October 30, 2014, 12:29:56 PM

Author Topic: Nikon's D800E 30% sharper than D800  (Read 8780 times)

privatebydesign

  • Canon EF 300mm f/2.8L IS II
  • *******
  • Posts: 2600
  • Ermintrude says "moo"
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #30 on: July 07, 2014, 10:18:04 AM »
That is laughable for several obvious reasons, first, they are saying the Ziess lens is perfect and causes zero resolution loss, that is impossible, it is either breaking the laws of physics, or their measurements are suspect yet again. And, just read any Nikon forum where people own both, and there are a surprising amount, they will tell you that is simply not true, yes the E does resolve slightly more, but 30% more, no.

Not exactly perfect.  just able to use the full resolution of the sensor.

If they sensor was 8 MP and a lens resolved 8, would you call that perfect?  Just pushes the limit of the sensor.

No that isn't how it works, there is a complex relationship between each individual elements efficiency and a systems efficiency. Pretty much all lenses can actually resolve way more than any sensor, just look at the difference between a lens optical bench tested lens and one that relies on a camera sensor, huge difference.

So if the sensor was a 20MP sensor and the lens and sensor were both perfect then you'd expect to get 20MP of resolution, this is what DXO are claiming for the D800E and Zeiss 135 combo. However if we ignore all other factors and the lens is only 99% perfect it can only possibly resolve 99% of a perfect sensors resolution, and no sensor/camera is perfect. So the perfect sensor and 99% perfect lens could equate to 19.8MP in a simplified form.

For the full equations look here under "System Resolution": http://en.wikipedia.org/wiki/Optical_resolution
Never look down on anybody, unless you are helping them up.

canon rumors FORUM

Re: Nikon's D800E 30% sharper than D800
« Reply #30 on: July 07, 2014, 10:18:04 AM »

jrista

  • Canon EF 400mm f/2.8L IS II
  • *******
  • Posts: 4492
  • EOL
    • View Profile
    • Nature Photography
Re: Nikon's D800E 30% sharper than D800
« Reply #31 on: July 07, 2014, 10:58:25 AM »
That is laughable for several obvious reasons, first, they are saying the Ziess lens is perfect and causes zero resolution loss, that is impossible, it is either breaking the laws of physics, or their measurements are suspect yet again. And, just read any Nikon forum where people own both, and there are a surprising amount, they will tell you that is simply not true, yes the E does resolve slightly more, but 30% more, no.

Not exactly perfect.  just able to use the full resolution of the sensor.

If they sensor was 8 MP and a lens resolved 8, would you call that perfect?  Just pushes the limit of the sensor.

No that isn't how it works, there is a complex relationship between each individual elements efficiency and a systems efficiency. Pretty much all lenses can actually resolve way more than any sensor, just look at the difference between a lens optical bench tested lens and one that relies on a camera sensor, huge difference.

So if the sensor was a 20MP sensor and the lens and sensor were both perfect then you'd expect to get 20MP of resolution, this is what DXO are claiming for the D800E and Zeiss 135 combo. However if we ignore all other factors and the lens is only 99% perfect it can only possibly resolve 99% of a perfect sensors resolution, and no sensor/camera is perfect. So the perfect sensor and 99% perfect lens could equate to 19.8MP in a simplified form.

For the full equations look here under "System Resolution": http://en.wikipedia.org/wiki/Optical_resolution

(Note, this response is for the benefit of everyone, it is not just a reply to PBD):

I wouldn't necessarily say it's complicated, but total optical system resolving power is non-obvious.

One thing, "perfect" resolution actually means "infinite" resolution. The resolving power of an optical system (i.e. a whole camera with lens and sensor) is limited by the resolving power if the least capable component. If that is the sensor, then resolving power of the whole has an asymptotic relationship with the resolution of the sensor.

I think it's tough to say that a lens resolves 99% of "perfection"...since perfection requires infinite resolving power (at an infinite aperture, to be explicit). What is 99% of infinity? Lens resolving power is also non-linear...it falls off as the aperture is made smaller. Lens bench tests often test at max aperture and at f/8, but that is not guaranteed. So one must be specific when discussing resolving power of a system.

If we have a lens at f/4, and that lens achieves the maximum diffraction-limited resolving power, it resolves 173lp/mm. A theoretical 8mp APS-C sensor would resolve about 80lp/mm. The resolution of the whole camera, lens and sensor combined, can be closely approximated by taking the RMS of the minimum resolvable spot for each, and converting back to lp/mm. The lens resolves a spot of 2.9µm, the sensor a spot of 6.2µm. The two when working together convolve to produce a spot size of 6.85µm, or a system resolution of 73lp/mm. The two together resolve LESS than the resolving power of the least capable...in this case, the sensor.

If we dice all the pixels in our 8mp sensor into quarters (make the pixels half as large), we end up with a 32mp sensor capable of resolving 161.3lp/mm. Combined with the same lens, the system resolution is 117lp/mm. If we make a sensor with the same resolving power as the lens, we have a 39.7mp sensor. The resolving power of the system is 122lp/mm. We are still short of the 173lp/mm of the lens. We haven't actually achieved "perfect" resolving power, despite increasing the resolution of our sensor. You never can...as you increase the performance of one component or the other, the bar just keeps getting higher...the mechanisms that convolve the image signal into the final output are constantly working against you, keeping you from actually achieving the real potential of either component. You would have to RADICALLY increase the performance of one in order to approach the limit of the other. To actually resolve the 173lp/mm spatial resolution possible with an f/4 lens, you would need pixels smaller than 0.25µm, or 250nm in size. That is smaller than the wavelengths of all visible light! It's even smaller than near UV, getting into deep UV. A sensor with pixels that small would be a 5.34 GIGApixel sensor!  And that camera would still resolve 171.7lp/mm...it's still falling short of the 173lp/mm theoretical maximum of an ideal f/4 lens.

The only way to achieve perfect resolution is to have both a lens and a sensor with infinite resolving power. Obviously, such a lens does not exist. The best you can hope for is diffraction limited behavior at a lens' maximum aperture. Few lenses achieve diffraction limited behavior at f/4, most still have a small amount of optical aberrations, especially around the periphery. The Otus is one lens that approaches ideal performance pretty closely, though.

peterisviksna

  • Power Shot G7X
  • **
  • Posts: 14
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #32 on: July 07, 2014, 11:51:00 AM »
That is laughable for several obvious reasons, first, they are saying the Ziess lens is perfect and causes zero resolution loss, that is impossible, it is either breaking the laws of physics, or their measurements are suspect yet again. And, just read any Nikon forum where people own both, and there are a surprising amount, they will tell you that is simply not true, yes the E does resolve slightly more, but 30% more, no.

Lets see. On the one hand we have a company that makes image correction software and understands lens performance and characteristics probably better than anyone else. On the other hand we have people such as yourself that, well, what exactly are your qualifications to argue that they're wrong? (Aside from having a web browser that is.)

Check out the measurements for the Zeiss Sonnar T* 135mm/2:
http://www.dxomark.com/Lenses/Carl-Zeiss/Zeiss-Carl-Zeiss-Apo-Sonnar-T-Star-F2-135-ZF2-Nikon-mounted-on-Nikon-D800E---Measurements__814

... it is an amazing lens.


Wow, you must be quite silly to take something DXO says with such certainty. To this day I can't, for the life of me, understand why people read that site of Nikon fanboys and preach it like it was the one and only truth and everyone that owns a camera that doesn't do well on their very subjective tests should now throw their camera away because it's not worthy.
peterisviksna.com | facebook.com/peterisviksnaphoto

Aglet

  • 1D X
  • *******
  • Posts: 1044
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #33 on: July 07, 2014, 09:54:59 PM »
The resolving power of an optical system (i.e. a whole camera with lens and sensor) is limited by the resolving power if the least capable component. If that is the sensor, then resolving power of the whole has an asymptotic relationship with the resolution of the sensor.

