[jsherman999]

I want high ISO with good quality, 24MP gives better light sensitivity.

No.

At the same time it is probably the quality of the sesnors available at design/build time that makes the choise "for us".

Yes, exactly. And the two data points that would most directly pertain to Pentax - because those are probably the sensors available to Pentax - show that 24MP is no cleaner than 36MP, at the same display sizes.

Now, Sony (or someone) could come out with a less-than-36MP sensor that had some SNR advantage over the current 36MP sensor, but it's not going to have better SNR simply because it has less MP, it's going to be better because it's a newer sensor with better QE. And then, they would probably be able to make a 36 or even 54MP version with the same pixel tech, same QE that has the same amount of noise at the same display sizes - with better resolution (just as we find ourselves now, like the current situation, one sensor generation removed.)

Don't fear the MP!

.

[dtmateojr]

I'm not quite sure how to take this comment - are you saying that Roger's bench testing is incorrect, or that the lens is so 'bad' so it's not showing the effect you expected to see? The 50mm in question is a pretty standard lens, ie it resolves quite well stopped down and isn't going to be an outlier in any way. Roger's team tests hundreds of lenses per year as part of their business, so he's going to know if something is giving 'bad' results right away. That link also has a few other lenses tested, all showing the same thing - 36MP giving more resolution than the lower-MP sensors at the same apertures, and past your f/11 point.

Here's a statement by Joseph James : "A common myth is that smaller pixels suffer more from diffraction than larger pixels. On the contrary, for a given sensor size and lens, smaller pixels always result in more detail. That said, as we stop down and the DOF deepens, we reach a point where we begin to lose detail due to diffraction softening. As a consequence, photos made with more pixels will begin to lose their detail advantage earlier and quicker than images made with fewer pixels, but they will always retain more detail. Eventually, the additional detail afforded by the extra pixels becomes trivial (most certainly by f/32 on FF)."

The bolded sentence describes what you see when you plot the resolution as Roger Cicala did and DXOmark did in the earlier plot I posted - the slope of the curve gets a little steeper with the higher-MP sensor past the MTF peak (meaning it loses resolution slightly faster,) but it never intersects the lower-MP curve until very low apertures, and not at f/11 with any of those tested lenses. Meaning, as Joseph says - it always resolves more. Past about f/16 with most lenses, the difference is going to be pretty small, and there's probably no difference between any sensor + lens at f/32, as Joseph says.

If you want to imagine perfect lenses only** and concentrate on just the Airy disk diameter vs. pixel size, you still are not seeing anything bad at f/11 at even large prints (16'' on long side in example below) and pretty close viewing distances (25cm)


(from here)

** and by the way, how are you defining 'perfect lens'? A perfect lens has traditionally been described as one that is sharpest wide-open, that does not suffer from any aberrations at all wide open and thus is 'diffraction limited' at maximum aperture. You mention a 'perfect lens' being diffraction limited at some other aperture, and that breaks the definition of 'perfect lens'. Just an aside.


.

---------- Post added 02-24-14 at 06:23 PM ----------



I don't see anywhere where anyone said details completely lost to diffraction were somehow recoverable, if that's what you were implying, I saw where someone said MTF can be affected by sharpening. (You don't disagree with that, do you? )

By the way, that person's methods and data is available if you want to contact them on that link.

See that's the problem with graphs with no labels. 1000 lpmm is impossible. For a 36mp sensor, a perfect lens (no abberation) can only do 175 lpmm (READ the table YOU posted for rough estimates of a good lens -- 159 lpmm at f8 sounds right). The only conclusion is that 1000+ is lpcm which is a very poor lens. Doh?!

Of course higher Mp will resolve a bit more than lower Mp sensors because no lens is perfect (i.e. a 12mp sensor can't fully produce 12mp). Very elementary. Really simple.

Here's my challenge for you, find me any camera FF or smaller, with any lens combo that can do 95 lpmm at f16. That's a conservative limit for diffraction at that aperture. Come back when you find one.

Edited: 95 lpmm @ f16

[jsherman999]

See that's the problem with graphs with no labels. 1000 lpmm is impossible. For a 36mp sensor, a perfect lens (no abberation) can only do 175 lpmm (READ the table YOU posted for rough estimates of a good lens -- 159 lpmm at f8 sounds right). The only conclusion is that 1000+ is lpcm which is a very poor lens. Doh?!

It's MTF50 values in line pair/Image Height.

Here's what Roger actually said (you should really read the links I provided, not trying to be snarky here, just take the time to read carefully, it pays dividends.)

"Here are the MTF 50 values in line pairs / image height for the center point, weighted average of 13 points, and average of the 4 corner points. Please note that the plotted average is NOT just the average of center and corners, so if the ‘average’ value is near the center, you know the lens stays fairly sharp in the middle regions, while if it’s nearly as low as the corners the lens falls off rapidly away from the center point."

