Yet another ?pointless Canon vs Nikon debate

Having spent a lifetime in the field of scientific verification & validation I want to take a different approach. If there is inherent differences in the proprietary image &/or workflow then we need to eliminate the operator as a variable. Next time you are out together swap your gear and take a series of images that challenges the sensor in terms of dynamic range and color. Send them (make sure they are blinded) to a third party and let them apply consistent development parameters according to file type and then send them back to both parties (again blinded). Rank them according to your own criteria and then you may have something. Exciting stuff! No wonder I try to get away from work to photograph as often as possible.

Cheers,

Jim

Perhaps I can add to some of the things said above. As many know, I do sensor evaluations on the raw data before any raw conversion so I can see real performance. I have done mostly Canon sensors, and would love to do more Nikon and other cameras if people would take the data or lend a camera. I am almost complete with a D800 analysis and should have it posted in a couple of weeks; just need to finish the text and implications.

EJ is correct in that the Nikon system does show higher dynamic range at low ISO. Bur Karl is also correct that Nikon truncates the lows. On the D800 at ISO 100 some 86 to 87% of the pixels are truncated at the low end. For those analyses that do not track this problem correctly, that gives an erroneous higher dynamic range and lower shadow noise. When I correct for this issue, I measure a dynamic range for the D800 of 13.2 stops per pixel. For comparison, the Canon 1DX is 11.4 stops and the 6D 11.5 stops.

But in my opinion what impacts image quality of Canon dslrs is the pattern noise at low ISOs. It is the banding pattern noise that limits extracting shadow detail and visually the most annoying (compared to random noise). The Nikons also have pattern noise, and while it is a little better controlled, a lot of the problem is simply cut off at the low end. Nikon also appears to be cooking the raw data, as there are some intensity levels missing, and the noise in the raw files at low light levels seem to have a pattern to them that is not random, but is not as bad as the banding pattern noise in the Canon cameras.

At high ISO, e.g. 1600, the newer Canon dslrs are producing a better image in my opinion. The Nikon d800 for example, cuts off about 35% of the low end pixels at high ISOs, and is showing more pattern noise than cameras like the Canon 6D.

What I have yet to verify is if the Nikons still run a median filter on the raw data (from the odd patterns I suspect yes on the D800). If so, then some small details will be lost, e.g. stars in a nightscape. I'll be out with a friend who has a D800 to test this--I did it once, but my 50 mm lens on my Canon failed and didn't stop down the aperture, so I could not do a detailed comparison and the apertures were not the same on the two cameras. I will try again soon.

Because the Canon images get better in terms of pattern noise and apparent read noise as one increases ISO, the problem with Canon sensors is not (or mostly not) at the sensor level; it is downstream electronics. Scientific sensors using similar readout mechanisms as Canon, have had higher dynamic range for decades, but they use slow readout and 16-bit A/D converters (e.g. take a minute to read out a sensor). So there is nothing inherently bad/old with the Canon sensors. Indeed, their high ISO low light performance is superb; they just need to get better control of the low ISO pattern noise.

Bottom line in my opinion, the lens is the key to imaging. Concentrate on better lenses regardless of manufacturer, and concentrate on making images in great light. After all, we used to make great images on slide film with only 5 stops or so of dynamic range, we ought to be able to do as well today with 11+ stops.

I bet there is more variability in raw converter algorithms and how the image is rendered than in the sensor differences. I have observed huge differences in raw converters on the same image, so it wouldn't surprise me if the same scene made with a canon and a nikon would produce different raw conversions and thus different images.

For examples of the pattern noise I am taking about, see these two reviews:
Canon 5D mark II: band pattern noise even at high isos:
http://www.clarkvision.com/articles/eva ... anon-5dii/

Canon 6D: virtually no pattern noise at ISO 1600 and higher:
http://www.clarkvision.com/articles/eva ... -canon-6d/

Other sensors I have analyzed:
http://www.clarkvision.com/articles/ind ... r_analysis

Roger

hullyjr wrote:Having spent a lifetime in the field of scientific verification & validation I want to take a different approach.  If there is inherent differences in the proprietary image &/or workflow then we need to eliminate the operator as a variable.  ..........................
Cheers,

Jim

Thank you Jim.

