5DMk2 issues acknowledged in Japanese web page

Started Dec 19, 2008 | Discussions thread
David J. Littleboy
Veteran MemberPosts: 3,295
Like?
What do you put there...
In reply to rhlpetrus, Dec 19, 2008

rhlpetrus wrote:

It looks to me that they occru in the transition from blown sensels
to non-blown ones, in the transition while reading and/or converting.
The problem is what do you put there. Interpolate nearby sensels,
but this is before demosaic'ing, befre pixels, it's in RAW, right?

My take is that they occur in situations that are so badly blown that it really doesn't matter. They're always associated with gross amounts of flare and blooming. So, IMHO, it really doesn't matter, as long as they become white and not black.

If correctly exposed pixels were losing it, that would be a seriously major disaster.

But then RAW won't be exactly RAW but a somewhat processed file
already.

Raw is already "somewhat processed", since CDS and other sensitivity variation corrections are made to the raw data.

And exactly how do you guarantee it's done only where problem would
have occurred and not in some other harsh transitions?

One of the notes here claims that they are "blacker than black", that is, at a level that's lower than the lowest value pixels ever take on. So the phenomenon is easily recognizable. (I think I saw it with my 5DII in one test shot, but I didn't save it.)

To me this is a complicated situation and Canon would do better by
just getting back to 1DsIII's unit and replace parts for everybody.
People do that in cars all the time, why not cameras?

I'd be cool with that too. I just think that interpolating out the freaked pixels ought to work fine. (I could be wrong, of course.)

I'd be interested in knowing what causes the problem. But I'm not surprised that problems occur in extreme overexposure. My bet is that if you took the same shots with film, the film images would show even more infelicities (flare and fogging adjacent areas of the film) than the digital ones.

I'm much more worried about pattern noise than black pixels. Fortunately, my 5D2 seems very clean at ISO 100: I can boost the exposure 4 stops, inspect at 300%, and not see patterns. Haven't tested ISO 200 yet, though. (One ISO 200 test file I downloaded had patterns visible with a 3-stop boost (at 300%). Since I would prefer to shoot at ISO 200, that was irritating.)

-- hide signature --

David J. Littleboy
Tokyo, Japan

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow