Megapixels, megapixels and again megapixels

[Rainer SLP]

Sorry if this is not a new topic ...

Hi,

My head is quite dizzy due to reading about more, more and more pixels on the sensors.

Having done a lot of Astrophotography with high quality telescopes and dedicated Astrocameras, there is something called under sampling and over sampling and also some the " Nyquist " theorem or whatever it is called and have not really got to understand it more or less.

Now having read a bit about MTF charts and LP/mm and native aperture as well as f/8.0 aperture, meridional and saggittal lines etc. etc. etc.

Which lenses really do an adequate sampling on the nowadays existing High Mega Megapixel sensors like the 36mp or 50mp or the rumoured 56mp 35mm full frame sensor.

On the medium format sensor that is less a question for me due to the fact of the far bigger pixel size in microns as in the tiny pixel size of the 35mm sensors not to speak about the micro imaging chips of the smartphones 

¿ When is oversampling too much ?

If I take for example the Canon EOS 5D Mark III with its tiny 35mm and a pixel pitch of 0.00625mm which would equal to about  80 LP/mm.

All Canon MTF charts for normal lenses nowadays do talk about 10LP/mm and 30LP/mm and so if I take the tiny 22,118,400 pixel chip of the Canon 5D Marl III with a resolving power of 80LP/mm I am already over sampling by 2.6777x

Now if I take the resolving power of the soon available Canon 5Ds with 120LP/mm (pixel pitch of 0.00414mm) ¿ does it make sense ?, especially when I then afterwards, for printing a nice coffee table book, I go down to a printing pitch of perhaps 175 to 200 lines per inch (The printer house wants the images for this in perhaps 300 to 400 dpi)

Now a coffee table book (I have some nice books from the German Tecklenborg Verlag which are printed near the 200 lines per inch) has a size of 12" x 9.5" (31cm x 24cm) so that would mean a nice image in the size of 4800 x 3800 pixels is good enough for the printer ... (¿ a Canon 5D Mark III is already a good tool for a coffee table book ?  )

¿ Am I missing something ?

¿ Is all this not Overkill ?

.. or is it the " Faster, Higher, Fatter and Bigger etc. " Hype for selling more and more cameras from one certain brand ...

What can we expect in future in regard to better resolving lenses and at what prices

Sorry if this sounds like a rant 

[spidermike]

So what exactly is your question?

Since when was 35mm sensor 'tiny'? If you want 'tiny' check out this site on the performance of the APS-C 7D2

http://www.clarkvision.com/reviews/evaluation-canon-7dii/

[luxborealis]

Funny, I was sitting in my living room on the weekend having a quiet moment with coffee and the thought struck me of how unimportant megapixels are. Of the 15 photographs I have hanging (none of which are "family" photos):

4 were made with a 5mp Minolta 7Hi (3-16x20s and an 11x14);
3 were made with a 5mp Olympus E-1 (1-11x14, 1-16x20 and a 24"x48" canvas);
3 were made with a 12mp Olympus E-30 (2-11x14s and a 16x20);
2 were made with a 36mp Nikon D800E (both 16x20s);
2 were made with a Pentax 67 (film camera - both 16x20s); and
1 was made with a Zone VI 4x5 Field camera (11x14)

Now, I do have others that are "unhung" with a slightly greater proportion of 4x5, 12mp and 36mp, but the point is, it's the content of the work and its ability to conjure up emotions and connections that counts - not the megapixels.

If the end result is mostly books and the odd large print, then camera manufacturers have already plateaued, and, one could argue, surpassed the requirements of 3/4s of people who buy their DSLRs.

I must admit to absolutely loving the detail I get with the D800E, but for most people, I doesn't matter - again it's the emotional attachment that works for them. It is usually other photographers who pixel peep the framed works and they rarely buy other photographers work.

[Jim Kasson]

My head is quite dizzy due to reading about more, more and more pixels on the sensors.

Having done a lot of Astrophotography with high quality telescopes and dedicated Astrocameras, there is something called under sampling and over sampling and also some the " Nyquist " theorem or whatever it is called and have not really got to understand it more or less.

