Relationship between LoCA and focus shift?

Yesterday, I tested the Coastal 60/4 at 1:2 magnification for longitudinal chromatic aberration (LoCA) and focus shift.

I found the LoCA outstanding:


Subject distance vs raw channel MTF50 in cycles/picture height

And the focus shift mediocre:


Green channel MTF50 in cy/ph vs subject distance

I measured the Otus 85/1.4 a month or so ago, and found the same thing.

At 1:2 the Sony 90/2.8 macro has modest LoCA:



But little focus shift:



So here's my question for you optics experts out there. Is there something about reducing LoCA when designing a lens that increases focus shift?

Details on the testing methodology: http://blog.kasson.com/?p=14682

Thanks,

Jim

--

"Is there something about reducing LoCA when designing a lens that increases focus shift?"

Interesting question.

Focus shift depends on spherical aberration.

Longitudinal chromatic aberration depends of choices of element powers and the dispersion characteristics of the glasses chosen.

So in principle those two items should be quite independent of one another.

If they are related, it will be through some kind of convoluted interaction of elements in a particular design. Like a glass of a desirable dispersion characteristic being too fragile to withstand aspheric grinding or too hard to aspherize quickly with MRH (magnetorheological machining). Just throwing that out there as an example of annoying things that sometimes come up and ruin things.

OpticsEngineer said:

"Is there something about reducing LoCA when designing a lens that increases focus shift?"

Interesting question.

Focus shift depends on spherical aberration.

Longitudinal chromatic aberration depends of choices of element powers and the dispersion characteristics of the glasses chosen.

So in principle those two items should be quite independent of one another.

If they are related, it will be through some kind of convoluted interaction of elements in a particular design. Like a glass of a desirable dispersion characteristic being too fragile to withstand aspheric grinding or too hard to aspherize quickly with MRH (magnetorheological machining). Just throwing that out there as an example of annoying things that sometimes come up and ruin things.

Click to expand...

Thank you.

Can you think of a consumer lens that has LoCA as good as the Otus 85/1.4 that has negligible focus shift? I'm going to assume that the list of consumer lenses that have LoCA as good as the Coastal 60/4 (which some would not call a consumer lens) is zero items long, but let me know if that's not the case, too, please.

You are always helpful on this forum, and I'm grateful for that.

Jim

--
http://blog.kasson.com

A lens system with very well controlled LoCA may have quite a lot of focus shift (spherical aberration). Or not... Either can hold true even in a very sharp lens at larger apertures (SA<>LoCA is a weak and secondary coupled aberration pair...). This is true for most of the image field, but not necessarily for the outer angles of the image projection (due to other practical reasons).

So to answer your actual question - no, they're not related in that way, especially not in a way that would make decreasing one inherently increase the other. Making construction/manufacturing cost compromises though, will have that effect. As will any other trade-off for that matter. Better LoCA = worse barrel finish? Cheaper lens hood?Maybe... In this case, Brian seems to have traded some SA performace to be able to increase wideband LoCA performance.

.....

But you have to remember to mentally weigh in how much (little) the light transmitted through the lens at smaller apertures really affect the larger aperture tests.

Your f/11 test is 3 full stops down, which means that the light through the f/11 opening constitutes the central 11.25% area of the light bundle through the wide open f/4 condition. Since the shift is "not large", the real effect is much smaller than that worst case scenario. Your real losses at f/4 are theoretically (again, quite simplified...) about 12.5% of the MTF loss of the f/11 condition at the "best" f/4 condition focal distance. Say you get an MTF0.7 factor here, and then you lower the real impact of this loss by a factor of eight (three full stops) - this becomes an MTF reduction factor of 0.95, or a 5 percent loss in MTF50 lp/mm. I.e - the lens COULD have been 5% "sharper" wide open if the SA wasn't there. Most probably not even close to detectable in visual ABX tests on real images.

