40D hot pixel - take it back?

Yes, I don't think we are...

Gabor, please correct me if I misunderstand here...

But I'm not talking about the star-like noise that creeps in for longer exposures. Fro that, yes, I would minus out the other images, etc. (Or just use NeatImage, or whatever.)

I'm talking about 1 (exactly 1) specific pixel that was bright red, no matter what exposure-- long, short, light subject, dark subject. Always exactly the same pixel.

As I understand it, Canon can service the camera and "map" these kinds of pixels so that they are not included in the image, but rather that 1 pixel hole is filled in by averaging its neighbor pixels.

The problem is how to get it to discover those pixels. From what I understand of this trick, it triggers a function in the camera to attempt to discover these pixels and internally remembers them so for future pictures it replaces their value with a neighbor-average value.

Since the fix, I've taken additional pics and closely examined that former dead pixel and its surroundings. Don't see any sort of skew, black-hole, or anything else that looks odd.

If you're talking about the same thing as me, can you elaborate on what you mean by the dark/black-hole areas?

Thanks,

~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

lol...

I don't think you understand the OP's problem or the fix he described.

I had the same issue and the fix did work. No pixels 'fell off' in the process.

I thought exactly the same thing.

When I first read the tip, I thought they were talking about cleaning the sensor-- after all, to activate this tip, you go into "sensor cleaning" mode...

But it's not actually cleaning the sensor. It's nothing to do with the new Rebel/40D vibrating-sensor-for-dust-removal feature (as I first thought).

From what I can tell . . .

It appears it's just some firmware feature of the camera that keeps track of a little map of pixels to ignore. It's a bad pixel. It's like computer operating systems keeping track of which hard-drive sectors to ignore that are bad.

Instead of taking the light-information from that pixel (which is always bad, like pure red in my case), it ignores the input of that one pixel. When it writes the image it replaces that pixel's color information with the average of the pixels immediately around it.

Going into sensor-cleaning mode where the mirror is normally pulled up so you can manually clean the sensor, but instead doing it with a lens attached apparently is Canon's little firmware trick for activating the camera to take a test image to find these bad pixels.

Like I said, the original thread where I read about this it didn't explain it too well, so I tried (with no luck) the sensor-vibrate-dirt-removal numerous times.

But after trying it this way, once, bright red pixel is gone.

Sure, I'm not getting any new pixel data from that pixel. It's just being ignored. But I don't care about 1 or 10 or 30 pixels missing information in the 10,000,000 or so others that do work. I just don't like this little bright red spot always in that one spot on my images.

My one concern about doing this was: what if it falsely assumes a bunch of other pixels are bad and starts ignoring them too? I mean, what if my 10mp cam drops to like 8 or 9mp because all these okay ones are being ignored.

But I figure since this is only changing internal memory mapping and nothing physical, worst case I could always send it to Canon some day and get it professionally remapped.

As it is, I can't see any loss of detail, any weird artifacts around where the red-pixel used to be, etc. It seems to have done exactly as advertised: mapped that one pixel to no longer be counted.

But again... if you know this works differently and can shed light on any problems this might present, I am all ears . I had planned on exchanging the camera before finding this fix, so I still can do that next week... Please elaborate if you think I'm missing something here. I will really appreciate the information.

Cheers,

~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

where this stays in the manual?

I don't find anything like that in the online manual (thanks to Jim, who photographed the entire book and created a PDF from the shots).

The sensor cleaning is described on Page 127, nothing like the procedure you described is there. Noise reduction is described on Page 156, the same as the procedure with the 20D.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

Note: all these are screen-caps of 200% zoom view in PS of the exact same area.

Here's a photo of the inside of the lens-cap, so the dead pixel is most obvious. (I think it might actually be 2 dead pixels, but anyway...)



Here's a normal/darkish exposure of a card I use for WB shots:



Here's the same card after the fix. (Note, wasn't trying to match exposure exactly...)



Here are two crops of the same rectangle after the fix with more normal subjects. Just a quick out the front window shot and part of a blanket.





--
~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

It's not in the manual, as far as I can tell.

I got this tip from an older online forum where the OP apparently got an inside tip from some Canon support guy who explained how they fix this problem when cameras are serviced.

