Pages: [1]   Go Down

Author Topic: Diffraction, pixel density and DOF  (Read 4126 times)

AFairley

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1486
Diffraction, pixel density and DOF
« on: November 26, 2014, 11:18:16 am »

Kirk Tuck posted a blog entry about pixie density, diffraction and DOF which started me thinking.  I am currently shooting a D800e and often stopping down to f11 to get the DOF I need (often f16 would be even better).  Now I am wondering if I would get better overall image sharpness using a 24MP camera at this aperture (less diffraction from the lower density sensor - is this assumption correct?) but not being particularly technical can't puzzle out the answer.

Has anyone done any empirical testing along these lines - both in terms of what one sees pixel peeping and on the print?
Logged

SZRitter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 384
Re: Diffraction, pixel density and DOF
« Reply #1 on: November 26, 2014, 11:28:33 am »

While not backing this up by testing, my understanding was, even when the diffraction set in, your image would still be sharper than the less dense sensor's image at the same aperture.

So, and this isn't backed up by math, but just an example, a 24MP sensor at f16 will never resolve less detail than an 18MP sensor also at f16 (assuming the same lens and sensor size). Diffraction, by my understanding is a product of sensor size, not density. That said, if you pixel peep the 24MP sensor, the evidence of the diffraction is more visible, but that is because you have more resolving power.

So, given this assumption, I usually try to shoot micro 4/3 around f4-f5.6 for maximum sharpness. And anything past f11 is noticeably not as sharp. 645 can be shot much higher (f11 or f16, I think). 35mm should be optimal around f8, but I would need to double check that.

I would add, another way to check is to use two bodies with the same sensor size, lens, and settings, then downsize the larger image to match the smaller sensor's pixel dimensions. They should be at least equal in sharpness, with the downsized image being a bit sharper.
« Last Edit: November 26, 2014, 11:33:50 am by SZRitter »
Logged

AFairley

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1486
Re: Diffraction, pixel density and DOF
« Reply #2 on: November 26, 2014, 11:33:25 am »

Oops, I see I was confusing sensor size with pixel density, so I guess that's a "never mind."   (Blush.)  Thanks.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Diffraction, pixel density and DOF
« Reply #3 on: November 26, 2014, 11:42:12 am »

Hi,

You always get higher quality with smaller pixels. But with smaller pixels you have more to loose.


Check this document: http://www.smt.zeiss.com/C12567A8003B8B6F/EmbedTitelIntern/CLN_31_MTF_en/$File/CLN_MTF_Kurven_2_en.pdf

The discussion you would be interested in is on pages 23-24. The author Dr. Hubert Nasse, he is AFAIK head of the group evaluating lenses at Zeiss.

Best regards
Erik


Kirk Tuck posted a blog entry about pixie density, diffraction and DOF which started me thinking.  I am currently shooting a D800e and often stopping down to f11 to get the DOF I need (often f16 would be even better).  Now I am wondering if I would get better overall image sharpness using a 24MP camera at this aperture (less diffraction from the lower density sensor - is this assumption correct?) but not being particularly technical can't puzzle out the answer.

Has anyone done any empirical testing along these lines - both in terms of what one sees pixel peeping and on the print?
Logged
Erik Kaffehr
 

SZRitter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 384
Re: Diffraction, pixel density and DOF
« Reply #4 on: November 26, 2014, 11:46:05 am »

I believe you are referring to this article, right? http://visualsciencelab.blogspot.com/2014/11/id-like-to-open-floor-to-technical.html

I just read it, and to me it seems like he just didn't take it to the next step, which is where my explanation goes. He left it at an individual pixel level, saying that a pixel will show diffraction effects earlier as your sensor gets more dense. Which is true, sort of.

But, he didn't bother to ask what the overall resolution was at that point compared to the less dense image.

So, you weren't confusing anything, just looking at an article that failed to see the whole picture (no pun intended) in it's scope.
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Diffraction, pixel density and DOF
« Reply #5 on: November 26, 2014, 01:26:29 pm »

Kirk Tuck posted a blog entry about pixie density, diffraction and DOF which started me thinking.  I am currently shooting a D800e and often stopping down to f11 to get the DOF I need (often f16 would be even better).  Now I am wondering if I would get better overall image sharpness using a 24MP camera at this aperture (less diffraction from the lower density sensor - is this assumption correct?) but not being particularly technical can't puzzle out the answer.

Has anyone done any empirical testing along these lines - both in terms of what one sees pixel peeping and on the print?

Here are my results for the Nikon D800e with the Zeiss 135 f/2.0 APO. I used Imatest with focus bracketing to obtain the sharpest image in the brackets. Processing was in Adobe ACR with no sharpening, and MTF50 was calculated with Imatest. In general, one encounters diffraction limits when the Airy disc is about twice the pixel spacing of a Bayer array. The pixel pitch is 4.8 um and the diffraction Airy disc diameters with green light are as follows: f/4, 5.2; f/5.6, 7.2; f/8, 10.3; f/11, 14.2; f/16, 20.7, and f/22, 28.5.

Falloff in resolution with smaller apertures is gradual. Maximum resolution was at f/4, but resolution was maintained reasonably well up to f/8 and f/11 is usable with convolution sharpening. I think these conclusions are similar to what others have noted.



Here are representative images at f/4, f/16, and f/22 unsharpened and sharpened with FocusMagic. The circle was added in Photoshop to show the Nyquist limit.









For a more detailed analysis, see Jim Kasson's simulation with the Otus. Adding sensor resolution always benefits system resolution, but the curve is relatively flat at large pixel sizes.

Regards,

Bill
« Last Edit: November 26, 2014, 05:21:04 pm by bjanes »
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: Diffraction, pixel density and DOF
« Reply #6 on: November 26, 2014, 03:33:57 pm »


For a more detailed analysis, see Jim Kasson's simulation with the Otus. Adding sensor resolution always benefits system resolution, but the curve is relatively flat at large pixel sizes.


Bill, that link goes to a Wikipedia article.  Did you mean this?

http://blog.kasson.com/?p=5905

Or maybe this?

http://blog.kasson.com/?p=6078

Jim

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Diffraction, pixel density and DOF
« Reply #7 on: November 26, 2014, 05:22:57 pm »

Bill, that link goes to a Wikipedia article.  Did you mean this?

http://blog.kasson.com/?p=5905

Or maybe this?

http://blog.kasson.com/?p=6078

Jim

Jim,

Sorry about the link. I was frustrated since the LuLa site would not accept my images and I had to post them as links on Smugmug. I corrected the link.

Bill
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: Diffraction, pixel density and DOF
« Reply #8 on: November 26, 2014, 06:27:02 pm »

Sorry about the link.

No problem, Bill. The take-home lesson is, with contemporary high-quality lenses (some zooms excepted) and APS-C, FF, and MF sensors of today, a finer pixel pitch will result in better print-resolution results at normal to high photographic quality--  SNR's of 10 and above.

Jim

Dave Ellis

  • Full Member
  • ***
  • Offline Offline
  • Posts: 103
Re: Diffraction, pixel density and DOF
« Reply #9 on: November 27, 2014, 01:26:11 am »

Kirk Tuck posted a blog entry about pixie density, diffraction and DOF which started me thinking.  I am currently shooting a D800e and often stopping down to f11 to get the DOF I need (often f16 would be even better).  Now I am wondering if I would get better overall image sharpness using a 24MP camera at this aperture (less diffraction from the lower density sensor - is this assumption correct?) but not being particularly technical can't puzzle out the answer.

Has anyone done any empirical testing along these lines - both in terms of what one sees pixel peeping and on the print?

The posters above have provided some good solid information and I know they have a far more in-depth knowledge of this subject than I. Could I just add my simple summary though

    • The amount of diffraction (that occurs in the image presented to the sensor by the lens) is determined by the aperture setting (or effective aperture for macro work) of the lens and the wavelength but nothing else.

      How we see that image when captured by the sensor is determined by sensor size and sensor density.

      The sensor size determines how much optical reduction occurs between the object and image - the larger the sensor (and image produced by the lens), the less the reduction factor which means that less resolution is required by the lens to resolve a certain level of detail in the object (for the same field of view).

      The higher the pixel density, the less degradation of the image will occur due to the mtf of the sensor.
So, for a given sensor size, higher pixel density is better considering sharpness alone. ie if you are comparing FF cameras, sharpness should be better with the 36 MP sensor than the 24Mp sensor.

Dave
« Last Edit: November 27, 2014, 02:23:57 pm by Dave Ellis »
Logged

Abe R. Ration

  • Newbie
  • *
  • Offline Offline
  • Posts: 24
    • Abe R. Ration's lens and camera blog
Diffraction is an optical phenomenon
« Reply #10 on: November 27, 2014, 04:08:28 am »

Kirk Tuck posted a blog entry about pixie density, diffraction and DOF which started me thinking.  I am currently shooting a D800e and often stopping down to f11 to get the DOF I need (often f16 would be even better).  Now I am wondering if I would get better overall image sharpness using a 24MP camera at this aperture (less diffraction from the lower density sensor - is this assumption correct?) but not being particularly technical can't puzzle out the answer.

