my standards at all. Apparently the in camera processing that takes place to generate the DNG file introduces a large amount of noise.
To patch the images, surface blurring the individual RGB channels seems to work. Most of the noise, as usual, is in the red channel. I have also had just as good or slightly better success changing the image mode to LAB and treating the noise in the lightness channel.
The fact is, I really didn’t buy into this system for pixel numbers, I bought it for its color performance. Though it takes a little work, I can really get the colors to stand out and the skin tones to look great, even in the ISO 400 files, if I save them as X3F files. But it takes some work. Lesson learned, shoot as much as you can at ISO 100 and save the files only as X3F. You cannot get around using SPP, but you can use it only for conversion. Use your main image editing software for making things look good. You just have to remember, there are no shortcuts with this system.