5d mark II strange black dots?

[jeev]

Has any one read this any comments or input from POTN memebrs having 5d2 in there hand?

http://photo.net/canon-eos-digital-camera-forum/00RewZ
http://forums.dpreview.com/forums/re...ssage=30222755

UPDATE:
Canon has resolved this issue with EOS 5D MarkII Firmware Update Version 1.0.7
LINK:
http://web.canon.jp/imaging/eosd/fir.../firmware.html

[aIpha]

Yea, I found those threads earlier today as well. Very odd behavior, would like to see other people with the 5d2 test this out.

[sboerup]

Wonder if its the DPP software trying to correct for lens abberrations or something. They have the new optics enhancements software or something in the new DPPs.

[gjl711]

Originally Posted by sboerup

Wonder if its the DPP software trying to correct for lens abberrations or something. They have the new optics enhancements software or something in the new DPPs.

If so you would think that the 50D would see the same issues as it contains the same features, no?

[Familiaphoto]

Interesting, guess we will have to see what others find as well.

[proxes]

I got my 5D2 yesterday and I just happened to take some shots of the cat and the christmas tree. All the ones of the tree with the lights blown out I don't see anything. The ones with the cat I used my 580ex and on a couple shots I could see the black on the right side of the catch lights in her eyes; it's in the JPG and converted DNG.

[Panopeeper]

These pixels are present in the raw data, i.e. the problem has nothing to do with raw conversion.

I made a first analysis and passed it to Canon.

See how this looks like in the non-demosaiced raw data: http://www.panopeeper.com/Demo/Canon...lackPixels.GIF

However, these "crazy" pixels do not look the same, when seen up-close.

[gjl711]

Originally Posted by Panopeeper

These pixels are present in the raw data, i.e. the problem has nothing to do with raw conversion.

I made a first analysis and passed it to Canon.

See how this looks like in the non-demosaiced raw data: http://www.panopeeper.com/Demo/Canon...lackPixels.GIF

However, these "crazy" pixels do not look the same, when seen up-close.

They look like dead pixels.

[FlyingPhotog]

Crazy Idea Warning: It alomst looks like "Per Pixel CA" or what we used to call in the old days of TV .. "BLOOM" The pickup tubes in some of the early color TV cameras would do this same thing right next to highlights.

[SwingBopper]

That is some seriously bad news for Canon if this problem is widespread. Signal or A/D Processor errors?
I'm holding off on buying one until I learn more.

[Panopeeper]

Originally Posted by gjl711

They look like dead pixels.

No, they don't. Looking them up-close makes the issue even more strange; however, one needs to know more of the CR2 raw data to understand that.

Beside, dead pixels are dead. These pixels look so only in that particular shot, because they follow other, saturated pixels.

[evilr00t]

@panopeeper:
What is the value of the darkest pixel in the "dot" region?
Near 1024, or near 0? (first one suggests a sensor readout (ie. "bloom") problem, second one suggests a on-camera sensor data processing error)

For the record, from the way it looks from Panopeeper's screenshot (it's not a single pixel, but rather a *group* of pixels), it's probably an actual sensor problem.

[Bill Boehme]

I looked at one of the full size images and had a couple thoughts that are just SWAG's, but since since the camera has has highlight automatic gain control to help reduce the problem with blown highlights, it might be related to a timing lag in the AGC circuit that causes a one or two pixel lag in AGC returning to normal. The other thought, which is probably less likely is that it could be a gain control problem in the noise reduction algorithm in which the map of fixed pattern noise is subtracted from image pixel values. In astrophotography, dark frame subtraction can produce black dots, but it is generally much more random appearing that what I saw in the sample image.

[Panopeeper]

Originally Posted by gjl711

What is the value of the darkest pixel in the "dot" region?
Near 1024, or near 0?

I did have a reason to say, that they look strange. Some are absolute zero or 1, while some others are in the region 300, 500, 700, etc, but anyway far under the black level.

I was wondering already earlier about the thousands of pixels far out of the customary range, for example 35000 pixels under 800.

Something stinks. IMO it is a product of the DIGIC, some programming error. I am confident that this is not a sensor issue.

[Panopeeper]

Originally Posted by bill boehme

The other thought, which is probably less likely is that it could be a gain control problem in the noise reduction algorithm in which the map of fixed pattern noise is subtracted from image pixel values.

Sorry Bill, but you are on the wrong page. There is no such thing as "map of fixed pattern noise is subtracted from image pixel values" on raw level with with CR2 images.