Ephotozine Review: Z 100-400/4.5-5.6 S VR

Started Jan 25, 2022 | Discussions
anotherMike Forum Pro • Posts: 10,792
Re: Ephotozine Beware - Poor testing protocols…
1

I need to shoot with it a *lot* more to have a full understanding of its performance, but so far, at least at studio distances, where I ran some tests, it's quite sharp. I don't know about MFD

anotherMike Forum Pro • Posts: 10,792
Re: Ode to Slanted Edges (vs Bench)

Thanks Jack.

I've worked with Imatest software and have an (older) copy of MTFmapper somewhere.

I would disagree about 50X f/l being a proxy for infinity, and it's really infinity results that I'm interested in for landscape work, so we'll have to disagree on that one. I don't have the space in the residence for 40 feet of draw either, so that one is out of the question as well.

When time permits (I'm not retired, and balance an income-producing job and a photography sideline) I absolutely do need to look at MTFmapper again for the mid length lenses in the manner you speak of. Perhaps not to the Bernard Delley level of depth, but I do need to get back to it at some point.

I've done a *lot* of subjective testing in an extremely thorough way (multiple test runs on different days to root out test error, etc) and just haven't yet found that the chart based tests correlate as well to my set of experiences as the MTF based ones that Roger Cicala has done, particularly the field curvature charts he ran a few years back, but you are right - I do need to delve into it again when time permits to increase my knowledge in that area.

My primary interest in all of this isn't to prove "my" lens is better than "the other guys" lens but to begin to see if I can correlate things I see subjectively with objective measurements, a strange interest, one that I might play with more if/when I can get to retirement.

But ultimately, in dream world, I want a bench to play with....

-m

Leonard Shepherd Forum Pro • Posts: 24,270
Re: A picture is worth a 1000 words

JimKasson wrote:

Here is an example of a lens outresolving a sensor, with a Nikon 180-400/4 wide open, set to 180mm, on a Z7, with crops from the center, the edges, and the corners.

The lens outresolves the sensor in all 9 locations.

Can you explain how the optical formula for calculating resolution should be updated?

-- hide signature --

Leonard Shepherd
In lots of ways good photography is much more about how equipment is used rather than anything else.

 Leonard Shepherd's gear list:Leonard Shepherd's gear list
Nikon Z9 Nikon AF-S Micro-Nikkor 60mm F2.8G ED Nikon AF-S Nikkor 300mm f/2.8G ED-IF VR Nikon AF Micro-Nikkor 200mm f/4D ED-IF Nikon AF-S Teleconverter TC-17E II +22 more
JimKasson
JimKasson Forum Pro • Posts: 41,535
Re: A picture is worth a 1000 words

Leonard Shepherd wrote:

JimKasson wrote:

Here is an example of a lens outresolving a sensor, with a Nikon 180-400/4 wide open, set to 180mm, on a Z7, with crops from the center, the edges, and the corners.

The lens outresolves the sensor in all 9 locations.

Can you explain how the optical formula for calculating resolution should be updated?

I don’t think it needs updating, but I am not sure what formula you’re talking about. Please state it.

What I’m talking about is pretty much sampled imaging systems 101. Harry Nyquist made his discovery about 90 years ago.

Are you familiar with the concept of imaging systems Q? That informs the optimum matching of the resolutions of sensors and lenses.

https://blog.kasson.com/the-last-word/whats-your-q/

-- hide signature --
 JimKasson's gear list:JimKasson's gear list
Leica Q2 Monochrom Nikon Z7 Fujifilm GFX 100 Nikon Z9 Hasselblad X2D 100c +1 more
JimKasson
JimKasson Forum Pro • Posts: 41,535
Re: Ode to Slanted Edges (vs Bench)
2

Jack Hogan wrote:

all good points, let me provide a different take on them. A Modulation Transfer Function is just a relevant way to interpret how an imaging system responds to points of light (Point Spread Functions) throughout the field of view. Since the image is a collection of very many such points of light it gives a good indication of what one can expect.

If obtained with proper technique, MTF from say a set of slanted edge captures can in practice give you much of the information useful to photographers that can be extracted from an optical bench - and vice versa. It nicely complements just judging what we see in published images. It allows us to quantify generic statements like 'the Zeiss 135 is better than the 100-400 at 135mm at infinity' (is it? by how much? will most people notice it? or do we have to trust the conflicted influencer du jour?). Which is why people like Jim Kasson show pictures *and* MTF information when evaluating critical lenses on a particular platform.

