Pages: [1]   Go Down

Author Topic: Diffraction Limits and the 6 Micron Sensor Sweet Spot  (Read 7788 times)

TomStermitz

  • Newbie
  • *
  • Offline Offline
  • Posts: 6
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« on: May 27, 2010, 12:07:16 pm »

(not to beat a dead horse...)

Digital Camera resolution is limited by Sensor Size, Pixel Size, Lens Diffraction, ISO Sensitivity, and the IR/moire filter at the Sensor and Bayer filter in the software. However the two critical limiting factors are Pixel Size and Lens Diffraction, due to the laws of optical physics. Diffraction causes resolution to degrade for pixels smaller than 6 micron and apertures smaller than f4.

For practical purposes in print, 6 micron pixel size leads to 12 Mpix cameras at crop frame, 18 Mpix (M9) to 24 Mpix (Sony A900) at full Frame, and 40 Mpix at 645 format. Images on the Web are quite fine using 2 micron pixel-size P&S or cell phones.

Film cameras face approximately the same resolution limitations, with grain size in the 2-9 micron range. To complete the picture, the Nikon 9000ED scanner which scans at 4000 dpi works out to 6.4 micron; scanning at 140 dpmm is 7.1 micron.

My D-90 has 4288x2848 5.5 micron pixels. In terms of printing, 300dpi gives me 9.5x14.3 inch (native resolution) landscapes on a Fuji Frontier laserjet. In practice, I can get good prints up to 12x18 using up-rezzing, noise reduction and sharpening, because most people don't walk up to my wall mounts with a magnifying glass.

In terms of ultimate precision, my 5.5 micron pixels already are "blurred" or averaged, due to diffraction, moire filtering, Bayer filter, etc. Of course, my technique is never less than optimal!

I'm sure this isn't new information for anyone, but I'd never framed the problem for myself so concisely. It makes my decision matrix for choosing camera format straightforward:
  • To print full resolution at 12x18, I need FF.
  • To print full resolution at 18x24, I need 645 format.
  • To shoot darker scenes, I need a camera with larger pixel size
  • To maximize my resolution I need to reduce diffraction, which means using apertures at or larger than f4
  • To get more depth of field, I need to use a smaller aperture, which hurts diffraction which means I need larger pixels.
In summary, technology marches, but physics rules. We aren't going to get much higher resolution cameras than 12 Mpix crop-frame or 24 Mpix FF because 6 microns is the limit given diffraction at the f4 aperture. On the upside, technology could give us better ISO (dark performance), or cheaper sensors.

Roger Clark at http://clarkvision.com provides a long, detailed discussion of "Digital Cameras: Does Pixel Size Matter?". But after reading all the spilled pixels, the resolution question is best illustrated by one picture showing Diffraction Effects vs pixel size for various apertures and different cameras.
Logged

TomStermitz

  • Newbie
  • *
  • Offline Offline
  • Posts: 6
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #1 on: May 27, 2010, 12:17:08 pm »

Given all the above, I find the Leica engineering decision-making very optimal.

The M9 has 6.9 micron pixels and 18 Meg of them. Leica has optimized their lenses for use at f2 and better. They took off the anti-aliasing filter to remove that contribution to blurring at the sensor. The only remaining possibility in the FF format would be to change the RGBG Bayer pattern.
Logged

feppe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2906
  • Oh this shows up in here!
    • Harri Jahkola Photography
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #2 on: May 27, 2010, 12:23:48 pm »

Quote from: TomStermitz
The M9 has 6.9 micron pixels and 18 Meg of them. Leica has optimized their lenses for use at f2 and better. They took off the anti-aliasing filter to remove that contribution to blurring at the sensor.

Not to beat this dead horse, but can we get away from the ridiculous notion that camera manufacturers put AA filters to blur the image, rather than to improve it.

JonathanBenoit

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 414
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #3 on: May 27, 2010, 01:35:43 pm »

Quote from: feppe
Not to beat this dead horse, but can we get away from the ridiculous notion that camera manufacturers put AA filters to blur the image, rather than to improve it.

Certainly both with or without have their own tradeoffs. It depends what you shoot where the benefits are. Clearly there is a difference in sharpness. I don't think anyone would debate that.
Before someone says, "sharpen in post". Sharp pixels out of the camera are always better than manipulated ones.
« Last Edit: May 27, 2010, 01:37:26 pm by JonathanBenoit »
Logged

NikoJorj

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1082
    • http://nikojorj.free.fr/
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #4 on: May 27, 2010, 03:55:18 pm »

Quote from: TomStermitz
Diffraction causes resolution to degrade for pixels smaller than 6 micron and apertures smaller than f4.
I wouldn't want to make this thread look like the famous Kagemusha (Nagashino) battle scene, but I'd say that for such a pixel pitch diffraction wouldn't have real-world effects until f/8 or rather f/11, if there is an AA filter...
« Last Edit: May 27, 2010, 03:56:23 pm by NikoJorj »
Logged
Nicolas from Grenoble
A small gallery

TomStermitz

  • Newbie
  • *
  • Offline Offline
  • Posts: 6
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #5 on: May 27, 2010, 04:53:26 pm »

Quote from: NikoJorj
I wouldn't want to make this thread look like the famous Kagemusha (Nagashino) battle scene, but I'd say that for such a pixel pitch diffraction wouldn't have real-world effects until f/8 or rather f/11, if there is an AA filter...

I think if you read Roger Clark's discussion, he agrees with you that AA filter, Bayer filter and diffraction are approximately similar limiting factors somewhere around f/4 to f/8 when the sensor is 6 - 8 Microns. It's probably like adding up "sum of the squares" noise in an audio system. If one factor starts to dominate, you can quickly ignore the others. You don't even have to care about a particular factor until you fix the worst one.

Diffraction doesn't suddenly kick in; contrast slowly degrades until you can't distinguish two lines. So, how far does MTF due to diffraction effects have to drop before you consider the resolution degraded? 30% or 60%? Maybe that means f/8 instead of f/4. See the graphic at ClarkVision,com as it lays the parameters out very well.

Like I said above, while worrying about mega pixels and printing, I didn't realize that the simplest, underlying limitation for camera systems is pixel size. I also found it interesting that each piece of a camera system has been optimized right up to the 6 microns limit, and that applies to film or scanning. Not a coincidence, more of an evolution of optimization.

Also, I realized that there is not much hope or need to wait for higher megapixel cameras in Crop-factor format because 6 microns => 12 Mpix is about the limit unless I can always shoot at f/2 or better. It also clarifies why FF hits 18-24 Mpix and why the Phase or Hassy or Pentax 645D makes sense at 40 Mpix.
Logged

feppe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2906
  • Oh this shows up in here!
    • Harri Jahkola Photography
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #6 on: May 27, 2010, 05:09:42 pm »

While I have little interest on the topic, an alternate view is that maybe 400 megapixels is a more appropriate sweet spot.

elf

  • Full Member
  • ***
  • Offline Offline
  • Posts: 244
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #7 on: May 27, 2010, 09:18:44 pm »

Quote from: feppe
While I have little interest on the topic, an alternate view is that maybe 400 megapixels is a more appropriate sweet spot.

the_online_photographer:Another non-stitcher's view of the world  .  46 gigapixels is latest claimed world record.

OP: Consider what relevence the pixel size is if you downsize a 400megapixel image.  Will diffraction effects still be visible?
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #8 on: May 28, 2010, 03:25:07 pm »

Hi,

My experience/thinking is that:

- Diffraction sets in after f/8 but may not be really relevant before f/16
- Regarding noise/DR the sensor size matters more than the pixel size. A bigger sensor can capture more photons.
- The limitations above are theoretical, i  practice non optimal technique may limit
- I have made some tests using 12.5 MPixel DX and 24.6 MP FX cameras with A2 prints. The difference in prints is much less than I can see on files.

Check this: http://www.pbase.com/ekr/image/107619976/original
and this: http://www.pbase.com/ekr/image/107823207/original

BR
Erik Kaffehr

Quote from: TomStermitz
(not to beat a dead horse...)

Digital Camera resolution is limited by Sensor Size, Pixel Size, Lens Diffraction, ISO Sensitivity, and the IR/moire filter at the Sensor and Bayer filter in the software. However the two critical limiting factors are Pixel Size and Lens Diffraction, due to the laws of optical physics. Diffraction causes resolution to degrade for pixels smaller than 6 micron and apertures smaller than f4.

For practical purposes in print, 6 micron pixel size leads to 12 Mpix cameras at crop frame, 18 Mpix (M9) to 24 Mpix (Sony A900) at full Frame, and 40 Mpix at 645 format. Images on the Web are quite fine using 2 micron pixel-size P&S or cell phones.

Film cameras face approximately the same resolution limitations, with grain size in the 2-9 micron range. To complete the picture, the Nikon 9000ED scanner which scans at 4000 dpi works out to 6.4 micron; scanning at 140 dpmm is 7.1 micron.

My D-90 has 4288x2848 5.5 micron pixels. In terms of printing, 300dpi gives me 9.5x14.3 inch (native resolution) landscapes on a Fuji Frontier laserjet. In practice, I can get good prints up to 12x18 using up-rezzing, noise reduction and sharpening, because most people don't walk up to my wall mounts with a magnifying glass.

In terms of ultimate precision, my 5.5 micron pixels already are "blurred" or averaged, due to diffraction, moire filtering, Bayer filter, etc. Of course, my technique is never less than optimal!

I'm sure this isn't new information for anyone, but I'd never framed the problem for myself so concisely. It makes my decision matrix for choosing camera format straightforward:
  • To print full resolution at 12x18, I need FF.
  • To print full resolution at 18x24, I need 645 format.
  • To shoot darker scenes, I need a camera with larger pixel size
  • To maximize my resolution I need to reduce diffraction, which means using apertures at or larger than f4
  • To get more depth of field, I need to use a smaller aperture, which hurts diffraction which means I need larger pixels.
In summary, technology marches, but physics rules. We aren't going to get much higher resolution cameras than 12 Mpix crop-frame or 24 Mpix FF because 6 microns is the limit given diffraction at the f4 aperture. On the upside, technology could give us better ISO (dark performance), or cheaper sensors.

Roger Clark at http://clarkvision.com provides a long, detailed discussion of "Digital Cameras: Does Pixel Size Matter?". But after reading all the spilled pixels, the resolution question is best illustrated by one picture showing Diffraction Effects vs pixel size for various apertures and different cameras.
Logged
Erik Kaffehr
 

douglasf13

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 547
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #9 on: May 28, 2010, 04:26:21 pm »

Quote from: TomStermitz
(not to beat a dead horse...)

Digital Camera resolution is limited by Sensor Size, Pixel Size, Lens Diffraction, ISO Sensitivity, and the IR/moire filter at the Sensor and Bayer filter in the software. However the two critical limiting factors are Pixel Size and Lens Diffraction, due to the laws of optical physics. Diffraction causes resolution to degrade for pixels smaller than 6 micron and apertures smaller than f4.

For practical purposes in print, 6 micron pixel size leads to 12 Mpix cameras at crop frame, 18 Mpix (M9) to 24 Mpix (Sony A900) at full Frame, and 40 Mpix at 645 format. Images on the Web are quite fine using 2 micron pixel-size P&S or cell phones.

Film cameras face approximately the same resolution limitations, with grain size in the 2-9 micron range. To complete the picture, the Nikon 9000ED scanner which scans at 4000 dpi works out to 6.4 micron; scanning at 140 dpmm is 7.1 micron.

My D-90 has 4288x2848 5.5 micron pixels. In terms of printing, 300dpi gives me 9.5x14.3 inch (native resolution) landscapes on a Fuji Frontier laserjet. In practice, I can get good prints up to 12x18 using up-rezzing, noise reduction and sharpening, because most people don't walk up to my wall mounts with a magnifying glass.

In terms of ultimate precision, my 5.5 micron pixels already are "blurred" or averaged, due to diffraction, moire filtering, Bayer filter, etc. Of course, my technique is never less than optimal!

I'm sure this isn't new information for anyone, but I'd never framed the problem for myself so concisely. It makes my decision matrix for choosing camera format straightforward:
  • To print full resolution at 12x18, I need FF.
  • To print full resolution at 18x24, I need 645 format.
  • To shoot darker scenes, I need a camera with larger pixel size
  • To maximize my resolution I need to reduce diffraction, which means using apertures at or larger than f4
  • To get more depth of field, I need to use a smaller aperture, which hurts diffraction which means I need larger pixels.
In summary, technology marches, but physics rules. We aren't going to get much higher resolution cameras than 12 Mpix crop-frame or 24 Mpix FF because 6 microns is the limit given diffraction at the f4 aperture. On the upside, technology could give us better ISO (dark performance), or cheaper sensors.

Roger Clark at http://clarkvision.com provides a long, detailed discussion of "Digital Cameras: Does Pixel Size Matter?". But after reading all the spilled pixels, the resolution question is best illustrated by one picture showing Diffraction Effects vs pixel size for various apertures and different cameras.

  For those that shoot at wide apertures and/or have AA filters on their sensors, there will still be an advantage to megapixel numbers going up.  I've seen a few calculate what the limits are at f1.4, but I can't remember the number.  Was it 100 megapixels or so??





Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #10 on: June 03, 2010, 04:37:01 pm »

Quote from: NikoJorj
... I'd say that for such a pixel pitch diffraction wouldn't have real-world effects until f/8 or rather f/11, if there is an AA filter...
Yes, real world experience suggests this, from what I have seen, despite theoretical arguments for diffraction effects at lower f-stops.

But anyway, as has been said before, reducing pixel size does not make image resolution worse at any given f-stop, even when diffraction starts to be a significant factor: even at f/11 or f/16, reducing pixel size from 6 microns to 4 microns will increase resolution, not decrease it: all that diffraction does is reduce the amount of resolution improvement that you get from the smaller pixels. So long as there is significant interest in getting high resolution at aperture ratios of f/2.8 and below, pixels sizes can get far below 6 microns before diffraction makes further pixel size reductions pointless.

Also, smaller formats tend to have more use for lower f-stops like f/4, due to getting more DOF at a given f-stop than a larger format. So the "diffraction problem" sets in at smaller pixel sized for smaller formats. The idea that the same 6 micron pixel size is a natural lower limit for a wide range of format sizes does not make sense to me.
Logged

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #11 on: June 03, 2010, 06:16:09 pm »

Quote from: TomStermitz
  • To print full resolution at 12x18, I need FF.
  • To print full resolution at 18x24, I need 645 format.
  • To shoot darker scenes, I need a camera with larger pixel size
  • To maximize my resolution I need to reduce diffraction, which means using apertures at or larger than f4
  • To get more depth of field, I need to use a smaller aperture, which hurts diffraction which means I need larger pixels.

I don't buy the idea that for larger prints you need larger sensors because of diffraction: you are forgetting that to achieve the same DOF, the cropped sensor requieres a wider aperture, so in the end the limitation because of diffraction is independent of the sensor size.

I also disagree that for darker scenes you need larger pixel sizes, you need larger sensors. Smaller pixels mean worse per-pixel SNR, but smaller pixels at a given sensor size mean statistical SNR improvement once the image is rescaled to the desired print size. Just collecting many photons is needed here.

The tradeoff between achievable resolution vs DOF in a single shot, is totally correct, and is theoretically independent of sensor size. With several shots (panoramas and/or focus blending), there are no limitations and any resolution+DOF can be achieved.

Regards
« Last Edit: June 03, 2010, 06:17:27 pm by Guillermo Luijk »
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #12 on: June 03, 2010, 08:04:00 pm »

Quote from: Guillermo Luijk
With several shots (panoramas and/or focus blending), there are no limitations and any resolution+DOF can be achieved.

Indeed. That has been painfully clear since 4x5 days. DoF issues can only be solved with movements or focus stacking.

Besides this discussion is based on the assumption that it is always possible to reach the full potential of an imaging device. We know that this is often not the case due to non optimal focussing,...

So the choice of a system should be based on the resolution that can be achieved in real world circumpstances.

Cheers,
Bernard

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #13 on: June 03, 2010, 08:13:56 pm »

Quote from: feppe
While I have little interest on the topic, an alternate view is that maybe 400 megapixels is a more appropriate sweet spot.

Yep... 400 megapixel is actually a reasonnable target when stitching. I can easily be achieved with 4 rows of 7 images with a 24 megapixel camera.

Some images already fit the bill  



Cheers,
Bernard

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #14 on: June 03, 2010, 08:15:58 pm »

Quote from: douglasf13
For those that shoot at wide apertures and/or have AA filters on their sensors, there will still be an advantage to megapixel numbers going up.  I've seen a few calculate what the limits are at f1.4, but I can't remember the number.  Was it 100 megapixels or so??

That would be a theoretical calculation only taking diffraction into account. We know that real world lenses always show decreasing image quality from 2.8 or so because of technological limitations related to cost/weight constraints. So it is probably more reasonnable to compute max achievable resolution at f2.8.

Cheers,
Bernard

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #15 on: June 07, 2010, 11:39:36 am »

Quote from: TomStermitz
In summary, technology marches, but physics rules. We aren't going to get much higher resolution cameras than 12 Mpix crop-frame or 24 Mpix FF because 6 microns is the limit given diffraction at the f4 aperture. On the upside, technology could give us better ISO (dark performance), or cheaper sensors.

The diffraction limit oft quoted is based on the size of the diffraction blurring, but ignores many other effects in digital capture.  There was an interesting recent thread at DPR started by someone asking whether the recent article on diffraction limits on this site made sense.  The best response was this one by Doug Pardee:

Quote
I've seen this sort of analysis done over and over—Cambridge in Color is often referenced for sensor element size vs. diffraction.

And all of it is balderdash, because it's based on a fallacious assumption: that every Airy disk (or circle of confusion, when discussing DoF) is somehow beautifully centered in the middle of one sensor element or another.

It's really quite straightforward: diffraction, misfocus, optical distortion, low-pass (anti-aliasing) filters, spatial quantization by the sensor elements, and Bayer demosaicing all reduce the sharpness of the captured image by some amount. Improving any one of those will improve the sharpness of the captured image. It's possible that one of the factors is so overwhelming that improving the others gives a "drop in the ocean" effect on sharpness, but the improvement is still there.

And indeed, I went to the DPR review of the G11; their resolution chart measurment is shot at f4, for an Airy disk size of 5.4µ.  The review shows the resolution chart of the G10 as well.  The 1.7µ pixels of the G10 are indeed diffraction limited -- there are about 3 pixels spacing between resolution chart lines at the point where MTF tanks.  And yet, the camera outresolves the G11 and LX3 examples from the same review, by a substantial margin.  When lines get spaced near Nyquist, details are rather sensitive to phase shifts -- whether an object falls on a pixel or in between.  Oversampling fixes that.  Also, though green may be limited at some point, red and blue are more coarsely sampled and so there is yet more to be had in terms of color resolution by higher sampling rates than the conventional wisdom dictates.

Quote
Roger Clark at http://clarkvision.com provides a long, detailed discussion of "Digital Cameras: Does Pixel Size Matter?". But after reading all the spilled pixels, the resolution question is best illustrated by one picture showing Diffraction Effects vs pixel size for various apertures and different cameras.

I think it's misleading to present a diffraction "limit" as some sort of hard wall beyond which there is nothing to be gained.  Yes, there is no MTF to be had beyond the Rayleigh limit, but the interaction of diffraction with the discrete sampling of digital capture means that there are gains to be had from pixels smaller than the Airy disk radius.  See the resolution chart I mentioned at the DPR review of the G11 (BTW, note how color moire is improved by the higher sampling rate of the G10).
Logged
emil

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
Diffraction Limits and the 6 Micron Sensor Sweet Spot
« Reply #16 on: June 09, 2010, 06:10:22 am »

Thanks Emil for this good discussion of the potential benefits of some oversampling: having photosites small enough that even the sparser red and blue photosites are oversampling the signal enough to control moiré and such. And this at all apertures of interest, not just at a somewhat arbitrarily declared "typical" f-stop. And that numerical resolution measures are never hard lines.  I particularly like this part of your quote from Doug Pardee:
Quote from: ejmartin
... When lines get spaced near Nyquist, details are rather sensitive to phase shifts -- whether an object falls on a pixel or in between.  Oversampling fixes that.  Also, though green may be limited at some point, red and blue are more coarsely sampled and so there is yet more to be had in terms of color resolution by higher sampling rates than the conventional wisdom dictates.
Logged
Pages: [1]   Go Up