It's not sensor cleaning.
It's not long-exposure noise removal.
Nothing to do with either of those.

I'm still confused about "hot" or "dead" or "warm" or "bright" pixels. This is a pixel that always shows Red. Not random pixels that get 'hot' during long exposures. I believe this would be properly called a "dead" pixel (?). Either way, it's nothing about fixing long-exposure noise or sensor cleaning.

Hope that helps.

I also just posted some sample pics of what I saw (before and after). Hope that helps too.

~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

Following is a simple demo showing the black hole?crater issue. I cropped a small part of the image, actually from a TV, not turned on, in a dark room. (It was a provocative test just to this purpose).

The exposures were 4 minutes long (all three).

First, the hot pixel in the normal shot:



then the image corrected with in-camer long exposure noise reduction; note the black hole in the center:



Then I made a dark shot (lens covered, viewfinder glued down), the same time:



This one can be adjusted (brightness, levels, curves, whatever) and substracted from the first one; the result is much better than the one with in-camera substraction.

PhotoZulu said:

Addendum:

Click to expand...

I have not seen the previous demo, when I was writing this one (I had to create and upload the crops). It is obvious, that these are different issues.

I think Eric had to fight dead pixels; that's an exposure independent problem.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

It's highly interesting. I saved the post you linked to.

Though there is still one thing I don't get: if this is a "feature" ofsome models (someone mentioned, that this did not work on the 20D), then why is this not published?

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

Ah, pictures worth 1k words, eh?

Yes, sorry if I threw the whole thing off by using the wrong terms. I wasn't sure about "dead" versus "hot" pixels.

Between your example & pics and mine, it seems clear we're talking about 2 very different phenomena.

That's comforting because I was worried the fix for my dead ( ) pixel was going to have some other long-term negative effect. But seems it'll be fine.

~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

The older thread about it asks the same question. Seems yet another thing Canon could publish and save some S&H on returns for servicing.

I was literally driving back across town to return my 40D because of this problem. Only reason I didn't was the store called me back saying they were mistaken about still having another in stock. That's the only reason I dug my Google-heels in a bit more to try to find a solution.

It's probably one of those things that's hard to do right, doesn't always work, and if people aren't careful could probably hurt something. Don't know.

But I'm just glad it worked.

~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

Do you see the effect of this procedure

1. on JPEG created by the camera,
2. in raw, and if so, with which raw processor?

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

If you use ACR, it will likely remove hot/dead pixels during conversion.

Noticed in Jpg only at first. When trying to isolate the problem did RAW+JPG output for comparisons. Dead pixel was in both.

(Used DPP for the Raw processing.)

After the fix, it's fixed in both.

I was wondering how it was fixing it. My guess is that it records a bad-pixel-map in the camera memory and applies the nearest-neighbor averaging while creating the "raw" (I guess it wouldn't be technically raw...) image.

If anyone has insight into how this is really working, I'd be interested. I'm always suspicious of things that seem to work like magic.

Also interesting: after the fix there is a different pixel showing up as pure Green. But it's about 1/5 as bright-- have to really pixel-peep to find it. So I'm not going to worry about it.

My guess (again) is that doing this recipe replaces the in-memory bad-pixel-map and this other pixel was probably mapped as bad before, but didn't make the threshold calculation this time around so got promoted back to "good". Total hunch/guesswork at this point, but it would explain everything I'm seeing.

I'm not going to worry about this other pixel-- only noticed it because I was pouring over every pixel after the fix to make sure it was really fixed. The one that got fixed was noticeable. This new one isn't.

~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

True that ACR (and other SW) likely would work-around the problem. That was brought up in other discussions on it too.

And I thought about just letting it go. Sort of the zen approach. Probably would have if it were an older camera that developed the problem. (Like, you know, a week old! ) If it had been my 3-year-old 20D I would've just let it slide. But since the camera was 0 days old at the time, I didn't think I should have to work around some issue even if minor.

Another factor for me is I do a real mix of lower-quality family snaps versus shots I really care about. So I only shoot probably 30% RAW.

And, oh yeah, the camera was brand-new!