Being a physician for over 35 yrs I am very familiar with the scientific method -  double blind, placebo-controlled etc etc. 

Don't you think the same  person using both systems in similar conditions and using the same processing techniques comes pretty close to 'eliminating the operator'?

Which is why I am waiting to hear from those who have used both Canon and Nikon systems. So far, only one such person has responded. Sorry but the majority of the other replies have not been helpful.

Regards,

Pradeep

Roger, thanks so much for the detailed response. I agree completely that overall it is the lens that is so much more important with the modern sensors. The new Nikon 80-400 is so good (from what I've seen) that it pretty much eliminates most other long lenses for wildlife work. That is a huge thing considering how difficult it is to carry and hand-hold the bigger zooms. Just add a good 24-70 to another body and you're covered.

Thanks again, very helpful to know what's going on in the complex world of digital sensors.

Pradeep

Primus wrote:The new Nikon 80-400 is so good (from what I've seen) that it pretty much eliminates most other long lenses for wildlife work.

I find this very hard to believe.  Besides image sharpness, contrast, and bokeh, unless one wants to only photograph static subjects, AF speed is critical for getting action.  I bet the zoom can't come close to matching a prime lens like a 300 f/2.8 for AF speed, and I would bet the same in the other parameters too.

Roger

Karl Günter Wünsch wrote:

E.J. Peiker wrote:Most Nikons have about a two stop dynamic range advantage

False measurements that lead some to believe in this nonsense. The Nikon cameras still clip black values and that screws up the metric from DxO (they really should be a little scientific and investigate their claims when they turn out to be physically impossible). For the Sony sensors it's even worse, the recent crop of their sensors only write a heavily compressed RAW file - a lossy compression I may add, that incurs heavy losses of precision (and thus again lead to false signal to noise ratios, over-inflating the "measured" DR).
But let's assume (for the sake of the argument) that this is the reason for the differences the two photographers here are seeing, why then does the Nikon camera get described as more clean  and sharp - when a larger captured dynamic range would yield the exact opposite as a two stop larger dynamic range would yield more lack luster and less contrasty (and thus less sharp looking) images - unless the photographer has the balls to toss the excess dynamic range where it belongs: Into the bin. - In the important dynamic range of the mid tones both sensors are about as equal as they get.
So no, they are not seeing the differences in the sensor area, the reason IMHO must be sought outside the confines of the camera/lens system.

In my workshops we have on numerous occasions had 5D3 and D800 shooters side by side.  Expose a backlit scene with both cameras pushing both cameras exposure right to the threshold on the right side of the histogram.  Then when the resultant images are analyzed side by side on the computer, the D800 has massively more shadow detail where the 5D3 is blocked up.  Not believing our own eyes, we shot a sunrise scene at monument valley on these two cameras and we simply could not record any detail on the 5D3 in the rock formations without blowing the sky out and on the D800 we simply could not silhouette the rocks without leaving some room at the right side of the histogram unused.  Furthermore, one can raise the level of the shadow detail on the D800 while there is simply nothing there on the 5D3.  This isn't measurement hocus pocus, it is real and can be demonstrated easily at any time.  Try it for yourself, you will be surprised at how much more useful information is recorded on the D800 than the 5D3.  This is of course all at ISO 100, as I stated and others, the difference goes away as ISO's get higher.

As for Sony sensors, your information is a bit misleading.  Sony cameras apply a lossy RAW compression with their image processor but this has absolutely nothing to do with the sensor itself.  Nikon cameras using the same sensor but a different image processor do not apply this lossy compression algorithm despite using the same sensor design (only the micro lenses are different).  This compression has absolutely nothing to do with the sensor so injecting Sony cameras into this Nikon vs. Canon debate is misdirecting the argument.

