Interesting article on DxO about 5D III and D800 resolving power

Started Apr 4, 2013 | Discussions thread
Mike CH Veteran Member • Posts: 9,631
Re: Interesting article on DxO about 5D III and D800 resolving power

John Sheehy wrote:

Mike CH wrote:

John Sheehy wrote:

There was a time when resolution was often described in lines or line pairs per millimeter.  Remember millimeter?  How confusing was that?

Yes, I do remember. No, I don't find it confusing. The problem comes when it is misrepresented or in the wrong context.

Unqualified "per sensor" measurements are just guilty as unqualified "per mm" measurements.

To use an analogy from somewhere else: it is quite obvious, isn't it, that a cars fuel consumption can be measured in square miles, yes?

That's pure hyperbole.  That is an abstraction that has nothing to do with elemental concerns.  Yeah, you can multiply by the width of the car, or drive in a spiral where your left tires follow the tracks of your right tires and see how much of a circle you can cover with a unit of gas, but both are extremely unlikely a need, compared to sensor or film performance per mm.


Hint: What SI unit do you end with if you reduce liter per kilometer to the simpelst possible term?

IMO, the least-confused people are those who see the lens first, as an analog projection device with its own set of imperfections, the FOV determined by the sensor size, and then the imaging quality per unit of sensor area.  Every other way of looking at things can be extended from this, with a minimal amount of confusion.

Ehm, yes?!? So how do you get from there, in a minimal way, to a practical comparison of (whatever measure you favour of) resolution at same FOV?

And why not go there directly?

Nothing wrong with quoting such things directly, but one should understand what is going on to create the frame, if one is to avoid coming to false conclusions extrapolating from that information, such as when a lens is not available to create the FOV needed for one sensor size.

I've seen so many people come to false conclusions over the years, because they do not understand the elemental issues of image capture, and can only grasp simple, high abstractions, such as this new "perceptual megapixel" nonsense.  Quoting resolution at a single contrast threshold may have been fine for film, since the contrast roll-off tail would be fairly predictable, but for discrete gridded sampling, it is very misleading, IMO, because you do not know what the contrast vs resolution curve does on either side of the chosen contrast threshold.  You could have a lot of lower-contrast resolution in a low-ISO capture that is good enough to use with sharpening, giving more resolution at a target contrast level in the output, at resolution levels impossible with a system that scored higher PMP.

Yes. I actually quite agree with your reasoning. I just cannot see the practical use of it.

Or, in other words, given your favourite measurements for two different systems, one FF, one APS-C, with lenses selected to give the same FOV, how do I get from your system of measurements to one which will answer my question about which system would have given me the better result? (Same framing, composition, FOV etc in both cases).

Could you walk us through the derivation?

starting from a position that fails to distinguish between relatives and absolutes.

Are you saying that there is no absolute measurement of resolution which can be used at the same FOV?

No.  Both the sensor alone, and the sensor with a given lens, aperture, and focus distance have their own potentials which can be measured, relative only to their frames, regardless of size.  PMP is not a very thorough measure of that, however.

Ah, so let us not use PMP. Let us use another measurement as long as it can be used to compare across systems with same FOV. Which would be your recommendation?

You might have been thinking that I was arguing in favor of PMP?

If so, then no, sorry, that was not the intention. My intention was to argue for a need to compare at same FOV, as a reaction to my feeling that several people in this thread seemed to indicate that a measurement of the sensors ability to handle an image on the focal plane was the end-all-be-all of sensor measurements.

My practical interest is simple. Say I've taken an image with my current system of a given scene with a given composition, framing and FOV. Could I have gotten a better result IQ-wise (sharper, more detail, better colour rendition) with a different system? Measures which can't answer that question are of, at most, secondary interest to.

Then they are of secondary interest to you.  They are of secondary interest to me, too, if I need a wide FOV and focal length limitations are not an issue.  That tells just half of the story though, as far as all of my interests go.

Different interests, different needs.

If I have to take an image of a bird, it had better be stuffed

And, I think I would rather have a 50MP sensor and lens that scores 32PMP than a 36MP sensor that scores 35PMP with the same FOV.  In fact, any sensor that performs PMP close to their true MP count has potential aliasing, and is a sign that the sensor has far too low a pixel density for that lens (or the AA filter is too weak for that lens, but more pixels is better solution than a weaker AA filter).

That makes sense to me.

Regards, Mike

-- hide signature --

Wait and see...

 Mike CH's gear list:Mike CH's gear list
Canon G7 X II Canon EOS 5D Mark III Canon EOS 5D Mark IV Canon EF 16-35mm F2.8L II USM Canon EF 100mm F2.8L Macro IS USM +14 more
Post (hide subjects) Posted by
MOD Mako2011
MOD Mako2011
MOD Mako2011
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow