GordonBGood
Veteran Member
Yesterday I went to TheCameraStore in Calgary where I am vacationing and got to handle a production version K-7 as well as take away some test shots on my SD card. As far as handling and feature set compared to my K200D - WOW - everything one could possibly look at is times two, except for the price which is times three ;- . There is only one item that is not times two, and in fact is pretty much the same as my K200D or the K20D and that is Dynamic Range (DR), and this may have even gotten a little worse. My approach in determining this was as follows:
1) Although I shot two images each of 7 stop underexposed and normal exposures of a gray card (supplied by the store) at all ISO's from 100 to 6400 in DNG raw mode, I ended up basing my conclusions on mostly just the black areas of the sensor that are masked to light that the processing uses to determine the black level compensation. These images were shot with all High ISO Noise Reduction turned off in the Custom Menu and Extended Dynamic Range was not used.
2) I used my conversion software to output a Colour Filter Array (CFA) map and looked at the standard deviation of the green photosites in this black unexposed area. The standard "bell curve" has of course been clipped to zero at about the median, but I determined the standard deviation by using the fact that about 84& of the readings will be below the higher inflection point.
3) These standard deviations were compared to those found by the same method for similar K20D images.
Given that DR is limited by the noise "floor" at the black end of the response (as per the DxOMark definition), the low ISO DR for the production K-7 is about the same as that of the K20D, and the same as that of the beta K-7 images I analysed a month or two ago. The higher ISO DR for the K-7 is in fact worse than that of the K20D but that may be due to a different start point for the Noise Reduction (NR) that is applied. I think that any perceived improvement in high ISO image quality will be only due to possibly improve NR techniques, but these will likely also tend to smear details, reduce resolution, and make the noise "grain" to be coarser.
So I am somewhat disappointed that the K-7's image quality in this one respect is not up to the standards of the rest of the camera and that I will have to wait at least another camera generation in order to see Pentax work on competitive image quality as to DR.
Now I will qualify this that those who were happy with the K20D as to image quality will likely also be happy with the image quality from the K-7, and those who are willing to use NR on their images in order to obtain reasonable results at higher ISO's such as ISO 1600 (or even ISO 800) and up may be reasonably satisfied although definitely not ecstatic.
A high DR specification is of use to people who want a wide exposure latitude or the ability to underexpose their images and recover them by EV boosting in post processing to a good image quality standard without NR, and also the ability to compress the DR by boosting the brightness of the deep shadows to reveal shadow detail without noticeable noise artifacts. Those who only make use of the top 7 or 8 stops of DR with well exposure bright images will never notice its lack. However, there have been many who have found even the lower ISO's of the K20D to be somewhat noisier than the competitition and therefore somewhat limiting. In this respect, the K-7 is no better, and may even be somewhat worse at the higher ISO's.
I did not want to believe these results and checked them using the raw DNG K-7 files from a production camera posted on Imaging Resource. The results were essentially the same. So unless you want to believe that two random samples in Calgary, Canada and another in Atlanta GA both have the same characteristics that are contrary to the norm, it looks like this is real.
To those that pooh-pooh technical results without images, I assure you that you will see more noise in push processed deep shadows for the K-7 samples than in equivalent push processed other cameras with a wide DR such as the newer Nikon CMOS sensors at say ISO 200 (since Nikon doesn't have a true ISO 100). For now I'll leave that exercise for the student. Note that you will have to look into the very darkest part of the images for this test with about a +4 EV exposure boost such that the dark part of the images are less then about 64/255 JPEG levels and preferably about 32/255 JPEG levels even after the 4 EV boost in the raw conversion.
It is much more difficult to compare high ISO images due to the variances in Bayer processing and Noise Reduction techniques.
Regards, GordonBGood
1) Although I shot two images each of 7 stop underexposed and normal exposures of a gray card (supplied by the store) at all ISO's from 100 to 6400 in DNG raw mode, I ended up basing my conclusions on mostly just the black areas of the sensor that are masked to light that the processing uses to determine the black level compensation. These images were shot with all High ISO Noise Reduction turned off in the Custom Menu and Extended Dynamic Range was not used.
2) I used my conversion software to output a Colour Filter Array (CFA) map and looked at the standard deviation of the green photosites in this black unexposed area. The standard "bell curve" has of course been clipped to zero at about the median, but I determined the standard deviation by using the fact that about 84& of the readings will be below the higher inflection point.
3) These standard deviations were compared to those found by the same method for similar K20D images.
Given that DR is limited by the noise "floor" at the black end of the response (as per the DxOMark definition), the low ISO DR for the production K-7 is about the same as that of the K20D, and the same as that of the beta K-7 images I analysed a month or two ago. The higher ISO DR for the K-7 is in fact worse than that of the K20D but that may be due to a different start point for the Noise Reduction (NR) that is applied. I think that any perceived improvement in high ISO image quality will be only due to possibly improve NR techniques, but these will likely also tend to smear details, reduce resolution, and make the noise "grain" to be coarser.
So I am somewhat disappointed that the K-7's image quality in this one respect is not up to the standards of the rest of the camera and that I will have to wait at least another camera generation in order to see Pentax work on competitive image quality as to DR.
Now I will qualify this that those who were happy with the K20D as to image quality will likely also be happy with the image quality from the K-7, and those who are willing to use NR on their images in order to obtain reasonable results at higher ISO's such as ISO 1600 (or even ISO 800) and up may be reasonably satisfied although definitely not ecstatic.
A high DR specification is of use to people who want a wide exposure latitude or the ability to underexpose their images and recover them by EV boosting in post processing to a good image quality standard without NR, and also the ability to compress the DR by boosting the brightness of the deep shadows to reveal shadow detail without noticeable noise artifacts. Those who only make use of the top 7 or 8 stops of DR with well exposure bright images will never notice its lack. However, there have been many who have found even the lower ISO's of the K20D to be somewhat noisier than the competitition and therefore somewhat limiting. In this respect, the K-7 is no better, and may even be somewhat worse at the higher ISO's.
I did not want to believe these results and checked them using the raw DNG K-7 files from a production camera posted on Imaging Resource. The results were essentially the same. So unless you want to believe that two random samples in Calgary, Canada and another in Atlanta GA both have the same characteristics that are contrary to the norm, it looks like this is real.
To those that pooh-pooh technical results without images, I assure you that you will see more noise in push processed deep shadows for the K-7 samples than in equivalent push processed other cameras with a wide DR such as the newer Nikon CMOS sensors at say ISO 200 (since Nikon doesn't have a true ISO 100). For now I'll leave that exercise for the student. Note that you will have to look into the very darkest part of the images for this test with about a +4 EV exposure boost such that the dark part of the images are less then about 64/255 JPEG levels and preferably about 32/255 JPEG levels even after the 4 EV boost in the raw conversion.
It is much more difficult to compare high ISO images due to the variances in Bayer processing and Noise Reduction techniques.
Regards, GordonBGood
'