If this is hard to visualize, then think about what would happen if you could place a light shield that's the same size as the f/11 opening in the center of the f/4 opening - when shooting at f/4. What would happen to the shot? Well, the T-stop would increase by 0.2, and OOF blur discs would look funky - but except for that? In the region of best focus -Not much...

The apertures that REALLY take a hit here are actually the f/5.6 and f/8 stops, since the first and second derivate of the SA line start out higher here than at f/4. At f/11, the SA loss starts to get overpowered by diffraction, so again the losses get drowned out by other, higher power contrast losses. I'm still only talking about the paraxial here, not the field - where a model analysis gets a lot more complicated rather quickly.

.......

BTW, I find it strange that you get so much higher lp/mm MTF50's at f/8 than at f/5.6. And the numbers seem quite low overall... Measurement artifact? I've never tested it at 1:2 though... Both on test sensors and on a bench I got significantly better results at 5.6 than at 8, both at real infinity (Zeiss K-8) and at 1:40 (which was my normal on-camera test distance). I would not recommend using it over the specified 1:4 magnification, that's where the floating element's effective working range stops as far as I remember from talking to Brian earlier.. I think he's a member here, so maybe ask him about the near-macro performance? Also, the lack of through-focus MTF50 curve flattening (DoF increase) looks suspicious. Your DoF at f/4 and f/8 look suspiciously similar...

"Can you think of a consumer lens that has LoCA as good as the Otus 85/1.4 that has negligible focus shift?"

I don't know of one. But if I were trying to track down a consumer lens of similar capability, I would ask Roger Cicala at LenRentals. I see they carry the Coastal 60/4 so he probably has an idea of similar lenses. Maybe some Schneider lenses for instance.

The_Suede said:

BTW, I find it strange that you get so much higher lp/mm MTF50's at f/8 than at f/5.6. And the numbers seem quite low overall... Measurement artifact?

I've never tested it at 1:2 though... Both on test sensors and on a bench I got significantly better results at 5.6 than at 8, both at real infinity (Zeiss K-8) and at 1:40 (which was my normal on-camera test distance). I would not recommend using it over the specified 1:4 magnification, that's where the floating element's effective working range stops as far as I remember from talking to Brian earlier..

Click to expand...

Just plugging the numbers into VWDOF at those two magnifications (0.5 vs 0.025) at an assumed pupil ratio of 1.0 yields an effective aperture of f/8.4 at 1:2 versus an effective aperture of f/5.74 at 1:40.

Plugging those effective aperture numbers into the diffraction+photosite aperture model, at a wavelength of 550 nm and a photosite pitch of 4.5 micron (assumed 100% fill factor square photosite), gives me the following numbers:

at 1:2, MTF50 is about 0.317 cycles/pixel, or 1684 line pairs / picture height, and

at 1:40, MTF50 is about 0.393 cycles/pixel, or 2088 line pairs / picture height.

This is a 20% drop in linear resolution, so that would explain part of the apparently low figures Jim measured.

The rest of the difference (measured ~1100 lp/ph vs model prediction of 1684 lp/ph) would require further explanation. Would it be fair to say that there is enough Spherical Aberration left at f/5.6 to reduce the resolution? Or is it possible that the actual aperture is smaller than f/5.6 (even before applying the correction for the 1:2 magnification) ?

I also see that Jim measured about 1600 lp/ph at f/5.6 using the Sony 90 mm f/4 lens at 1:2 magnification, which is pretty close to the 1684 predicted by the simple model. (Again, assuming a pupil ratio of 1.0).

fvdbergh2501 said:

The_Suede said:

BTW, I find it strange that you get so much higher lp/mm MTF50's at f/8 than at f/5.6. And the numbers seem quite low overall... Measurement artifact?

