EOS 7D Mark II User experience Thread

700mm - f8 - 1/2000 - ISO640 - HH

shadow areas can be lifted quite nicely

I would like to see whether that's true in practice.

John

rnclark wrote:

Mike in O wrote:Roger, are you saying that Canon is purposely skewing the dxo numbers so that Nikon looks better?

Ha Ha!  Nikon, whether they realize it or not, is modifying the raw data that significantly makes their results look better from image metrics like those in dxo.  Now to be fair, their results are actually pretty impressive from a visual impact of the images out of the camera.  It seems to me if we had the same algorithms, we should be able to improve the canon results just as much.

Roger

Newer Nikon bodies like the D810, D750, and D5300 (all Sony Exmor sensors) no longer clip black to zero yet exhibit the same read noise performance as previous Nikon bodies which do clip. This implies DxO's measurements have been properly accounting for the clipping in their measurements of the Sony Exmor sensors on Nikon bodies.

crw816 wrote: It's great to be able to set SS and Aperture, set a Maximum ISO and then let the camera set ISO.

Totally agree, this is by far the most ueful shooting mode for me. A feature that is on Nikon bodies since at least the D300, and that Canon has been reluctant to add to their bodies (still not on the 6D for instance). Not sure how the maximum iso setting interferes, though. Do the image get underexposed beyond this limit ?

Corey, those crops do not look that sharp indeed, especially if you were using an L lens (?). I am wondering wether there is a bit of front-focusing on the swan ?

Vertigo wrote:

crw816 wrote: It's great to be able to set SS and Aperture, set a Maximum ISO and then let the camera set ISO.

Totally agree, this is by far the most ueful shooting mode for me. A feature that is on Nikon bodies since at least the D300, and that Canon has been reluctant to add to their bodies (still not on the 6D for instance). Not sure how the maximum iso setting interferes, though. Do the image get underexposed beyond this limit ?

Yes, that is correct.  But being in Manual Mode you naturally have very fast access to both SS and Ap, so with a quick flick you can adjust one or the other to get back into a close exposure range and still rely on the camera to meter and adjust ISO.  To change the exposure compensation I programmed the little lever (new button near the joystick) so that when engaged it activates the top wheel and makes the necessary adjustment.  I'm loving it more and more and really wish that canon would include the ability to do this in the next 5D3 firmware.

Vertigo wrote:Corey, those crops do not look that sharp indeed, especially if you were using an L lens (?). I am wondering wether there is a bit of front-focusing on the swan ?

They do look quite soft, but remember that this is a huge crop.  The cropped swan is only about 0.4 megapixels.  If this image had been made with a 1DX with the same lens it would only be 0.13 megapixels, or on the 5D3 it would be 0.18 megapixels. (estimating)  Hard to imagine it looking any better on either of those cameras if you took a massive crop like this.

The Jury is still out for me on this softness idea.  I'm shooting with a 300 f2.8is II and a 70-200 f2.8is II and would have expected better results than I am seeing.  My lenses are not calibrated to this body so I hope to take some time to do that with FoCal this weekend.

Also, after shooting FF for a couple of years I'm trying to improve my technique on shooting with a crop sensor.  I remember reading something from EJ about how he had to increase his shutter speed dramatically on the D800 to get sharp images and that it had to do with pixel size and density.  That would make the 7D2 equivalent to a 51.7 megapixel FF camera with this respect, so I imagine that technique is even more critical than the D800 to produce sharp images at the pixel level.

Corey, could you update your post with tech specs for these images?

You can also expect some atmospheric degradation in such crops.

I had a look back at my 7D images, of distant bird that need similar amount of cropping. With default LR import (sharpness gain 25), they look sharper. But when pushing back the slider to zero, they indeed look more or less like Corey's. My mistake, sorry Corey ;o)

Now, the soon to be announced (it is a rumor so pinch of salt but the image looks genuine!) 100-400 would make the 7DII a very appealing travel/lightweight option. I currently tend to ignore my 100-400 as it could not handle the resolution of the old 7D and on a FF camera as it is too short, mostly. Plus the 7DII gives the option of adding a TC for one hell of a range. If it is as good as the Nikon 80-400 it will be a winning combo for many people.

I hope Corey doesn't mind, but here's an image he posted on his flickr account yesterday...

https://www.flickr.com/photos/corey-hayes/15102275183/

Obviously a closeup, minor 25% crop. PP limited to 25% sharpening in LR w/no extra noise reduction
EXIF indicates 7DII 500mm EF f5.6 1/800 1000ISO

For ISO 1000, no extra noise reduction and limited sharpening it looks pretty good to me.

I was mentioning to Corey that I wonder if some Canon shooters that migrated from DX to FF over the past few years will have to tune-up their technique in order to get the best out of the 7DII -- cropping heavy with a dense sensor can be challenging.

Vertigo wrote:I had a look back at my 7D images, of distant bird that need similar amount of cropping. With default LR import (sharpness gain 25), they look sharper. But when pushing back the slider to zero, they indeed look more or less like Corey's. My mistake, sorry Corey ;o)

I'm ultimately going to make up my mind on this camera after Adobe starts supporting the RAW file.  I know what my 5D3 looks like when I convert a file to DNG and open in LR, but the DPP conversion to TIF first might not be a fair comparison.  

Gary Irwin wrote:I wonder if some Canon shooters that migrated from DX to FF over the past few years will have to tune-up their technique in order to get the best out of the 7DII

I had a somewhat similar thought : people using a 5D3 are probably now used to super-clean, super-sharp images at 100%. On a 7D or 7DII, images don't look so nice at 100%, but the higher pixel density should imply to look at them at about 60% for a fair comparison. 

I often use 50% (1:2) display on LR to evaluate the global quality of 7D images, rather than 100%.

horshack wrote:

rnclark wrote:

Mike in O wrote:Roger, are you saying that Canon is purposely skewing the dxo numbers so that Nikon looks better?

Ha Ha!  Nikon, whether they realize it or not, is modifying the raw data that significantly makes their results look better from image metrics like those in dxo.  Now to be fair, their results are actually pretty impressive from a visual impact of the images out of the camera.  It seems to me if we had the same algorithms, we should be able to improve the canon results just as much.

Roger

Newer Nikon bodies like the D810, D750, and D5300 (all Sony Exmor sensors) no longer clip black to zero yet exhibit the same read noise performance as previous Nikon bodies which do clip. This implies DxO's measurements have been properly accounting for the clipping in their measurements of the Sony Exmor sensors on Nikon bodies.

The D810 still clips.  But Nikon adds an offset after the clip.  Seems sneaky to me.  Examples at: http://blog.kasson.com/?p=6544#comment-184578
Note too the dropping of values, making gaps in the intensity range.

Roger

rnclark wrote:

horshack wrote:

rnclark wrote:

Mike in O wrote:Roger, are you saying that Canon is purposely skewing the dxo numbers so that Nikon looks better?

Ha Ha!  Nikon, whether they realize it or not, is modifying the raw data that significantly makes their results look better from image metrics like those in dxo.  Now to be fair, their results are actually pretty impressive from a visual impact of the images out of the camera.  It seems to me if we had the same algorithms, we should be able to improve the canon results just as much.

Roger

Newer Nikon bodies like the D810, D750, and D5300 (all Sony Exmor sensors) no longer clip black to zero yet exhibit the same read noise performance as previous Nikon bodies which do clip. This implies DxO's measurements have been properly accounting for the clipping in their measurements of the Sony Exmor sensors on Nikon bodies.

The D810 still clips.  But Nikon adds an offset after the clip.  Seems sneaky to me.  Examples at: http://blog.kasson.com/?p=6544#comment-184578
Note too the dropping of values, making gaps in the intensity range.

Roger

The link you provided is evidence of Nikon's WB preconditioning rather than clipping and rebiasing. Also, DxO measured the D750's DR at close to 14.5EV, which is actually better previous bodies that do actually clip such as the D800/E. The thesis that clipped blacks are the source of Exmor's extra dynamic range is often repeated on message boards yet has been disproved before by independent raw testers who know how to bias the levels to account for the clipping (as DxO's is doing as well). There have been inconsistencies in NEF ADUs due to poor processing decisions by Nikon, including white-balance preconditioning going back several generations of bodies. What some may attribute to sneaky is really just poor engineering. If you'd like some advice on how to properly analyze NEFs I would suggest PM'ing Marianne Oelund or Iliah Borg at their dpreview hangouts.

horshack wrote: Roger
The link you provided is evidence of Nikon's WB preconditioning rather than clipping and rebiasing. Also, DxO measured the D750's DR at close to 14.5EV, which is actually better previous bodies that do actually clip such as the D800/E. The thesis that clipped blacks are the source of Exmor's extra dynamic range is often repeated on message boards yet has been disproved before by independent raw testers who know how to bias the levels to account for the clipping (as DxO's is doing as well). There have been inconsistencies in NEF ADUs due to poor processing decisions by Nikon, including white-balance preconditioning going back several generations of bodies. What some may attribute to sneaky is really just poor engineering. If you'd like some advice on how to properly analyze NEFs I would suggest PM'ing Marianne Oelund or Iliah Borg at their dpreview hangouts.

It matters not what you call it, white balance preconditioning, or whatever.  When values are set to one value or clipped at zero, that biases noise statistics, making the noise appear lower and dynamic range appear higher.  It matters what is recorded in the raw file that is not really raw data, that then biases scores in testing if not compensated for.  Probably no camera is truly raw in their raw file, but it is a mater of degree.  I do know how to analyze raw data; I've been doing it for 40+ years, including evaluating NASA sensors.  I have my own custom code that I wrote.  See my web site.

Roger

rnclark wrote:

horshack wrote: Roger
The link you provided is evidence of Nikon's WB preconditioning rather than clipping and rebiasing. Also, DxO measured the D750's DR at close to 14.5EV, which is actually better previous bodies that do actually clip such as the D800/E. The thesis that clipped blacks are the source of Exmor's extra dynamic range is often repeated on message boards yet has been disproved before by independent raw testers who know how to bias the levels to account for the clipping (as DxO's is doing as well). There have been inconsistencies in NEF ADUs due to poor processing decisions by Nikon, including white-balance preconditioning going back several generations of bodies. What some may attribute to sneaky is really just poor engineering. If you'd like some advice on how to properly analyze NEFs I would suggest PM'ing Marianne Oelund or Iliah Borg at their dpreview hangouts.

It matters not what you call it, white balance preconditioning, or whatever.  When values are set to one value or clipped at zero, that biases noise statistics, making the noise appear lower and dynamic range appear higher.  It matters what is recorded in the raw file that is not really raw data, that then biases scores in testing if not compensated for.  Probably no camera is truly raw in their raw file, but it is a mater of degree.  I do know how to analyze raw data; I've been doing it for 40+ years, including evaluating NASA sensors.  I have my own custom code that I wrote.  See my web site.

Roger

The ADU gaps are the result of coarse digital scaling from Nikon's WB preconditioning rather than from black clipping. I'm familiar with your site - it's one of the seminal sources for sensor analysis and has been appreciated by many including myself. NEFs require special handling to analyze correctly and I offered those references to help you avoid some of the pitfalls others have fallen into analyzing them.

horshack wrote: The ADU gaps are the result of coarse digital scaling from Nikon's WB preconditioning rather than from black clipping. I'm familiar with your site - it's one of the seminal sources for sensor analysis and has been appreciated by many including myself. NEFs require special handling to analyze correctly and I offered those references to help you avoid some of the pitfalls others have fallen into analyzing them.

The gaps don't really concern my regarding noise statistics.  In the link I provided. the sharp peak on the left side of the histogram and no data below it is the problem that will bias statistics the most.  That is the evidence of data getting clipped to a constant value.

rnclark wrote:

horshack wrote: The ADU gaps are the result of coarse digital scaling from Nikon's WB preconditioning rather than from black clipping. I'm familiar with your site - it's one of the seminal sources for sensor analysis and has been appreciated by many including myself. NEFs require special handling to analyze correctly and I offered those references to help you avoid some of the pitfalls others have fallen into analyzing them.

The gaps don't really concern my regarding noise statistics.  In the link I provided. the sharp peak on the left side of the histogram and no data below it is the problem that will bias statistics the most.  That is the evidence of data getting clipped to a constant value.

The left peak in Jim's Rawdigger graphs don't start until ISO 1250 - the DR we've been discussing is at base and low ISO. As for the peaks starting at ISO 1250 those are likely the result of coarse digital scaling as well - Nikon switches from analog gain to digital multiplication at around ISO 1000 on their Exmor-based cameras.

This is fascinating and all, but way off topic. Perhaps a new thread discussing sensor physics would be more appropriate??

Vertigo wrote:

crw816 wrote: It's great to be able to set SS and Aperture, set a Maximum ISO and then let the camera set ISO.

Totally agree, this is by far the most ueful shooting mode for me. A feature that is on Nikon bodies since at least the D300, and that Canon has been reluctant to add to their bodies (still not on the 6D for instance). Not sure how the maximum iso setting interferes, though. Do the image get underexposed beyond this limit ?

Corey, those crops do not look that sharp indeed, especially if you were using an L lens (?). I am wondering wether there is a bit of front-focusing on the swan ?

My focus may still not be 100% and in the case of the swan I dont no if I picked the sharpest one, but in the case of the birds there was some haze from the water because they were so far away, looking at my 1dmkiv shots from the same location they look the same most of the time.