5D4 Sensor Defect Discovered

[romanr74]

Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

I cannot quite agree with that statement. Isn't the question "when" it brakes and how hard you have to test?

 

[Orangutan]

Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

I cannot quite agree with that statement. Isn't the question "when" it brakes and how hard you have to test?

True. "When" is outside legitimate exposure settings, and "how hard" is before you push the LSBs of the pixel data into the mid-tones.

I'll add my agreement to what's been said before:it's OK to test-to-failure, but don't claim a defect exists if failure is well outside legitimate exposure and camera settings. It's OK to label such things as "limitations;" e.g., the low-ISO DR of Canon's best sensor is not as wide as that for Sony. That's a limitation, not a defect. The same holds for banding: if it occurs significantly beyond real-world, legitimate exposure, then it's not a defect.

 

[tron]

Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

I cannot quite agree with that statement. Isn't the question "when" it brakes and how hard you have to test?

True. "When" is outside legitimate exposure settings, and "how hard" is before you push the LSBs of the pixel data into the mid-tones.

I'll add my agreement to what's been said before:it's OK to test-to-failure, but don't claim a defect exists if failure is well outside legitimate exposure and camera settings. It's OK to label such things as "limitations;" e.g., the low-ISO DR of Canon's best sensor is not as wide as that for Sony. That's a limitation, not a defect. The same holds for banding: if it occurs significantly beyond real-world, legitimate exposure, then it's not a defect.

What you say is reasonable. However, I would like to point you to DPReview report on 5D4 who pushed it more and did not report any issues. Only random noise a little higher than sonikon which to tell the truth could be expected. So in practice they set a standard and a deviation from that could imply that a body is probably defective.

 

[Orangutan]

Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

I cannot quite agree with that statement. Isn't the question "when" it brakes and how hard you have to test?

True. "When" is outside legitimate exposure settings, and "how hard" is before you push the LSBs of the pixel data into the mid-tones.

I'll add my agreement to what's been said before:it's OK to test-to-failure, but don't claim a defect exists if failure is well outside legitimate exposure and camera settings. It's OK to label such things as "limitations;" e.g., the low-ISO DR of Canon's best sensor is not as wide as that for Sony. That's a limitation, not a defect. The same holds for banding: if it occurs significantly beyond real-world, legitimate exposure, then it's not a defect.

What you say is reasonable. However, I would like to point you to DPReview report on 5D4 who pushed it more and did not report any issues. Only random noise a little higher than sonikon which to tell the truth could be expected. So in practice they set a standard and a deviation from that could imply that a body is probably defective.

In that case, a person could claim their own copy of the 5D4 is defective, rather than that the 5D4 sensor, in general, is defective.

 

[tron]

Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

I cannot quite agree with that statement. Isn't the question "when" it brakes and how hard you have to test?

True. "When" is outside legitimate exposure settings, and "how hard" is before you push the LSBs of the pixel data into the mid-tones.

I'll add my agreement to what's been said before:it's OK to test-to-failure, but don't claim a defect exists if failure is well outside legitimate exposure and camera settings. It's OK to label such things as "limitations;" e.g., the low-ISO DR of Canon's best sensor is not as wide as that for Sony. That's a limitation, not a defect. The same holds for banding: if it occurs significantly beyond real-world, legitimate exposure, then it's not a defect.

What you say is reasonable. However, I would like to point you to DPReview report on 5D4 who pushed it more and did not report any issues. Only random noise a little higher than sonikon which to tell the truth could be expected. So in practice they set a standard and a deviation from that could imply that a body is probably defective.

In that case, a person could claim their own copy of the 5D4 is defective, rather than that the 5D4 sensor, in general, is defective.

On that I totally agree...

 

[Loswr]

Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

I cannot quite agree with that statement. Isn't the question "when" it brakes and how hard you have to test?

True. "When" is outside legitimate exposure settings, and "how hard" is before you push the LSBs of the pixel data into the mid-tones.

I'll add my agreement to what's been said before:it's OK to test-to-failure, but don't claim a defect exists if failure is well outside legitimate exposure and camera settings. It's OK to label such things as "limitations;" e.g., the low-ISO DR of Canon's best sensor is not as wide as that for Sony. That's a limitation, not a defect. The same holds for banding: if it occurs significantly beyond real-world, legitimate exposure, then it's not a defect.