The main advantage of running slanted edge tests vs a bench is that if you are somewhat familiar with a command line interpreter you can do it for free at home, right now. On the other hand, you have to have a dedicated space not to mention a degree and a full time job in the optics industry to set up, keep calibrated and run an optical bench. Not to mention the cost of a bench: impractically unaffordable for lens review purposes (I wish I had a Hubble to replace the telescope in the yard but ...).

So we take a chunk of an afternoon to set up and take captures of slanted edges at various positions, orientations and typical usage set ups. A large utility knife works well, not perfect but often close enough. A flat chart has advantages and disadvantages. And, contrary to a bench, we get system results that also include the effect of the camera/sensor on resolution , which as photographers is useful. Want through-focus? Got it. Then we feed the captures to excellent open source MTF Mapper by DPR contributor Frans van den Bergh -and out come MTF curves and all sorts of fantastic information relevant to the spatial resolution performance of the camera and lens.

Z7+50mm/1.8S at f/4, G ch, from utility knife slanted about 4 degrees captured at 'infinity' (10m away).

As for testing at infinity: there is no such thing, since we are never able to test at actual infinity. Even optical benches are never *perfectly* collimated everywhere. So it's a question of diminishing returns and good enough. For most practical purposes 50x focal length or more is good enough, that would be about 22' for the 135mm above. I can do about twice that with minimal inconvenience in my house (I would have to hit the backyard for the 400mm; there is a gentleman who wrote a paper on this subject working in his condo's parking lot :).

Contrary to popular belief the slanted edge method does provide a full MTF curve, including *the higher (finer structure) spatial frequencies*, which however are often less than useful at working f-numbers because drowned in aliasing when pixel peeping a typical monitor at 100%.

Since most people would not know how to properly interpret an MTF curve, testers have tried to come up with a single figure that summarizes much of what the curve has to say as far as the spatial frequency response of a system is concerned. They have names like SQF, SQRI, CMT Acutance and, yes, MTF50. It turns out that in many practical cases MTF50 correlates well with the others (if you are interested in the reason why post a question in the PST forum or take a peak here ). They are quantitative and useful: for instance a 5% difference in MTF50 is hard to spot but by 10% most people notice it.

Perfect? Of course not, just another tool in the evaluation bag. But don't discount MTF from a slanted edge off hand just because it is not from an optical bench: you can perform the 135mm experiment up top yourself and get a decent quantitative answer useful to photographers - today.

Jack
PS On the other hand comparing results from different sites is not a good idea and a subject for another day.

Good summary, Jack. Mike touched on wanting MTFs at particular frequencies, rather than the frequencies for particular contrasts. Slanted edge MTFs can be reported that way. These days I usually compute a metric that I call microcontrast, which I define as contrast at 0.25 cy/px. The frequency could just as easily be in cy/mm, but I don't find that as useful.

-- hide signature --
 JimKasson's gear list:JimKasson's gear list
Leica Q2 Monochrom Nikon Z7 Fujifilm GFX 100 Nikon Z9 Hasselblad X2D 100c +1 more
Leonard Shepherd Forum Pro • Posts: 24,270
Re: A picture is worth a 1000 words

JimKasson wrote:

I am not sure what formula you’re talking about. Please state it.

The usual one - image resolution equals 1/sensor resolution plus 1/image resolution.

-- hide signature --

Leonard Shepherd
In lots of ways good photography is much more about how equipment is used rather than anything else.

 Leonard Shepherd's gear list:Leonard Shepherd's gear list
Nikon Z9 Nikon AF-S Micro-Nikkor 60mm F2.8G ED Nikon AF-S Nikkor 300mm f/2.8G ED-IF VR Nikon AF Micro-Nikkor 200mm f/4D ED-IF Nikon AF-S Teleconverter TC-17E II +22 more
JimKasson
JimKasson Forum Pro • Posts: 41,535
Re: A picture is worth a 1000 words
3

Leonard Shepherd wrote:

JimKasson wrote:

I am not sure what formula you’re talking about. Please state it.

The usual one - image resolution equals 1/sensor resolution plus 1/image resolution.

That is a loosely phrased (especially with image resolution on both sides of the equation!) approximation to another approximation, I believe. The other approximation is System MTF = Lens MTF * Camera MTF.

