Pages: 1 [2]   Go Down

Author Topic: Is the "MP race" a good thing?  (Read 9857 times)

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8914
Re: Is the "MP race" a good thing?
« Reply #20 on: August 22, 2016, 04:31:39 pm »

Hi Bart,

Could you elaborate on that one?

Sure. Microlenses refract their part of the cone of light onto the light sensitive part of the individual sensor elements, thus creating fewer wasted photons that might fall on transistors, 'wiring' (gates, connectors), or even spill to the neighboring sensels. If the angle of incident light gets too oblique, they can no longer focus on the correct part of the sensel. For lenses with a short distance between the exit pupil and the sensor, those angles are already oblique near the corners of full frame sensors. That's why the physical arrangement of the microlenses is offset a bit in modern designs, progressively more offset (less centered) towards the corners of the sensor (no longer centered above the sensel).

With such an optimized system, decentering or shifting will mis-focus the image rays and reduce the efficiency (leading to vignetting, rippling, and other artifacts).

When gapless microlenses are used, the apparent fill-factor of the sensel is increased, which not only increases the amount of light that is utilized, but the sensor also starts behaving more like an area sampling device, instead of a point sampling device. That improves its MTF roll-off towards the Nyquist frequency, and thus also reduces aliasing. The recorded signal is more true to the input signal.

Quote
Another point:

Let's say that I want a world class lens, like the Otus 85/1.4. For me it would make some sense to have a world class sensor like the Canon 5DsR or the Sony A7rII, rather than a sensor of more limited resolution like the A7II.

With the A7rII is still see some colour aliasing artefacts even with middle of the road lenses. So it is in no way immune to aliasing. With P45+ it is more of a generic problem. One of the worst issues I have seen are rippled water surfaces. But, I would say that issues with the A7rII are almost a magnitude less than with the P45+.

Yes, the OTUS and Sigma ART lenses will create a high contrast low residual aberration input signal at wider apertures. Narrower apertures will have diffraction taking over, and that somewhat reduces the differences with lesser lenses.

The sensel pitch differences (denser sampling) of the A7rII versus the large P45+ sensel pitch will reduce the aliasing, because the Nyquist frequency of the A7rII is much higher, and there will be fewer small details that will cause aliasing, if the same focal length is used. If a shorter focal length is used, then the image magnification is reduced and there will be an increase of small detail in the input projection, maybe cancelling the benefit to some degree, but diffraction per pixel can be more (depending on aperture number used), acting a bit like an (inefficient) AA-filter.

Quote
My guess is that fill factor plays a very major role, so modern sensors with large fill factors are less prone to aliasing than older sensors with a smaller fill factor.

That's correct.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Is the "MP race" a good thing?
« Reply #21 on: August 22, 2016, 04:33:00 pm »

Thanks!

Erik

I'll prove my case with my Sony sensor 2006 6MP Pentax K100D and 18-55mm $80 kit lens with screenshot courthouse below. Both shot Raw taken at slightly different angles, distances and focal length. The one on the left was taken at 50mm, the one on the right 55mm. This is real world shooting situation involving looking and adjusting to fit subject in frame as any photographer would employ. Both were processed and sharpened in CS5/ACR 6.7.
Logged
Erik Kaffehr
 

Telecaster

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3686
Re: Is the "MP race" a good thing?
« Reply #22 on: August 22, 2016, 05:27:20 pm »

I have no problem with collecting as much info as I can, then throwing away what I don't need.

4096x3072 images viewed on my iPad (at 2048x1536 pixels, its native res) look more coherent than 2048x1536 images. I was surprised when I first noticed this but it's been borne out over time. The same is true, to a lesser extent, when viewing 8K images on my 4K TV.

Also, 4K video downsampled to 1080p and viewed on my 4K TV looks more real to me than when viewed at full res. I've never seen 4K video viewed at full res that doesn't look like a CGI construct.

-Dave-
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Is the "MP race" a good thing?
« Reply #23 on: August 23, 2016, 12:17:33 am »

Hi Bart,

My question was about the negative influence of the AA-filter, which is not clear to me.

The issues you mention with the microlenses is something I am aware of, but thanks for the explanation anyway. Your explanations are always a good read.

Best regards
Erik


Sure. Microlenses refract their part of the cone of light onto the light sensitive part of the individual sensor elements, thus creating fewer wasted photons that might fall on transistors, 'wiring' (gates, connectors), or even spill to the neighboring sensels. If the angle of incident light gets too oblique, they can no longer focus on the correct part of the sensel. For lenses with a short distance between the exit pupil and the sensor, those angles are already oblique near the corners of full frame sensors. That's why the physical arrangement of the microlenses is offset a bit in modern designs, progressively more offset (less centered) towards the corners of the sensor (no longer centered above the sensel).

Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Is the "MP race" a good thing?
« Reply #24 on: August 23, 2016, 12:31:48 am »

Hi Bart,

It is a bit of my experience that if there is a lot of aliasing it doesn't matter a lot if I use Capture One or LR. C1 does do a better job on keeping away aliasing but it cannot work miracles.

Both C1 and LR have "moiré reduction filters". The one in LR is more heavy handed. So demosaic certainly is a win for Capture One.

With the P45+, aliasing issues are common. About the worst offender is rippled water. The question is how disturbing these things are in prints.

I would assume that an OLP filter that is designed for the job is more efficient in holding colour artefacts at bay than stopping down and having diffraction doing the job. Just to say, I see less aliasing on the A7rII than I was expecting.

Best regards
Erik

Yes, its resolution is a bit less of a mess, but indeed, nothing beats more high quality pixels for a cleaner result.

In fact, such fat pixels could use an AA-filter, but they are usually unavailable in such large sizes (and counterproductive for large amounts of lens shift). With more and denser sampling sensels, the need for using AA-filters is reduced, or the havoc that an unfiltered sensor wreaks is less obtrusive (due to diffraction and the presence of fewer smaller features).

Cheers,
Bart
Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Is the "MP race" a good thing?
« Reply #25 on: August 31, 2016, 11:37:44 am »

Hi Bart,

A few years ago there was a very good posting by "Chrismuc" who shot the same studio setup with four widely different cameras, Contax ND (6MP full frame OLP), Leica DMR (10 MP, 1.3X crop non OLP), Canon D5II (21 MP, full frame, OLP) and Phase One I 180 (80 MP, FF 645 non OLP), see that thread here: http://forum.luminous-landscape.com/index.php?topic=73668.0

At that time I have found that this image was smooth and natural: http://forum.luminous-landscape.com/index.php?action=dlattach;topic=73668.0;attach=72119;image while this one was coarse: http://forum.luminous-landscape.com/index.php?action=dlattach;topic=73668.0;attach=72119;image

Both these images were downscaled to 6MP, by the author.

If you check the attached screen dump I think that A.jpg stands out, while the other three have a much less articulated and more natural texture. What is your take?

A) Leica DMR
B) Iq 180
C) Contax N
D) Canon 5DII

Best regards
Erik








What's often underestimated, is that with proper software (e.g. Topaz 'Detail') we can boost the visibility of the finest detail that our eyes can see (with low contrast), and that will add a lot of 'realism' to surface structure. We will be able to more clearly see what that stuff is made of; e.g. is it cardboard or is it leather?

In many art objects, surface structure and material choice are part of the expression of the creative intent, and I think photography is no exception. Of course, like in painting, there are impressionistic styles and there are realistic styles, but without enough pixels, realistic becomes hard if at all possible.

Cheers,
Bart
Logged
Erik Kaffehr
 

Telecaster

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3686
Re: Is the "MP race" a good thing?
« Reply #26 on: August 31, 2016, 03:18:15 pm »

My takeaway from the two 6mp images is that the Contax G2 system, occasional AF hiccups aside, was very nice indeed.  :D  Lovely lenses, including the prettiest rendering 28mm I've ever owned.

-Dave-
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Is the "MP race" a good thing?
« Reply #27 on: September 02, 2016, 03:10:22 pm »

Hi,

The Contax here was a 6 MP DSLR. All the images were scaled to 6MP.

The key issue may have been that the Contax had a correct OLP filter and thus could achieve a smooth reproduction, albeit probably giving up a bit on micro contrast in spite of it's large pixels.

For me, the Leica DMR image was far to "gritty".

Both the I-180 and he Canon had very smooth rendition. Showing the images at 6MP resolution obviously removed the greatest advantage of the 80 MP back.

Some of the Loxia lenses for Sony are derived from Contax G system designs, but redesigned taking the cover glass into account.

Best regards
Erik


My takeaway from the two 6mp images is that the Contax G2 system, occasional AF hiccups aside, was very nice indeed.  :D  Lovely lenses, including the prettiest rendering 28mm I've ever owned.

-Dave-
Logged
Erik Kaffehr
 

Telecaster

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3686
Re: Is the "MP race" a good thing?
« Reply #28 on: September 02, 2016, 03:24:24 pm »

Erik, the camera being photographed, rather than the one doing the photographing, is a G2. I was just indulging in a bit of nostalgia. :) As for the N Digital, I remember at the time being impressed by its smooth tonality. I think it had a "low ISO" (like 32 or even 25) mode that multisampled.

I like the later M-mount Zeisses, and what I've seen of the Loxias too, but I don't feel the ZM 28mm has the same rendering as the G version.

-Dave-
« Last Edit: September 02, 2016, 03:29:15 pm by Telecaster »
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Is the "MP race" a good thing?
« Reply #29 on: September 02, 2016, 03:49:17 pm »

Just to say, I have bought a couple of Contax Y/C mount lenses to use on the A7rII. The main reason I bought them was that they have manual aperture and give me tilts on the HCam TSII/A7rII combo. Both are zooms, a 28-85/3.3-4 and a 35-135/3.3-4.5.

I like those lenses very much.

Best regards
Erik


Erik, the camera being photographed, rather than the one doing the photographing, is a G2. I was just indulging in a bit of nostalgia. :) As for the N Digital, I remember at the time being impressed by its smooth tonality. I think it had a "low ISO" (like 32 or even 25) mode that multisampled.

I like the later M-mount Zeisses, and what I've seen of the Loxias too, but I don't feel the ZM 28mm has the same rendering as the G version.

-Dave-
Logged
Erik Kaffehr
 
Pages: 1 [2]   Go Up