No. Diffraction is an optical phenomenon. Camera sensor just samples the image then lens has draws. Regardless of diffraction, the more pixels you have, the higher the resolution, though if the aperture number is very large, the difference can be absolutely insignificant.

There is more on this here on my blog.
Logged
Abe R. Ration
amateur photographer, amateur scientist, amaterur camera buff
http://aberration43mm.wordpress.com/

Abe R. Ration

  • Newbie
  • *
  • Offline Offline
  • Posts: 24
    • Abe R. Ration's lens and camera blog
Re: Diffraction, pixel density and DOF
« Reply #11 on: November 27, 2014, 04:16:41 am »

Here's a small example image on diffraction, deconvolution/sharpening and full frame sensor sampling 24MP. For the last image quick&dirty deconvolution was used.



As diffraction provides well defined blur, it is relatively simple one to try to deconvolve.
Logged
Abe R. Ration
amateur photographer, amateur scientist, amaterur camera buff
http://aberration43mm.wordpress.com/

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Diffraction, pixel density and DOF
« Reply #12 on: November 27, 2014, 08:35:21 am »

Abe,

A very nice demonstration of the utility of deconvolution. Your f/16 image resolves at or very close to Nyquist and the contrast is markedly improved by deconvolution.

For a 24 MP full frame sensor, the Nyquist is 79 lp/mm. From Table 1 of the article by Osuna and García, resolution at f/16 is 93 lp/mm at Rayleigh and this gives adequate information for deconvolution to restore the image, provided that noise is not excessive.

With the 36 MP full frame image at f/22 that I posted previously, Nyquist is 103 lp/mm and diffraction limited resolution at f/22 is 68 lp/mm at Rayleigh and 83 lp/mm at Dawes. Deconvolution restores contrast, but there is insufficient information at Nyquist for restoration, as indicated by failure of the spokes of the star target to extend to the Nyquist circle.

Although deconvolution works its magic at f/16 in your image, would image quality be equal to that obtained at f/4 by an excellent lens?

Bill
Logged

Abe R. Ration

  • Newbie
  • *
  • Offline Offline
  • Posts: 24
    • Abe R. Ration's lens and camera blog
Re: Diffraction, pixel density and DOF
« Reply #13 on: November 27, 2014, 10:56:42 am »

Abe,

A very nice demonstration of the utility of deconvolution. Your f/16 image resolves at or very close to Nyquist and the contrast is markedly improved by deconvolution.

For a 24 MP full frame sensor, the Nyquist is 79 lp/mm. From Table 1 of the article by Osuna and García, resolution at f/16 is 93 lp/mm at Rayleigh and this gives adequate information for deconvolution to restore the image, provided that noise is not excessive.

With the 36 MP full frame image at f/22 that I posted previously, Nyquist is 103 lp/mm and diffraction limited resolution at f/22 is 68 lp/mm at Rayleigh and 83 lp/mm at Dawes. Deconvolution restores contrast, but there is insufficient information at Nyquist for restoration, as indicated by failure of the spokes of the star target to extend to the Nyquist circle.

Although deconvolution works its magic at f/16 in your image, would image quality be equal to that obtained at f/4 by an excellent lens?

Bill
The SNR of the data and the used point spread function and deconvolution algorithm are the variables we need to consider.
In the example I used, if I recall right, simply Lightrooms simple deconvolution (i.e. 100% details), though I might have used Topaz Detail 3's deconvolution instead. I don't know why I didn't make a note of that  :( In principle deconvolving diffracion can be quite simple - it would be simpler still if the aperture were perfcetly circular.
Using a better (i.e. likely slower) deconvolution algorithm with a well defined PSF could do better job with less noise.
In this sample the f/16 (+one set off AA-filters) resoleved to Nyquist for much of the cases inspite of using suboptimal PSF etc. I don't think there is much reason to think why, if SNR is high enough and PSF well defined why one could not get resolving upto Nyquist. At f/4 on APO Summicron only Nyquist can also be reached ;)

But I am sure there are others here with much deeper understanding on this and I'd be delighted to hear of them.


 
Logged
Abe R. Ration
amateur photographer, amateur scientist, amaterur camera buff
http://aberration43mm.wordpress.com/
Pages: [1]   Go Up