Your loosely phrased one doesn't work for extinction resolution.

Neither of those say anything about aliasing, which is what happens when the lens is too sharp for the sensor.

Jim

-- hide signature --
 JimKasson's gear list:JimKasson's gear list
Leica Q2 Monochrom Nikon Z7 Fujifilm GFX 100 Nikon Z9 Hasselblad X2D 100c +1 more
Lance B
Lance B Forum Pro • Posts: 34,340
Re: I agree, it is an excellent lens
1

Ernie Misner wrote:

"At closer focusing distances the 100-400 is very sharp."

Hi Lance, do you think it looses much sharpness approaching MFD (or enough to be concerning?) I asked anotherMike as well. I'm hoping it is actually usable at MFD. Thanks.

As I stated in my post, the 100-400 seems sharper at closer distances than at farther distance. As for Minimum Focus Distance that jury is still out.

 Lance B's gear list:Lance B's gear list
Nikon Z7 II Nikon Z9 Nikon AF-S Teleconverter TC-17E II Nikon AF-S Teleconverter TC-20E III Nikon AF-S Nikkor 80-400mm f/4.5-5.6G ED VR +19 more
aclo Contributing Member • Posts: 554
Re: A picture is worth a 1000 words
2

JimKasson wrote:

Leonard Shepherd wrote:

JimKasson wrote:

I am not sure what formula you’re talking about. Please state it.

The usual one - image resolution equals 1/sensor resolution plus 1/image resolution.

That is a loosely phrased (especially with image resolution on both sides of the equation!) approximation to another approximation, I believe. The other approximation is System MTF = Lens MTF * Camera MTF.

Your loosely phrased one doesn't work for extinction resolution.

Neither of those say anything about aliasing, which is what happens when the lens is too sharp for the sensor.

Jim

Even if this was an exact relation, it relates the system resolution to those of the lens and the sensor. And it's consistent with saying the lens can outresolve the sensor: if the lens resolution is rL, the sensor resolution is rS and the total system resolution is rT then we have 1/rT=1/rL+1/rS. If say rS is much larger than the rL, so the sensor outresolves the lens, ie rS>>rL, then 1/rS<<1/rL so 1/rT=1/rS+1/rL is approximately equal to 1/rL or in other words rT=rL, ie, the lens resolution limits the total resolution (or, increasing the sensor resolution won't increase the system resolution). I'm not sure how this equation implies a lens can't outresolve a sensor, the equation itself tells you that if one of the rL or rS is much larger than the other then the lower one determines the overall resolution.

I guess this approximation comes from System MTF = Lens MTF * Camera MTF and using that the MTF is (as far as I can tell) the FT of the image of a delta function, and that the MTF's width is roughly inversely proportional to the width of the image of the delta fn then it sort of makes sense?

JimKasson
JimKasson Forum Pro • Posts: 41,535
Re: A picture is worth a 1000 words
2

aclo wrote:

JimKasson wrote:

Leonard Shepherd wrote:

JimKasson wrote:

I am not sure what formula you’re talking about. Please state it.

The usual one - image resolution equals 1/sensor resolution plus 1/image resolution.

That is a loosely phrased (especially with image resolution on both sides of the equation!) approximation to another approximation, I believe. The other approximation is System MTF = Lens MTF * Camera MTF.

Your loosely phrased one doesn't work for extinction resolution.

Neither of those say anything about aliasing, which is what happens when the lens is too sharp for the sensor.

Jim

Even if this was an exact relation, it relates the system resolution to those of the lens and the sensor. And it's consistent with saying the lens can outresolve the sensor: if the lens resolution is rL, the sensor resolution is rS and the total system resolution is rT then we have 1/rT=1/rL+1/rS. If say rS is much larger than the rL, so the sensor outresolves the lens, ie rS>>rL, then 1/rS<<1/rL so 1/rT=1/rS+1/rL is approximately equal to 1/rL or in other words rT=rL, ie, the lens resolution limits the total resolution (or, increasing the sensor resolution won't increase the system resolution). I'm not sure how this equation implies a lens can't outresolve a sensor, the equation itself tells you that if one of the rL or rS is much larger than the other then the lower one determines the overall resolution.

Define “resolution” as you’re using it above, please. But the above ignores the sampling frequency, doesn’t it? And the sampling frequency is key to knowing whether the sensor is out resolved by the lens.