.

Of course higher Mp will resolve a bit more than lower Mp sensors because no lens is perfect (i.e. a 12mp sensor can't fully produce 12mp). Very elementary. Really simple.

So... are you backing away from your previous point, then, that at f/11 and smaller there's no resolution benefit to the 36MP with any lens?

Here's my challenge for you, find me any camera FF or smaller, with any lens combo that can do 95 lpmm at f16. That's a conservative limit for diffraction at that aperture. Come back when you find one.

Edited: 95 lpmm @ f16

Why? What do you think that proves, and how does it relate to what you've said before?

.

[dtmateojr]

It's MTF50 values in line pair/Image Height.

Here's what Roger actually said (you should really read the links I provided, not trying to be snarky here, just take the time to read carefully, it pays dividends.)

"Here are the MTF 50 values in line pairs / image height for the center point, weighted average of 13 points, and average of the 4 corner points. Please note that the plotted average is NOT just the average of center and corners, so if the ‘average’ value is near the center, you know the lens stays fairly sharp in the middle regions, while if it’s nearly as low as the corners the lens falls off rapidly away from the center point."

.



So... are you backing away from your previous point, then, that at f/11 and smaller there's no resolution benefit to the 36MP with any lens?



Why? What do you think that proves, and how does it relate to what you've said before?

.

If that is MTF50 then it is even more impossible. MTF50 for 36mp is 87 lpmm or 870 lpcm for a PERFECT lens. Considering that a 36mp is not really 36mp (thanks to bayer interpolation) how can a lens go past the sensor resolution? ROFL!!!

[jsherman999]

If that is MTF50 then it is even more impossible. MTF50 for 36mp is 87 lpmm or 870 lpcm for a PERFECT lens. Considering that a 36mp is not really 36mp (thanks to bayer interpolation) how can a lens go past the sensor resolution? ROFL!!!

Please, read carefully. MTF50 values in LP/IH.

By the way, if you question Roger's methods re anything, he's a nice guy who responds to emails and probably would be glad to explain anything that's not clear to you, like Y-axis labels. For that matter, (in all seriousness,) Joseph James responds to emails too. I don't know how much real access you have to your professors (I sometimes had to schedule weeks in advance when I was an undergrad, depending on the prof) but you could and should talk to engineers and physicists already in the field who have been doing and explaining these things for years.

[dtmateojr]

Please, read carefully. MTF50 values in LP/IH.



By the way, if you question Roger's methods re anything, he's a nice guy who responds to emails and probably would be glad to explain anything that's not clear to you, like Y-axis labels. For that matter, (in all seriousness,) Joseph James responds to emails too. I don't know how much real access you have to your professors (I sometimes had to schedule weeks in advance when I was an undergrad, depending on the prof) but you could and should talk to engineers and physicists already in the field who have been doing and explaining these things for years.

Image height for full frame is 24mm right? The D800 is 4192 pixels "tall" which gives you 175 pixels/mm. MTF50 is roughly half of that so give or take 90 lpmm MAX.

So it's impossible. There, I showed you the simple algebra.

Edit:

Ok here's another angle to his test. It could be that he is able to detect approx 1000+ lines in a 24mm image. In that case he is still not maximizing the sensor which is capable of 2000+ lines at MTF50 for a 24mm image height.

The lens isn't at fault here but the methodology. His image does not have enough lines to max out the sensor.

[jsherman999]

Image height for full frame is 24mm right? The D800 is 4192 pixels "tall" which gives you 175 pixels/mm. MTF50 is roughly half of that so give or take 90 lpmm MAX.

So it's impossible. There, I showed you the simple algebra.

Edit:

Ok here's another angle to his test. It could be that he is able to detect approx 1000+ lines in a 24mm image. In that case he is still not maximizing the sensor which is capable of 2000+ lines at MTF50 for a 24mm image height.

The lens isn't at fault here but the methodology. His image does not have enough lines to max out the sensor.

Not quite sure what your point is - are you imagining the shape of the curve will change in either case? ( Or the curves will all of a sudden meet at f/11? )

LensRentals presents it in a form that simplifies cross-format comparisons better using that arbitrary scale (I think 500-1300 IIRC) - lp/mm is just as good but requires a conversion to extrapolate results between formats. That's all. The curves don't change, only the units do, they're both derived from Imatest, and it tracks DXOMark, SLRgear and photozone results.

Also, the methodology used in the "sharpening Joker's" tables is here. Make sure you look at the PDF linked near the bottom of the page. Looks like they use actual DxOmark data, and a Matlab function from Norman Koren. (you may have heard of him.)


.

---

Last edited by jsherman999; 02-24-2014 at 11:08 PM.