I've never tested it at 1:2 though... Both on test sensors and on a bench I got significantly better results at 5.6 than at 8, both at real infinity (Zeiss K-8) and at 1:40 (which was my normal on-camera test distance). I would not recommend using it over the specified 1:4 magnification, that's where the floating element's effective working range stops as far as I remember from talking to Brian earlier..

Click to expand...

Just plugging the numbers into VWDOF at those two magnifications (0.5 vs 0.025) at an assumed pupil ratio of 1.0 yields an effective aperture of f/8.4 at 1:2 versus an effective aperture of f/5.74 at 1:40.

Plugging those effective aperture numbers into the diffraction+photosite aperture model, at a wavelength of 550 nm and a photosite pitch of 4.5 micron (assumed 100% fill factor square photosite), gives me the following numbers:

at 1:2, MTF50 is about 0.317 cycles/pixel, or 1684 line pairs / picture height, and

at 1:40, MTF50 is about 0.393 cycles/pixel, or 2088 line pairs / picture height.

This is a 20% drop in linear resolution, so that would explain part of the apparently low figures Jim measured.

The rest of the difference (measured ~1100 lp/ph vs model prediction of 1684 lp/ph) would require further explanation. Would it be fair to say that there is enough Spherical Aberration left at f/5.6 to reduce the resolution? Or is it possible that the actual aperture is smaller than f/5.6 (even before applying the correction for the 1:2 magnification) ?

I also see that Jim measured about 1600 lp/ph at f/5.6 using the Sony 90 mm f/4 lens at 1:2 magnification, which is pretty close to the 1684 predicted by the simple model. (Again, assuming a pupil ratio of 1.0).

Click to expand...

Same thing, but evaluating the measured (Coastal Optics) vs simulated results at f/8, using the effective aperture at f/8, which turns out to be f/12 at 1:2, and f/8.2 at 1:40 (according to VWDOF):

at 1:2, model MTF50 is about 0.245, or 1301 lp/ph, and

at 1:40, model MTF50 is about 0.322, or 1711 lp/ph.

This means that Jim's measurement at f/8 (green channel), coming in at just over 1400 lp/ph, is a bit higher than the model result. MTF Mapper's measurement error should be below 5% under those conditions (typically with a tendency to underestimate, assuming we are talking about version 0.5.1), which puts the measured value of 1400 lp/ph outside of the expected measurement error.

Again, I wonder what the true (geometric) aperture of the Coastal Optics lens is when set to f/8; if the real aperture is slightly larger than indicated (say, around f/7.25 in stead of dead on f/8), it could explain the higher-than-expected MTF50 measured at f/8.

fvdbergh2501 said:

fvdbergh2501 said:

The_Suede said:

BTW, I find it strange that you get so much higher lp/mm MTF50's at f/8 than at f/5.6. And the numbers seem quite low overall... Measurement artifact?

I've never tested it at 1:2 though... Both on test sensors and on a bench I got significantly better results at 5.6 than at 8, both at real infinity (Zeiss K-8) and at 1:40 (which was my normal on-camera test distance). I would not recommend using it over the specified 1:4 magnification, that's where the floating element's effective working range stops as far as I remember from talking to Brian earlier..

Click to expand...

Just plugging the numbers into VWDOF at those two magnifications (0.5 vs 0.025) at an assumed pupil ratio of 1.0 yields an effective aperture of f/8.4 at 1:2 versus an effective aperture of f/5.74 at 1:40.

Plugging those effective aperture numbers into the diffraction+photosite aperture model, at a wavelength of 550 nm and a photosite pitch of 4.5 micron (assumed 100% fill factor square photosite), gives me the following numbers:

at 1:2, MTF50 is about 0.317 cycles/pixel, or 1684 line pairs / picture height, and

at 1:40, MTF50 is about 0.393 cycles/pixel, or 2088 line pairs / picture height.

This is a 20% drop in linear resolution, so that would explain part of the apparently low figures Jim measured.

The rest of the difference (measured ~1100 lp/ph vs model prediction of 1684 lp/ph) would require further explanation. Would it be fair to say that there is enough Spherical Aberration left at f/5.6 to reduce the resolution? Or is it possible that the actual aperture is smaller than f/5.6 (even before applying the correction for the 1:2 magnification) ?

I also see that Jim measured about 1600 lp/ph at f/5.6 using the Sony 90 mm f/4 lens at 1:2 magnification, which is pretty close to the 1684 predicted by the simple model. (Again, assuming a pupil ratio of 1.0).

Click to expand...

Same thing, but evaluating the measured (Coastal Optics) vs simulated results at f/8, using the effective aperture at f/8, which turns out to be f/12 at 1:2, and f/8.2 at 1:40 (according to VWDOF):

at 1:2, model MTF50 is about 0.245, or 1301 lp/ph, and

at 1:40, model MTF50 is about 0.322, or 1711 lp/ph.

This means that Jim's measurement at f/8 (green channel), coming in at just over 1400 lp/ph, is a bit higher than the model result. MTF Mapper's measurement error should be below 5% under those conditions (typically with a tendency to underestimate, assuming we are talking about version 0.5.1), which puts the measured value of 1400 lp/ph outside of the expected measurement error.

Again, I wonder what the true (geometric) aperture of the Coastal Optics lens is when set to f/8; if the real aperture is slightly larger than indicated (say, around f/7.25 in stead of dead on f/8), it could explain the higher-than-expected MTF50 measured at f/8.

Click to expand...

Yes, very true - I didn't even weigh in the geometric aperture change for the high magnification. IMO this makes the measurement series even more strange. Especially since another known good (the Sony lens) shows a behavior that matches my expectations from earlier experiences quite closely.

Since I don't know the working method of the floating element in the CO, the real aperture is an unknown (even if we think we know the reproduction ratio is 1:2).

The SA effects on paraxial WO MTF can be modeled accurately if you have the SA line, but it takes a bit of effort. You have to integrate the CoC's given for a fixed image projection distance.

You can however approximate it by assuming a second order line and taking two or three aperture samples, as Jim did. Take the CoC profile from the best focus distance from the WO condition at the reference distance. Do a linear subtraction of the [CoC]x[defocus blur] weighted by total light transmission you get from the higher number apertures at the same distance. Since the higher apertures should have a wider and flatter CoC, the subtraction has a smaller relative effect on the WO CoC peak than the WO CoC field, effectively sharpening the CoC by convolution. The difference is (approximately) the SA loss. The easiest way to do this is to calculate the full MTF of each aperture setting for the distance set, and model the CoC from the MTF.

The method is however very susceptible to accuracy losses though, normally you overestimate the SA loss. The measurements at higher aperture numbers are diffracted, where the real WO ray integration is not - and the diffraction widens the CoC profile of the subtracted CoC's, sharpening the baseline more than the theoretically needed amount. Additionally, the SA line is most often a third or fourth power line, so it doesn't behave as smoothly as our approximation.

fvdbergh2501 said:

The_Suede said:

BTW, I find it strange that you get so much higher lp/mm MTF50's at f/8 than at f/5.6. And the numbers seem quite low overall... Measurement artifact?

I've never tested it at 1:2 though... Both on test sensors and on a bench I got significantly better results at 5.6 than at 8, both at real infinity (Zeiss K-8) and at 1:40 (which was my normal on-camera test distance). I would not recommend using it over the specified 1:4 magnification, that's where the floating element's effective working range stops as far as I remember from talking to Brian earlier..

Click to expand...

Just plugging the numbers into VWDOF at those two magnifications (0.5 vs 0.025) at an assumed pupil ratio of 1.0 yields an effective aperture of f/8.4 at 1:2 versus an effective aperture of f/5.74 at 1:40.

Plugging those effective aperture numbers into the diffraction+photosite aperture model, at a wavelength of 550 nm and a photosite pitch of 4.5 micron (assumed 100% fill factor square photosite), gives me the following numbers:

at 1:2, MTF50 is about 0.317 cycles/pixel, or 1684 line pairs / picture height, and

at 1:40, MTF50 is about 0.393 cycles/pixel, or 2088 line pairs / picture height.

This is a 20% drop in linear resolution, so that would explain part of the apparently low figures Jim measured.

The rest of the difference (measured ~1100 lp/ph vs model prediction of 1684 lp/ph) would require further explanation. Would it be fair to say that there is enough Spherical Aberration left at f/5.6 to reduce the resolution? Or is it possible that the actual aperture is smaller than f/5.6 (even before applying the correction for the 1:2 magnification) ?

I also see that Jim measured about 1600 lp/ph at f/5.6 using the Sony 90 mm f/4 lens at 1:2 magnification, which is pretty close to the 1684 predicted by the simple model. (Again, assuming a pupil ratio of 1.0).

Click to expand...

I would expect lower than theoretical errors if the lens axis were not exactly aligned orthogonal to the razor blade. Is that right?

I can test this by rerunning the Coastal curves with the same 800 sample exposures and a smaller ROI. If there is misalignment, that should produce higher MTF50s, right?

Jim

--
http://blog.kasson.com

JimKasson said:

I would expect lower than theoretical errors if the lens axis were not exactly aligned orthogonal to the razor blade. Is that right?

I can test this by rerunning the Coastal curves with the same 800 sample exposures and a smaller ROI. If there is misalignment, that should produce higher MTF50s, right?

Click to expand...

That's not it. I reduced the ROI from 600x400 to 300x200 and the MTF50 numbers went up by about 2%.

Jim

JimKasson said:

JimKasson said:

I would expect lower than theoretical errors if the lens axis were not exactly aligned orthogonal to the razor blade. Is that right?

I can test this by rerunning the Coastal curves with the same 800 sample exposures and a smaller ROI. If there is misalignment, that should produce higher MTF50s, right?

Click to expand...

That's not it. I reduced the ROI from 600x400 to 300x200 and the MTF50 numbers went up by about 2%.

Jim

--
http://blog.kasson.com

Click to expand...

If you reduced the ROI and the numbers went up it can mean one of two things: 1) lens distortion; 2) noise. In the second case if you take a few captures and take the mean of their results they should average themselves out. In the first case, not.

Jack

Is this about Longitudinal Chromatic Aberration or normal Chromatic Aberration?

As far as I understand it, you measure the focus shift between the various colors (Chart 1 and 3). This says something about the focal length for these colors. But LoCA is a color dispersion that takes place in the blur in front of and behind the focus point even if all the colors come nearly together in that focus point.

See an example here (almost below)

I'm no a authority in optics, but have tried to draw what is the difference to me.



I think what you measure is not a value of Longitudinal Chromatic Aberration.

snellius said:

Is this about Longitudinal Chromatic Aberration or normal Chromatic Aberration?

Click to expand...

LoCA. What's "normal" CA? LaCA?

snellius said:

As far as I understand it, you measure the focus shift between the various colors (Chart 1 and 3).

Click to expand...

From Wikipedia: There are two types of chromatic aberration: axial (longitudinal), and transverse (lateral). Axial aberration occurs when different wavelengths of light are focused at different distances from the lens, i.e., different points on the optical axis (focus shift).

https://en.wikipedia.org/wiki/Chromatic_aberration

snellius said:

This says something about the focal length for these colors.

Click to expand...

It says something about whether some colors are focused in front of or behind other colors.

snellius said:

But LoCA is a color dispersion that takes place in the blur in front of and behind the focus point even if all the colors come nearly together in that focus point.

Click to expand...

That I don't buy. If the colors all focused at the same axial location, none would be blurred.

snellius said:

See an example here (almost below)

I'm no a authority in optics, but have tried to draw what is the difference to me.