Since I never had the problem on 5 other digi-cams, I wasn't too willing to just live with it. It was going to bug me...

~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

I am interested on which level the correction is applied. The question is, if raw processors have to figure it out from some data in the raw file (like the dust delete information), or if the raw data is already corrected.

If you have not deleted it yet, pls send me or upload somewhere a raw file with the visible red pixel, and one after the "secret procedure", I would like to analyze them (in separate emails pls because of mailbox limitation).

Thanks

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

I first read about this on this forum, and I believe the claim was that the 5D had this feature.

The idea was that by following that protocol, you induced the camera to perform a self-hot-pixel remapping of its own. That would be a very good feature to have.

The Oly DSLRs have had this for years. Any time you want to map out the stuck pixels in the camera, you just go into the menu and activate the feature.

The legend went that the 5D had this same capability but that it was an undocumented bonus feature. I don't have a 5D, so I was unable to test it.

Then later, I read that people were claiming that the 30D did it too, and that even the 20D did it.

I did have a 20D so I tried it. But when I did, it didn't do anything for it. But maybe it only works for grossly stuck pixels and not for just "leakers" that show up at longer exposure times.

It's possible that the 40D has this feature, but again, if this really is true, why in the world wouldn't Canon announce it to the world? It's a selling point. It's something (like the micro lens adjustment) that would make customers happy and potentially cause people to buy their cameras. So I just don't see why Canon would fail to trumpet this if it was real.

But I've read enough of people's accounts claiming success with it that I'm just not sure. As I've said before: I'd like for this to be true. But I'm still skeptical. If it's true, then it's a good thing for sure

--
Jim H.

First, I believe that PhotoZulu's assessment of what's supposedly going on is exactly correct.

For the Olys and other cameras that have built-in auto-stuck-pixel remapping, that's exactly how they do it.

And if, when you did your remapping procedure, that green pixel did not meet the threshold of detection for the internal program to catch it, then exactly as you suggest, it would not have been mapped out after you were done with that. As you said, it was "promoted" to being a good pixel at that point.

In the Olys that do this, the remapping is done at the RAW level. So even if you shoot in RAW, and use whatever RAW processor you prefer, the stuck pixels are gone!

I'd imagine that Canon would do it the same way (again, assuming they're really doing it). I just can't help being skeptical, but it sure sounds like it's working for PhotoZulu.

I do not believe that this does anything like the dust mapping thing which requires that the RAW converter be aware of it all. Instead, I believe the stuck pixels are being mapped out the same way they've always been mapped out at the factory.

When these sensor ICs are made, they have defects. There are almost certainly none that don't have at least a few stuck pixels. So in order to achieve a reasonable yield, they simply map the dead and stuck pixels out of the cameras at the factory. But this has always been believed to be something that only the factory or an authorized repair place could do (for the Canons). And really, that's just as silly as not having micro adjustment for lens AF.

I've always felt that Oly was smart to put this in their cameras. Maybe Canon is now doing it. But why don't they document it if it's true?

I sound like a broken record.

--
Jim H.

When I shoot long exposures with my 20D, I use ACR to process the RAWs because it automagically removes hot and stuck pixels. Why DPP can't do this, I don't understand. ACR also does a much more thorough job of analyzing the row black pixels and thus doesn't show nearly the "banding" (aka pattern noise) that DPP does. And that's just silly. Canon should have the edge, wouldn't you think?

But maybe the newer DPP versions do this better.

--
Jim H.

Looks like the raw ones I took were all late last night, which I've already deleted. (I didn't think to make sure I had a raw copy before trying out the fix.) After it looked like it was fixed, I re-took some in raw just to make sure it was "really" fixed and not some raw--> jpg conversion thing. But my before ones are gone.

As I mentioned, though, my post-fix ones have the less noticeable green dead pixel in a different location. I'll try the fix again and see if it changes this green one this time; making sure to shoot raw before/after ones.

Will update once done.

~~~~~~~~~~~~~~~
Eric M. Wilson
photozulu.smugmug.com

a JPEG version, so that one can exactly identify the location of that pixel?

--
Gabor

http://www.panopeeper.com/panorama/pano.htm