Now having read a bit about MTF charts and LP/mm and native aperture as well as f/8.0 aperture, meridional and saggittal lines etc. etc. etc.

Which lenses really do an adequate sampling on the nowadays existing High Mega Megapixel sensors like the 36mp or 50mp or the rumoured 56mp 35mm full frame sensor.

On the medium format sensor that is less a question for me due to the fact of the far bigger pixel size in microns as in the tiny pixel size of the 35mm sensors not to speak about the micro imaging chips of the smartphones 

¿ When is oversampling too much ?

If I take for example the Canon EOS 5D Mark III with its tiny 35mm and a pixel pitch of 0.00625mm which would equal to about  80 LP/mm.

All Canon MTF charts for normal lenses nowadays do talk about 10LP/mm and 30LP/mm and so if I take the tiny 22,118,400 pixel chip of the Canon 5D Marl III with a resolving power of 80LP/mm I am already over sampling by 2.6777x

Now if I take the resolving power of the soon available Canon 5Ds with 120LP/mm (pixel pitch of 0.00414mm) ¿ does it make sense ?, especially when I then afterwards, for printing a nice coffee table book, I go down to a printing pitch of perhaps 175 to 200 lines per inch (The printer house wants the images for this in perhaps 300 to 400 dpi)

Now a coffee table book (I have some nice books from the German Tecklenborg Verlag which are printed near the 200 lines per inch) has a size of 12" x 9.5" (31cm x 24cm) so that would mean a nice image in the size of 4800 x 3800 pixels is good enough for the printer ... (¿ a Canon 5D Mark III is already a good tool for a coffee table book ?  )

¿ Am I missing something ?

¿ Is all this not Overkill ?

IMHO, we're a long way from having sensor pixel pitches fine enough that they can extract most everything our lenses can deliver.

http://blog.kasson.com/?p=5905

http://blog.kasson.com/?p=5920

If you want more context, feel free to ask questions, and I'll do my best.

Why are things different in astrophotography? I'm not an astrophotographer, but the big telescope sensors that I've seen don't use Bayer CFAs, and resolution is for the most part (without tricks) limited by atmospheric turbulence.

Jim

[fdisilvestro]

We are nowhere near oversampling. There are plenty of lenses that out-resolve 36 Mpixel sensors. 
Moire issues would be nonexistent if that were true.

[BernardLanguillier]

The increase of MP is IMHO an order of magnitude less important than the improvements in dynamic range.

This being said, more MP is always better everything else being equal. Many lenses are good enough not to bottleneck the image quality delivered by the system.

Now, many applications are adressed perfectly fine by mid range cameras from an image quality standpoint, which may indeed be an excellent reason to stop funding camera companies and spent money on something else instead!

Then there is the desire to create the best possible craft which remains a trait of amateur practicioners in the noble sense of the word. It would in fact make perfect sense moving forward for pros to use increasingly lower end equipment compared to amateurs.

Cheers,
Bernard

[hjulenissen]

IMHO, we're a long way from having sensor pixel pitches fine enough that they can extract most everything our lenses can deliver.

http://blog.kasson.com/?p=5905

http://blog.kasson.com/?p=5920

If you want more context, feel free to ask questions, and I'll do my best.

Why are things different in astrophotography? I'm not an astrophotographer, but the big telescope sensors that I've seen don't use Bayer CFAs, and resolution is for the most part (without tricks) limited by atmospheric turbulence.

Jim

You plots are for a hypothetical CFA-less sensor, yes? So the point where increasing resolution has close to zero benefit would be even higher than suggested by your nice simulations?

I think that increased spatial resolution and increased DR goes hand in hand. The big question is if there will be enough interest and money in camera tech to support the (no doubt) large investements needed to get there.

-h

[Diego Pigozzo]

When is oversampling too much?

"Too much" in relation to what end?

[Bart_van_der_Wolf]

You plots are for a hypothetical CFA-less sensor, yes? So the point where increasing resolution has close to zero benefit would be even higher than suggested by your nice simulations?

Demosaicing will cost a bit of resolution, but not that much. Besides, we're seeing multi-sampling sensor designs with full RGB color sampling being introduced. It'll take a while before we reach the 2.0 to 2.5 micron pitch size that seems to suggest a point of diminishing returns with current sensor and lens technology. The half pitch multi-sampling sensors are a good step in the right direction (for stationary subjects and rock-solid tripods and subject supports), but they effectively have an up to 200% fill factor thus reducing the modulation at the Nyquist frequency pretty fast.

I think that increased spatial resolution and increased DR goes hand in hand.

If everything else would stay the same, yes. But sofar we're seeing increasing dynamic range despite shrinking photosite sizes, so clearly things are not staying the same. It looks like 'well-depth' has increased from some 1500  to 3000 electrons per square micron. Besides that, there are also new technologies under development that act more as single photon counters and use a kind of binning (either temporal or physical) to achieve dynamic range.

The big question is if there will be enough interest and money in camera tech to support the (no doubt) large investements needed to get there.

Well, nothing new under the sun there...

Cheers,
Bart

[hjulenissen]

The half pitch multi-sampling sensors are a good step in the right direction (for stationary subjects and rock-solid tripods and subject supports), but they effectively have an up to 200% fill factor thus reducing the modulation at the Nyquist frequency pretty fast.

A seemingly obvious compromise would be to do (e.g.) 16MP Bayer CFA sans OLPF in a IBIS M4/3 camera. Use the IBIS mechanics to simulate OLPF in hand-held mode (shaking the sensor) and to methodically step the sensor around in tripod mode.

That way you would have "better modulation at Nyquist" when doing multishot (or in other words: access to aliasing that can be contructively combined for stationary targets to do super-resolution) AND you get low-aliasing day-to-day usage hand-held. It might be that the simulated OLPF has somewhat worse characteristics than physical OLPF though:
http://www.dpreview.com/forums/post/53492844
http://www.dpreview.com/forums/post/51894737

Of course, when the point of diminishing returns in terms of MP is approaching, stuff like OLPF, color filtering/demosaicing artifacts etc should be pretty much moot: the optical system (along with camera shake etc) will be the prime limiting factors wrgt spatial resolution/artifacts.

-h

[kers]

If I take for example the Canon EOS 5D Mark III with its tiny 35mm and a pixel pitch of 0.00625mm which would equal to about  80 LP/mm.
All Canon MTF charts for normal lenses nowadays do talk about 10LP/mm and 30LP/mm and so if I take the tiny 22,118,400 pixel chip of the Canon 5D Marl III with a resolving power of 80LP/mm I am already over sampling by 2.6777x

If i understand it correctly you have to include the bayer pattern so the resolution will be less.
(please correct me if i am wrong, technical photographers on this forum)

According to this site the best lens for 35mm is the Otus 85mm 1.4 and its
"Maximum resolving power is 107MP at an aperture of f/4.1, resolving power wide open is 85MP. "
source :     http://www.lenscore.org      on the otus 85mm page.

So that leaves some room for progress on the sensor side to show the true lens quality.


What i notice myself on 36MP is that most lenses do fine in the centre, but lack 36MP performance in the corners.
Also you see - especially with zoomlenses- some parts on the image that have less quality than others and also some unexpected sharp-unsharp areas due to field curvature with wide angles.

That said, i agree with Terry McDonald that for most use the amount of technical quality is not relevant at all. But i do architecture and then you like to see those fine details ... in the corners as well.

At the same time with increasing MP's the main platform to publish has shrunk from magazines to the internet.  

[Jim Kasson]

You plots are for a hypothetical CFA-less sensor, yes? So the point where increasing resolution has close to zero benefit would be even higher than suggested by your nice simulations?

No, they are for a Bayer CFA sensor with RGGB layout.

Jim

[Iluvmycam]

OP, for your work go with the highest MP you can get.