As Roger pointed out, this may not be what he is seeing but your statements claiming what I said to be nonsense is not correct.  The difference he is seeing might just be lens to camera autofocus calibration related for all I know.  The Canon 70-200 f/4 lens is certainly very capable of excellent image quality.

E.J. Peiker wrote: As Roger pointed out, this may not be what he is seeing but your statements claiming what I said to be nonsense is not correct.  The difference he is seeing might just be lens to camera autofocus calibration related for all I know.  The Canon 70-200 f/4 lens is certainly very capable of excellent image quality.

??  My sensor analyses have nothing to do with any lens.     I was doing one test with a lens to see if the nikon  d800 deleted faint stars with a median filter as earlier nikons did that; it has nothing to do with autofocus calibration (I focus star image with live view).

My sensor analyses confirm what you are seeing with higher dynamic range of the Nikons.  But 5d3 shadows should not be "blocked up."  If so, that is a problem with the raw converter.  I have noticed a default on ACR that subtracts a black level from Canon raw files.  That clips the lows.  Perhaps that is what you saw.

Like I said earlier, I see more issues and differences with raw conversion than I do with real raw sensor data.

Roger

Although I have not used the most recent "top of the line" from Canon and Nikon, I have switched back and forth both brands since 2009, using mostly used equipment, so as a "serial switcher" it seems my answer is awaited.

Concerning bodies, I have used a D300 for 3 years or so, then a 7D for 2 more years, and now I use a D3s. More punctually I have used and processed pictures from D7000, D7100, 5D2, 1dIV, 100D also.
Probably comparing apples to oranges here, however not worse than comparing a 80-400/5.6 to a 70-200/4 ;o)

With the workflow I am using (RAW through LR, 2 then 4), my feeling is that Nikon bodies tend to produce more "commited" colors and tones (now that is an unscientific, however first-hand assesment). Sometimes it looks just right out of the box, as if pp was done for you. However when it does not fit, it can be pretty hard to correct in post.
Canon bodies, through LR, seem to yield more predictable and reliable results. I have written a few posts here on how good I found the 7D is, even compared to apparently less noisy sensors. I like the clarity of the images I get with it, and I find files are easilly processable, whereas d3s files can look gorgeous, but when not, need you to scratch you head to get what you want. Annoying sometimes.

Another point is that for an unknown reason, Nikon files usually look great on the camera LCD, but worse in LR, whereas Canon file are, again, more consistent and WYSIWYG in this regard.

But as other have said above, the lens matters more. Nikkor 300/2.8 Ai, 500/4 Aip, 500/4 AF-S mkII, 300/4 AF-S, Canon 300/2.8L IS, 500/4 L IS, 400/5.6L are among the used glass I owned, and to be honest, Canon lenses have impressed me the most. I am in the "shoot that heavy lens wide open" thing, and the 400 and 500L have yielded jaw-dropping results. I found Nikkor teles are greatly constructed but often need to be closed 2/3 of a stop to give best results, which makes using them a tad more complicated in the field (until Canon and Nikon provide us with a true "TAv" mode as Pentax does).

Also, I must agree with other posters that nowadays the post-processing software is the prism through which we see things, and all the above only holds with LR … Or not, I can't say, I did not switch processing softwares (although it would be a cheaper switch for sure)

Manu.

Primus wrote:Roger, thanks so much for the detailed response. I agree completely that overall it is the lens that is so much more important with the modern sensors. The new Nikon 80-400 is so good (from what I've seen) that it pretty much eliminates most other long lenses for wildlife work. That is a huge thing considering how difficult it is to carry and hand-hold  the bigger zooms. Just add a good 24-70 to another body and you're covered.

Thanks again, very helpful to know what's going on in the complex world of digital sensors.

