Canon 7D Mark II - DXOMark Review

[Crapking]

The flickering phenomenon is still real / valid for many of us, and I can say that shooting the 7d II with anti-flicker feature has significantly eased my post-processing headaches in this gym; now if ACR would just come out, then the headache of DPP to TIFF to LR would also disappear....

 

[jrista]

Re: The 7D2 gets pooped on by DXO... to no one's surprise.

Just to put the record strait, for everyone reading Neuro's post, at least in regards to myself. I care about sensor IQ as a whole, not just low ISO, not just high ISO, not just one thing or another. I care about it overall, top to bottom, and I see sensor IQ as the one single area that Canon could, if they would invest the resources, realize very significant gains.

Significant gains would be nice, but it's not like they're way behind as many people (including DxO) thinks they are. Read what I wrote above.

http://www.canonrumors.com/forum/index.php?topic=23547.msg459899#msg459899

I did read what you wrote. I think it's a lot of personal opinion...subjectivity. It wouldn't be any different if I performed the same exercise, or anyone else performed it. We all see what we see, and we don't all see the same things. We all have different computers, different screens, calibrated differently, with different pixel densities, etc. etc.


There is a need for objective testing. I'm not a fan of DXO. I don't like their largely black box methodology in the way they report. I think their underlying data, however, which can be gleaned from sensorgen.info, is pretty sound (since it contains none of DXO's bias, just their measurements.)


I am very happy that DPR updated their studio sample comparison engine to support visual normalized results as well as full-size results. That allows the best and broadest range of comparisons.


Subjective opinions based on your own visual observations never go down well. I see what I see, and I've proclaimed what I see on these forums...which generally, on my screens calibrated the way they are in the lighting I have, Canon files show up quite blotchy with high color noise, while Exmor (and in most cases Toshiba) sensors have finer grained color noise and are not blotchy as Canon files. Clearly, on your screens the way you have them calibrated, you see something different. We could argue for days about who is right, and in the end, neither of us would be.


Objective measurements are important to having a reasonable basis for comparison. Algorithms can give us comparable results that exclude personally biased observation from the formula. Anyone who has done such objective measurements inevitably comes to the same conclusions: Canon sensors have fallen behind the competition, and are continuing to fall farther behind. It used to just be that at ISO 100, Exmor and Toshiba had about a two stop or greater advantage. Now, with the A7s, that advantage has been extended to the high ISO end (and not just high ISO, but ultra high ISO). Things keep progressing in competitor camps, with significant gains across the board...and, well, marginal progress at best on the Canon front...marginal at best.


If Canon could bring some of really modern innovations to bear, including things they have already patented (like power-disconnected reads, which could effectively eliminate dark current noise), I don't think it is unreasonable to think that 15 stops or more of real DR is possible. Canon could not just catch up with the competition, but actually leapfrog them, if they would invest some real resources into not just innovating in the sensor arena, but turning those innovations into actual consumer products.

 

[scyrene]

Re: The 7D2 gets pooped on by DXO... to no one's surprise.

I'm neither pro sony nor anti-canon or low iso dr fanatic. Of course that my needs are all that counts for my photography. Thats why I only talk about what I would like to have and not judge anybody else's preferences.

Your needs should be all that counts for your photography. But frequently on these forums, people seem to believe that Canon has a responsibility to meet their specific personal needs, and go on to claim that Canon is 'doomed' if their personal needs aren't met (presumably because they assume their personal needs represent those of the majority). Those who disagree, or point out Canon's many benefits and innovations (lenses, top AF, DPAF, anti-flicker, etc.) are accused of being fanboys trying to stifle innovation (innovation being narrowly defined by those people specifically as improved low ISO DR).

Just to set the record strait, here. For everyone reading this. (i.e. this is not just a direct reply to Neuro...it's informational...boggles my mind that I have to caveat like this, but hey, this community is just so wonderful about dissenting viewpoints.)


First, in my case, it's not just about low ISO DR. It's about improved sensor IQ across the board, increased DR across the board, low and high ISO (because it's already been done.) Increased color fidelity across the board, low and high. Increased resolution. Increased sharpness, yet not at the cost of false detail (which, IMO, is best achieved by pushing sensor resolution to the point where you generally oversample the lens, and are thus legitimately able to drop the AA filter entirely, and simply resport to downsampling to "sharpen.)


