After looking at the banding problem in godfrog's GX20 I decided to try to eliminate the variable of whether I need to support the K-7 for these fringing corrections. I quickly applied the algorithm to my available high ISO 6400 K-7 DNG files. The good news is based on the K-7 raw samples from Imaging Resource (IR) ( www.imaging-resource.com ), the algorithm for compensating for the black offset levels seems to be completely changed from that of the K20D such that so far I can detect absolutely no Horizontal/Vertical Pattern Noise (HPN/VPN), nor is there any significant offset shift across the sensor in either direction according to the correction algorithm, and although the overall offset compensation could be slightly tweaked to give slightly blacker blacks, this would barely improve the Dynamic Range (DR) in the deep shadows which is mostly limited by the level of the noise anyway (about 6 stops for only a moderate image quality at ISO 6400).

It looks like the PRIME2 imaging engine has greatly improved algorithms for the scanning and on-the-fly offset compensation and scaling which may be modeled along the lines of one I suggested back with the problems of VPN on the K10D. I don't have a lens-cap shot for the K-7 at ISO 6400 but from what I see in the masked black borders Pentax is doing a great job of processing given the sensor signals with which they have to work.

Based on this analysis and unless I hear otherwise, I will stop referring to the K-7 for the rest of this project and stick to the K20D and GX20.

Regards, GordonBGood