The Next Generation?

[regken]

Originally Posted by falconeye

Funny to see the word "arouse" in this context. As a non native speaker, I thought it would have some reserved other meaning
..

Yeah well when you are my age you get aroused any way you can. I found this site, http://www.takinami.com/yoshihiko/ph...procedure.html, that shows the resolution test done on some of my old glass. What I'm trying to figure out is will my old lenses work on a FF sensor IF Pentax ever comes out with one.

I think the short answer is try them on it IF it ever comes out.

Thanks,

Ken

[ricardobeat]

Originally Posted by falconeye

Hi Ben, my view on this is this (my physicists's long-term view, not my photographic need's view for the next generation -- so you may want to ignore it ):

A pixel site should'n get smaller than about 3 micron (because of noise and the optical limits of fine glass like the FA 31mm which is still good at 170 Lp/mm). But not larger either, because our human eyes are so amazingly sharp at their point of sharpest vision (~81 MPixels when we are allowed to point them wherever we want on an image...).

So, we end up with FF and 96 MPixels. For rounding and marketing reasons, this means:

The DSLR pixel race will go up to 100 MPixels and stop there.

For what print size? :/

That doesn't make much sense to me. Above 220 dpi almost no one can tell the difference. And a FF sensor with 81mp is way beyond the resolution of the FA 31mm or any other lens (more than 500Lp/mm, if i'm not wrong).

And for those wondering, 96mp is just a little more than double 14mp resolution, it's not that big of a step (9.600px X 10.000px). Sooner than we think this mega-pixel measurement will be left behind, it's just marketing speak.

[falconeye]

Originally Posted by ricardobeat

For what print size? :/

That doesn't make much sense to me. Above 220 dpi almost no one can tell the difference. And a FF sensor with 81mp is way beyond the resolution of the FA 31mm or any other lens (more than 500Lp/mm, if i'm not wrong).

And for those wondering, 96mp is just a little more than double 14mp resolution, it's not that big of a step (9.600px X 10.000px). Sooner than we think this mega-pixel measurement will be left behind, it's just marketing speak.

What you say is common sense and may be plain correct. Which is why I made my post to articulate a different possibility. It was based on educated guesses, though:

"Above 220 dpi almost no one can tell the difference":
- The resolving power of the human eye is a measured fact. If above 220 dpi printers show no additional detail then you need to print large enough
- The eye (9k x 9k) at 220dpi needs 40" or 1m printing size (A0), viewed from a distance giving a 120° view (i.e., standing 23" or 60cm away).

An image printed smaller or viewed such that the borders are kept in view will not fool the eye into the original scenery. Which is not possible with a today's camera. But which was the base of my guess.


And a FF sensor with 81mp is way beyond the resolution of the FA 31mm:
- It is one of the finest glass around and should not stand behind Zeiss FF glass for the K mount. The latter is specified by Zeiss at 300 Lp/mm which is just the figure required to meet my spec. Of course, software needs to push the low microcontrast (5%) then which means that sensor noise needs to be very low.


96mp is just a little more than double 14mp resolution, it's not that big of a step:
- Exactly. It just means to go FF and to shrink the sensor cells from 5 to 3 microns. P&S sensor cells have already passed this size which is the smallest reasonable figure with today's technology.

[falconeye]

Originally Posted by regken

http://www.takinami.com/yoshihiko/ph...procedure.html, that shows the resolution test done on some of my old glass. What I'm trying to figure out is will my old lenses work on a FF sensor IF Pentax ever comes out with one.

This test used the USAF Lens Resolution Test Chart on Kodak T-MAX 100 film. It is based on observing the point where Siemens stars fuse together. Unfortunately, this is mostly limitated by the film and its grain at very low contrast. Not like modern laser beam MTF measurements.

In this particular test, I would consider any figure of 98 Lp/mm excellent and probably more than required for a future high end sensor. 98 is just 1 step below the tests maximum score (110), only reached by the FA50/2.8macro at f/11.

[thibs]

Originally Posted by rparmar

I do find this interesting but in the nearer term there are more practical concerns. The K20D already shows up the limitations of bad or low resolution lenses. A potential camera with more megapixels or a larger sensor would need even better glass. Consider the cost of your kit when the 31/43/77 is only the starting point -- not everyone can afford this.

Mmm but low end lenses(e.g. FAJ line) were really consummer low end grade and they showed a lot less performance / shrapness (+ ... just add what you want) than mid-grade or high grade lenses on film.

And film went better and better all over the years as well. How is the situation of sensors any different? It may go fatser, OK, but that's it.

[gkopeliadis]

Originally Posted by benjikan

...
I want to see a 25 to 30 mega pixel DSLR with noise levels similar to the Dalsa digital sensors at 100 iso. ...