But that aside, Neuro's post here conveniently misses the point (that I have tried to make on many occasions) that Canon already excels in every area except sensors (as relative to current, modern technology.) The only area that Canon can realize significant improvements to overall output IQ is their sensor technology. Canon already has excellent, if not superior, AF technology. They are clearly distributing that technology, which premiered with the 1D X, to the rest of their cameras (or at least the rest of their pro-grade cameras, which is, IMO, good enough.) Canon already has superior glass (yes, I really do believe superior in most cases, although there are outliers that fall behind the competition, usually on the shorter end). Canon already has excellent ergonomics (won't say superior...it's a matter of taste/preference here.)


Canon currently excels in most areas...sensor IQ is the one area they do not excel. They haven't really excelled there for years, even back in the 5D II days, Canon's read noise was already a problem, and they had already reached the general limits of what their sensor technology could do...~11 stops (give or take) DR at best.


Just to put the record strait, for everyone reading Neuro's post, at least in regards to myself. I care about sensor IQ as a whole, not just low ISO, not just high ISO, not just one thing or another. I care about it overall, top to bottom, and I see sensor IQ as the one single area that Canon could, if they would invest the resources, realize very significant gains. (And maybe they have...there are rumors about a radical new layered sensor coming from Canon in 2015... Personally, given how Canon excels in every other area I care about, I truly hope the rumors are true, and that come 2015, Canon trounces the competition with something mindblowing. That's my hope. I'm still a skeptic. )

We've read your posts long enough to know you're not a one-issue guy, and that you care about many things - for good reason. Your contributions have been overwhelmingly positive, I think.

As for DR etc, I think the proof of the pudding would come with a camera especially aimed at landscapes/studio work, but we'll see. I'm still saving up for whatever comes

 

[jrista]

Re: The 7D2 gets pooped on by DXO... to no one's surprise.

We've read your posts long enough to know you're not a one-issue guy, and that you care about many things - for good reason. Your contributions have been overwhelmingly positive, I think.

Well, thanks.

As for DR etc, I think the proof of the pudding would come with a camera especially aimed at landscapes/studio work, but we'll see. I'm still saving up for whatever comes

I hope Canon does something amazing. I want to see photo quality that just blows my mind. I just hope it doesn't cost $8999. I could buy a LOT of astrophotography gear, even rather expensive CCD cameras and a bunch of really high quality filters, for that much money.

 

[scyrene]

Re: The 7D2 gets pooped on by DXO... to no one's surprise.

We've read your posts long enough to know you're not a one-issue guy, and that you care about many things - for good reason. Your contributions have been overwhelmingly positive, I think.

Well, thanks.

As for DR etc, I think the proof of the pudding would come with a camera especially aimed at landscapes/studio work, but we'll see. I'm still saving up for whatever comes

I hope Canon does something amazing. I want to see photo quality that just blows my mind. I just hope it doesn't cost $8999. I could buy a LOT of astrophotography gear, even rather expensive CCD cameras and a bunch of really high quality filters, for that much money.

I'm sure they'll set the price where they know it'll make them most money. They haven't stayed profitable this long without knowing how to do that - but personally I'd like everything to cost 1/10th what it does now

 

[AndreSilva]

There is some interesting analysis here, about the score and comments DxO mark gave to the 7d2:

https://www.youtube.com/watch?v=FTuBr0W0Zhw

 

[lo lite]

From Roger, the respected sensor reviewer of Clarkvision:

"A note on DXO numbers. It appears that DXO is not correcting Nikons truncating of the raw data, which artificially improves dynamic range by about a stop. Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good.

Previous to the 7D2 and 6D, pattern noise was a real limitation in Canon cameras (the 1D4 and 1DX are also pretty good, but not a good as the 7D2 and 6D). That pattern noise produced poor shadow areas compared to what could be extracted from Nikon sensors, especially at some ISOs (like 200 and 400 on many canon cameras). But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." -
http://www.naturescapes.net/forums/viewtopic.php?f=57&t=249565&start=112

True words. I remember having heard it somewhere that Nikon's RAW isn't actually RAW but somewhat preprocessed. I wonder if their "lead in DR" which is often claimed here, origins from this preprocessing.