Pradeep

Light-quality, regardless of brand name:

It is the lens and the quality of the light.  It has occurred to me that very high light-quality is so overwhelming to the eye/brain, that it tends to cause me to overlook the other parameters of an image.    

Regarding the 80-400G AFS….I agree with you:

1.  I have owned 3 Nikon 300 2.8's since 1984.  I love the 300 2.8's.   I now own the 80-400G AFS 

2.   The 80-400G AFS is sharp.  Ok, my 80-400G AFS is not quite as sharp as my 300 2.8's were…...but it is sharp enough for the pros and that's at all focal lengths.  

3.   Autofocus speed and bokeh do go to the 2.8's.  But the autofocus speed of the 80-400 has been sufficient in all situations for me thus far.  Thus far, I have a couple of images of medium-sized flying bird that this lens has nailed for me.   So I am not relegated to static images with this zoom.  Nope.  And if the background is far enough away……isn't bokeh a mute point?  

4.  Both lenses are compact, but when you add the portability and especially the versatility of the 80-400G AFS, this new zoom just does it for me.  I now favor my Nikon 80-400G AFS.  It stays attached to my D610.

5.  The lens just makes me .

Robert King
www.itsaboutnature.net

rnclark wrote:

E.J. Peiker wrote: As Roger pointed out, this may not be what he is seeing but your statements claiming what I said to be nonsense is not correct.  The difference he is seeing might just be lens to camera autofocus calibration related for all I know.  The Canon 70-200 f/4 lens is certainly very capable of excellent image quality.

??  My sensor analyses have nothing to do with any lens.     I was doing one test with a lens to see if the nikon  d800 deleted faint stars with a median filter as earlier nikons did that; it has nothing to do with autofocus calibration (I focus star image with live view).

My sensor analyses confirm what you are seeing with higher dynamic range of the Nikons.  But 5d3 shadows should not be "blocked up."  If so, that is a problem with the raw converter.  I have noticed a default on ACR that subtracts a black level from Canon raw files.  That clips the lows.  Perhaps that is what you saw.

Like I said earlier, I see more issues and differences with raw conversion than I do with real raw sensor data.

Roger

You may have missed my point with that last sentence.  I never said, nor meant, that your analysis of sensors had anything to do with lenses.  My point was just that the OP could be seeing a difference in image quality between the Nikon and Canon simply due to lens AF calibration and nothing more. 

On the other part of that, the shadows will be blocked up if the scene is outside the dynamic range of the sensor and no part of the image is overexposed.  Finally,I don't think the OP's question has anything to do with photographing stars.

E.J. Peiker wrote: On the other part of that, the shadows will be blocked up if the scene is outside the dynamic range of the sensor and no part of the image is overexposed.  Finally,I don't think the OP's question has anything to do with photographing stars.

Hi EJ,
This may be semantics, but technically on the canon raw data, the low end "never gets blocked up,"   I take blocked up to mean hitting a limit, as in the data are zero with no information.  In Canon raw files, there is an offset in the data, so the low end drops into the noise, but does not go below zero.  At low ISO the downstream electronics noise on Canon is higher than Nikons, so the nikons see deeper, but then nikons do get blocked up (data truncates at zero).  The advantage with the canon method is one can average frames and keep extracting fainter signals; that can be important in astrophotography or night photography.  With Nikons, with the signal being truncated to zero, one is limited when averaging multiple frames. 

The question about stars is that it is a good test to show what the median filter does on low-end details.  In the astrophotography world, people found early on that Nikon runs a median filter on the raw data, and that filter was deleting faint stars in images.  That would also delete fine faint detail in shadows.  In practice, at low ISO, while it would compromise shadow detail, that would be occurring at a level a stop or so below where the canons lose signals in the noise at low ISO.  At high ISO, the tables are reversed and the dynamic range of Nikon and Canon are close for similar pixel sizes, and the Canon's produce better shadow/low level detail, and the Nikons block up some data at zero.  The question is do modern Nikons still have the median filter?  There definitely appears to be some cooking of the raw files on the D800.

