5D4 Sensor Defect Discovered

[tron]

Hello.

What does a 100 increase in shadows equate to in Exposure push terms.

Just that 2 + 100 could equal +6 EV

Except that many times the problem shows with shadows at 0.

 

[colorblinded]

I just tested the "lens cap" shots from @PixelTrawler

The difference between FW 1.01 and 1.02 shows a massive improvement in streaking. FW 1.02 is producing random noise like you'd expect to see after a push.

That's encouraging to hear.

 

[tron]

Allow me to deviate from the thread topic a little. Attached is a screeshot from the latest DPP of a photo a 100% magnification. ISO was 8000. Not bad

 

[Alex_M]

Hallelujah! eventually we can put this Lens Cap Shooter's Diary saga away and return to our business of taking meaningful photos )
Thanks Canon for listening and standing by your products. I will be looking to acquire 5D Mark IV later in 2017 as soon as prices were settled a bit.

For now, I purchased second 6D body and that will do me in the interim.

I just tested the "lens cap" shots from @PixelTrawler

The difference between FW 1.01 and 1.02 shows a massive improvement in streaking. FW 1.02 is producing random noise like you'd expect to see after a push.

 

[AlanF]

Not trying to be a jerk, but after 34 pages of this, if someone claims their 5D4 doesn't have streaks I kind of want a .CR2 file to prove it.

I believe Alan (and I thank him too). He could ignore us but he contributed his findings. Don't forget LSXPhotog had mentioned his rental 5D4 didn't exhibit this behavior. Of course now I do not know what to do with my 5D4...

@AlanF: How many stops have you underexposed?

Thanks tron
What I did was to use flash in the dark at night focussed on a nearby plant in the garden. The rest of the photo appeared completely black. After the push, the background was a nice uniform green, as it should have been. I used iso640.

To do a test more akin to yours, I just now set to iso 100, photographed through a bright window, which occupied 1/3 rd of the frame and the other 2/3rd was a darkened wall. The "through the window shot" was in correct exposure at 1/13s and I so used 1/200s, which is a 4-stop underexposure. I then pulled +3ev and +5 shadows in DPP, and found no banding or artefacts. The same is true for pulling the 4ev under exposure up 4ev in DxO, PRIME or regular noise reduction. I added a further +100 shadows (probably 2ev, equivalent to a 6 ev push). There were no artefacts with the regular noise reduction but bands appeared with PRIME, presumably from the noise reduction algorithm. So, I think my sensor is perfectly satisfactory.

 

[PixelTrawler]

Hallelujah! eventually we can put this Lens Cap Shooter's Diary saga away and return to our business of taking meaningful photos )
Thanks Canon for listening and standing by your products. I will be looking to acquire 5D Mark IV later in 2017 as soon as prices were settled a bit.

For now, I purchased second 6D body and that will do me in the interim.

I just tested the "lens cap" shots from @PixelTrawler

The difference between FW 1.01 and 1.02 shows a massive improvement in streaking. FW 1.02 is producing random noise like you'd expect to see after a push.

Good news, Diary over.
My replacement 5d4 is better (did anyone else have dust on the viewfinder and sensor out of the box? Mine did and the replacement. I posted on a fb 5d4 group and one man said he is on his 4th replacement for same. The viewfinder dust is tiny this time - its under the focusing screen, and the sensor I can clean so its staying, but Canon need to clean their lines).

The streaks are more random and dont appear until a higher push whereas the first body had slightly wavy lines that were brighter and appeared more easily.

Ill post a before and after of a badly underexposed long exposure where I hit the trigger at 60s by accident instead of 180s. It was very impressive how much I can lift. I could use this image whereas I would have binned it on my 5d3

One other thing Ive discovered is the colour of the dark area has an impact. If its green, trees or grass, you can push like crazy and wont see it. If its brown/dark like sand/rocks you can push less. It seems the wavelength of the light has an impact on the data captured.

 

[JukkaS]

Not trying to be a jerk, but after 34 pages of this, if someone claims their 5D4 doesn't have streaks I kind of want a .CR2 file to prove it.

