Your "Black Dots" article - question

[duranash]

In your article, you said "Needless to say, on a normal size print it's going to be quite hard to see "black dots", even if you are looking for them!"

That comment, combined with a similar discussion from Ken Rockwell gave me some comfort that I should keep my newly arrived 5DII and not return it.  But in reading the following thread from the Infamous, DPReview, now I'm a bit more concerned.  The poster attached a full resolution image...not a pixel peeped crop, and the dots are clearly visible.  Now I don't take too many shots of Christmas trees, but none-the-less.......  I just won second place in my photo club's annual competition with a shot of two bridges at night with lights reflecting in the river (film capture using Provia 100F) and I wonder if the 5DII could have handled the task.

http://forums.dpreview.com/forums/read.asp?forum=1032&message=30331125

Any thoughts on the image posted in the attached thread?  I plan to take a shot of my christmas tree this evening (lots of tiny white lights) and see what the image would look like at a print size of 16"x24" which is a size that I print too quite often with my Epson 3800.

Thanks for your thoughts

[Bob Atkins]

Far be it from me to agree with Ken Rockwell, who often seems to be deliberately "over the top" in what he writes in order to atrtract attention. However in this case he's not 100% wrong (nor is he 100% right).

It's an issue which could be a problem for some users, though I suspect unless the effect was pointed out to them, most users wouldn't ever notice it! It's also not clear if it happens with every camera, just some cameras, just some cameras at some settings etc.

I'm certain that Canon will address the issue, probably with a firmware update. Now there are those that say you can't fix a hardware problem with firmware, but that's not really true either. Most cameras map out bad pixels in firmware. All sensors have some bad pixels, but since they are interpolated in via firmware, users can't find them and so can't worry about them! It's also possible that the firmware controls some amplifier parameters. It can certainly control gain (since that's what sets the ISO). I don't know what other hardware features are under software control.

By now Canon should have figured out what the cause is and be working on a fix. Obviously they can't afford to wait long on this one given the volume of negative press (mostly on the dpreview forums, but other forums too). I suspect it may be due to pixel amplifier undershoot ("ringing") under some conditions. That could be cured by modifying the rise/fall time of the amplifier, but I don't know if that's directly under firmware control.

The firmware fix would be to look for pixels below the black level which occured just after after a saturated highlight set of pixels was read out of a row. Those pixels would then be replaced by ones with values interpolated from the surrounding pixels (just as dust and bad pixels can be mapped out). That really shoudn't be too difficult and it would be impossible to detect since it's typically only affectiing groups of a few pixels.

In your position I'd test out my new EOS 5D II (I can't afford one yet, so I haven't done this myself!) with Xmas tree lights or streetlights, or (even better) a shot that I'd actually want to take, such as a bridge at night, and see if the "black dot" issue bothered me. If it didn't, I'd keep the camera and wait for Canon to update the firmware (or provide some other solution). If I found that my particular camera showed the black dots all or most of the time, and they were noticable on prints, I might be tempted to think about returning the camera before my time to do so was up. If I had a 14 day return period, I might wait 10-12 days just to see what Canon do, and if they haven't done anything by then at least I'd have the option of still returning the camera.

I'd be interested to hear what you find. At some point I should be getting my hands on a 5D MkII to test, but it doesn't look like that will be happening anytime soon, and I suspect the issue will have been addressed by the time it does happen.

[duranash]

I just took several shots of our Christmas tree in the living room which is decorated with the small white bulbs.  All large JPG's.  On one of the very underexposed shots I was only able to definitely identify one of the Black Dot like I'm seeing on DPReview.  Literally one instance on the whole frame - looking at 100% and carefully panning across the image.  This was at f16 and 1600 ISO.  I took several other shots at varying ISO's, but at wider f-stops.  Those images were exposed better in that you could see the ornaments, etc.---no Black Dots that I can see.  Since I'm shooting in my living room, the lights are pretty big in the image, but there are lots of very tiny reflections too.  I think I'll try some outdoor shots tomorrow night - its only 3 degrees out there tonight...brrrr.

[duranash]

........but when I took some outdoor shots where the points of light are small, I get exactly the same effect as many others are reporting.  Very noticable at 100%.  I sized one of the images to 16x22 and printed it.  I will concede that most people would never notice the black dots, but they are definitely there on the print.  Of course, once you see them, then they stick out like a sore thumb!
I also tried taking a couple of shots with backlight on snow...which gives you lots of bright points of light reflecting from the granules of ice.  The effect is noticable there also, but really slight.  One has to really look in the slightly shaded areas where the contrast between the bright reflections and background are greatest.  Not worth worrying about in the snow shots.

[KeithB]

The online photographer saw it during his evaluation of the A900, the D700 and the 5DII.

http://theonlinephotographer.typepad.com/the_online_photographer/2008/12/sony-vs-nikon-v.html

If you are right, Bob and it is overshoot, then it is caused by the engineers reading the chip too fast.  It might be able to be fixed in firmware if that kind of thing is programmable.

[Bob Atkins]

If it is amplifier overshhot (or in this case "undershoot"), it's related to the damping factor of the amplifier, which is in turn related to readout speed. If you want to read the signal out fast, you need a fast response time, so the ouput signal rise and fall time must be short, and the shorter you make it, the more likely it is to have overshoot. The amplifier rise/fall time is usually determined by hardware, whereas the readout speed could well be under software control. They're reading out a lot of data pretty fast so they may not be able to slow the readout process without slowing down the camara.

Removing the black dots by software may not be too hard. It's just like removing dust. If you can come up with some way of identifying them automatically, it would be pretty easy to clone them out automatically too.

[duranash]

I ran across a Photoshop action that seems to do a fairly decent job of removing the black dots.

I guess my greatest concern is that Canon's "fix" will in turn screw up something else and/or that they really need a hardware change and will make said hardware change silently in later versions of the camera....and those of us dumb enough to buy early might just simply get stuck.

I have definitely, without doubt, and absolutely decided to keep the camera......or return it!

[duranash]

.....so I DID absolutely decide to keep the camera.

Bob, I haven't yet bothered to load firmware v1.07 and I'll be anxious to hear your opinion.  As I mentioned before, my greatest concern is that the "fix" might mess up something else.  Most of what I'm reading on other Forums indicate the fix works well though.

[Bob Atkins]

I haven't heard any problems with the v1.0.7 firmware. Everyone seems to think it works great.

I'm looking at a 5D mkII right now, but I've had a lot of trouble getting it to show the black dots. I can just about see them in JPEGs (though they aren't very black), but I can't find them at all in the RAW files. Maybe I just got one that doesn't show the problem much (others have also reported difficulty reproducing the problem), or maye I just haven't found the right conditions yet.

If I had a 5D MkII which showed black dots, I don't think I'd have any hesitation in doing the firmware update though.