What you say is reasonable. However, I would like to point you to DPReview report on 5D4 who pushed it more and did not report any issues. Only random noise a little higher than sonikon which to tell the truth could be expected. So in practice they set a standard and a deviation from that could imply that a body is probably defective.

Perhaps. But several of these reported 'issues' indicate the problem occurs when there's something bright on one side of the frame, and the remainder of the frame is dark. If the DPR scenes don't meet those criteria, then your conclusion may not apply.

 

[JukkaS]

Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

I cannot quite agree with that statement. Isn't the question "when" it brakes and how hard you have to test?

True. "When" is outside legitimate exposure settings, and "how hard" is before you push the LSBs of the pixel data into the mid-tones.

I'll add my agreement to what's been said before:it's OK to test-to-failure, but don't claim a defect exists if failure is well outside legitimate exposure and camera settings. It's OK to label such things as "limitations;" e.g., the low-ISO DR of Canon's best sensor is not as wide as that for Sony. That's a limitation, not a defect. The same holds for banding: if it occurs significantly beyond real-world, legitimate exposure, then it's not a defect.

What you say is reasonable. However, I would like to point you to DPReview report on 5D4 who pushed it more and did not report any issues. Only random noise a little higher than sonikon which to tell the truth could be expected. So in practice they set a standard and a deviation from that could imply that a body is probably defective.

Perhaps. But several of these reported 'issues' indicate the problem occurs when there's something bright on one side of the frame, and the remainder of the frame is dark. If the DPR scenes don't meet those criteria, then your conclusion may not apply.

The picture if the girl on last pages is from DPReview test. So their test body also shows the same phenomenon when push over the limits. Is it then a problem that should not occur, I guess that is the whole topic here.

 

[Mikehit]

The picture if the girl on last pages is from DPReview test. So their test body also shows the same phenomenon when push over the limits. Is it then a problem that should not occur, I guess that is the whole topic here.

It depends on how you see 'the limits'.
I may be mistaken here but from what I have seen, people see banding when having a high-contrast scene and try to make the deep shadows have as much detail as a moderately-well lit scene. I am not sure many people would try to do that, because you would be aiming for shadows to be shadows albeit with a bit of detail to keep the viewer's interest.

 

[TimoV]

In my case, there are visible stripes in the night sky by adjusting exposure +1 in Lightroom (same happen with other programs).

Today I get information from Canon (Finland and Sweden RCC)...

That the camera meets all the criteria and there is no any fault of the camera.
It is therefore a feature of the product and the restriction.

...they (Canon service in Finland) see the stripes in RAW file of my photo, but it seem to be ok for Canon.
https://drive.google.com/open?id=0B1kKzKzpj3PieXZzTWVVSDUtWGc
(link to jpg where exposure is adjusted +1)

For now, I choose to continue with my old good 6D, it's have a little more noise in the night photos but at least it don't have disturbing stripes.

 

[Orangutan]

In my case, there are visible stripes in the night sky by adjusting exposure +1 in Lightroom (same happen with other programs).

Please post a raw sample.

 

[JukkaS]

In my case, there are visible stripes in the night sky by adjusting exposure +1 in Lightroom (same happen with other programs).

Today I get information from Canon (Finland and Sweden RCC)...

That the camera meets all the criteria and there is no any fault of the camera.
It is therefore a feature of the product and the restriction.

...they (Canon service in Finland) see the stripes in RAW file of my photo, but it seem to be ok for Canon.
https://drive.google.com/open?id=0B1kKzKzpj3PieXZzTWVVSDUtWGc
(link to jpg where exposure is adjusted +1)

For now, I choose to continue with my old good 6D, it's have a little more noise in the night photos but at least it don't have disturbing stripes.

Sorry to hear that -this was a bit what we expected Timo. Not nice. Mostly the stripes are showing only in frames where the shadows are clipping, but these night shots (with clear sky) are an exception in my case as well as we have discussed plenty with Timo.

 

[TimoV]

RAW file https://drive.google.com/open?id=0B1kKzKzpj3PianlHLTczYTl1c2M

Date information may be wrong in that file becoz I succesfully messy date settings in the dark, cold and windy place near to Baltic Sea

 

[romanr74]

Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

I cannot quite agree with that statement. Isn't the question "when" it brakes and how hard you have to test?