I believe Alan (and I thank him too). He could ignore us but he contributed his findings. Don't forget LSXPhotog had mentioned his rental 5D4 didn't exhibit this behavior. Of course now I do not know what to do with my 5D4...

@AlanF: How many stops have you underexposed?

Thanks tron
What I did was to use flash in the dark at night focussed on a nearby plant in the garden. The rest of the photo appeared completely black. After the push, the background was a nice uniform green, as it should have been. I used iso640.

To do a test more akin to yours, I just now set to iso 100, photographed through a bright window, which occupied 1/3 rd of the frame and the other 2/3rd was a darkened wall. The "through the window shot" was in correct exposure at 1/13s and I so used 1/200s, which is a 4-stop underexposure. I then pulled +3ev and +5 shadows in DPP, and found no banding or artefacts. The same is true for pulling the 4ev under exposure up 4ev in DxO, PRIME or regular noise reduction. I added a further +100 shadows (probably 2ev, equivalent to a 6 ev push). There were no artefacts with the regular noise reduction but bands appeared with PRIME, presumably from the noise reduction algorithm. So, I think my sensor is perfectly satisfactory.

Thanks for this analysis. I did the same kind if test at first (now 200 tests more) and wondered what all the fuzz was about. Most likely your shadows include anyhow too much detail to show the problem. Your body would be the first one which would clearly be free of this "feature". For some reason this is hard to believe. How badly are the shadows clipping in your latter test?

 

[happysnapper99]

Hallelujah! eventually we can put this Lens Cap Shooter's Diary saga away and return to our business of taking meaningful photos )
Thanks Canon for listening and standing by your products. I will be looking to acquire 5D Mark IV later in 2017 as soon as prices were settled a bit.

For now, I purchased second 6D body and that will do me in the interim.

I just tested the "lens cap" shots from @PixelTrawler

The difference between FW 1.01 and 1.02 shows a massive improvement in streaking. FW 1.02 is producing random noise like you'd expect to see after a push.

Good news, Diary over.
My replacement 5d4 is better (did anyone else have dust on the viewfinder and sensor out of the box? Mine did and the replacement. I posted on a fb 5d4 group and one man said he is on his 4th replacement for same. The viewfinder dust is tiny this time - its under the focusing screen, and the sensor I can clean so its staying, but Canon need to clean their lines).

The streaks are more random and dont appear until a higher push whereas the first body had slightly wavy lines that were brighter and appeared more easily.

Ill post a before and after of a badly underexposed long exposure where I hit the trigger at 60s by accident instead of 180s. It was very impressive how much I can lift. I could use this image whereas I would have binned it on my 5d3

One other thing Ive discovered is the colour of the dark area has an impact. If its green, trees or grass, you can push like crazy and wont see it. If its brown/dark like sand/rocks you can push less. It seems the wavelength of the light has an impact on the data captured.

I did! Mine was terrible in terms of dust. Had two big bits bottom left and right. Got my replacement on Thursday and it's a lot cleaner.

 

[tron]

Both my bodies (earlier and current) didn't have any dust...

 

[Loswr]

To do a test more akin to yours, I just now set to iso 100, photographed through a bright window, which occupied 1/3 rd of the frame and the other 2/3rd was a darkened wall.

Thanks for this analysis. I did the same kind if test at first (now 200 tests more) and wondered what all the fuzz was about. Most likely your shadows include anyhow too much detail to show the problem. Your body would be the first one which would clearly be free of this "feature". For some reason this is hard to believe. How badly are the shadows clipping in your latter test?

The first out of how many? The tens of thousands of 5DIVs already sold? Lol, right.

 

[JukkaS]

To do a test more akin to yours, I just now set to iso 100, photographed through a bright window, which occupied 1/3 rd of the frame and the other 2/3rd was a darkened wall.

Thanks for this analysis. I did the same kind if test at first (now 200 tests more) and wondered what all the fuzz was about. Most likely your shadows include anyhow too much detail to show the problem. Your body would be the first one which would clearly be free of this "feature". For some reason this is hard to believe. How badly are the shadows clipping in your latter test?

The first out of how many? The tens of thousands of 5DIVs already sold? Lol, right.

Well, here we have quite many users, I have checked now 4 bodies, there are many other forums as well. So far not a single body without this striking which reflects to the brightest parts. The problem is that people test with photos with lots of detail and it doesn´t show, others are not using proper equipment and playing wise by flicking from iPhones ;-).

 

[PixelTrawler]

Here is a 3.3ev lift. At this point there are faint lines visible in the lower left hand side.
But its a very big lift on a very dark underexposed image. I'm very happy with this example. Thats very impressive detail recovery on a scene of what I found so far to be the worst for showing any signs of banding.

This is with a 10ND B&W filter that murders detail in the foreground if the exposure is wrong.

Thats a total write off image thats almost usable. If I spent time on that image I could get something out of it.
I would bin a similar image on my 5D3.

I'm also pretty sure its better than my first 5D4, I was less impressed with that one. I'm convinced they are not all exactly equal.

If anyone wants that raw to screw around with PM me.

 

[Loswr]

To do a test more akin to yours, I just now set to iso 100, photographed through a bright window, which occupied 1/3 rd of the frame and the other 2/3rd was a darkened wall.

Thanks for this analysis. I did the same kind if test at first (now 200 tests more) and wondered what all the fuzz was about. Most likely your shadows include anyhow too much detail to show the problem. Your body would be the first one which would clearly be free of this "feature". For some reason this is hard to believe. How badly are the shadows clipping in your latter test?

The first out of how many? The tens of thousands of 5DIVs already sold? Lol, right.

Well, here we have quite many users, I have checked now 4 bodies, there are many other forums as well. So far not a single body without this striking which reflects to the brightest parts. The problem is that people test with photos with lots of detail and it doesn´t show, others are not using proper equipment and playing wise by flicking from iPhones ;-).

Four bodies, wow, that must be a huge fraction of the number of 5DIVs sold to date. Convenient that anyone who doesn't see the 'defect' is just doing it wrong.

Of course, it's quite possible that this 'defect' affects all 5DIVs because it's not a defect at all, but rather a simple case of pushing beyond limitations. Start your car, put it in neutral, then put a brick on the gas pedal and leave it for an hour or so. If it overheats and/or the engine seizes, is your car defective? Or is the real problem the person who put the brick on the gas pedal? Hint: it's the latter.

 

[JukkaS]

To do a test more akin to yours, I just now set to iso 100, photographed through a bright window, which occupied 1/3 rd of the frame and the other 2/3rd was a darkened wall.

Thanks for this analysis. I did the same kind if test at first (now 200 tests more) and wondered what all the fuzz was about. Most likely your shadows include anyhow too much detail to show the problem. Your body would be the first one which would clearly be free of this "feature". For some reason this is hard to believe. How badly are the shadows clipping in your latter test?

The first out of how many? The tens of thousands of 5DIVs already sold? Lol, right.

Well, here we have quite many users, I have checked now 4 bodies, there are many other forums as well. So far not a single body without this striking which reflects to the brightest parts. The problem is that people test with photos with lots of detail and it doesn´t show, others are not using proper equipment and playing wise by flicking from iPhones ;-).

Four bodies, wow, that must be a huge fraction of the number of 5DIVs sold to date. Convenient that anyone who doesn't see the 'defect' is just doing it wrong.

Of course, it's quite possible that this 'defect' affects all 5DIVs because it's not a defect at all, but rather a simple case of pushing beyond limitations. Start your car, put it in neutral, then put a brick on the gas pedal and leave it for an hour or so. If it overheats and/or the engine seizes, is your car defective? Or is the real problem the person who put the brick on the gas pedal? Hint: it's the latter.

These point gets us nowhere. The core questions is all the time, is this really something we have to live with even though it affects some quite normal pictures. If there really is working bunch of cameras, I want one or will this and swich camp. If you push Nikon 810 like this, you get only smooth noice in most of the cases, where this camera struggles. Go Nikon then? I might well do it.

 

[ZachOly]

member when everyone was saying "it's a hardware limitation, you're using it wrong" and then a firmware update improved the streaking?

Good times.

It'll be interesting to see if any improvements are made in future FW releases.

 

[JukkaS]

member when everyone was saying "it's a hardware limitation, you're using it wrong" and then a firmware update improved the streaking?

Good times.

It'll be interesting to see if any improvements are made in future FW releases.

Word!

 

[Loswr]

These point gets us nowhere. The core questions is all the time, is this really something we have to live with even though it affects some quite normal pictures. If there really is working bunch of cameras, I want one or will this and swich camp. If you push Nikon 810 like this, you get only smooth noice in most of the cases, where this camera struggles. Go Nikon then? I might well do it.

No, those points are directly relevant. If this was such a pervasive problem as you suggest, affecting every 5DIV and affecting normal shooting, there would be a far bigger outcry than a handful of people posting on Internet forums (and often the same people on multiple forums, so it's not even as many complainers as it appears).

DPReview is pretty thorough, including 6-stop underexposures pushed back up, and they have no love of Canon, so if they found this 'defect' they'd be shouting it from the virtual rooftops. (I know, you've got the answer for that, DPR is not doing their tests right.)

As for Nikon, if that system will meet your needs better, don't posture and threaten uselessly on the Internet, just switch. Given that the Sony sensors (in Nikon or Sony bodies) have had a big advantage over Canon sensors in terms of low ISO DR and hard shadow pushing for several years now, if smooth noise with hard shadow pushing is critical for you, then honestly you're a fool to be still using Canon bodies.

 

[JukkaS]

These point gets us nowhere. The core questions is all the time, is this really something we have to live with even though it affects some quite normal pictures. If there really is working bunch of cameras, I want one or will this and swich camp. If you push Nikon 810 like this, you get only smooth noice in most of the cases, where this camera struggles. Go Nikon then? I might well do it.

No, those points are directly relevant. If this was such a pervasive problem as you suggest, affecting every 5DIV and affecting normal shooting, there would be a far bigger outcry than a handful of people posting on Internet forums (and often the same people on multiple forums, so it's not even as many complainers as it appears).

DPReview is pretty thorough, including 6-stop underexposures pushed back up, and they have no love of Canon, so if they found this 'defect' they'd be shouting it from the virtual rooftops. (I know, you've got the answer for that, DPR is not doing their tests right.)

As for Nikon, if that system will meet your needs better, don't posture and threaten uselessly on the Internet, just switch. Given that the Sony sensors (in Nikon or Sony bodies) have had a big advantage over Canon sensors in terms of low ISO DR and hard shadow pushing for several years now, if smooth noise with hard shadow pushing is critical for you, then honestly you're a fool to be still using Canon bodies.

We are many fools here. I used to be a happy fool with my 6D. If the picture of the girl in the former pages exposed by the highlights (from DPReview) + Timos Shopping mall frame is everything but normal, I get your point. And if you are not able to see a problem in those frames, I recommend a visit to doctor. It is obvious that Canon does not admit this to be a problem, they just cannot. Let´s see how the upcoming firmwares will be, not to talk about the later bodies. I suppose, this problem will "magically" disappear, at least minimized.

 

[unfocused]

I have a new idea. Everyone who finds this "defect" should immediately return their cameras and ask for a new one. When that one has the same "defect" do it again and again and again. This will build up stock at the Canon refurbished store and enable more photographers who aren't as worried about lens cap photography to buy the 5D IV at a bargain price.

That should make many people happy.

 

[Alex_M]

sound like a plan! right.. where do we start? 8)

I have a new idea. Everyone who finds this "defect" should immediately return their cameras and ask for a new one. When that one has the same "defect" do it again and again and again. This will build up stock at the Canon refurbished store and enable more photographers who aren't as worried about lens cap photography to buy the 5D IV at a bargain price.

That should make many people happy.