I think what you measure is not a value of Longitudinal Chromatic Aberration.

Click to expand...

I can't understand the point you're making with the chart.

Jim

--
http://blog.kasson.com

JimKasson said:

snellius said:

Is this about Longitudinal Chromatic Aberration or normal Chromatic Aberration?

Click to expand...

I can't understand the point you're making with the chart.

Jim

Click to expand...

Just my third-hand interpretation (and, I should add, optics is waaay out of my wheelhouse), but the chart shows that LoCA relates to the rate of defocus being different near the focus plane, i.e., all wavelengths are perfectly focused right on the same focus plane, but the rate at which the wavelengths "defocus" as we move away from the focus plane differs.

With the "plain" CA, the rate of defocus between the wavelengths is roughly constant, but the focus plane for each wavelength is offset relative to the others.

This explanation makes sense to me, since "plain" CA can be reduced by scaling the red and blue channels to match the size of the green plane (this option is implemented in dcraw, and is typically included in a lens correction profile such as used in ACR/LightRoom). Rescaling the red and blue channels cannot fix their degree of defocus, but it does remove the colour fringes around edges, and since we are more sensitive to green sharpness anyway, this seems like a reasonable fix. I suppose one could even apply more sharpening to red and blue relative to the scaling factor required to bring them in line with green. The scaling factor between the red/blue/green wavelengths does not change (much?) as a function of distance from the focus plane.

LoCA, on the other hand, is not easy to correct for because the degree of defocus is a function of distance from the focus plane and wavelength, i.e., without knowing a pixel's true distance (along optical axis) from the focus plane, it is not possible to correct for defocus (which is still wavelength dependent).

If this interpretation is correct, then the shift the MTF50 peak between the wavelengths (as a function of focus distance) that you measure is probably a combination of CA and LoCA. The LoCA component can probably be quantified by looking at the derivative of your MTF50 by distance by wavelength data, after shifting the red and blue peaks to line up with the green peak (for example).

I have a 1970's Nikkor 105 mm f/4 AI macro lens that displays a very strong magenta / green (yellow?) shift as you move from in front of to behind the focus plane at f/4, which according to my understanding is typical of LoCA. This might be a good candidate for testing. I have a manual focus rail with a very fine adjustment screw (something like 100 turns per inch thread) with a 5 cm travel on the fine thread (and an independent coarse position adjustment). I am a little reluctant to capture 1000 images (50 micron steps) manually, though.

The_Suede said:

fvdbergh2501 said:

The_Suede said:

fvdbergh2501 wrote: [snip]

Click to expand...

Click to expand...

Click to expand...

The_Suede said:

Yes, very true - I didn't even weigh in the geometric aperture change for the high magnification. IMO this makes the measurement series even more strange. Especially since another known good (the Sony lens) shows a behavior that matches my expectations from earlier experiences quite closely.

Since I don't know the working method of the floating element in the CO, the real aperture is an unknown (even if we think we know the reproduction ratio is 1:2).

The SA effects on paraxial WO MTF can be modeled accurately if you have the SA line, but it takes a bit of effort. You have to integrate the CoC's given for a fixed image projection distance.

You can however approximate it by assuming a second order line and taking two or three aperture samples, as Jim did. Take the CoC profile from the best focus distance from the WO condition at the reference distance. Do a linear subtraction of the [CoC]x[defocus blur] weighted by total light transmission you get from the higher number apertures at the same distance. Since the higher apertures should have a wider and flatter CoC, the subtraction has a smaller relative effect on the WO CoC peak than the WO CoC field, effectively sharpening the CoC by convolution. The difference is (approximately) the SA loss. The easiest way to do this is to calculate the full MTF of each aperture setting for the distance set, and model the CoC from the MTF.

The method is however very susceptible to accuracy losses though, normally you overestimate the SA loss. The measurements at higher aperture numbers are diffracted, where the real WO ray integration is not - and the diffraction widens the CoC profile of the subtracted CoC's, sharpening the baseline more than the theoretically needed amount. Additionally, the SA line is most often a third or fourth power line, so it doesn't behave as smoothly as our approximation.

Click to expand...

Thanks for the detailed reply. I think I understand the essence of what you are describing, but I must confess that I will have to read up a bit more before I will understand it all.

Quick question: WO = "Wavefront Optics" ?

JimKasson said:

snellius said:

I'm no a authority in optics, but have tried to draw what is the difference to me.

Click to expand...

Click to expand...

I think what you measure is not a value of Longitudinal Chromatic Aberration.

JimKasson said:

I can't understand the point you're making with the chart.

Click to expand...

My English is not so good (even bad), but my illustrations are in general quite clear.

Like I said, I'm not a scientist but a technician. I'm no authority in optics but know the terms as used in photography.

In photography I know the term Chromatic Aberration as the color dispersion which is easy to solve by scaling the color channels (the lens has for each color a different focal length and thus for each color a different magnification).

The Longitudinal Chromatic Aberration is something quite different. As far as I know this term in photography is applied to the color dispersion in front of and behind the focus point. In the focus point the colors come almost together, but before and after this point the colors shifting apart. The effect of this color dispersion increases from the focus point either forwards or backwards, but the further you go the blur is also increasing and blurring the color effects.



There may be a difference in the use of the term LoCA within science and photography.

(I think Loca is the longitudinal aspect or effect of Lateral Chromatic Aberration.)

snellius said:

snellius wrote: I think what you measure is not a value of Longitudinal Chromatic Aberration.

snellius said:

JimKasson wrote: I can't understand the point you're making with the chart.

Click to expand...

In photography I know the term Chromatic Aberration as the color dispersion which is easy to solve by scaling the color channels (the lens has for each color a different focal length and thus for each color a different magnification).

Click to expand...

Perhaps it would help to review and agree on the terminology as used in Photography (there are no question marks but assume these are questions):

1) Longitudinal Chromatic Aberrations are the result of different wavelengths focusing on different planes perpendicular to the lens (axial direction), even in the center of the field of view: by measuring the distance from the subject at which each raw color channel's MTF peaks Jim is able to quantify how well corrected a lens is for LoCA;

2) Even with zero LoCa and in the center of the field of view a lens may produce different sized blur spots with light of different wavelengths (ignoring diffraction for now): this is mostly due to Spherical Aberrations (and in particular to Spherochromatism - Brandon Dube). In addition, in the presence of a color image and a lens with LoCA, the size of the sensor's CFA bandwidths will contribute to averaging wavelength responses together, producing some blur of their own; Diffraction, SA and CFA wavelength spreads can explain the relative absolute peak MTF value differences that Jim measures.

3) Lateral Chromatic Aberrations 'represent a change in the [oblique] focal length of the lens with color, so that the size of the image in one color is greater than in another color. The effect if present vanishes in the center of the field and becomes progressively worse as the obliquity increases' (Kingslake, 1992). This, I believe, is what dcraw and most raw converters correct. I don't think that Jim measures this (if not accidentally).

Makes sense? There is a bit more here .

Jack

fvdbergh2501 said:

JimKasson said:

snellius said:

Is this about Longitudinal Chromatic Aberration or normal Chromatic Aberration?

Click to expand...

I can't understand the point you're making with the chart.

Jim

Click to expand...

Just my third-hand interpretation (and, I should add, optics is waaay out of my wheelhouse), but the chart shows that LoCA relates to the rate of defocus being different near the focus plane, i.e., all wavelengths are perfectly focused right on the same focus plane, but the rate at which the wavelengths "defocus" as we move away from the focus plane differs.

With the "plain" CA, the rate of defocus between the wavelengths is roughly constant, but the focus plane for each wavelength is offset relative to the others.

Click to expand...