And I have another link regarding the 7DII's sensor performance: http://www.clarkvision.com/reviews/evaluation-canon-7dii/index.html

 

[jrista]

From Roger, the respected sensor reviewer of Clarkvision:

"A note on DXO numbers. It appears that DXO is not correcting Nikons truncating of the raw data, which artificially improves dynamic range by about a stop. Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good.

Previous to the 7D2 and 6D, pattern noise was a real limitation in Canon cameras (the 1D4 and 1DX are also pretty good, but not a good as the 7D2 and 6D). That pattern noise produced poor shadow areas compared to what could be extracted from Nikon sensors, especially at some ISOs (like 200 and 400 on many canon cameras). But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." -
http://www.naturescapes.net/forums/viewtopic.php?f=57&t=249565&start=112

True words. I remember having heard it somewhere that Nikon's RAW isn't actually RAW but somewhat preprocessed. I wonder if their "lead in DR" which is often claimed here, origins from this preprocessing.

And I have another link regarding the 7DII's sensor performance: http://www.clarkvision.com/reviews/evaluation-canon-7dii/index.html

Hate to say it, but this is simply not true. Clipping to the black point in a Canon file would cut out a lot of data, and DR would not meaningfully improve (it might improve numerically, but there are caveats). Clipping marginally improved DR at best. A group of Nikon hackers actually removed the black point clipping in Exmor cameras a while back. In their testing, the difference in terms of noise with the bias offset restored was minimal...at most, ISO 100 RN was 6e-, and in some models lower than that. That is in comparison to the base RN of around 3e- in Exmor-based cameras at ISO 100. That resulted in less than a stop loss in DR overall, sometimes less than half a stop (depending on the camera).


That still means that a bias offset restored Exmor camera still has around 13 stops of DR, to a Canon cameras 11 stops. Operating in exactly the same way...with a bias offset rather than black point clipping. Why so little change? Because Canon cameras as much as 25-40 electrons worth of read noise at ISO 100...that is orders of magnitude greater than what Exmor has. You would have to "clip" a very significant amount of the signal in a Canon file...so while DR may improve, instead of discarding ~3e- worth of potentially usable data, you would be throwing away 22-37e- worth of potentially usable data. That's a LOT of data...that's a lot of potentially recoverable shadow data that is now simply GONE. Now...it's barely there to begin with in a Canon camera, because the read noise is so bad. So, maybe it doesn't matter so much that the information is clipped out.


However, you have to recognize what's actually happening to the signal. Canon's DR numbers may improve...but they are not improving because of a reduction in read noise (which is what happened with Exmor sensors), thereby increasing the amount of recoverable data. They are improving because of a removal of information, thereby DECREASING the amount of recoverable data.

 

[lo lite]

From Roger, the respected sensor reviewer of Clarkvision:

"A note on DXO numbers. It appears that DXO is not correcting Nikons truncating of the raw data, which artificially improves dynamic range by about a stop. Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good.

Previous to the 7D2 and 6D, pattern noise was a real limitation in Canon cameras (the 1D4 and 1DX are also pretty good, but not a good as the 7D2 and 6D). That pattern noise produced poor shadow areas compared to what could be extracted from Nikon sensors, especially at some ISOs (like 200 and 400 on many canon cameras). But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." -
http://www.naturescapes.net/forums/viewtopic.php?f=57&t=249565&start=112

True words. I remember having heard it somewhere that Nikon's RAW isn't actually RAW but somewhat preprocessed. I wonder if their "lead in DR" which is often claimed here, origins from this preprocessing.

And I have another link regarding the 7DII's sensor performance: http://www.clarkvision.com/reviews/evaluation-canon-7dii/index.html

Hate to say it, but this is simply not true. Clipping to the black point in a Canon file would cut out a lot of data, and DR would not meaningfully improve (it might improve numerically, but there are caveats). Clipping marginally improved DR at best. A group of Nikon hackers actually removed the black point clipping in Exmor cameras a while back. In their testing, the difference in terms of noise with the bias offset restored was minimal...at most, ISO 100 RN was 6e-, and in some models lower than that. That is in comparison to the base RN of around 3e- in Exmor-based cameras at ISO 100. That resulted in less than a stop loss in DR overall, sometimes less than half a stop (depending on the camera).


That still means that a bias offset restored Exmor camera still has around 13 stops of DR, to a Canon cameras 11 stops. Operating in exactly the same way...with a bias offset rather than black point clipping. Why so little change? Because Canon cameras as much as 25-40 electrons worth of read noise at ISO 100...that is orders of magnitude greater than what Exmor has. You would have to "clip" a very significant amount of the signal in a Canon file...so while DR may improve, instead of discarding ~3e- worth of potentially usable data, you would be throwing away 22-37e- worth of potentially usable data. That's a LOT of data...that's a lot of potentially recoverable shadow data that is now simply GONE. Now...it's barely there to begin with in a Canon camera, because the read noise is so bad. So, maybe it doesn't matter so much that the information is clipped out.


However, you have to recognize what's actually happening to the signal. Canon's DR numbers may improve...but they are not improving because of a reduction in read noise (which is what happened with Exmor sensors), thereby increasing the amount of recoverable data. They are improving because of a removal of information, thereby DECREASING the amount of recoverable data.

jrista, did you actually read what Roger N. Clark is saying in his writings? If not, here are the links again:

http://www.naturescapes.net/forums/viewtopic.php?f=57&t=249565&start=112

here Clark basically states that "Nikons truncating of the raw data, which artificially improves dynamic range by about a stop" which is basically consistent with what you're stating. But he doesn't stop there, he also states that Nikons RAW is not actually RAW data but has been tinkered with: "Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good. … But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." You completely ignored that fact.

You also never commented on http://www.clarkvision.com/reviews/evaluation-canon-7dii/index.html

And finally I have an astro image taken with the 7DII for you to show what is possible: http://www.clarkvision.com/galleries/gallery.astrophoto-1/web/horsehead.rclark.c11.22.2014.0J6A1680-1750-sugav70.f-bin4x4s.html

Whatever, Roger N. Clark is much more trustworthy to me (since he maintains an independent website with actual measurements taken in a competent way) than somebody, who believes that 21 steps of DR would be possible with an 16 bit digitizer in a public forum. http://www.canonrumors.com/forum/index.php?topic=23877.msg467462#msg467462

 

[dgatwood]

here Clark basically states that "Nikons truncating of the raw data, which artificially improves dynamic range by about a stop" which is basically consistent with what you're stating. But he doesn't stop there, he also states that Nikons RAW is not actually RAW data but has been tinkered with: "Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good. … But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." You completely ignored that fact.

I'd be curious to see an example of this "wormy" look. The D800 shots I've seen blown up look like random noise with a Gaussian distribution, much like my 6D's shots do. The only pics I've seen that look wormy have been low-quality JPEG copies of RAW files, with the usual artifacts around high frequency components.

Now the 5D Mark III... its shadow noise looks wormy, but that's because of banding.

Whatever, Roger N. Clark is much more trustworthy to me (since he maintains an independent website with actual measurements taken in a competent way) than somebody, who believes that 21 steps of DR would be possible with an 16 bit digitizer in a public forum. http://www.canonrumors.com/forum/index.php?topic=23877.msg467462#msg467462

The way I interpreted what he said was that they might do some interesting compression of the image by using a 16-bit converter nonlinearly, which is plausible, though it would be pretty bizarre.

 

[lo lite]

here Clark basically states that "Nikons truncating of the raw data, which artificially improves dynamic range by about a stop" which is basically consistent with what you're stating. But he doesn't stop there, he also states that Nikons RAW is not actually RAW data but has been tinkered with: "Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good. … But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." You completely ignored that fact.

I'd be curious to see an example of this "wormy" look. The D800 shots I've seen blown up look like random noise with a Gaussian distribution, much like my 6D's shots do.

The issue is discussed here: https://nikonhacker.com/viewtopic.php?t=85 I wonder that jrista is not aware of that "star eating quality" of the Nikon bodies given that he always claims to be such a great astro-photographer.

You can see the wormy look here (and understand why he called it "wormy"): http://www.astrosurf.com/buil/nikon_test/test.htm the site is in french but I assume you know how to use google translate.

 

[Lee Jay]

Re: The 7D2 gets pooped on by DXO... to no one's surprise.