The bottom line is that both camera systems produce amazing images.  The Nikon raw strategy produces higher dynamic range at low ISO.  The Canon strategy produces better data for high ISO/low light imaging at the low end.  But in the scheme of things, the differences are small and only seen in extreme cases.  Whether or not one encounters such cases depends on the photographer and their imaging style and subject.  With canon, the low ISO dynamic range issue is easily compensated for by obtaining two or more frames for an HDR.  For nikon, the zero clipping issue is migigated by doing multiple exposures, one longer.  Both cases can work with static subjects, but fail if the scene is changing too fast.

Bottom bottom line: both system do great imaging, neither are perfect.   Better to concentrate on the subject and the light, and not on the pixels.

Roger

You are right, just semantics, we are actually thinking the same thing

Agree with everything in your last post. I am very much a low ISO shooter, rarely venturing much beyond 400 so my perspective is a bit different than yours

Roger,

Most people do not do Astro Photography and would not be limited by the way Nikon handles noise. The vast majority are more concerned with differences in highlights versus shadow detail needed to see detail in the shadows of sunlit images. Nikon and Sony are far better at this type of work than Canon in terms of absolute dynamic range.

For Astro Photography, to my knowledge, Nikon is still "cooking" the raw files with a pass that gets rid of, what it considers, stray points of light. That process can get rid of noise and it can also get rid of star detail that is real, not noise. For artistic purposes, this doesn't impact most night sky images as that level of detail isn't usually needed. For scientific work, it kills things that it should not be making the camera less than useful there.

rnclark wrote:

Primus wrote:The new Nikon 80-400 is so good (from what I've seen) that it pretty much eliminates most other long lenses for wildlife work.

I find this very hard to believe.  Besides image sharpness, contrast, and bokeh, unless one wants to only photograph static subjects, AF speed is critical for getting action.  I bet the zoom can't come close to matching a prime lens like a 300 f/2.8 for AF speed, and I would bet the same in the other parameters too.

Roger

Roger, sorry, I didn't mean to say that it is sharper than a 300 2.8. I used to own one of those and it was the sharpest lens I have ever used.

My point was that it is so sharp from 80mm to 400mm that one does not need to carry several different lenses to cover that range. And it is no bigger or heavier than the usual 70-200 f2.8 from either stable. If Canon were to come up with a replacement for the old 100-400 push-pull lens, this is what I am hoping it will be. However, I doubt if it will happen soon as that will kill off the 200-400 1.4x sales. 

As far as the AF speed is concerned, we were shooting wildlife that was running either towards us or across the field and my friend got tack sharp pictures everytime with AF-servo (whatever Nikon calls it).

Pradeep

rnclark wrote:

E.J. Peiker wrote: On the other part of that, the shadows will be blocked up if the scene is outside the dynamic range of the sensor and no part of the image is overexposed.  Finally,I don't think the OP's question has anything to do with photographing stars.

Hi EJ,
This may be semantics, but technically on the canon raw data, the low end "never gets blocked up,"   I take blocked up to mean hitting a limit, as in the data are zero with no information.  In Canon raw files, there is an offset in the data, so the low end drops into the noise, but does not go below zero.  At low ISO the downstream electronics noise on Canon is higher than Nikons, so the nikons see deeper, but then nikons do get blocked up (data truncates at zero).  The advantage with the canon method is one can average frames and keep extracting fainter signals; that can be important in astrophotography or night photography.  With Nikons, with the signal being truncated to zero, one is limited when averaging multiple frames. 
Roger

Roger,

Thought you would like to know that with Nikon's newest Expeed 4 processor, black is no longer set to 0.  So far that would be the D5300, D3300, D4s and D810 bodies.  Wish I could find the specific discussion, but one knowledgeable reference says "One difference is that the raw data no longer has a Black Level of 0"   http://www.dpreview.com/forums/post/53244080.  As I recall, black is now set around 600.

John

Primus wrote:
Am I seeing something that is really not there? Switching to Nikon would be quite painful given how much I've invested in Canon so far. Given the lack of development in the 100-400mm range (I doubt they will come up with a new one because it will kill off sales of the 200-400 1.4x) I am a bit disappointed.

I have done a bit of shooting with Mark Picard over the last few years.  He's a Nikon guy and I shoot canon.  I've always noticed that his images look much nicer on the camera LCD as well as on the PC and in print than mine do... even when we are shooting the same subject and positioned right next to each other. 

I'm pretty comfortable that if we switched cameras I's still see the same end result.

Is Nikon better than Canon?  Nah... he's just an amazing photographer.  

Hey, one local Nikon guy sometimes has me shaking my head sometimes. Especially when he tells me he shoots Jpeg only and does little to no post processing except adjusting contrast. I have seen some of his images taken in really contrasty conditions and come out looking great like the lighting was more even (sure, he may pick and choose which ones he wants to print but still, it's been fairly consistent).

crw816 wrote:

Primus wrote:
Am I seeing something that is really not there? Switching to Nikon would be quite painful given how much I've invested in Canon so far. Given the lack of development in the 100-400mm range (I doubt they will come up with a new one because it will kill off sales of the 200-400 1.4x) I am a bit disappointed.

I have done a bit of shooting with Mark Picard over the last few years.  He's a Nikon guy and I shoot canon.  I've always noticed that his images look much nicer on the camera LCD as well as on the PC and in print than mine do... even when we are shooting the same subject and positioned right next to each other. 

I'm pretty comfortable that if we switched cameras I's still see the same end result.

Is Nikon better than Canon?  Nah... he's just an amazing photographer.  

You are right, the photographer makes all the difference - usually.

Which is why in my OP I stated- I  want to hear from people who've used BOTH systems, which would eliminate the human factor - largely.

Pradeep

Anthony Medici wrote:Roger,

Most people do not do Astro Photography and would not be limited by the way Nikon handles noise. The vast majority are more concerned with differences in highlights versus shadow detail needed to see detail in the shadows of sunlit images. Nikon and Sony are far better at this type of work than Canon in terms of absolute dynamic range.

For Astro Photography, to my knowledge, Nikon is still "cooking" the raw files with a pass that gets rid of, what it considers, stray points of light. That process can get rid of noise and it can also get rid of star detail that is real, not noise. For artistic purposes, this doesn't impact most night sky images as that level of detail isn't usually needed. For scientific work, it kills things that it should not be making the camera less than useful there.

My point wasn't just for for astrophotography.  Astrophotography is a good test to show what happens at the low end.  But if Nikon does run filters on raw data, it will destroy fine detail, and particularly low level details in shadows in all images, not just astrophotos.  I see many nightscape images that have an amazing lack of faint stars; quite unnatural.  This is sometimes due to running filters.  But if the camera is running filters, I would not want that--I would rather choose and tune the filter for the application.  Running filters on raw data and clipping the blacks can not be good for shadow detail

It is good news if Nikon has changed in camera raw processing and now is not clipping the blacks in the D810.

Roger

The experiments I have seen with what Nikon is doing for noise reduction only removes a bright pixel (as in single pixel) where it is surrounded by dark pixels. To my knowledge, this rarely if every happens outside of Astrophotography. The detail in the shadow area of a daylight or inside photo never has a single bright pixel next to dark ones so the filter Nikon is using would not impact that.

Night sky images without faint stars can easily be explained by the photographer using an after market noise reduction tool on the image. So unless those images you have seen are "right out of the camera" images, I would think that it was the processing that caused this, not the camera.

Also, unless you shoot both systems side by side, your statements about what the Nikon system can or can't do is all based on second hand knowledge. Maybe at some point in the future we can shoot side by side and process the images side by side. Then we would have the data to discuss just what advantages one system has over the other.