This was shot with an old 6mp cam. it is in a number of museum collections. For street work almost any MP will work...under good lighting.


https://danielteolijr.files.wordpress.com/2015/03/carla-and-babydoll-copyright-2012-daniel-d-teoli-jr-mr.jpg


For tough lighting, higher MP is welcome. This was with a 16MP Fuji, but my 24 MP Leica would not even do as good. (Nor would I even risk using it in the Red Light District in this circumstance.)


https://danielteolijr.wordpress.com/2015/04/13/an-example-of-push-processing-a-digital-image/

My ideal street cam would be a Leica knockoff by Fuji with a 36 MP organic sensor in it. Camera priced at $2000. Then it could produce in tough light and be cheap enough to risk using in dangerous areas.

[PeterAit]

Funny, I was sitting in my living room on the weekend having a quiet moment with coffee and the thought struck me of how unimportant megapixels are.

Bravo!! My sentiments exactly. IMHO, there is way too much focus on megapixels and other tech aspects of photography here on LuLa. Yes, we need to be conversant with those things, but they are fundamentally boring and, let's face it, not at all challenging. If you are a bad photographer with a 5 MP camera, you will still be a bad photographer with a 50 MP camera.

[Rainer SLP]

Hi Everybody,

Thanks for the great answers and as I see there is a certain consensus about too much megapixels and not too much megapixels and that some existing lenses do outperform the resolving power of the camera sensors and so there are others which do not do it.

Having mostly Canon lenses from the L serie, and so based on the answers above, I can buy the new Canon EOS 5Ds and my lenses will give more resolution on the 50megapixel chip and I will be able to crop out interesting parts from my 50mp images.

Also this will give me better prints for crops of the original images.

¿ Right ?

[Diego Pigozzo]

...
I can buy the new Canon EOS 5Ds and my lenses will give more resolution on the 50megapixel chip and I will be able to crop out interesting parts from my 50mp images.
Also this will give me better prints for crops of the original images.
¿ Right ?

You're the only one knowing if your current camera is a limiting factor for your photography.

[Rainer SLP]

We are nowhere near oversampling. There are plenty of lenses that out-resolve 36 Mpixel sensors. 
Moire issues would be nonexistent if that were true.

Hi Francisco,

¿ Is there somewhere a page where I can take a look of the resolving power of lenses ?

Yesterday I found LenScore but did not find how they measure the sesolving power of a lens.

Thanks

[Rainer SLP]

"Too much" in relation to what end?

Hi Diego,

In relation to printing ...

[Rainer SLP]

If i understand it correctly you have to include the bayer pattern so the resolution will be less.
(please correct me if i am wrong, technical photographers on this forum)

According to this site the best lens for 35mm is the Otus 85mm 1.4 and its
"Maximum resolving power is 107MP at an aperture of f/4.1, resolving power wide open is 85MP. "
source :     http://www.lenscore.org      on the otus 85mm page.

So that leaves some room for progress on the sensor side to show the true lens quality.


What i notice myself on 36MP is that most lenses do fine in the centre, but lack 36MP performance in the corners.
Also you see - especially with zoomlenses- some parts on the image that have less quality than others and also some unexpected sharp-unsharp areas due to field curvature with wide angles.

That said, i agree with Terry McDonald that for most use the amount of technical quality is not relevant at all. But i do architecture and then you like to see those fine details ... in the corners as well.

At the same time with increasing MP's the main platform to publish has shrunk from magazines to the internet.  

Hi Kers,

But i do architecture and then you like to see those fine details ... in the corners as well.

And here we need both, High pixel count and high resolving lenses ...

¿ correct ?

and now comes the question again for architecture. Normal to long focal lenses do not have those problems but it gets critical for wide angle and ultra wide angle lenses ...

I would like to see an image taken inside a house or a church with strong decorated (painted) ceiling with the Canon EOS 5Ds and the new EF 11-24mm f4L zoom lens as well as with the new EF 8-15mm f4L ... be it circular or full frame fish eye

 

[Rainer SLP]

You're the only one knowing if your current camera is a limiting factor for your photography.

Yes ...