[dtmateojr]

Not quite sure what your point is - are you imagining the shape of the curve will change in either case? ( Or the curves will all of a sudden meet at f/11? )



LensRentals presents it in a form that simplifies cross-format comparisons better using that arbitrary scale (I think 500-1300 IIRC) - lp/mm is just as good but requires a conversion to extrapolate results between formats. That's all. The curves don't change, only the units do, and they're both derived from Imatest.



Also, the methodology used in the "sharpening Joker's" tables is here. Make sure you look at the PDF linked near the bottom of the page. Looks like they use actual DxOmark data, and a Matlab function from Norman Koren. (you may have heard of him... or not.)





.

I expect a sharp drop from f5.6 to f11 by approx 2:1 if the lens is good enough. Will check the other docos later.

Edit:

I posted a comment on that link hoping that he would reply. My point is that if there were only 100 lines in the image then f5.6 to f16 will show a flat graph. Increase it to 500 lines and you start seeing a curve. Increase it to 2000 lines to max out the sensor and you will see the huge difference between f5.6 and f11

[konquerian]

I voted for 24mp...
Long ago, 6mp is just enough, then suddenly upgraded to 14mp and to 16mp, which are really huge files especially that I always shoot in RAW format... I work as a photojournalist (also shoots for commercial) and photography is my life where I literally go out and shoot everyday - as in everyday... So, with my workflow and (imagine) with a 24mp sensor, that would take up a disk space fast more so with a 36mp...
Hmmm... Two system would be good, 24mp for everyday or fast-phase shooting, and 36mp for those commercial shoots...
If Pentax ever brings out a FF camera, two choices would be good just like what Sony did - a7 for speed and a7r for detail...

I think pentax is definitely capable of bringing a well all-around performance full frame camera to the table, combining 36mp for detail, speed on fps, and other never before seen technologies into one package. I loved the nikon d800e, but the fps was just too slow for sport and I would have loved an AA filter simulator with the camera, plus in body shake reduction. However, the pictures from that nikon camera did have amazing breath taking details. I would really want to see pentax coming out with a pro sport 36mp FF camera. In a heartbeat, I would sell all my junks in my garage and max out my credit to get one.

[Winder]

between the current 24MP and the 36MP, I'll take the 36MP. If there is a new 24MP that has better DR and S/N than the D4s, then I'll take the 24MP. I think the crop mode of the 36MP is a very important feature for Pentax since they have a limited FF lens line up.

I'm not overly concerned with sensor tech or MPs at this point. I'm more interested in AF speed/accuracy and quality glass.

[K David]

Can I say none? I'd like a Pentax FF to have a 16-MP sensor. That would give Pentax the ability with their best engine to have a ridiculous frame rate and burst mode that keeps going like a Nevada highway.

[normhead]

I'm a physics major. I eat gaussian curves. I major in optics and dsp. I know what I'm talking about I don't think you understand diffraction.

Ha ha, you've been ElJayed..

I like my 24 Mp K-3. It's a little disconcerting Nikon had to go all the way back to 16 Mp to get a significantly better high ISO performance. My next camera won't be purchased for Mp, it will be purchased for better high ISO performance...specifically lower noise. With the K-3 I'm cropping to 16 mp, and I can live with my K-5 by framing tighter. But i really am enjoying the ability to crop. So can I have 24 Mp FF super low noise ff sensor? No not whatever sensor is in the A7, D600, or D800. I'm not looking for a one stop difference. Anyone can fake a stop. I'd go for 3 though.

[ElJamoquio]

Ha ha, you've been ElJayed..

You really have no idea what that means.

[Wired]

16 or 36 for me. since 16 is not listed, 36.

I was not overly impressed with the sensor in my D600 when I owned it. But, the dynamic range and general look of the files the D800 are producing for me are much nicer than the D600 files and give me a lot more room to work. Even then I find less noise from my D800 than my D600 when I compare similar images. It also depends on crop and how you scale the final image too... of course hard drive space is a huge one and I'm buying hard drives as fast as possible because of those massive raw files!

I think in a Pentax DSLR FF I would prefer the 16mp Sony sensor (D4 sensor?) and see what Pentax could do with it. With what Nikon have been able to get from it, I assume Pentax could get even more out of it as they usually do! Then slap in a 9fps shutter. Make something like a D4 competitor, but for D600 price!

[jsherman999]

I think in a Pentax DSLR FF I would prefer the 16mp Sony sensor (D4 sensor?) and see what Pentax could do with it. With what Nikon have been able to get from it, I assume Pentax could get even more out of it as they usually do! Then slap in a 9fps shutter. Make something like a D4 competitor, but for D600 price!

Problem is that it doesn't look like the sensor used in the D4/D4s/Df is a Sony sensor, it's a NIkon-designed sensor, according to Chipworks. Something of equal capability might be available from Sony if there was demand.