Thank-you.  I think you've nicely answered my question from the first page of this thread.

Still... there's something about the explanation that seems to be missing for me...  Is there a contrast limit or something in place when making these calculations?
Otherwise it only seems logical that a lens capable of resolving 150 lp/mm (at some contrast ratio) should be able to resolve 100 lp/mm (eg. ~d800e) which is roughly the max resolving power of a sensor, even if it's at some reduced contrast ratio.  :-\

jrista

  • Canon EF 400mm f/2.8L IS II
  • *******
  • Posts: 4492
  • EOL
    • View Profile
    • Nature Photography
Re: Nikon's D800E 30% sharper than D800
« Reply #34 on: July 08, 2014, 01:12:25 AM »
The resolving power of an optical system (i.e. a whole camera with lens and sensor) is limited by the resolving power if the least capable component. If that is the sensor, then resolving power of the whole has an asymptotic relationship with the resolution of the sensor.

Thank-you.  I think you've nicely answered my question from the first page of this thread.

Still... there's something about the explanation that seems to be missing for me...  Is there a contrast limit or something in place when making these calculations?
Otherwise it only seems logical that a lens capable of resolving 150 lp/mm (at some contrast ratio) should be able to resolve 100 lp/mm (eg. ~d800e) which is roughly the max resolving power of a sensor, even if it's at some reduced contrast ratio.  :-\

Sorry, I should have stated, those numbers are all for MTF50, or 50% contrast. It is indeed possible to use a lower contrast level, however it's pretty standard to use MTF50 for photography. As we get smaller pixels and AA filters are weakened or dropped, it'll certainly be possible for lenses to meaningfully resolve detail at lower contrast levels, so maybe MTF30 could be used instead. I think by around the Rayleigh limit, contrast is too low for most modern camera sensors to pick up at higher frequencies. The human eye can barely discern contrast at that level, and we have a biological supercomputer processing what our eyes see.

I would actually have to derive all the resolutions for MTF30, I don't have that memorized for key apertures. :P

The D800 has 4.9µm pixels, or 102lp/mm. A 150lp/mm lens, at, what, around f/4.2 or so I guess, has a spot size of 1/(150 * 2), or 2.2µm (To convert from line pairs to spot size, you multiply by two to get lines, and take the reciprocal to get line thickness/spot size). Take the RMS of those two sizes, SQRT(0.0049^2mm + 0.0033^2mm), and you get 0.0059mm (5.9µm). Take the reciprocal of that and divide by two to get line pairs again: (1/0.0059) / 2 = 84.6lp/mm. That's eh, getting there. You can get very close...something like the Otus at a wider aperture probably resolves enough detail to get much closer to the 102lp/mm resolving power of the D800. The point, though, was that getting close-ish isn't too difficult...it's getting very close, i.e. maximizing the potential of a sensor, that actually gets REALLY tough...in fact, its' impossible to fully realize the full resolving power of a sensor with a high resolution lens, and to get very close, you rapidly run into diminishing returns. You get to the point where doubling your lens resolution gets you a few line pairs closer.

It's an asymptotic relationship...system resolving power is asymptotically related to the resolving power of the lowest common denominator of the system. For diffraction limited (or close to it) lenses at fast apertures, that's usually the sensor. For diffraction limited lenses at narrower apertures, that's usually the lens.

3kramd5

  • Canon 7D MK II
  • *****
  • Posts: 457
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #35 on: July 08, 2014, 09:36:18 AM »
its' impossible to fully realize the full resolving power of a sensor with a high resolution lens, and to get very close, you rapidly run into diminishing returns. You get to the point where doubling your lens resolution gets you a few line pairs closer.

It's an asymptotic relationship...system resolving power is asymptotically related to the resolving power of the lowest common denominator of the system.

Thus the moral of the story is: if your primary goal is to maximize system resolution (and for the majority of people, that's likely not the case, regardless of what one may read on some fringe forums), always upgrade the weakest component.
5D3, 5D2, 40D; Various lenses

jrista

  • Canon EF 400mm f/2.8L IS II
  • *******
  • Posts: 4492
  • EOL
    • View Profile
    • Nature Photography
Re: Nikon's D800E 30% sharper than D800
« Reply #36 on: July 08, 2014, 01:37:50 PM »
its' impossible to fully realize the full resolving power of a sensor with a high resolution lens, and to get very close, you rapidly run into diminishing returns. You get to the point where doubling your lens resolution gets you a few line pairs closer.

It's an asymptotic relationship...system resolving power is asymptotically related to the resolving power of the lowest common denominator of the system.

Thus the moral of the story is: if your primary goal is to maximize system resolution (and for the majority of people, that's likely not the case, regardless of what one may read on some fringe forums), always upgrade the weakest component.

Yeah, pretty much. Although that can become prohibitively expensive at some point.

The 7D is a good camera, but it doesn't perform terribly well with the telephoto lens it's most often paired with, the 100-400. However if you move up to one of the Canon great white primes, the 7D becomes a stellar performer at ISO settings 1600 and lower, and becomes viable at ISO settings higher than that in the evenings. It's just that you have to spend a LOT of money on those lenses to maximize the potential of the 7D.

On the flip side, if you upgrade the camera itself, to one with a higher resolution sensor (which the 7D II should have), then instead of spending $6000 to $13,000, you spend maybe $2500-3000. It's still a large chunk of change, but not necessarily prohibitively expensive.

canon rumors FORUM

Re: Nikon's D800E 30% sharper than D800
« Reply #36 on: July 08, 2014, 01:37:50 PM »

3kramd5

  • Canon 7D MK II
  • *****
  • Posts: 457
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #37 on: July 08, 2014, 11:44:47 PM »
its' impossible to fully realize the full resolving power of a sensor with a high resolution lens, and to get very close, you rapidly run into diminishing returns. You get to the point where doubling your lens resolution gets you a few line pairs closer.

It's an asymptotic relationship...system resolving power is asymptotically related to the resolving power of the lowest common denominator of the system.

Thus the moral of the story is: if your primary goal is to maximize system resolution (and for the majority of people, that's likely not the case, regardless of what one may read on some fringe forums), always upgrade the weakest component.

Yeah, pretty much. Although that can become prohibitively expensive at some point.

The 7D is a good camera, but it doesn't perform terribly well with the telephoto lens it's most often paired with, the 100-400. However if you move up to one of the Canon great white primes, the 7D becomes a stellar performer at ISO settings 1600 and lower, and becomes viable at ISO settings higher than that in the evenings. It's just that you have to spend a LOT of money on those lenses to maximize the potential of the 7D.

On the flip side, if you upgrade the camera itself, to one with a higher resolution sensor (which the 7D II should have), then instead of spending $6000 to $13,000, you spend maybe $2500-3000. It's still a large chunk of change, but not necessarily prohibitively expensive.

Most definitely. I'd love to see a chart of total resolving power versus dollars. Gotta pay to play.
5D3, 5D2, 40D; Various lenses

privatebydesign

  • Canon EF 300mm f/2.8L IS II
  • *******
  • Posts: 2600
  • Ermintrude says "moo"
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #38 on: July 09, 2014, 12:33:35 AM »
Most definitely. I'd love to see a chart of total resolving power versus dollars. Gotta pay to play.

What I know is you need more than a $2,199 Zeiss to resolve 100% of a 36MP 135 format sensor. DXO just did it again, they threw any "scientific" credibility they had out the window, not that they had much left anyway.

A result made all the more comical when you look at the $4,000 Otus results on the D800E, down to 33MP, a drop of over 8% for what is regarded as one of the finest photography camera lenses ever made. I think DXO have two teams of testers and the Canon team, who clearly multiply all their results by 0.9, mistakenly did the Nikon Otus when they did the Canon Otus as well. No that's not true, I believe the Canon testers are OK, it is the Nikon test team that multiply all their results by 1.15.

As for not being able to have 99% of perfection or it being a strange way to look at it, I understand that, I was just trying to illustrate that anybody claiming perfect anything is farcical and it isn't as simple as the lens being capable of resolving more than the sensor. Like the >14 stops of DR in a 14 bit file, extrapolate to ridiculous figures all you want (DXO) but if I can't actually realise that shadow lifting capability it is of no practical use.
Never look down on anybody, unless you are helping them up.

jrista

  • Canon EF 400mm f/2.8L IS II
  • *******
  • Posts: 4492
  • EOL
    • View Profile
    • Nature Photography
Re: Nikon's D800E 30% sharper than D800
« Reply #39 on: July 09, 2014, 01:24:58 AM »
Most definitely. I'd love to see a chart of total resolving power versus dollars. Gotta pay to play.

What I know is you need more than a $2,199 Zeiss to resolve 100% of a 36MP 135 format sensor. DXO just did it again, they threw any "scientific" credibility they had out the window, not that they had much left anyway.

A result made all the more comical when you look at the $4,000 Otus results on the D800E, down to 33MP, a drop of over 8% for what is regarded as one of the finest photography camera lenses ever made. I think DXO have two teams of testers and the Canon team, who clearly multiply all their results by 0.9, mistakenly did the Nikon Otus when they did the Canon Otus as well. No that's not true, I believe the Canon testers are OK, it is the Nikon test team that multiply all their results by 1.15.

As for not being able to have 99% of perfection or it being a strange way to look at it, I understand that, I was just trying to illustrate that anybody claiming perfect anything is farcical and it isn't as simple as the lens being capable of resolving more than the sensor. Like the >14 stops of DR in a 14 bit file, extrapolate to ridiculous figures all you want (DXO) but if I can't actually realise that shadow lifting capability it is of no practical use.

Totally agree with everything here. DXO has some really wacko shite goin on with their lens tests. They have the weirdest lens test results I've ever seen...anywhere.

LetTheRightLensIn

  • Canon EF 300mm f/2.8L IS II
  • *******
  • Posts: 3939
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #40 on: July 09, 2014, 01:50:38 AM »
Most definitely. I'd love to see a chart of total resolving power versus dollars. Gotta pay to play.

What I know is you need more than a $2,199 Zeiss to resolve 100% of a 36MP 135 format sensor. DXO just did it again, they threw any "scientific" credibility they had out the window, not that they had much left anyway.

A result made all the more comical when you look at the $4,000 Otus results on the D800E, down to 33MP, a drop of over 8% for what is regarded as one of the finest photography camera lenses ever made. I think DXO have two teams of testers and the Canon team, who clearly multiply all their results by 0.9, mistakenly did the Nikon Otus when they did the Canon Otus as well. No that's not true, I believe the Canon testers are OK, it is the Nikon test team that multiply all their results by 1.15.

As for not being able to have 99% of perfection or it being a strange way to look at it, I understand that, I was just trying to illustrate that anybody claiming perfect anything is farcical and it isn't as simple as the lens being capable of resolving more than the sensor. Like the >14 stops of DR in a 14 bit file, extrapolate to ridiculous figures all you want (DXO) but if I can't actually realise that shadow lifting capability it is of no practical use.

They probably scored the Otus wide open so losing 8% wouldn't be surprising even for that lens.