Hi Frans, I don't even know what a wheelhouse is, so take the following with a grain of salt.

I agree that Longitudinal CA is effectively axial defocus as a function of wavelength. For large amounts of defocus (say optical path differences greater than about 1 wavelength) the geometric ray tracing model seems to work fairly well. The rate of change of the blur disc size can be considered 'roughly constant' with this model.

However that's a LOT of defocus. For the amount of defocus to be expected from LoCA for a typical lens (say less than 0.5 lambda OPD, approaching the in-focus criterion of 1/4 lambda) a more complex model is needed which takes diffraction into consideration, such as Hopkins (1955). The rate of change of the blur disc is not constant in this model though, nor are the blur discs in each CFA channel a differently sized version of the other wavelengths.

I believe that one cannot correct for LoCA, just as one cannot correct for defocus, by simple means like resizing or deconvolution. One can always fudge it, but that's what it is. In fact no raw converter I know knows how to correct for axial CA automatically and none do.

I understand that DCRAW and other raw converters correct for Lateral CA, as described in my message to Snellius. Perhaps this is what he means by 'plain' CA.

Jack

Much easier than that
"Wide Open"

Jack Hogan said:

fvdbergh2501 said:

JimKasson said:

snellius said:

Is this about Longitudinal Chromatic Aberration or normal Chromatic Aberration?

Click to expand...

I can't understand the point you're making with the chart.

Jim

Click to expand...

Just my third-hand interpretation (and, I should add, optics is waaay out of my wheelhouse), but the chart shows that LoCA relates to the rate of defocus being different near the focus plane, i.e., all wavelengths are perfectly focused right on the same focus plane, but the rate at which the wavelengths "defocus" as we move away from the focus plane differs.

With the "plain" CA, the rate of defocus between the wavelengths is roughly constant, but the focus plane for each wavelength is offset relative to the others.

Click to expand...

Hi Frans, I don't even know what a wheelhouse is, so take the following with a grain of salt.

I agree that Longitudinal CA is effectively axial defocus as a function of wavelength. For large amounts of defocus (say optical path differences greater than about 1 wavelength) the geometric ray tracing model seems to work fairly well. The rate of change of the blur disc size can be considered 'roughly constant' with this model.

However that's a LOT of defocus. For the amount of defocus to be expected from LoCA for a typical lens (say less than 0.5 lambda OPD, approaching the in-focus criterion of 1/4 lambda) a more complex model is needed which takes diffraction into consideration, such as Hopkins (1955). The rate of change of the blur disc is not constant in this model though, nor are the blur discs in each CFA channel a differently sized version of the other wavelengths.

I believe that one cannot correct for LoCA, just as one cannot correct for defocus, by simple means like resizing or deconvolution. One can always fudge it, but that's what it is. In fact no raw converter I know knows how to correct for axial CA automatically and none do.

I understand that DCRAW and other raw converters correct for Lateral CA, as described in my message to Snellius. Perhaps this is what he means by 'plain' CA.

Jack

Click to expand...

My point is: What Jim is measuring seems to me Chromatic Aberration or Lateral Chromatic Aberration, but no Longitudinal Chromatic Aberration. He measures in the focal plane A small differences in distance, but for Longitudinal Chromatic Aberration is mainly the difference in B major.



A lens could be theoretically perfect corrected for Chromatic Aberration. There is therefore no Lateral Chromatic Aberration, but possibly he stil shows Longitudinal Chromatic Aberration.



As an example, a picture with the Micro Nikkor 105/2.8 VR. The picture has only been in Nikon Capture NX2 to turn off all the fixes for the different types of Chromatic Aberration and save the RAW as JPG file. This picture clearly shows that the lens barely has lateral chromatic aberration (see far left and right in the plane of focus), but has serious Longitudinal Chromatic Aberration. This lens would score well in the subject distance shift measurement for the three colors (pretty equal), but he has very serious LoCA.