Web Analytics RiceHigh's Pentax Blog: Issue of Hot/Dead Pixels of the K20D

Friday, May 09, 2008

Issue of Hot/Dead Pixels of the K20D

Reports of the hot/dead pixels found in images produced by different K20D units continue and lately it has been reported by a DPR technical writer, who is responsible to review the K20D:-

http://forums.dpreview.com/forums/readflat.asp?forum=1036&thread=27852469

"During testing we have encountered the same problems with hot pixels that have been reported elsewhere. We have continued to experience the same difficulties with a second production body and are currently awaiting a response from Pentax before proceeding with further testing."

So, where's the "elsewhere" as mentioned? Here it is:-

http://www.optyczne.pl/69.1-Test_aparatu-Pentax_K20D_Wst%EAp.html

Since the above review is in Polish, someone had summarised some of the key points and posted it below:-

K20D not recommended, inconsistent hot pixels issue

And, some of the key statements are quoted:-

"main issue was inconsistent quality between K20D samples, which they consider to be 'unusual development for Pentax products, and sad to see '. They tried a few different K20Ds, all had hot pixels, but was was particularly worrying was that hot pixels changed in each sessions, disappearing and appearing in different places. This appears to indicate production issues with the latest 14.6 megapixels sensor (made by Samsung)."

"See the above URL and go to section 7 for samples of noise and hot pixels. The site also found that there are many hot pixels (more than in other equipment they tested), and that the hot pixels are permanently mapped and excluded in JPEGs during in camera processing, thus there is more and more excluded hot pixels due to their unpredictable and changing appearance. It was unclear is there a way of resetting this mapping, but this would be fixable (if Pentax decides to do so) in future firmware."

For those who has limited bandwidth in accessing the Internet, a direct link of a sample picture showing many random hot pixels is here:-

http://www.optyczne.pl/upload2/7758_k20d_hoty.png

"The test had many other criticisms and discovered inconsistent quality issues. The very first K20D they purchased produced loud focusing noises, exposed badly, and was assembled badly -- it was replaced by a local distributor, and many issues were not present in the replacement... but some came back in the third K20D they tested."

Well, is there any report from real K20D end-users (not reviewers) for the hot pixel issue then? Yes, I've seen two new K20D reported about the hot pixels found in pictures produced by their K20D in a thread opened one month ago:-

http://www.dchome.net/viewthread.php?tid=491778&extra=page%3D1

(Post in Traditional Chinese, use BabelFish is required)

Direct link to a sample "hot" picture can be seen here.

The thread open person also reported that he had tried the K20D "pixel re-mapping" function, which is essentially a method to rule out known hot or dead pixel(s) by disabling particular pixel with problem and instructing the camera to use the neighbourhood pixel values to calculate (aka interpolate) a new value to insert into that dead pixel. However, he wrote that the problem could not be resolved even using this available function. So, finally, as the local Pentax service centre was unable to resolve the problem, they sent his unit back to Pentax Japan. The user has finally got back his K20D unit and he has made a quick post in telling us that the problem "has been resolved", with a small downsized sample picture posted too. But however, I am not sure if he has thoroughly tested his unit or not as he has not reported back further in details since then, despite he promised that there would be a more detailed "final" report ...

Afterall, it seems to me that it is just a hardware bug of the Samsung new CMOS sensor (for which Pentax was naive enough to hope that the firmware could eliminate the problem all). It might be that the pixel density are just too high as too many pixels are packed in a too small area with a too large photo site per pixel and with too little isolation/separation between pixels too such that the whole sensor/image may become just too hot for the inside temperature when it is working. As such, the accumulated thermal energy, which possibly could not be drained away properly physically, has caused electrical signal noise got out of control somehow at certain areas and finally those physically heated up pixels are recorded in the final image as hot pixels or even dead ones (if it constantly appeared red or green, usually).

6 comments:

  1. Anonymous10/5/08 13:21

    What happened with PENTAX after it has been acquired by HOYA? Problem again.

    I don't think viewfinder tilted in K10D has not been resolved yet, now their new CMOS.

    PENTAX, you are going to killing me softly.

    ReplyDelete
  2. Anonymous10/5/08 22:03

    Hey, to fix one small point, DEAD pixels are "off" not "stuck in one color" as you imply.
    Again your lack of knowledge is showing...The below statement is incorrect
    has caused electrical signal noise got out of control somehow at certain areas and finally those physically heated up pixels are recorded in the final image as hot pixels or even dead ones (if it constantly appeared red or green, usually).

    ReplyDelete
  3. Anonymous11/5/08 00:49

    The "user" picture from the Chinese forum doesn't show any hot pixels, but a few stuck pixels in a row - but that's a good reason for replacement.

    Hot pixels are WHITE, and show up sprayed around an area, not concentrated...

    ReplyDelete
  4. Anonymous12/5/08 18:41

    :) I guess Pentax decided to used the same sample k20d units that where deployed in the desert of saudia not long ago for the press .. they said.. ok lets give the reviewer the same test models :)

    ReplyDelete
  5. Great work.

    ReplyDelete
  6. Guess I'll have to post here since you didn't update this blog w/ my testing results for this.
    Just FYI, I'm OCD about testing and pixel peeping but I tested my K20D when I first got it quite a bit (including testing for sensor/viewfinder tilt, low temperature banding testing, etc.).
    The hot pixels are really weird because they look like pixels with rings around them and there were quite a few that moved around. Higher ISOs made it worse, but long 30sec exposures had no problems (probably because dark frame subtraction removed them). I took probably 50 photos to test this.
    The kicker is that all this was fixed by running the built-in pixel mapping feature in the menu. I pixel peeped the next set of images all the way up to the useless ISO6400 and the hot pixels were gone.
    I've since put it through a photoshoot of over 100 images w/o problems and had to use shake reduction in low light for 1/4 of the shoot.
    So, yes, it's weird that they move around but the pixel mapping seems to fix it so that it's really a non-issue...

    ReplyDelete