Magenta edge tinging at high ISO and a possible solution...

Started Aug 17, 2009 | Discussions thread
GordonBGood Veteran Member • Posts: 6,308
Re: More on What about K10d raw files?

solarider wrote:

GordonBGood wrote:

It looks to me that is in the order of the size of the problem that I am fixing here with the K20D, especially in its worst case scenario. I don't suppose many K10D owners will read this thread given the K20D in the title, so maybe we need to start a new thread to poll them?

I recall popphoto mentioning this magenta cast with their K10D test and looked it up: "Magenta color casts in deep shadows" - found in What's (Not) Hot list.

Also below this list in the same test review is mentioned this: "There are some minor issues with magenta color casting in shadows, especially in JPEG files. These issues are less of a concern in RAW/DNG capture. We were also working with extremely early production samples, and Pentax may well do some firmware fixes before shipping units to retailers."

I'm fairly sure that popphoto was clear in conveying these findings to Pentax... just don't know why this has not been resolved in three of their DSLR's.

Nic, although the magnitude of the colour tingeing may be the same between the K10D and the K20D (two camera models), there is really nothing in common as to the cause of the tingeing.

For the K10D, it was caused by the the camera designers in thinking that there would be some benefit to scan the red/blue from the one channel and the greens with the other channel in opposite directions. This was hardwired into the sensor, the NuCORE data acquisition chips, and the way the PRIME imaging engine controlled them and could not be fixed with a firmware update. The K200D, which I own, and the K-m/K2000D use a similar sensor to that of the K10D but a different data acquisition system controlled by the PRIME imaging engine and do not have this problem so the problem was fixed when Pentax could - in the next model(s) hardwiring, which also have much less Vertical Pattern Noise (VPN).

The K20D, which has a similar data acquisition system and PRIME imaging engine to the K200D but a different sensor which is hardwired to be scanned and corrected slightly differently, doesn't generally have so much VPN other than for cases of some faulty photosites in a key location (like godfrog's K20D). However, the K20D has this different colour tingeing problem due to some characteristic of the sensor that is not being compensated. Again, this can not be corrected with the K20D using firmware, but is fixed in the next model.

The K-7 (the third camera model) does not appear to have these colour tingeing problems, and appears to have yet another algorithm as to the types of "black offset compensation" that have to do with causing it, which also seems to improve Horizontal and Vertical Pattern Noise (HPN/VPN). However, the K-7 has yet another problem related to this in the "thin green vertical line for high sensor temperatures" problem for some cameras, which again is likely due to a design bug/oversight in the black offset compensation or a related function (stitching of channels).

There is a common misunderstanding across the Internet forums as to what kind of magic can be accomplished through firmware updates, especially as related to low level data acquisition functions. The only way that the typical 0.1 GHz camera CPU's can manage to acquire these large MP images is with the help of the dedicated imaging engines, or co-processors if you will. Even when Pentax understand the problems, as I am sure they do, once the camera is in production there is very little they can change if it takes very much processing and there are very limited sorts of things that they can do with the hardwiring of the imaging engine. For instance, given that they can determine the appropriate reversing gradients to apply across the K10D sensor or use an algorithm similar to as I use in this program to compensate for the the vertical colour tingeing to the 14.6 MP K20D, the time required to actually apply the correction smudging all of those photosites is 10's of seconds using the CPU in firmware! This correction could be done very quickly using the imaging engine in the pass that writes the raw data out to SD flash memory and during the raw conversion to JPEG's but these algorithms are hardwired with the thinking that black level compensation has already been taken care of, so no further compensation is possible. So no, Pentax won't release a firmware fix for this. The "thin green line" problem with the K-7 could be fixed in a timely fashion as it doesn't require adjusting all of the MP and just a limited number of values, but the problem is likely more related to digging out how big a correction to apply as the "step" is buried in the background noise.

The common problem we see here with these three cameras is not so much related to quality control/assurance but rather having an independent design review process that identifies these types of problems earlier in the design cycle so they can be corrected. Within days of pre-production and/or production image samples being published, we usually identify these types of problems - but by then it is usually too late. Thus, while we can be assured that the successor to the K-7 won't have the "thin green line" problem, it very likely will use yet another sensor design, whether from Samsung or from someone else - perhaps Sony, and adaptations made to PRIME2/3 for that sensor may well introduce yet another type of problem.

Regards, GordonBGood

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow