5D4 Sensor Defect Discovered

[Alex_M]

Sorry, I could not respond earlier due to time difference between our countries.

the image softening in the bottom section of the photo is the result of new DXO "miniature" filter enabled by default - likely bug in the software. the feature is very new. anyhoo, I have switched it off and re-run export. see the image attached. looks reasonably good to me. )

... A person from this forum tested it and first he said it looked like it was better, but it was because of some default settings of DXO which softens the picture. When he disabled those the picture is exactly the same with the striking. At least in his picture with a night sky.
...

not sure what is that person from the forum is talking about but you can see the resulting image - the only thing that has been changed - "Miniature filter" switched off ( Tilt and Shift simulation).

I have very good news:

Stop the press​

Melbourne, Australia, 20.11.2016, 3:15pm. Canon Inc NYSE: CAJ value shoot through the roof following this announcement from Melbourne, Australia. buy!

Same image processed in DXO Optics Pro 11.3.0 at +3EV and +50 shadows, -100 highlights, standard (non-prime) rendering - comes out almost clear of banding!

i can see some barely visible faint meager streak to the right of the person's fist but that's nothing!!!

I start noticing banding more clearly if the image is pushed crazy +4EV, +100 shadows but it was unnecessary as even with such a serious level of underexposure and +3EV exposure compensation I had to pull back highlights by -100 to avoid clipping.

same image pushed +2EV and +50 shadows in LR resulted in some nasty banding as per original image posted.
so there we go.

You can push the same image 2EV further with DXO Optics Pro than with LR. what a crock

Photo taken from:

https://www.dpreview.com/samples/0574215952/canon-eos-5d-mark-iv-real-world-samples-gallery

Vignetting correction, + 1EV, + 100 shadow recovery.
Sorry, but this is not normal. My old 5dmk3 is way better.

I'm a little confused - is there two almost identical shots in that gallery of the girl sitting there that I am missing? Because it looks to me like areas that are in focus when I open this in ACR are -obliterated- in the DXO screenshot you posted. So did I miss another copy of this image taken at a wider aperture, or is DXO just doing heavy heavy noise reduction to render without the banding?

 

[PixelTrawler]

I will post some images later but I can see the issues clearly in mine. And it impacts the exact type of images I do.
I tend to do a lot of 2-5 min landscape exposures. I hoped with the upgrade from the mark 3 to 4 that I could lift the foreground detail more.

Even a lift of a digital gradient filter of 3 stops and no shadow push shows the horizontal purpleish streaks.

I dont see it if I do similar with a fast exposure, underexposed to the same degree and apply similar.

It gets worse the longer the exposure which seems to indicate its hardware/read noise.

Its really horrible banding thats impossible to get rid of.

Update - I need to do more testing - I had the shadow slider up when I saw the streaks at their worst - the rest of the above is true, it seems to get worse with exposure length. The streaks are strongest on the left and there is a bright / dark pattern to them

So far... with the Mark 3
-- A fast exposure has no issue.
-- A long exposure with underexposed foreground - pushed to 3 stops and 20% shadow - no issue
-- A long exposure with underexposed foreground - pushed to 3 stops and 50% shadow - horrible streaks appear on the lower left side is strongest and they fade across the frame (but this a pretty extreme boost but it was something I was hoping to be able to do with the Mark IV as a 10ND filter tends to really darken foreground rocks and beach etc)

However, similar photos on my Mark 3
-- A long exposure with underexposed foreground - pushed to 3 stops and 0% shadow - the whole frame is covered in vertical streaks
-- A long exposure with underexposed foreground - pushed to 3 stops and 20% shadow - the whole frame is covered in vertical streaks and its a total mess
-- A long exposure with underexposed foreground - pushed to 3 stops and 50% shadow - the whole frame is covered in vertical streaks and it looks like someone drew the image in pencil its basically just vertical streaks!

So the Mark IV is definitely better than the 3 but maybe thats just the limits. I still have the mark 3 and I will attempt to get a side by side shot comparison to really see...

But the killer question is whether the 5D3 files would be usable.
It is one thing to be 'better' but another to still be 'useable'

According to this PixelTrawlers experience 5D4 would be quite a lot better.

Its really hard to quantify. Unfortunately I dont have the exact same images to compare. But I have similar.
And I can only go on the two units I have. And my experience.

But... a similar ish image (ive a load of shoreline long exposures taken over the years) on the 4 can take a 3ev push and 50% shadows and the mark 3 degrades at 3ev and 0% shadow.

Its not in anyway scientific, its not the same images... but im getting a sense it is a decent bit better.

When you go over the limits of the 4, its banding you can do nothing with. You just cant exceed that level. Its impossible to deal with bar dial back.

But I got some really nice images this morning. Im really happy with the files im getting.

 

[tron]

I just used DXO prime 2016 and HQ(fast) noise reduction on my two 5D4 test shots and there was no difference from DPP and Adobe DNG. They had the same banding

 

[Alex_M]

at how many stops of exposure compensation and how much extra you had to push the shadows?

I just used DXO prime 2016 and HQ(fast) noise reduction on my two 5D4 test shots and there was no difference from DPP and Adobe DNG. They had the same banding

 

[tron]

at how many stops of exposure compensation and how much extra shadows you had to push?

I just used DXO prime 2016 and HQ(fast) noise reduction on my two 5D4 test shots and there was no difference from DPP and Adobe DNG. They had the same banding

The first was shot metered manual at about -4 EV. The second was shot at -3 2/3 EV. At -3EV it barely shows so this is the 5D4 limit for me. The only thing is that the target was dominating the frame so metering from a dark object at -3EV puts it in zone II (metering normally would put it in zone V so 3EV would put it in II)
That is fine (no banding) unless someone wants to uplift shadows a lot. Even so it should be OK.It will have to do for the next 4 or 5 years. There is always the possibility of a future ML support and HDR methods...

 

[Alex_M]

yup, +3EV and +50 shadows push produces reasonably clean push with DXO Optics Pro. My argument is that the same push with LR will be not as clean. That certainly depends on DXO settings used. I suggest having DXO Smart Lighting disabled for starters.

at how many stops of exposure compensation and how much extra shadows you had to push?

I just used DXO prime 2016 and HQ(fast) noise reduction on my two 5D4 test shots and there was no difference from DPP and Adobe DNG. They had the same banding

The first was shot metered manual at about -4 EV. The second was shot at -3 2/3 EV. At -3EV it barely shows so this is the 5D4 limit for me.

 

[jrista]

Sorry, I could not respond earlier due to time difference between our countries.

the image softening in the bottom section of the photo is the result of new DXO "miniature" filter enabled by default - likely bug in the software. the feature is very new. anyhoo, I have switched it off and re-run export. see the image attached. looks reasonably good to me. )

... A person from this forum tested it and first he said it looked like it was better, but it was because of some default settings of DXO which softens the picture. When he disabled those the picture is exactly the same with the striking. At least in his picture with a night sky.
...

not sure what is that person from the forum is talking about but you can see the resulting image - the only thing that has been changed - "Miniature filter" switched off ( Tilt and Shift simulation).

There is still obvious posterization and false colors in there. It seems like some of the banding was traded for another problem. Half dozen of one, six of the other.

 

[Alex_M]

I am not saying it is perfect. I am sane enough (hopefully) not to expect perfect IQ after insane exposure push in post.
To my eyes the photo proceeded in LR looked worse at the same level of processing. Personally, I see no value in heavy exposure compensation in post. Anything that requires over +1.5EV in post has low to no value to me. I am sorry for the confusion.

P.S. I had to also pull highlights back by crazy -100 to eliminate overexposure on parts of the person's face. It's a miracle that the image was not turned to pixel dust following all this post processing madness

There is still obvious posterization and false colors in there. It seems like some of the banding was traded for another problem. Half dozen of one, six of the other.

 

[Loswr]

I am not saying it is perfect. I am sane enough (hopefully) not to expect perfect IQ after insane exposure push in post.
To my eyes the photo proceeded in LR looked worse at the same level of processing. Personally, I see no value in heavy exposure compensation in post. Anything that requires over +1.5EV in post has low to no value to me. I am sorry for the confusion.

Please cease and desist with all this pragmatism and practicality. Can't you see that the inability to intentionally underexpose by 6 stops and push the exposure back up in post is just one more example of Canon's poor, sub-par, unacceptable IQ? How can you – or anyone – possibly be satisfied, much less happy, with this sensor tripe that Canon continues to serve up? How dare you even suggest that the 5DIV can produce usable images?!?!?!?!?!?!?

;D

 

[jrista]

I am not saying it is perfect. I am sane enough (hopefully) not to expect perfect IQ after insane exposure push in post.
To my eyes the photo proceeded in LR looked worse at the same level of processing. Personally, I see no value in heavy exposure compensation in post. Anything that requires over +1.5EV in post has low to no value to me. I am sorry for the confusion.

There is still obvious posterization and false colors in there. It seems like some of the banding was traded for another problem. Half dozen of one, six of the other.

And being able to recover shadows deeper than 2-3 stops is very important to others. Canon has certainly made strides, the 5D IV is without question better than the 5D III, but they still have a ways to go as well.

I just love how this forum constantly reduces everything to "If you have to push the shadows more than a stop, you don't know how to expose." What baloney. Exposure latitude expands the capabilities of the camera. It's a very simple equation. This forum is a singularly unique place in the universe where exposure latitude, and being able to make full use of every last scrap of it if you need to (and there absolutely ARE reasons to need more exposure latitude in the real world, by experienced and professional photographers), is shunned as something only idiots could possibly need.

: I'm so glad I don't spend more than a few minutes a month on this forum...

 

[Alex_M]

I am sorry for making you angry. I said "personally" that hopefully describes as in my own case and needs?
I take some care and _usually_ get my photos exposed more or less correctly. Therefor extreme shadow push is not on my agenda. I am not saying that wider DR would not be beneficial. It would of course. But at the same time for type of photography i am into, I can happily get away with what I have now available to me in my humble Canon 6D.

I am not saying it is perfect. I am sane enough (hopefully) not to expect perfect IQ after insane exposure push in post.
To my eyes the photo proceeded in LR looked worse at the same level of processing. Personally, I see no value in heavy exposure compensation in post. Anything that requires over +1.5EV in post has low to no value to me. I am sorry for the confusion.

There is still obvious posterization and false colors in there. It seems like some of the banding was traded for another problem. Half dozen of one, six of the other.

And being able to recover shadows deeper than 2-3 stops is very important to others. Canon has certainly made strides, the 5D IV is without question better than the 5D III, but they still have a ways to go as well.

I just love how this forum constantly reduces everything to "If you have to push the shadows more than a stop, you don't know how to expose." What baloney. Exposure latitude expands the capabilities of the camera. It's a very simple equation. This forum is a singularly unique place in the universe where exposure latitude, and being able to make full use of every last scrap of it if you need to (and there absolutely ARE reasons to need more exposure latitude in the real world, by experienced and professional photographers), is shunned as something only idiots could possibly need.

: I'm so glad I don't spend more than a few minutes a month on this forum...

 

[JukkaS]

I am not saying it is perfect. I am sane enough (hopefully) not to expect perfect IQ after insane exposure push in post.
To my eyes the photo proceeded in LR looked worse at the same level of processing. Personally, I see no value in heavy exposure compensation in post. Anything that requires over +1.5EV in post has low to no value to me. I am sorry for the confusion.

There is still obvious posterization and false colors in there. It seems like some of the banding was traded for another problem. Half dozen of one, six of the other.

And being able to recover shadows deeper than 2-3 stops is very important to others. Canon has certainly made strides, the 5D IV is without question better than the 5D III, but they still have a ways to go as well.

I just love how this forum constantly reduces everything to "If you have to push the shadows more than a stop, you don't know how to expose." What baloney. Exposure latitude expands the capabilities of the camera. It's a very simple equation. This forum is a singularly unique place in the universe where exposure latitude, and being able to make full use of every last scrap of it if you need to (and there absolutely ARE reasons to need more exposure latitude in the real world, by experienced and professional photographers), is shunned as something only idiots could possibly need.

: I'm so glad I don't spend more than a few minutes a month on this forum...

Seems to also be forgotten that originally this striking has been happening in clear night sky shots without any pushing. It just is not so colourful, but clearly seen in my own and other images (not all but many). As said then you need a good monitor with black levels. So the evolution to only talking about insanely pushing is out of context.

 

[tron]

I am not saying it is perfect. I am sane enough (hopefully) not to expect perfect IQ after insane exposure push in post.
To my eyes the photo proceeded in LR looked worse at the same level of processing. Personally, I see no value in heavy exposure compensation in post. Anything that requires over +1.5EV in post has low to no value to me. I am sorry for the confusion.

There is still obvious posterization and false colors in there. It seems like some of the banding was traded for another problem. Half dozen of one, six of the other.

And being able to recover shadows deeper than 2-3 stops is very important to others. Canon has certainly made strides, the 5D IV is without question better than the 5D III, but they still have a ways to go as well.

I just love how this forum constantly reduces everything to "If you have to push the shadows more than a stop, you don't know how to expose." What baloney. Exposure latitude expands the capabilities of the camera. It's a very simple equation. This forum is a singularly unique place in the universe where exposure latitude, and being able to make full use of every last scrap of it if you need to (and there absolutely ARE reasons to need more exposure latitude in the real world, by experienced and professional photographers), is shunned as something only idiots could possibly need.

: I'm so glad I don't spend more than a few minutes a month on this forum...

This is a summary of what has happened in this and other threads. The criticism comes mainly from non 5D4 owners. Using Canon since 1988 (EOS620) and owing lots of lenses I can be considered as a Canon fan. But this is different than assuming Canon cameras (hmmm make that mostly sensors) are near perfect as they are and they do not need improvement.

The main culprit of disappointment BELIEVE IT or not is ................. dpreview!

Yes you read correctly. In spite of accusing Canon 24 hours a day they wrote about 5D4 that it is a significant improvement which may be BUT in addition they included examples of underexposing and then pushing up to 6EV. It goes without saying that we should expect the same level of performance (at least at 5EV levels) - with the reported noise of course - and not some random color bands. It is only natural to assume then that our cameras are problematic and should be replaced!

 

[Loswr]

I just love how this forum constantly reduces everything to "If you have to push the shadows more than a stop, you don't know how to expose." What baloney.

As opposed to reducing everything to, "If you can't push the shadows at least four stops, the camera has poor, sub-par, unacceptable IQ." Which is baloney left out in the sun until the stench of it rotting induces nausea.

 

[wockawocka]

This thread is a great example of why Canon ignore a lot of what people say and do their own thing. Perhaps if we could come together and have a unified view on what really matters we'd all be better off.

Watching the 5D4 release and subsequent comments unfold has been hilarious with no real bearing on the art of photography.

 

[tron]

This thread is a great example of why Canon ignore a lot of what people say and do their own thing. Perhaps if we could come together and have a unified view on what really matters we'd all be better off.

Watching the 5D4 release and subsequent comments unfold has been hilarious with no real bearing on the art of photography.

Last time I checked this is not a forum about the " art of photography"...

Let me guess: You do not have a 5DMkIV...

 

[Alex_M]

I don't recall Canon claiming DR improvements for 5D IV over predecessor. Have I missed that announcement? where do I look for information regarding 5D IV capable of +5EV push?

I can see about +1EV improvement in DR over my old trusty Canon 6D.

I am happy to assist further to quantify DR improvements in more scientific way.

I need one of the forum members that owns 5D IV and X-Rite colorchecker Passport take a few photos for me. I will advise on the procedure over personal messages. I will run DR analysis on the resulting files and will provide forum with more or less accurate information on what we are looking at.

 

[Refurb7]

I shot a few dozen weddings with the worse-than-5D4 old and crusty 5D3 and 6D this year under all sorts of lighting conditions and never saw any banding. My biggest push was around +2EV. I must be doing something wrong.

 

[wockawocka]

This thread is a great example of why Canon ignore a lot of what people say and do their own thing. Perhaps if we could come together and have a unified view on what really matters we'd all be better off.

Watching the 5D4 release and subsequent comments unfold has been hilarious with no real bearing on the art of photography.

Last time I checked this is not a forum about the " art of photography"...

Let me guess: You do not have a 5DMkIV...

I've 3 x 5DSr's and 2 x 5D4's

This thread is essentially full of people arguing if when driving a car at 200mph, if it's normal behaviour for the pistons to fly out when you drop it into 2nd.

Not all sensors are made equal, like with cpu's. Testing something until it breaks, then claiming foul when it breaks is at best a stupid thing to be doing.

 

[Alex_M]

[sarc] +2EV is so last century. +5EV is all the rage for the year 2017. Do not risk your business over this issue - get yourself up to speed... now! [/sarc]

I shot a few dozen weddings with the worse-than-5D4 old and crusty 5D3 and 6D this year under all sorts of lighting conditions and never saw any banding. My biggest push was around +2EV. I must be doing something wrong.