True. "When" is outside legitimate exposure settings, and "how hard" is before you push the LSBs of the pixel data into the mid-tones.

I'll add my agreement to what's been said before:it's OK to test-to-failure, but don't claim a defect exists if failure is well outside legitimate exposure and camera settings. It's OK to label such things as "limitations;" e.g., the low-ISO DR of Canon's best sensor is not as wide as that for Sony. That's a limitation, not a defect. The same holds for banding: if it occurs significantly beyond real-world, legitimate exposure, then it's not a defect.

What you say is reasonable. However, I would like to point you to DPReview report on 5D4 who pushed it more and did not report any issues. Only random noise a little higher than sonikon which to tell the truth could be expected. So in practice they set a standard and a deviation from that could imply that a body is probably defective.

In that case, a person could claim their own copy of the 5D4 is defective, rather than that the 5D4 sensor, in general, is defective.

This is what OP did, isn't it? Maybe the "subject title" is too catchy...

 

[Orangutan]

RAW file https://drive.google.com/open?id=0B1kKzKzpj3PianlHLTczYTl1c2M

Date information may be wrong in that file becoz I succesfully messy date settings in the dark, cold and windy place near to Baltic Sea

On first look in LR it doesn't look bad to me, considering the dark areas have <2% exposure in 2 or more channels. To me this looks like a DR limitation: you've exposed for brighter elements, and the dark areas suffered. I'll have a closer look later.

 

[ZachOly]

RAW file https://drive.google.com/open?id=0B1kKzKzpj3PianlHLTczYTl1c2M

Date information may be wrong in that file becoz I succesfully messy date settings in the dark, cold and windy place near to Baltic Sea

I can see streaking/banding at as low as +1.5 exposure in Photoshop. Just look near the top left.

I can also see streaking at +1 exposure and +50 shadows.

Anyone who says this is acceptable is out to lunch. Sorry.

It's a defect, at least on this camera.

 

[Loswr]

I can see streaking/banding at as low as +1.5 exposure in Photoshop. Just look near the top left.

I can also see streaking at +1 exposure and +50 shadows.

Anyone who says this is acceptable is out to lunch. Sorry.

It's a defect, at least on this camera.

Wait, you mean the sensor can't capture information where none exists and the software can't fabricate clean signal from noise? Crap, I guess my 1D X is defective too. Bummer.

 

[ZachOly]

I can see streaking/banding at as low as +1.5 exposure in Photoshop. Just look near the top left.

I can also see streaking at +1 exposure and +50 shadows.

Anyone who says this is acceptable is out to lunch. Sorry.

It's a defect, at least on this camera.

Wait, you mean the sensor can't capture information where none exists and the software can't fabricate clean signal from noise? Crap, I guess my 1D X is defective too. Bummer.

Did you even open the file? There's almost nothing clipping.

And when pushed slightly (+1 exposure, +50 shadows), I'm seeing lots of horizontal lines right through the sky.

 

[TimoV]

I'm more than happy to accept if there are just noise appearing from non information (near black level) , but that extra streaks... My EOS 6D show up only more noise in same case and that noise is smooth and no streaks

 

[Orangutan]

I can see streaking/banding at as low as +1.5 exposure in Photoshop. Just look near the top left.

I can also see streaking at +1 exposure and +50 shadows.

Anyone who says this is acceptable is out to lunch. Sorry.

It's a defect, at least on this camera.

Wait, you mean the sensor can't capture information where none exists and the software can't fabricate clean signal from noise? Crap, I guess my 1D X is defective too. Bummer.

Did you even open the file? There's almost nothing clipping.

And when pushed slightly (+1 exposure, +50 shadows), I'm seeing lots of horizontal lines right through the sky.

Looking in the top-most band I'm able to see, it's 1.1% in two of the channels. How is this enough data to push exposure as you've done, without some kind of banding? I'd love to see an identical shot for comparison that does allow this.

Again, your test shot must show 1.1% in two channels in the dark sky, then push it.

 

[ZachOly]

I can crank the exposure on my 5D3 when shoot astro stuff. I get random color noise in the pitch black sky. Which is totally acceptable!

What I'm seeing on the 5D4 are several thick horizontal discoloured streaks.