April 21, 2014, 09:04:29 AM

Author Topic: 46mp sensor useless for landscape?  (Read 20055 times)

Mt Spokane Photography

  • 1D X
  • *******
  • Posts: 7714
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #15 on: October 07, 2012, 12:56:08 PM »
The total MTF of a system is the product of the individual mtf's of the components.  The final MTF will never be greater than the lowest MTF in the system.
Thus, a camera with a lens MTF of 0.9, and a body of 0.7 will be 0.63.  Increase either one, and the value goes up.  Suppose, we have a body with a very large number of MP, say a billion, with a impossibly high MTF of 0.99.  The system MTF will then be basically that of the lens.  (0.891).
What all this means is that increasing one component or the other is going to raise the mtf of a real world camera / lens system. Of course, at some point, there will be enough MTF in a camera body to cause the lens to be the limiting factor, but that is well out in the future, we are not near reaching it.

canon rumors FORUM

Re: 46mp sensor useless for landscape?
« Reply #15 on: October 07, 2012, 12:56:08 PM »

neuroanatomist

  • CR GEEK
  • *******
  • Posts: 12794
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #16 on: October 07, 2012, 01:19:22 PM »
Ok, I think this is too technical for me. So 46 MP is same as 36 MP for landscap?

No, 46 MP is higher resolution than 36 MP, for any application.
EOS 1D X, EOS M, and lots of lenses
______________________________
Flickr | TDP Profile/Gear List

friedmud

  • EOS M2
  • ****
  • Posts: 217
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #17 on: October 07, 2012, 01:59:10 PM »
Of course, at some point, there will be enough MTF in a camera body to cause the lens to be the limiting factor, but that is well out in the future, we are not near reaching it.

I wouldn't necessarily say that.  Go read about the lengths D800E users have to go to in order to get all the resolution out of their sensors.  You have to use the _very_ best glass and hit the perfect f/stop with the perfect focusing (hyperfocal generally).  It is already getting quite fiddly to feed 36mp... if Canon is going to do 46mp things are definitely going to get interesting.

BTW - It's a pet peeve of mine that many "landscape photographers" don't truly understand diffraction, DOF and hyperfocal focusing.  If you think "all landscape shots are at f/8 or smaller" then you need to go do some reading and shooting.

Just go get a really good camera and lens, stick it on a good tripod, use manual focusing and mirror lockup and go through a series of shots from f/4 to f/22.  Choose a good landscape scene with foreground interest (although your camera doesn't have to be right up against the foreground interest) and go through the aperture series while focusing at 3 different points:

1.  At the foreground interest
2.  Halfway to the distant object (like mountains or a far off building)
3.  At the distant object

Compare your results.

The results will be pretty damn obvious.  You'll get the best sharpness focusing halfway between (which is just an approximation of the hyperfocal distance that will be good enough for you to see what's going on) the near and far subjects and with an aperture that is just on the large side of the diffraction limit (generally around f/5 to f/8 for most modern sensors and good glass).

By focusing better (right at the true hyperfocal distance) you can get everything sharp with larger apertures (I'm using f/5.6 more and more often lately in my own work).

If you use a tilt-shift you can do even better than that because you tilt the focal plane so that it more closely approximates your scene (instead of just being perpendicular to your lens).

All of this is a way of saying that as we get higher MP sensors... we (the landscapers) are going to need to do more work and be more diligent to get the very best possible image.

I read an article not too long ago where a pro (really?) landscape photographer was advocating shooting everything at f/22 focused at infinity.  I nearly lost my lunch.  How could a "pro" never have even done the above testing?  Because when you're using film or only 10mp it simply doesn't matter that much....

friedmud

  • EOS M2
  • ****
  • Posts: 217
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #18 on: October 07, 2012, 02:31:45 PM »
Beyond shooting, I also definitely recommend using a "DOF Calculator" for a while (like this one http://www.dofmaster.com/dofjs.html ).  You can also get really good DOF calculators for iOS and Android so you can have them in the field.

Just playing with one will give you a good idea of aperture sizes to use.

Note that according to that dofmaster website with a D800 at 24mm at f/5.6 I can focus at 11.2 ft (the hyperfocal distance) and get everything from 5.6 ft to infinity in focus.

Even at f/4 I can focus at ~16ft and get everything from 8ft to infinity in focus!  If your camera is on a tripod at eye level, it's going to be rare to have something closer than 8 ft...

Once you get a feel for this you can "wing it" in the field... I generally just focus a bit beyond my foreground interest and choose an aperture around f/5 to f/8 depending on if there is something closer than that.  Beyond f/8 on my 7D diffraction starts to kill any gains made in DOF...

TonyY

  • Guest
Re: 46mp sensor useless for landscape?
« Reply #19 on: October 07, 2012, 02:36:49 PM »
Just hoping Sigma can make a 35mm full frame 48MP (3 x 16 MP) sensor, some of the Sigma D2P Merrill sample picture are really astonishing, shows tons of detail and color compare with my 5D Mark II.


TonyY

  • Guest
Re: 46mp sensor useless for landscape?
« Reply #20 on: October 07, 2012, 02:58:26 PM »
very interesting, all the Nikon D800/D800E sample landscap images are taken at f/8, still no detail:

http://imaging.nikon.com/lineup/dslr/d800/sample02.htm
http://imaging.nikon.com/lineup/dslr/d800/sample01.htm

Canon 5D Mark III is even worth:

Positron

  • Rebel T5i
  • ****
  • Posts: 129
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #21 on: October 07, 2012, 03:22:59 PM »
The source article seems very in-depth and while I'd love to spend enough time reading it to understand everything it's saying, I have actual work to do today, so I will shamelessly ask the question on my mind instead: assuming that what's said in the article is correct, how does it affect lenses that can tilt? For example, if you have a TS-E 24mm, can you leave it wide open or nearly wide open to maximize the effective resolution, and use only tilt to gain the DoF you need? Put more directly, will tilting provide more absolute resolution than stopping down, or is there some other loss from tilt?

canon rumors FORUM

Re: 46mp sensor useless for landscape?
« Reply #21 on: October 07, 2012, 03:22:59 PM »

dtaylor

  • 7D
  • *****
  • Posts: 443
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #22 on: October 07, 2012, 04:02:39 PM »
I wouldn't necessarily say that.  Go read about the lengths D800E users have to go to in order to get all the resolution out of their sensors.  You have to use the _very_ best glass and hit the perfect f/stop with the perfect focusing (hyperfocal generally).  It is already getting quite fiddly to feed 36mp... if Canon is going to do 46mp things are definitely going to get interesting.

I don't consider a 7D that difficult and it has a higher pixel pitch than the D800. There are plenty of lenses that can provide 46 MP of detail FF (18 MP crop).

Diffraction is not the hard, fast limit people think it is. There's a gradual loss of detail contrast. That's something people don't understand when discussing diffraction or lens testing. They will, for example, read that a lens "only" provides XY lpmm at 50% MTF, or that you can "only" achieve XY lpmm at Z aperture, and assume no sensor with smaller pixels can benefit. What they don't realize is that the lens or aperture setting can still saturate any sensor at 20% MTF. And in our digital world, PS can bring back much of the lost detail contrast.

You run into similar nonsense comparing FF and APS-C sensors, and especially comparing lens resolution on the two formats because of the way the major sites compute the resolution numbers.

heptagon

  • Rebel T5i
  • ****
  • Posts: 147
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #23 on: October 07, 2012, 04:18:50 PM »
Ok, I think this is too technical for me. So 46 MP is same as 36 MP for landscap?

No, 46 MP is higher resolution than 36 MP, for any application.

This is not ultimately true. According to the Nyquist theorem you get the whole information about a bandwidth limited signal if your samples are spaced close enough. Making more samples provides you with no new information. While the airy function is no sinc function, it is rather similar. So there is supposed to be a pixel spacing below which the quality gains become marginal. However i suspect this isn't the 50% contrast assumed in the calculations and if you consider the effects of the bayer pattern there should be a lot more pixels than we have today.

Radiating

  • EOS M2
  • ****
  • Posts: 323
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #24 on: October 07, 2012, 05:11:40 PM »
Color resolution and black and white resolution are different. It will still look sharper Even if you're not getting any more red redolution. The maximum resolution of the best full frame prime lenses is just under 30 mp (28.3 mp). Add a low pass filter into the mix and you can justify at least 31 mp as being usable. With 36mp bring usable in around 3 years.

46mp is just useless thoug for the next decade probably. But hey it's better than 20 mp even if its overkill.

neuroanatomist

  • CR GEEK
  • *******
  • Posts: 12794
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #25 on: October 07, 2012, 06:12:13 PM »
Ok, I think this is too technical for me. So 46 MP is same as 36 MP for landscap?

No, 46 MP is higher resolution than 36 MP, for any application.

This is not ultimately true. According to the Nyquist theorem you get the whole information about a bandwidth limited signal if your samples are spaced close enough. Making more samples provides you with no new information. While the airy function is no sinc function, it is rather similar. So there is supposed to be a pixel spacing below which the quality gains become marginal.

Yes, returns diminish, but we're not at marginalized returns yet.  But note that I stated higher resolution, not 10 MP higher resolution.

Since you went and invoked Nyquist, :P I will ask - what is the physical phenomenon that we are sampling, and what property(ies) of that phenomenon provide the limits from which we determine the minimal frequency to adequately capture all information present, and the optimal oversampling frequency?

The maximum resolution of the best full frame prime lenses is just under 30 mp (28.3 mp). Add a low pass filter into the mix and you can justify at least 31 mp as being usable. With 36mp bring usable in around 3 years.

I'd love to see your evidence for this, especially since you post such a specific value as 28.3 MP.  Do you have a link or data to share?  Hopefully, the data will also explain how an APS-C sensor with a higher pixel density than the densest FF sensor on the market can resolve a higher level of local detail from those 'best full frame lenses' than a FF sensor.
EOS 1D X, EOS M, and lots of lenses
______________________________
Flickr | TDP Profile/Gear List

Policar

  • 7D
  • *****
  • Posts: 361
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #26 on: October 07, 2012, 06:17:38 PM »
So what if 19 megapixels is the most a FF red-sensitive sensor can resolve at f8? A 46 megapixel bayer sensor only has 11.5 million red-sensitive photosites. No problem at all...

At a certain point it does get kind of pointless, especially if you shoot landscapes without a T/S lens (in which case--what are you thinking?). We're not there yet. There's a test online between 80MP backs and 4x5 and 8x10 and 8x10 is still best by far, though 80MP is sharper with less fine detail than 4x5.

With film enlargements you're limited by the size of the grain. 4x5 isn't that much sharper than 6x7 when stopped down to equivalent DoF but the tonality and grain are better. Same with 8x10, you can make a wall-sized print (80''X100") that holds up to scrutiny even up close, but an equivalent enlargement of 135 (to 8''X12')' is sharper (just way, way smaller and with the same granularity and tonality). With digital, there's no grain, but when you can see the pixel structure, that's bad, so there's still that matter of tonality but to a much less significant extent.

And despite diffraction, landscape shooters have shot at f64 on 8x10 and even 4x5 for quite a while and made huge prints. We might not get sharper than that without stitching, but unless you're printing bigger than wall-sized in a venue where people will stand right next to the print (extremely rare, and obviously you would want to use stitching in these cases), then you'll be fine.

Tammy

  • Rebel T5i
  • ****
  • Posts: 110
    • View Profile
Re: 46mp sensor useless for landscape?
« Reply #27 on: October 08, 2012, 12:58:08 AM »
There are some really smart people on this board. I love reading these threads.
5D Mk III - 24-70L Mk II - 24L Mk II - 100L - 135L - 50mm F/1.4

canon rumors FORUM

Re: 46mp sensor useless for landscape?
« Reply #27 on: October 08, 2012, 12:58:08 AM »

jrista

  • 1D X
  • *******
  • Posts: 3254
  • POTATO
    • View Profile
    • Nature Photography
Re: 46mp sensor useless for landscape?
« Reply #28 on: October 08, 2012, 01:30:54 AM »
Ok, I think this is too technical for me. So 46 MP is same as 36 MP for landscap?

No, 46 MP is higher resolution than 36 MP, for any application.

This is not ultimately true. According to the Nyquist theorem you get the whole information about a bandwidth limited signal if your samples are spaced close enough. Making more samples provides you with no new information. While the airy function is no sinc function, it is rather similar. So there is supposed to be a pixel spacing below which the quality gains become marginal.

Yes, returns diminish, but we're not at marginalized returns yet.  But note that I stated higher resolution, not 10 MP higher resolution.

Since you went and invoked Nyquist, :P I will ask - what is the physical phenomenon that we are sampling, and what property(ies) of that phenomenon provide the limits from which we determine the minimal frequency to adequately capture all information present, and the optimal oversampling frequency?

The maximum resolution of the best full frame prime lenses is just under 30 mp (28.3 mp). Add a low pass filter into the mix and you can justify at least 31 mp as being usable. With 36mp bring usable in around 3 years.

I'd love to see your evidence for this, especially since you post such a specific value as 28.3 MP.  Do you have a link or data to share?  Hopefully, the data will also explain how an APS-C sensor with a higher pixel density than the densest FF sensor on the market can resolve a higher level of local detail from those 'best full frame lenses' than a FF sensor.

To further Neruo's argument (which is excellent), everyone has to keep in mind that the total "system" resolution of our combined camera+lens is considerably lower than the individual spatial resolutions of each system component. A high resolution sensor, such as the 7D (approx 116 lp/mm raw luminance spatial resolution) or the D800 (approx 100 lp/mm raw luminance spatial resolution) are theoretically capable of capturing FAR more resolution than we actually actually realize once a lens is attached. Total system resolution, when accounting for the intrinsic blur of each component (such as that caused by diffraction in a lens or the low-pass filter of a sensor) as well as non-ideal alignment of the projected image and the sensor pixel layout, is roughly the root mean square (RMS) of the blur of each component. On average, at best, were probably getting around 70 lp/mm at the most ideal aperture possible, and much closer to 40-50 lp/mm on average.

We would only be able to realize an approach to, say, the 116 lp/mm of raw spatial resolution the 7D (or a theoretical 47.6mp sensor, so 46.1mp is close enough) has to offer if we had an ideal, perfect, diffraction-limited lens at something around f/0.7. But thats just an approach...it would take an ideal, perfect, diffraction-limited lens of literally infinite dimension to actually achieve 116 lp/mm, which is flat out impossible (system resolution is asymptotically related to the spatial resolution of the highest resolution component in the system). We have a LONG way to go before the spatial resolution of a sensor actually becomes an issue at all, and then it will become an issue due to the diffraction cutoff frequency (the point where pixels become smaller than the wavelengths of light they are supposed to react to), not diffraction itself.
« Last Edit: October 08, 2012, 01:35:02 AM by jrista »
My Photography
Current Gear: Canon 7D | Canon EF 600mm f/4 L IS II | EF 100-400mm f/4.5-5.6 L IS | EF 16-35mm f/2.8 L | EF 100mm f/2.8 Macro | 50mm f/1.4
New Gear List: Canon 5D III/7D II | Canon EF 300mm f/2.8 L II

PackLight

  • Guest
Re: 46mp sensor useless for landscape?
« Reply #29 on: October 08, 2012, 02:19:13 AM »
If the original premiss of the OP is that the lens would be useless for landscapes, wouldn't it be just as useless for most all other types of photography. Green, Blue and Red come in other shades other than Trees, Sky and Soil.


canon rumors FORUM

Re: 46mp sensor useless for landscape?
« Reply #29 on: October 08, 2012, 02:19:13 AM »