I guess this approximation comes from System MTF = Lens MTF * Camera MTF and using that the MTF is (as far as I can tell) the FT of the image of a delta function,

The MTF is the absolute value of the FT of the point spread function, which is indeed the system response to a Delta function. The absolute value part is why multiplying MTFs is an approximation.

and that the MTF's width is roughly inversely proportional to the width of the image of the delta fn then it sort of makes sense?

One thing to remember is that sensor MTF is not determined only by pixel pitch. The sampling window is important.

-- hide signature --
 JimKasson's gear list:JimKasson's gear list
Leica Q2 Monochrom Nikon Z7 Fujifilm GFX 100 Nikon Z9 Hasselblad X2D 100c +1 more
aclo Contributing Member • Posts: 554
Re: A picture is worth a 1000 words

JimKasson wrote:

aclo wrote:

Even if this was an exact relation, it relates the system resolution to those of the lens and the sensor. And it's consistent with saying the lens can outresolve the sensor: if the lens resolution is rL, the sensor resolution is rS and the total system resolution is rT then we have 1/rT=1/rL+1/rS. If say rS is much larger than the rL, so the sensor outresolves the lens, ie rS>>rL, then 1/rS<<1/rL so 1/rT=1/rS+1/rL is approximately equal to 1/rL or in other words rT=rL, ie, the lens resolution limits the total resolution (or, increasing the sensor resolution won't increase the system resolution). I'm not sure how this equation implies a lens can't outresolve a sensor, the equation itself tells you that if one of the rL or rS is much larger than the other then the lower one determines the overall resolution.

Define “resolution” as you’re using it above, please. But the above ignores the sampling frequency, doesn’t it? And the sampling frequency is key to knowing whether the sensor is out resolved by the lens.

I can't, as I don't know where the relation came from and am not defending the equation. I am saying that the qualitative picture that equation gives is that if the "resolution" of one of (lens, sensor) is much smaller than the other then that limits the total "resolution," no matter how much you increase the other. So I am unsure why it would be at odds with saying one of the two components can "outresolve" the other, as I interpret "outresolve" as "have much more resolution which then becomes useless as it does not contribute any more to the system resolution".

I guess this approximation comes from System MTF = Lens MTF * Camera MTF and using that the MTF is (as far as I can tell) the FT of the image of a delta function,

The MTF is the absolute value of the FT of the point spread function, which is indeed the system response to a Delta function. The absolute value part is why multiplying MTFs is an approximation.

and that the MTF's width is roughly inversely proportional to the width of the image of the delta fn then it sort of makes sense?

One thing to remember is that sensor MTF is not determined only by pixel pitch. The sampling window is important.

Right, true, it depends on many implementation details. Thanks, I'd never thought about this in any depth.

JimKasson
JimKasson Forum Pro • Posts: 41,535
Re: A picture is worth a 1000 words

aclo wrote:

JimKasson wrote:

aclo wrote:

Even if this was an exact relation, it relates the system resolution to those of the lens and the sensor. And it's consistent with saying the lens can outresolve the sensor: if the lens resolution is rL, the sensor resolution is rS and the total system resolution is rT then we have 1/rT=1/rL+1/rS. If say rS is much larger than the rL, so the sensor outresolves the lens, ie rS>>rL, then 1/rS<<1/rL so 1/rT=1/rS+1/rL is approximately equal to 1/rL or in other words rT=rL, ie, the lens resolution limits the total resolution (or, increasing the sensor resolution won't increase the system resolution). I'm not sure how this equation implies a lens can't outresolve a sensor, the equation itself tells you that if one of the rL or rS is much larger than the other then the lower one determines the overall resolution.

Define “resolution” as you’re using it above, please. But the above ignores the sampling frequency, doesn’t it? And the sampling frequency is key to knowing whether the sensor is out resolved by the lens.

I can't, as I don't know where the relation came from and am not defending the equation. I am saying that the qualitative picture that equation gives is that if the "resolution" of one of (lens, sensor) is much smaller than the other then that limits the total "resolution," no matter how much you increase the other. So I am unsure why it would be at odds with saying one of the two components can "outresolve" the other, as I interpret "outresolve" as "have much more resolution which then becomes useless as it does not contribute any more to the system resolution".

But I’m talking about something else entirely, and that’s aliasing. That’s a clear indication that the lens is out resolving the sensor, because the system mtf at the Nyquist frequency is too high.

I guess this approximation comes from System MTF = Lens MTF * Camera MTF and using that the MTF is (as far as I can tell) the FT of the image of a delta function,

The MTF is the absolute value of the FT of the point spread function, which is indeed the system response to a Delta function. The absolute value part is why multiplying MTFs is an approximation.

and that the MTF's width is roughly inversely proportional to the width of the image of the delta fn then it sort of makes sense?

One thing to remember is that sensor MTF is not determined only by pixel pitch. The sampling window is important.

Right, true, it depends on many implementation details. Thanks, I'd never thought about this in any depth.

-- hide signature --
 JimKasson's gear list:JimKasson's gear list
Leica Q2 Monochrom Nikon Z7 Fujifilm GFX 100 Nikon Z9 Hasselblad X2D 100c +1 more
aclo Contributing Member • Posts: 554
Re: A picture is worth a 1000 words

JimKasson wrote:

aclo wrote:

JimKasson wrote:

aclo wrote:

Even if this was an exact relation, it relates the system resolution to those of the lens and the sensor. And it's consistent with saying the lens can outresolve the sensor: if the lens resolution is rL, the sensor resolution is rS and the total system resolution is rT then we have 1/rT=1/rL+1/rS. If say rS is much larger than the rL, so the sensor outresolves the lens, ie rS>>rL, then 1/rS<<1/rL so 1/rT=1/rS+1/rL is approximately equal to 1/rL or in other words rT=rL, ie, the lens resolution limits the total resolution (or, increasing the sensor resolution won't increase the system resolution). I'm not sure how this equation implies a lens can't outresolve a sensor, the equation itself tells you that if one of the rL or rS is much larger than the other then the lower one determines the overall resolution.

Define “resolution” as you’re using it above, please. But the above ignores the sampling frequency, doesn’t it? And the sampling frequency is key to knowing whether the sensor is out resolved by the lens.

I can't, as I don't know where the relation came from and am not defending the equation. I am saying that the qualitative picture that equation gives is that if the "resolution" of one of (lens, sensor) is much smaller than the other then that limits the total "resolution," no matter how much you increase the other. So I am unsure why it would be at odds with saying one of the two components can "outresolve" the other, as I interpret "outresolve" as "have much more resolution which then becomes useless as it does not contribute any more to the system resolution".

But I’m talking about something else entirely, and that’s aliasing. That’s a clear indication that the lens is out resolving the sensor, because the system mtf at the Nyquist frequency is too high.

I appreciate that if you have detail in the image at high enough freqs it aliases down and in this instance causes visual artifacts in the frequencies reproduce. So that is a real sense in which you "outresolve" the sensor. I was simply pointing out that the equation itself supports the notion that either of (lens, sensor) can have too much resolution which contributes nothing to the result.

However I now realise that possibly the earlier rhetorical question as to whether the equation needs updating meant that the eqn says that too much lens resolution does nothing, rather than actually worsen things (which I think is what you are focussing on, correct? or am I still missing the point here?)

BasilG Forum Pro • Posts: 10,016
Re: A picture is worth a 1000 words

aclo wrote:

JimKasson wrote:

Leonard Shepherd wrote:

JimKasson wrote:

I am not sure what formula you’re talking about. Please state it.

The usual one - image resolution equals 1/sensor resolution plus 1/image resolution.

That is a loosely phrased (especially with image resolution on both sides of the equation!) approximation to another approximation, I believe. The other approximation is System MTF = Lens MTF * Camera MTF.

Your loosely phrased one doesn't work for extinction resolution.

Neither of those say anything about aliasing, which is what happens when the lens is too sharp for the sensor.

Jim

Even if this was an exact relation, it relates the system resolution to those of the lens and the sensor. And it's consistent with saying the lens can outresolve the sensor: if the lens resolution is rL, the sensor resolution is rS and the total system resolution is rT then we have 1/rT=1/rL+1/rS. If say rS is much larger than the rL, so the sensor outresolves the lens, ie rS>>rL, then 1/rS<<1/rL so 1/rT=1/rS+1/rL is approximately equal to 1/rL or in other words rT=rL, ie, the lens resolution limits the total resolution (or, increasing the sensor resolution won't increase the system resolution). I'm not sure how this equation implies a lens can't outresolve a sensor, the equation itself tells you that if one of the rL or rS is much larger than the other then the lower one determines the overall resolution.