Just to put the record strait, for everyone reading Neuro's post, at least in regards to myself. I care about sensor IQ as a whole, not just low ISO, not just high ISO, not just one thing or another. I care about it overall, top to bottom, and I see sensor IQ as the one single area that Canon could, if they would invest the resources, realize very significant gains.

Significant gains would be nice, but it's not like they're way behind as many people (including DxO) thinks they are. Read what I wrote above.

http://www.canonrumors.com/forum/index.php?topic=23547.msg459899#msg459899

I did read what you wrote. I think it's a lot of personal opinion...subjectivity. It wouldn't be any different if I performed the same exercise, or anyone else performed it. We all see what we see, and we don't all see the same things. We all have different computers, different screens, calibrated differently, with different pixel densities, etc. etc.


There is a need for objective testing. I'm not a fan of DXO. I don't like their largely black box methodology in the way they report. I think their underlying data, however, which can be gleaned from sensorgen.info, is pretty sound (since it contains none of DXO's bias, just their measurements.)


I am very happy that DPR updated their studio sample comparison engine to support visual normalized results as well as full-size results. That allows the best and broadest range of comparisons.


Subjective opinions based on your own visual observations never go down well. I see what I see, and I've proclaimed what I see on these forums...which generally, on my screens calibrated the way they are in the lighting I have, Canon files show up quite blotchy with high color noise, while Exmor (and in most cases Toshiba) sensors have finer grained color noise and are not blotchy as Canon files. Clearly, on your screens the way you have them calibrated, you see something different. We could argue for days about who is right, and in the end, neither of us would be.


Objective measurements are important to having a reasonable basis for comparison. Algorithms can give us comparable results that exclude personally biased observation from the formula. Anyone who has done such objective measurements inevitably comes to the same conclusions: Canon sensors have fallen behind the competition, and are continuing to fall farther behind. It used to just be that at ISO 100, Exmor and Toshiba had about a two stop or greater advantage. Now, with the A7s, that advantage has been extended to the high ISO end (and not just high ISO, but ultra high ISO). Things keep progressing in competitor camps, with significant gains across the board...and, well, marginal progress at best on the Canon front...marginal at best.


If Canon could bring some of really modern innovations to bear, including things they have already patented (like power-disconnected reads, which could effectively eliminate dark current noise), I don't think it is unreasonable to think that 15 stops or more of real DR is possible. Canon could not just catch up with the competition, but actually leapfrog them, if they would invest some real resources into not just innovating in the sensor arena, but turning those innovations into actual consumer products.

Here's the problem. I went looking through my shots (over a quarter million of them) for high DR environments. It turns out that it's exceptionally rare for me to need more than 9 stops of DR. The thing is, I need those 9 stops regardless of what ISO I'm using. In fact, I often need more DR at ISO 3200 than at ISO 100 because of the harshness of low artificial light.

It gets worse.

When I'm shooting at base ISO it's because shutter speed isn't limiting my shots. In those cases, it's usually no problem to shoot a quick +/-2 or 3 stop bracketed burst and, from those, create an 18 stop HDR. When I'm shooting at high ISO, it's because shutter speed does matter and an HDR is therefore not possible.

Conclusion: Canon's 10-12 stop base ISO DR is plenty for me. What I need more of is high ISO DR and few technologies available now seem to provide that.

 

[Lawliet]

You can see the wormy look here (and understand why he called it "wormy"): .

Once you understand that it refers to the D3/D300(still analog readout, pre on sensor ADC), it becomes obvious why it doesn't matter.

 

[Lee Jay]

Here's another problem. If you download the sensoegen CSV file and look at it closely, you'll discover than the DxO data isn't consistent with itself. QE can vary greatly with ISO acoording to the data when, in reality, it doesn't vary at all with ISO. It also appears that other data is inconsistent. How can we trust data that can't possibly be correct?

 

[scyrene]

here Clark basically states that "Nikons truncating of the raw data, which artificially improves dynamic range by about a stop" which is basically consistent with what you're stating. But he doesn't stop there, he also states that Nikons RAW is not actually RAW data but has been tinkered with: "Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good. … But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." You completely ignored that fact.

I'd be curious to see an example of this "wormy" look. The D800 shots I've seen blown up look like random noise with a Gaussian distribution, much like my 6D's shots do.

The issue is discussed here: https://nikonhacker.com/viewtopic.php?t=85 I wonder that jrista is not aware of that "star eating quality" of the Nikon bodies given that he always claims to be such a great astro-photographer.

You can see the wormy look here (and understand why he called it "wormy"): http://www.astrosurf.com/buil/nikon_test/test.htm the site is in french but I assume you know how to use google translate.

In fairness (and whatever the rights and wrongs here, I have no idea), I don't think jrista claims to be a great astrophotographer. He posts good astrophotographs and very detailed, considered posts on the topic.

 

[dgatwood]

You can see the wormy look here (and understand why he called it "wormy"): .

Once you understand that it refers to the D3/D300(still analog readout, pre on sensor ADC), it becomes obvious why it doesn't matter.

Pretty much. Also, what I see there looks an awful lot like 5D Mark III banding, not DNR.

 

[jrista]

From Roger, the respected sensor reviewer of Clarkvision:

"A note on DXO numbers. It appears that DXO is not correcting Nikons truncating of the raw data, which artificially improves dynamic range by about a stop. Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good.

Previous to the 7D2 and 6D, pattern noise was a real limitation in Canon cameras (the 1D4 and 1DX are also pretty good, but not a good as the 7D2 and 6D). That pattern noise produced poor shadow areas compared to what could be extracted from Nikon sensors, especially at some ISOs (like 200 and 400 on many canon cameras). But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." -
http://www.naturescapes.net/forums/viewtopic.php?f=57&t=249565&start=112

True words. I remember having heard it somewhere that Nikon's RAW isn't actually RAW but somewhat preprocessed. I wonder if their "lead in DR" which is often claimed here, origins from this preprocessing.

And I have another link regarding the 7DII's sensor performance: http://www.clarkvision.com/reviews/evaluation-canon-7dii/index.html

Hate to say it, but this is simply not true. Clipping to the black point in a Canon file would cut out a lot of data, and DR would not meaningfully improve (it might improve numerically, but there are caveats). Clipping marginally improved DR at best. A group of Nikon hackers actually removed the black point clipping in Exmor cameras a while back. In their testing, the difference in terms of noise with the bias offset restored was minimal...at most, ISO 100 RN was 6e-, and in some models lower than that. That is in comparison to the base RN of around 3e- in Exmor-based cameras at ISO 100. That resulted in less than a stop loss in DR overall, sometimes less than half a stop (depending on the camera).


That still means that a bias offset restored Exmor camera still has around 13 stops of DR, to a Canon cameras 11 stops. Operating in exactly the same way...with a bias offset rather than black point clipping. Why so little change? Because Canon cameras as much as 25-40 electrons worth of read noise at ISO 100...that is orders of magnitude greater than what Exmor has. You would have to "clip" a very significant amount of the signal in a Canon file...so while DR may improve, instead of discarding ~3e- worth of potentially usable data, you would be throwing away 22-37e- worth of potentially usable data. That's a LOT of data...that's a lot of potentially recoverable shadow data that is now simply GONE. Now...it's barely there to begin with in a Canon camera, because the read noise is so bad. So, maybe it doesn't matter so much that the information is clipped out.


However, you have to recognize what's actually happening to the signal. Canon's DR numbers may improve...but they are not improving because of a reduction in read noise (which is what happened with Exmor sensors), thereby increasing the amount of recoverable data. They are improving because of a removal of information, thereby DECREASING the amount of recoverable data.

jrista, did you actually read what Roger N. Clark is saying in his writings? If not, here are the links again:

http://www.naturescapes.net/forums/viewtopic.php?f=57&t=249565&start=112

here Clark basically states that "Nikons truncating of the raw data, which artificially improves dynamic range by about a stop" which is basically consistent with what you're stating. But he doesn't stop there, he also states that Nikons RAW is not actually RAW data but has been tinkered with: "Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good. … But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." You completely ignored that fact.

I read everything, and I never ignored anything. The simple fact of the matter is, Sony Exmor has far lower REAL read noise at ISO 100 than Canon sensors. Without clipping, it doesn't matter how you slice it...the Exmor is better. That's my point. Some 5.6e- RN Sony vs. 33e- or 38e- RN Canon...that's the difference here. Canon can clip all they want. All they are doing by clipping is discarding potentially useful signal. For the record, THE SAME GOES FOR NIKON! Why else do you think someone hacked Nikon cameras in order to remove the clipping and restore the bias offset? The difference is, Nikon threw away around 3e- worth of signal, whereas if Canon clipped, they would be throwing away 25e- or more worth of signal. That is a difference of 833% (25/3). That is very NON-trivial.

I am fully aware of what Roger Clark is saying. I am also aware that the improvement in DR comes at the cost of well capacity. You discard 33e- worth of charge, then your FWC drops as well. Lower FWC, then you counteract the gain in DR (not entirely, but you still counteract it). Your not really going to gain anything except a better number in the end...in other words, you wouldn't actually gain as much shadow pushing as an Exmor has...because the clipping is not what gives Exmor it's shadow pushing ability. Sony gained more DR in Exmor because they REDUCED read noise, vs. simply EXCLUDING read noise. One improves the signal, one cuts out signal. They didn't just take a pair of scissors and cut it out...they innovated new technology and actually reduced read noise. Actually, they significantly reduced read noise and dark current both. The dark current in Exmor is so low as to be effectively non-existent.

You also never commented on http://www.clarkvision.com/reviews/evaluation-canon-7dii/index.html

Was I supposed to? I've read the page...aside from improvements in banding, I don't believe that article tells us anything that any other review does not. Canon read noise in terms of overall levels, as a limiting factor, is still roughly the same. Not surprising, given that Canon's problem is most likely not their sensors, but their readout and ADC architecture. As I've said on countless occasions...until Canon increases the parallelism of their ADC, shortens the path length from pixel to ADC, and reduces ADC frequency...nothing is going to change on the Canon RN front.

And finally I have an astro image taken with the 7DII for you to show what is possible: http://www.clarkvision.com/galleries/gallery.astrophoto-1/web/horsehead.rclark.c11.22.2014.0J6A1680-1750-sugav70.f-bin4x4s.html

The astro image...well, it could be significantly better. I am honestly not sure of the quality of the sensor, as the processing is fairly poor (look at the horrible color gradients and transitions around flame...that's beginner work, and I know Roger has been doing AP for a LOT longer than I have...those kinds of mistakes shouldn't be something he's still doing...) Especially if the 7D II has better noise quality than my 5D III, which I've used to create these:






All of these images pushed my 5D III to the limits, with long exposures (particularly the elephant nebula one) at fast apertures...long exposures were necessary to get all the dim detail above the noise floor. Despite that, because of the high Canon noise levels (both the high RN, as well as very high dark current and dark current noise levels), I had to apply some very heavy NR on the Pleiades image...far more NR than I wanted to. In order to combat that, I am currently working on tripling the sub count. With a modded Nikon camera? I would never have had to do that...I've worked Nikon data on several occasions, provided by generous members on the CN forums. The difference in quality is night and day. Noise quality is PHENOMENAL on images from one particular member who uses a D5100...the data is incredibly responsive, like butter. I still have that data, I may be able to do a comparison.

Whatever, Roger N. Clark is much more trustworthy to me (since he maintains an independent website with actual measurements taken in a competent way) than somebody, who believes that 21 steps of DR would be possible with an 16 bit digitizer in a public forum. http://www.canonrumors.com/forum/index.php?topic=23877.msg467462#msg467462

I never said 21 stops of DR is possible with a 16-bit digitizer. I actually stated that it was because of the use of s-log for video that they were able to take advantage of the 21 stops of DR the sensor had to offer, but that in a stills camera they would want a 20-bit ADC (I say 20-bit, because it seems far more likely than a 21-bit ADC, and read noise would have to be LUDICROUSLY low to actually achieve 21 stops IRL.)

 

[jrista]

here Clark basically states that "Nikons truncating of the raw data, which artificially improves dynamic range by about a stop" which is basically consistent with what you're stating. But he doesn't stop there, he also states that Nikons RAW is not actually RAW data but has been tinkered with: "Also Nikon filters the raw data, improving noise and dynamic range further. I believe, based on some experiments, that if the canon data were treated similarly, it would result it numbers at least as good. … But Nikon's raw files look "wormy" in the shadow areas from the in camera filtering of the raw data. The Canon 7D2 raw data looks much more random, as it should be." You completely ignored that fact.