But their lens tests are often weird (lots of weird stuff even in their data plots, although some of the worst stuff they seem to have fixed) and their overall lens scores are nonsensical since how does a zoom turn into a single number? How do you compare one prime wide open to another partly stopped down? How do you compare one zoom at 24mm f/2.8 to another at 50mm f/5.6 to another at 24mm f/4 to another at.... it's apple to oranges because of their ridiculous T-stop, nearest to wide open obsession where they try to find the closest to wide that gives a T-stop they like that manages to just be sharp to a certain degree, what kind of a goal is that for comparison?

That said, their sensor plot data is generally pretty good and certainly better than you'll find at other review sites.

LetTheRightLensIn

  • Canon EF 300mm f/2.8L IS II
  • *******
  • Posts: 3939
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #41 on: July 09, 2014, 01:53:51 AM »
http://www.dxomark.com/Reviews/Best-lenses-for-the-Nikon-D800E-The-sharpest-full-frame-camera-ever-measured/Sharpness-analysis-D800E-vs.-D800

And how much of the sharpness is actually aliasing noise (fake detail)? And what about how it has like probably way more than 30% more aliasing and moire issues?

LetTheRightLensIn

  • Canon EF 300mm f/2.8L IS II
  • *******
  • Posts: 3939
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #42 on: July 09, 2014, 01:54:28 AM »
That is laughable for several obvious reasons, first, they are saying the Ziess lens is perfect and causes zero resolution loss, that is impossible, it is either breaking the laws of physics, or their measurements are suspect yet again. And, just read any Nikon forum where people own both, and there are a surprising amount, they will tell you that is simply not true, yes the E does resolve slightly more, but 30% more, no.

There could easily be some rounding done on that chart.

canon rumors FORUM

Re: Nikon's D800E 30% sharper than D800
« Reply #42 on: July 09, 2014, 01:54:28 AM »

dilbert

  • Canon EF 300mm f/2.8L IS II
  • *******
  • Posts: 3111
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #43 on: July 09, 2014, 02:11:45 AM »
Most definitely. I'd love to see a chart of total resolving power versus dollars. Gotta pay to play.

What I know is you need more than a $2,199 Zeiss to resolve 100% of a 36MP 135 format sensor. DXO just did it again, they threw any "scientific" credibility they had out the window, not that they had much left anyway.

What is the basis for your knowledge here?

Quote
Quote
A result made all the more comical when you look at the $4,000 Otus results on the D800E, down to 33MP, a drop of over 8% for what is regarded as one of the finest photography camera lenses ever made. I think DXO have two teams of testers and the Canon team, who clearly multiply all their results by 0.9, mistakenly did the Nikon Otus when they did the Canon Otus as well. No that's not true, I believe the Canon testers are OK, it is the Nikon test team that multiply all their results by 1.15.

As for not being able to have 99% of perfection or it being a strange way to look at it, I understand that, I was just trying to illustrate that anybody claiming perfect anything is farcical and it isn't as simple as the lens being capable of resolving more than the sensor. Like the >14 stops of DR in a 14 bit file, extrapolate to ridiculous figures all you want (DXO) but if I can't actually realise that shadow lifting capability it is of no practical use.

Totally agree with everything here. DXO has some really wacko shite goin on with their lens tests. They have the weirdest lens test results I've ever seen...anywhere.

But that wacko S___ produces arguably the best RAW conversion results when they feed it into DxO Optics.

9VIII

  • 5D Mark III
  • ******
  • Posts: 658
    • View Profile
Re: Nikon's D800E 30% sharper than D800
« Reply #44 on: July 09, 2014, 05:30:31 AM »
This reminds me or multiple hot button topics in popular culture.
The "establishment" says "x" is true, then people with a little bit of logical thinking point out inconsistencies, and the best comeback usually boils down to "I trust the establishment more than you".

All I can say is there's way too much marketing in science these days.
-100% RAW-

canon rumors FORUM

Re: Nikon's D800E 30% sharper than D800
« Reply #44 on: July 09, 2014, 05:30:31 AM »