Though I like the simple mathematical argument, I guess there's a practical question here - what does rL>>rS mean? Even a 24 MP FX sensor has approx. 160 pixels/mm, approx. 80 lp/mm if you want. If ">>" means "5x" or "10x", then you will struggle to find a lens that can resolve that much, and below that, it may be hard to argue that the other component can be entirely neglected.

A while ago, I came up with the following chart for lens vs. system resolution for various sensor pixel counts (using the formulae detailed here:  https://photo.blogoverflow.com/2012/06/the-realities-of-resolution/ ):

Things would appear to break down when approaching and exceeding the 24 MP FX sensor's pixel pitch because I would expect aliasing to do something to these curves, but I am not sure how that would be incorporated here. Maybe Jim can weigh in?

JimKasson
JimKasson Forum Pro • Posts: 41,535
Re: A picture is worth a 1000 words

aclo wrote:

JimKasson wrote:

aclo wrote:

JimKasson wrote:

aclo wrote:

Even if this was an exact relation, it relates the system resolution to those of the lens and the sensor. And it's consistent with saying the lens can outresolve the sensor: if the lens resolution is rL, the sensor resolution is rS and the total system resolution is rT then we have 1/rT=1/rL+1/rS. If say rS is much larger than the rL, so the sensor outresolves the lens, ie rS>>rL, then 1/rS<<1/rL so 1/rT=1/rS+1/rL is approximately equal to 1/rL or in other words rT=rL, ie, the lens resolution limits the total resolution (or, increasing the sensor resolution won't increase the system resolution). I'm not sure how this equation implies a lens can't outresolve a sensor, the equation itself tells you that if one of the rL or rS is much larger than the other then the lower one determines the overall resolution.

Define “resolution” as you’re using it above, please. But the above ignores the sampling frequency, doesn’t it? And the sampling frequency is key to knowing whether the sensor is out resolved by the lens.

I can't, as I don't know where the relation came from and am not defending the equation. I am saying that the qualitative picture that equation gives is that if the "resolution" of one of (lens, sensor) is much smaller than the other then that limits the total "resolution," no matter how much you increase the other. So I am unsure why it would be at odds with saying one of the two components can "outresolve" the other, as I interpret "outresolve" as "have much more resolution which then becomes useless as it does not contribute any more to the system resolution".

But I’m talking about something else entirely, and that’s aliasing. That’s a clear indication that the lens is out resolving the sensor, because the system mtf at the Nyquist frequency is too high.

I appreciate that if you have detail in the image at high enough freqs it aliases down and in this instance causes visual artifacts in the frequencies reproduce. So that is a real sense in which you "outresolve" the sensor. I was simply pointing out that the equation itself supports the notion that either of (lens, sensor) can have too much resolution which contributes nothing to the result.

However I now realise that possibly the earlier rhetorical question as to whether the equation needs updating meant that the eqn says that too much lens resolution does nothing, rather than actually worsen things (which I think is what you are focussing on, correct? or am I still missing the point here?)

I think you’ve got it. In today’s commercial world, with good lenses, they always out resolve the sensors at some f stops and some places in the frame.

-- hide signature --
 JimKasson's gear list:JimKasson's gear list
Leica Q2 Monochrom Nikon Z7 Fujifilm GFX 100 Nikon Z9 Hasselblad X2D 100c +1 more
aclo Contributing Member • Posts: 554
Re: A picture is worth a 1000 words
1

JimKasson wrote:

I think you’ve got it. In today’s commercial world, with good lenses, they always out resolve the sensors at some f stops and some places in the frame.

Thanks, should have probably have had the entire discussion in my head instead of out loud and saved your, Leonard's and BasilG's time!

JimKasson
JimKasson Forum Pro • Posts: 41,535
Re: A picture is worth a 1000 words

BasilG wrote:

aclo wrote:

JimKasson wrote:

Leonard Shepherd wrote:

JimKasson wrote:

I am not sure what formula you’re talking about. Please state it.

The usual one - image resolution equals 1/sensor resolution plus 1/image resolution.

That is a loosely phrased (especially with image resolution on both sides of the equation!) approximation to another approximation, I believe. The other approximation is System MTF = Lens MTF * Camera MTF.

Your loosely phrased one doesn't work for extinction resolution.

Neither of those say anything about aliasing, which is what happens when the lens is too sharp for the sensor.

