Diffraction

[DanLehman]

Yes, the DOF vs diffraction trade-offs are independent of format; you just get a given combination of OOF blurring and diffraction blurring (when viewing at the same size etc.) with a different aperture ratio: from f/64 in 10"x8" to f/32 on 5"x4" to f/8 in 35mm to f/4 in 4/3" to about f/1.2 on a phone.  (And so the shutter speed goes up and/or the ISO speed goes down as the format gets smaller.)  I have not pushed beyond f/11 in 4/3" format, just as I did not go beyond f/22 with 35mm film cameras, but some macros might benefit from f/16 plus suitable sharpening.

Interesting, as Ming Thein has it that diffraction turns (primarily, at least) on pixel pitch (and what is THAT for film?!) --to wit : "diffraction limits vary with pixel pitch, not sensor size".  So, 16mpx APS-C ~= 36mpx 135, 16mpx m4/3 worse.   (Though I recall Ming's asserting that HE could detect diffraction in m4/3 16mpx at f/5.6, I see also him saying that at this same aperture the diffraction "starts to kick in" for the smaller yet higher-resolution 1" sensor'd RX100 --when I'd think one should expect then f/4.)  (My eyes resolve more of these arguments than my brain.)

-d.

[bjanes]

Interesting, as Ming Thein has it that diffraction turns (primarily, at least) on pixel pitch (and what is THAT for film?!) --to wit : "diffraction limits vary with pixel pitch, not sensor size".  So, 16mpx APS-C ~= 36mpx 135, 16mpx m4/3 worse.   (Though I recall Ming's asserting that HE could detect diffraction in m4/3 16mpx at f/5.6, I see also him saying that at this same aperture the diffraction "starts to kick in" for the smaller yet higher-resolution 1" sensor'd RX100 --when I'd think one should expect then f/4.)  (My eyes resolve more of these arguments than my brain.)

I would say that sensor size and pixel pitch both must be considered. A perfect lens will render a point of light (such as a star) as an Airy disc, whose diameter in microns is approximately equal to the reciprocal of the f/number. A binary star or other closely spaced points of light will be projected on the sensor or film as an Airy disc whose diameter is approximately equal to the reciprocal of the f/number. When the discs overlap to a significant degree (defined as the Rayeigh limit) the stars can no longer be resolved. When the Airy disc exceeds about 1.4 to 2 times the pixel pitch, the sensor is diffraction limited. However, its resolution will be no worse than that of a less dense sensor. If you increase the size of the sensor or film, more non-overlapping Airy discs can be accommodated and lead to higher resolution in terms of line pairs per picture height.

Bill

[MattBurt]

I was reading the book Understanding Exposure recently and the author Bryan Peterson uses f/22 on his 35mm camera all the time. His photos look very sharp and he seems to have sold many. Is this just an acceptable amount or would he compensate somehow with better lenses or bodies?
I have seen it (I think) at f/22 on APS-C but haven't really noticed on MF and I don't have anything 35mm besides film now.
Based on this it seems the problem could be overblown for many situations.

[ErikKaffehr]

Hi,

Diffraction is essentially a function of aperture. Pixel pitch has no effect on diffraction at all. Clearly, diffraction will be more noticable with a high resolving sensor than with a low resolving sensor. It will also be more noticeable with a good less than with a less good lens. That is because the more you have the more you loose.

If you have the same lens and shoot say f/11 instead of f/5.6 that may be the optimum aperture on a camera having say 16 or 36 MP the 36 MP camera will always held some advantage over the 16 MP camera.

The easiest way to see this is that each component of the lens/sensor system has an MTF and the MTF of the system is the product of the MTF of it's components.

A high resolution sensor has a higher MTF than a lower resolution sensor. So the system MTF will be higher with the higher resolution sensor. This holds valid until MTF of one component is zero.

Now, if you look at a picture on the screen at actual pixels, the picture from the sensor having smaller pixels will be more magnified than the image from the sensor with the larger pixels.

Now, if we assume a larger sensor and make say an A2-size print than we need less magnification than would be make the same size print from a smaller sensor. A2 is around 16"x23".

• 37x49 would need 11x
• 24x26mm would need 17x magnification
• 4/3 would need 34x magnification

So the diffraction circle would be magnified 11x on the MFD, 17x on the 24x36 mm and 34x on APS-C.

That means that you could shoot f/11 on MFD, f/8 on 24x36 and f/4 on 4/3 For the same diffraction effect in print. On the other hand, those apertures would also give the same depth of field.

I have both MFD and 24-36 mm, my best MFD lenses reach optimum performance at f/5.6 and loose some sharpness stopped down to f/8, but that is more measurable than visible. The best 24x36mm lens I have reaches optimum at f/4.

An important parameter in imaging is sharpening. Digital imaging should always include some sharpening and better sharpening is possible with more pixels. Sensors with less noise also allow more sharpening.

All things equal and ignoring depth of field a larger sensor will always deliver better image quality than a smaller sensor. 

But smaller sensors need less stopping down for a given DoF and a lens designed for a smaller sensor can often have better resolution and MTF than a lens designed for a larger sensor.

Best regards
Erik

Interesting, as Ming Thein has it that diffraction turns (primarily, at least) on pixel pitch (and what is THAT for film?!) --to wit : "diffraction limits vary with pixel pitch, not sensor size".  So, 16mpx APS-C ~= 36mpx 135, 16mpx m4/3 worse.   (Though I recall Ming's asserting that HE could detect diffraction in m4/3 16mpx at f/5.6, I see also him saying that at this same aperture the diffraction "starts to kick in" for the smaller yet higher-resolution 1" sensor'd RX100 --when I'd think one should expect then f/4.)  (My eyes resolve more of these arguments than my brain.)

-d.

[DanLehman]

• 37x49 would need 11x
• 24x26mm would need 17x magnification
• 4/3 would need 34x magnification

So the diffraction circle would be magnified 11x on the MFD, 17x on the 24x36 mm and 24x on APS-C.

That means that you could shoot f/11 on MFD, f/8 on 24x36 and f/4 on APS-C. For the same diffraction effect in print. On the other hand, those apertures would also give the same depth of field.

In my digesting this, I think you slipped from 4/3 to APS-C (and '34' to '24'--though that might be nearly right for APS-C!--, but then f/5.6 re DoF)?

Thanks,
-d.

[ErikKaffehr]

Thanks!

Mistakes when typing, thanks for pointing out the errors.

Best regards
Erik

In my digesting this, I think you slipped from 4/3 to APS-C (and '34' to '24'--though that might be nearly right for APS-C!--, but then f/5.6 re DoF)?

Thanks,
-d.