Diffraction mitigation on the Canon PowerShot G7 X Mark II

Doug Kerr

Well-known member
On the site Canon Watch there is an interview with the two Canon engineers responsible for the design of the new Canon PowerShot G7 X Mark II.

I found the following comment especially interesting:

Kota Terayama: Correction of diffraction blurring, even for a small aperture, has been added.

We have, I think, tended to think of the deleterious effects of diffraction as being intractable.

It will be interesting the learn how this new feature performs, and perhaps eventually to learn how it works.

Best regards,

Doug

Asher Kelman

OPF Owner/Editor-in-Chief
Do they refer to the overlapping of the "ripples" waves of light due to "scraping" of the wave front on the edge of the aperture, assuming that my description of events is itself a reflection of reality.

Asher

Doug Kerr

Well-known member
Hi, Asher,
Do they refer to the overlapping of the "ripples" waves of light due to "scraping" of the wave front on the edge of the aperture, assuming that my description of events is itself a reflection of reality.

Are you referring to the classical mechanism of diffraction itself?

The diffraction phenomenon results from interference between various portions of the "beam" of light passing through the aperture, leading to different intensity of the "beam" at different angles. When the diameter of the aperture is small, the geometry is such that this difference in intensity with angle is substantial. With a larger aperture, the geometry is such that the phenomenon of differing intensity at different exit angles is less.

I do not consider that the word "scraping" is useful in discussing this phenomenon.

Best regards,

Doug

Doug Kerr

Well-known member
In the following discussion I take certain liberties with the precise facts in the interest of fostering understanding of the mechanism involved in diffraction.

The mechanism of diffraction can be better understood if we note that when the wavefront of a light "beam" passes through an aperture, in effect at each infinitesimal point across the aperture the luminous flux that arrives there in effect powers a tiny "point source". (The wave is in effect "reconstructed" at such a place.) The luminous flux that emerges from each of these point sources spreads out in sort of a cone.

Next, imagine a very tiny aperture. Across it there are just a very few of these infinitesimal point sources, and the "cones" of light they each emit essentially completely overlap in their "wave" natures. They therefore collaborate in their effect, and thus the overall "output" of the aperture is a cone. When (after being converged by the following lens elements) that falls on the film or a sensor, it illuminates a rather large diameter spot. This is diffraction at its "worst".

Next imagine a quite large aperture. Across it there are a large number of these infinitesimal point sources, each emitting a tiny amount of luminous flux in a cone.

In these little cones, flux is of course in the form of a wave. If we consider the parts of these cones that are propagating "straight ahead", their wave natures remain "in phase", and so they cooperate in making a "straight ahead" aspect of the whole beam.

But the, if we consider the parts of all these cones that are propagating at a slight angle to "straight ahead, their wave natures are not quite "in phase". The reason is that the in at any given place in their travel they have traveled different distances from the plane of the aperture, their place of birth (like two runners who start off together at positions along a slanted starting line). Thus those parts of the respective cones "cancel out" to some extent, reducing their collective potency.

As we consider this effect for all the little cones, at each angle from the axis, we find that overall, the result of this is that the preponderance of the luminous flux traveling at an any angle to the axis is canceled out, while the preponderance of the flux traveling "straight ahead" works "in concert".

This almost totally straight-ahead only "beam", after passing through the remaining lens elements, creates almost a point on the film or sensor. This is diffraction in a negligible degree.

The larger the aperture, the more does this happen (the less is the spot on the film "spread" by diffraction). And the smaller the aperture, the closer is the operation to my first scenario, that of an infinitesimal-sized aperture, in which the spot on the film is greatly spread by diffraction.

************

I close by calling attention to a phenomenon that is not an exact parallel but which nevertheless my be of some value to those trying to understand diffraction.

Consider a "dish-type" microwave radio transmitting antenna. The larger its diameter (in fact spoken of as the diameter of its "aperture"), the narrower will be its "beamwidth". A very large diameter antenna will create a beam so narrow that its "footprint" on a distant surface will be very small in diameter. An antenna of lesser diameter will create a wider beam such that its "footprint" on a distant surface will be much larger diameter.

This all happens by the way that the radio waves reflected from different parts of the dish cancel out with regard to their effect at an angle, but "cooperate" with respect to their effect "straight ahead". The larger the diameter of the dish, the more prominent is this phenomenon.

Time now for breakfast.

Best regards,

Doug

Doug Kerr

Well-known member
[Breakfast has been a little delayed.]

In connection with diffraction, it is common to say that it occurs at the "edge" of the aperture (thus the common locution of "scraping" the edge), or the statement that diffraction will occur when a light beam encounters a plate with its edge in the middle of the beam.

That is, the "edge" is accused of creating the diffraction.

It does, but not in the way that the descriptions suggest. The role of the edge is only that it is the beginning of a region through which the light cannot pass.

If we wish, we can consider, at any arbitrary plane in the travel of a light mean, that the beam is there "reconstructed" by the mechanism of an infinity of infinitesimal "point emitters" located across that plane. If we consider the totality of this, we find that all the emissions from the point course not precisely "straight ahead" cancel out in phase, and thus we only have the emissions that travel straight ahead.

In other words, nothing has happened to the beam by passing through this hypothetical array of point sources.

But now consider that, at a certain plane, where we imagine this scenario to exist, there is a straight-edged opaque plate intruding into the beam. In effect, perhaps half of these little "point sources" are disabled.

The result is that the "cancellation and reinforcement" process I described above at some arbitrary plane does not fully take place. The result is that the beam is somewhat "spread out" in the direction perpendicular to the edge of the intruding plate. This is the famous "edge diffraction."

Now, is it the edge of the plate that causes this (as is so often said, even as suggested by the name of the phenomenon)? No. It is the plate itself, which only begins at that edge.

Best regards,

Doug

Doug Kerr

Well-known member
Now, regarding the mitigation of diffraction in a photographic system. Because diffraction is a deterministic process (its effect can be predicted precisely), it would seem that we could counteract the impact of diffraction on an image by deconvolution. That is, if we treat the diffraction mechanism as a spatial filter, we can subject the resulting image to the "inverse filter" to "back out" the effect of diffraction.

So perhaps this is already available to us in various applications that offer deconvolution to counteract such other ills as misfocus blurring. Others here are more familiar with these applications than I.

Here is one interesting discussion of that matter:

In any case, perhaps the Canon PowerShot G7 X Mark II utilizes, internally, such an approach.

Best regards,

Doug

James Newman

Member
And all of this before breakfast. Very fascinating. I always read your technically informative pieces, usually numerous times. I find it interesting just to see how much of it I can actually understand. Unfortunately that is most often, not too much. It does, however, make me want to look into it further and see if a can understand it better. Thank you for that.
James

Doug Kerr

Well-known member
Hi, James,

And all of this before breakfast. Very fascinating. I always read your technically informative pieces, usually numerous times. I find it interesting just to see how much of it I can actually understand. Unfortunately that is most often, not too much. It does, however, make me want to look into it further and see if a can understand it better. Thank you for that.

Thank you so much.

Best regards,

Doug

Doug Kerr

Well-known member
I said:

So perhaps this is already available to us in various applications that offer deconvolution to counteract such other ills as misfocus blurring. Others here are more familiar with these applications than I.

It turns out that (in 2010) I began a discussion here of a simulation done by Bart van der Wolf of that very possibility:

The thread also contain some follow-on work done by (then) new member David Ellsworth.

Best regards,

Doug

Jerome Marot

Well-known member
Deconvolution works best, but simple sharpening works almost as well and is considerably less processor intensive. Maybe Canon uses that instead.

Doug Kerr

Well-known member
This paper:

is also of interest in connection with this general matter.

Best regards,

Doug

Doug Kerr

Well-known member
Hi, Jerome,

Deconvolution works best, but simple sharpening works almost as well and is considerably less processor intensive. Maybe Canon uses that instead.

Well, that might well be.

It may be that the in-camera sharpening customarily used is effective in mitigating the effects of diffraction, and Canon has just decided to "market" that.

Best regards,

Doug

Asher Kelman

OPF Owner/Editor-in-Chief
In the following discussion I take certain liberties with the precise facts in the interest of fostering understanding of the mechanism involved in diffraction.

The mechanism of diffraction can be better understood if we note that when the wavefront of a light "beam" passes through an aperture, in effect at each infinitesimal point across the aperture the luminous flux that arrives there in effect powers a tiny "point source". (The wave is in effect "reconstructed" at such a place.) The luminous flux that emerges from each of these point sources spreads out in sort of a cone.

That's the sense for the metaphor of scraping of the light waves against the edge.

So what is happening on a physics stand point. We have a light wave part of which is very close to a thin metal surface edge. What forces are at work on the atomic level? What is the sequence of interactions between the wave/photon package and the surface.

Asher

Michael Nagel

Active member
Kota Terayama: Correction of diffraction blurring, even for a small aperture, has been added.

I think that there are some E-M1 users here, who are quite familiar with diffraction compensation.

What's the difference in the approach used for the G7 X Mark II?

Best regards,
Michael

Doug Kerr

Well-known member
Hi, Michael,

I think that there are some E-M1 users here, who are quite familiar with diffraction compensation.

Thank you for that link. I was not aware of that feature.

What's the difference in the approach used for the G7 X Mark II?

I don't yet have any info on what is done there. All I have is that quotation that calls attention to the feature.

Best regards,

Doug

Doug Kerr

Well-known member
This morning I read (rather superficially, I must confess) the extensive review of the Canon EOS 5Ds R body (don'tcha just love Canon's "scheme" of notation - there is method to the madness, but madness to that method), by Dustin Abbot, published on Canon Rumors. This body has a native resolution on the order of 50 Mpx, and (effectively) no antialising lowpass filter.

There was an interesting observation related to the matter of the impact of diffraction, namely that diffraction "set in earlier" [as we decreased aperture] than on predecessor bodies with smaller pixel count (and thus, because all the bodies being compared had "full-frame 35-mm" size sensors, predecessor bodies with greater pixel pitch). This was pointed out as one of the "cons" a photographer might consider when contemplating purchasing a 5Ds R.

How does this happen? Well, it merely means that the greater resolution of this body means that a very "small" impact of diffraction (i.e., a small diameter blur figure) can be perceived in the image delivered by the 5Ds R, whereas the effects of such a small blur figure wouild not be [so] visible on a lower-resolution image.

So to say that this is a "disadvantage" of the 5Ds R is a little bit like saying, "No, we never reconcile the company's books to pennies; that often gives us a bad result", meaning, "when we do, they more often fail to balance."

But it certainly is, as it is popular to say these days, "a thing".

Best regards,

Doug