Jim

Even if this was an exact relation, it relates the system resolution to those of the lens and the sensor. And it's consistent with saying the lens can outresolve the sensor: if the lens resolution is rL, the sensor resolution is rS and the total system resolution is rT then we have 1/rT=1/rL+1/rS. If say rS is much larger than the rL, so the sensor outresolves the lens, ie rS>>rL, then 1/rS<<1/rL so 1/rT=1/rS+1/rL is approximately equal to 1/rL or in other words rT=rL, ie, the lens resolution limits the total resolution (or, increasing the sensor resolution won't increase the system resolution). I'm not sure how this equation implies a lens can't outresolve a sensor, the equation itself tells you that if one of the rL or rS is much larger than the other then the lower one determines the overall resolution.

Though I like the simple mathematical argument, I guess there's a practical question here - what does rL>>rS mean? Even a 24 MP FX sensor has approx. 160 pixels/mm, approx. 80 lp/mm if you want. If ">>" means "5x" or "10x", then you will struggle to find a lens that can resolve that much, and below that, it may be hard to argue that the other component can be entirely neglected.

A while ago, I came up with the following chart for lens vs. system resolution for various sensor pixel counts (using the formulae detailed here: https://photo.blogoverflow.com/2012/06/the-realities-of-resolution/ ):

Things would appear to break down when approaching and exceeding the 24 MP FX sensor's pixel pitch because I would expect aliasing to do something to these curves, but I am not sure how that would be incorporated here. Maybe Jim can weigh in?

Why don’t you just stop plotting each curve where the aliasing starts. By the way, your assumption of 100% fill factor doesn’t hold for some cameras, like the GFX 50x ones.

-- hide signature --
 JimKasson's gear list:JimKasson's gear list
Leica Q2 Monochrom Nikon Z7 Fujifilm GFX 100 Nikon Z9 Hasselblad X2D 100c +1 more
aclo Contributing Member • Posts: 554
Re: A picture is worth a 1000 words

BasilG wrote:

Though I like the simple mathematical argument, I guess there's a practical question here - what does rL>>rS mean? Even a 24 MP FX sensor has approx. 160 pixels/mm, approx. 80 lp/mm if you want. If ">>" means "5x" or "10x", then you will struggle to find a lens that can resolve that much, and below that, it may be hard to argue that the other component can be entirely neglected.

Yes good point, what I said is too soggy if you go beyond qualitative arguments.

A while ago, I came up with the following chart for lens vs. system resolution for various sensor pixel counts (using the formulae detailed here: https://photo.blogoverflow.com/2012/06/the-realities-of-resolution/ ):

Things would appear to break down when approaching and exceeding the 24 MP FX sensor's pixel pitch because I would expect aliasing to do something to these curves, but I am not sure how that would be incorporated here. Maybe Jim can weigh in?

Also a good question.

BasilG Forum Pro • Posts: 10,016
Re: A picture is worth a 1000 words
1

JimKasson wrote:

BasilG wrote:

Though I like the simple mathematical argument, I guess there's a practical question here - what does rL>>rS mean? Even a 24 MP FX sensor has approx. 160 pixels/mm, approx. 80 lp/mm if you want. If ">>" means "5x" or "10x", then you will struggle to find a lens that can resolve that much, and below that, it may be hard to argue that the other component can be entirely neglected.

A while ago, I came up with the following chart for lens vs. system resolution for various sensor pixel counts (using the formulae detailed here: https://photo.blogoverflow.com/2012/06/the-realities-of-resolution/ ):

Things would appear to break down when approaching and exceeding the 24 MP FX sensor's pixel pitch because I would expect aliasing to do something to these curves, but I am not sure how that would be incorporated here. Maybe Jim can weigh in?

Why don’t you just stop plotting each curve where the aliasing starts.

That would be the most pragmatic approach probably.

By the way, your assumption of 100% fill factor doesn’t hold for some cameras, like the GFX 50x ones.

Good point.

BasilG Forum Pro • Posts: 10,016
Re: A picture is worth a 1000 words
1

aclo wrote:

JimKasson wrote:

I think you’ve got it. In today’s commercial world, with good lenses, they always out resolve the sensors at some f stops and some places in the frame.

Thanks, should have probably have had the entire discussion in my head instead of out loud and saved your, Leonard's and BasilG's time!

Not to worry. I enjoy this type of discussion.

Keyboard shortcuts:
FForum MMy threads