I'd be curious to see an example of this "wormy" look. The D800 shots I've seen blown up look like random noise with a Gaussian distribution, much like my 6D's shots do.

The issue is discussed here: https://nikonhacker.com/viewtopic.php?t=85 I wonder that jrista is not aware of that "star eating quality" of the Nikon bodies given that he always claims to be such a great astro-photographer.

You can see the wormy look here (and understand why he called it "wormy"): http://www.astrosurf.com/buil/nikon_test/test.htm the site is in french but I assume you know how to use google translate.

In fairness (and whatever the rights and wrongs here, I have no idea), I don't think jrista claims to be a great astrophotographer. He posts good astrophotographs and very detailed, considered posts on the topic.

It should be noted that all of the Nikon sensors that were tested to have that "wormy" look were Nikon sensors, not Sony sensors. Today, having used astro image data from Nikon cameras with Sony sensors myself, I can attest to the quality. Noise quality and noise levels are an order of magnitude better than Canon quality and noise levels...at least an order of magnitude. I have a post on CN somewhere where I worked on a guys image produced with a Nikon DSLR...I was amazed at how responsive the data was to edits, how little NR the data required, etc. It was truly phenomenal...something I yearn for.


I probably will not be getting a Nikon for AP...I have higher ambitions than that. (I do not call myself a great astrophotographer, however I do have aspirations to become one. ;P I am also very disappointed in Roger's astrophotography...for a guy who's been at it far longer than I have, he should really have far, far better processing skills... :-\ The harsh color shifts and poor quality gradients in his recent horse head nebula...he should be well beyond that now.) My next camera for AP will very likely be a QSI, either the QSI 683 or 690, with the WSG-8 FW and IGP option, along with some astrodon filters. Nikon cameras are good, but during the summer, there is no substitute for a thermally regulated CCD. (Plus, QSI puts more effort into ensuring a pure gaussian read noise characteristic than anyone other than FLI.)

 

[candc]

[/quote]

The astro image...well, it could be significantly better. I am honestly not sure of the quality of the sensor, as the processing is fairly poor (look at the horrible color gradients and transitions around flame...that's beginner work, and I know Roger has been doing AP for a LOT longer than I have...those kinds of mistakes shouldn't be something he's still doing...) Especially if the 7D II has better noise quality than my 5D III, which I've used to create these:






All of these images pushed my 5D III to the limits, with long exposures (particularly the elephant nebula one) at fast apertures...long exposures were necessary to get all the dim detail above the noise floor. Despite that, because of the high Canon noise levels (both the high RN, as well as very high dark current and dark current noise levels), I had to apply some very heavy NR on the Pleiades image...far more NR than I wanted to. In order to combat that, I am currently working on tripling the sub count. With a modded Nikon camera? I would never have had to do that...I've worked Nikon data on several occasions, provided by generous members on the CN forums. The difference in quality is night and day. Noise quality is PHENOMENAL on images from one particular member who uses a D5100...the data is incredibly responsive, like butter. I still have that data, I may be able to do a comparison.

[/quote]

all of your images on that site are fantastic. i especially like the "orion and running man nebula"

 

[jrista]

all of your images on that site are fantastic. i especially like the "orion and running man nebula"

Thanks. I see LOTS of problems with all of them myself. Part of that is skill...I haven't been doing this even one full year yet. Part of it is equipment. Canon sensors are NOI-SY!! I say that a lot, and it's not just me being anti-Canon...I experience and deal with Canon's high noise levels constantly. It is rather frustrating.


For my Pleiades image, I now have 208 subs, or 8h 40m. I am currently working on it again...even THAT much exposure time is not enough. Even after averaging that many subs (which is a 14.4x reduction in noise!), it still has far, far too much:

People seem to think I'm just full of shit now because I want more DR and demand Canon step up their game. Just for the record...I speak from lots of personal experience digging the faintest details possible out of the read noise from image stacks produced from high end CCD cameras, Nikon cameras and Canon cameras. I speak from a lot of experience, it isn't just me being anti-Canon or spouting bullshit. How much this matters ultimately depends on what you do and how you process...but whether it matters to you or not does not change the actual facts.