I think that 15-18Mpixel is the upper limit in resolution that OPTICS can take advantage of at resonable prices. Lens are analog devices and we have reached the limits of that technology a long-long-long time ago (see how many ancient age lens are still holding up to the same optic level as today's top designs)

[falconeye]

Originally Posted by gkopeliadis

I think that 15-18Mpixel is the upper limit in resolution that OPTICS can take advantage of at resonable prices. Lens are analog devices and we have reached the limits of that technology a long-long-long time ago (see how many ancient age lens are still holding up to the same optic level as today's top designs)

The difference in resolution performance between sensor and film is just going to develop now. There was no need to have glass sharper than film in the past.

And even then, good ancient glass resolves at 300 Lp/mm still not reached by sensors.

Where utmost need for resolution existed (astronomy, espionage satellites) the optics are only bound by the limits of diffraction which is FAR beyond today's limits.

[gkopeliadis]

Originally Posted by falconeye

...Where utmost need for resolution existed (astronomy, espionage satellites) the optics are only bound by the limits of diffraction which is FAR beyond today's limits.

The key phrase is "reasonable prices" and I may add "reasonable size and weight".
A misalignment of a few micrometers between the lens elements is enough to destroy any fine resolution. Think now that photography lenses have a lot of moving parts (a telescope or spy satellites lens none) and can’t be constructed with exotic materials and methods and … voila!
Most today’s photographic lenses can’t even reach the 10Mpixel sensor’s Nyquist frequency on APS-C sensor size (on full frame sensors the theoretical limit is obviously higher for the same lens) except from the very center area. And that’s for perfectly still, flat targets with cameras perfectly focused on tripods with vibration free methods.

[mermozjg]

Originally Posted by falconeye

The difference in resolution performance between sensor and film is just going to develop now. There was no need to have glass sharper than film in the past.

And even then, good ancient glass resolves at 300 Lp/mm still not reached by sensors.

Where utmost need for resolution existed (astronomy, espionage satellites) the optics are only bound by the limits of diffraction which is FAR beyond today's limits.

a good article worth to be read about lens resolution and diffraction limits at luminous-landscape.com:
Do Sensors Outresolve Lenses?

[gkopeliadis]

Originally Posted by mermozjg

a good article worth to be read about lens resolution and diffraction limits at luminous-landscape.com:
Do Sensors Outresolve Lenses?

Exactly !

[rparmar]

Originally Posted by mermozjg

a good article worth to be read about lens resolution and diffraction limits at luminous-landscape.com:
Do Sensors Outresolve Lenses?

Um. I think I understood about one-fifth of that. And I have a degree in physics -- just not much exercised I suppose!

But I did follow the conclusion: "Only for highly corrected lenses (with better performance at f/5.6 than f/8) do higher sensor resolutions make sense." To be specific, for the APS-C sensor 13MP is enough, or 27MP for those "highly corrected lenses" of which the authors speak. The K20D, with 15MP, has already reached the first of these theoretical limits.

So even though I may not know my MTF from my MDF I believe I was justified in writing "The K20D already shows up the limitations of bad or low resolution lenses. A potential camera with more megapixels or a larger sensor would need even better glass. Consider the cost of your kit when the 31/43/77 is only the starting point -- not everyone can afford this."

That said, according to Photozone the FA31, FA43 and FA77 show best performance at f/4. I interpret the results to mean that these lenses will still be useful on a future 52MP camera.

(The FA50 is in the same league whereas the DA21 and DA16-45 correspond more with the f/5.6 to f/8 maximum resolution class. This seems to correspond well with common experience.)

With the K20D we have already approached the summit of performance for all but the most expensive and well-constructed lenses.

[falconeye]

Originally Posted by gkopeliadis

The key phrase is "reasonable prices" and I may add "reasonable size and weight".
A misalignment of a few micrometers between the lens elements is enough to destroy any fine resolution.

Yes, you may be right. But then, progress has always come from people completely ignoring such lines of argument A few engineers will accept the challenge implied by your concerns.

Originally Posted by mermozjg

Do Sensors Outresolve Lenses?

Good article!

The K20D is at 100 lp/mm.

I may add that existing resolution (above the Rayleigh criterion) can always be used to restore an undistorted, higher-contrast image (if noise is low), using computational optics. So, MTF50 isn't the only figure to look at.

Originally Posted by rparmar

I interpret the results to mean that these lenses will still be useful on a future 52MP camera. [...] With the K20D we have already approached the summit of performance for all but the most expensive and well-constructed lenses.

I agree with both statements.

But a 100 MPixel FF sensor wouldn't be that much more demanding (170 lp/mm, just 70% more).