[normhead]

With a raft of new cheaper Full Frame cameras coming out the age old question is going to be asked. DO I need the 35 MP of a Nikon D800 or can I get by with a Cnaon 6D. But if you're going to start comparing the latest offerings or even Pentax's latest strategy then you need to understand some of the concepts here. One of the big questions of course "why do I need more mega pixels and another would be is bigger better in MP?

Do Sensors ?Outresolve? Lenses?

If you're not familiar with the issue here's a brief synopsis.

At this point it is theoretically likely that sensor resolution has surpassed lens resolution. Read the above article for a more detail explanation. This isn't the only place I've read this by the way. I've read it from enough different sources including peer reviewed research papers, I'm comfortable with the findings. At that point adding more MP doesn't add more detail it ads more pixels. So in essence you could produce the same picture by putting the image in photoshop and increasing it's size. Same resolution, larger image.

Flip down through the charts graphs and formulas and at some point you'll find the theoretical limits.

APS-c 16 MP
FF- 35 MP
MF 70 Mp

That means that after those numbers it's not theoretically possible to increase image detail by adding more pixels, all you add is more pixels. Now you could say great… the Pentax K-5 is 16 MP, I don't need anymore. The Nikon D800 is 36 MP, I don't need anymore. Theoretically. And this can be seen with the current Oplympus 24 Mp APS-c camera. There is a link to one of the reviews above. Apparently, and I have to take the reviewers word for it, the increase in MP does not lead to a corresponding link in detail reproduction.

OK, so that's the theory. What's the pracitce.

Most of us don't have the technical ability to test anything to do with optical theory. I wouldn't even know how to test the accuracy of the f-stop on a lens etc. But somewhere in the back of my mind is a quote from Falconeye , who has done way more testing than most of us on this site will ever even conceive of doing.. suggested the practical limits of APS-c was 6 Mp. I've heard enough people suggest that a D700 gave them such good images they had no need to upgrade past 12 MP for FF. The only reason it stuck with me was it seemed like such an outrageous statement at the time.

So, I'm sure by this point you've read enough and you're saying, well where are you going with this?

IF the maximum detail your camera system can capture is 6 MP, what it the advantage to 16 MP. IN terms of detail, none, So essentially the sensor wars , in terms of MP should be over. It's quite possible that the *ist and the D700 had already captured every thing all but the most sophisticated, expensive lenses had to offer. From the theory you can deduce that both 16 MP APS_c and 35 MP FF are overkill. They are expressing a theoretical max that for various other reasons, lens distortions , different wave lengths of light, diffraction etc, can never be achieved.

SO having achieved that, where is a camera maker to go to improve IQ?

Well obviously from this analysis, the first place to go if improving your image is paramount is to a dynamic range and color depth. Honestly when I look at the images from my 14 MP k20D and my K-5 the improvement in the K-5 is color depth. Richer color. Deeper color. Color that draws you right into the picture. Pentax, realizing that their sensor is already achieving the theoretical max in detail with available pixel size and the resolving power of current lenses, has turned their attention to the other parts of IQ. Noise reduction, dynamic range, low light capacity and color depth. Looking a the latest offerings from Canon and Nikon, the 6D and D600, you can see the same trends. The days of increasing sensor IQ by increasing image size are past for the moment. Camera makers have to find other ways to make their cameras better.

But for the FF frame shopper , at this moment. one has to ask.. are the D700 owners right? Do the pracitical limits of lens resolution mean there is no additional advantage to increasing sensor size. And are the cameras they are releasing well over the practical limit for image detail. Or have both Nikon and Canon calculated the practical limit for inclusion in their FF cameras. Is there any point in being a Nikon D800 owner if you aren't going to pay for one of this out of sight costly diffraction reduced lenses. Will you get just as much detail, better low noise and low light performance from a D600 or a 6D? It certainly looks like Canon went for best low noise, low light sensor in theirs., and if the 6D sensor is better than the practical limit of the lenses, then would you achieve a better result expanding a 20 Mp 6D image to 35 Mp, than you would taking the same picture with a D800? Should Pentax listen to falc and go back to say a 10 MP sensor with better low noise and low light capability. IN a practical sense, it's pointless building a sensor that exceeds it's usefulness according to the theoretical limit of the lens. What we need to know is the practical max. With it's excellent low light performance and dynamic range, that 6D could be the bargain of the day. There's just no way we'll ever probably ever know. WIth both Canon and Nikon bringing out FFs in the low 20s, I'm willing to bet the practical max for FF is somewhere in that range. IN fact with low, light performance, low noise and color depth being the sleeted areas of improvement, you have to ask, will camera manufacturers roll their MP back to lower levels in order to achieve in these suddenly way more important areas of competition.

I know if I was a camera manufacturer, that's where I'd be looking. Looking at the Pentax K-5 II, my guess is contrary to what many on here have suggested, they are staying ahead of the game. The game has changed, though it may take the consumer a while to realize it.

---

Last edited by normhead; 09-22-2012 at 09:45 AM.

[JinDesu]

One must also consider the AA filter for bayer sensors. This reduces the resolution at the sensor.

The k-5 and the D800 both have a sensor pitch of about 4.8 (and it makes sense, the D800 has about the same amount of pixels per area as the k-5). If considering the APS-C sensor of the k-5 has the greatest high ISO performance and DR of all APS-C sensors, would it not make sense for that sensor to be used as the FF sensor as well? And as far as I can see, the D800 sensor does perform as such - someone (possibly Falk) did an extrapolation of how the k-5 sensor would perform at FF and it matched what the DXO mark of the D800 is.

[jsherman999]

One of your more well-reasoned posts, Norm! I think there are just a few misconceptions and I have thoughts on a lot of points but no time right now, so I'l quickly start with: Don't forget about cropping, and the positive effects of downsampling.


.

[Rondec]

The question is really whether the numbers you have quoted are accurate, Norm. My understanding is that good lenses can out resolve even current 24 megapixel APS-C sensors (can't tell how Pentax lenses would do since there isn't a camera available with such a sensor). So, that would mean that they could resolve up to 50 megapixels on a full frame camera.

The problem is obviously that the higher the pixel density, the less likely you will see benefit. Little problems in technique, going up in iso, basically anything that can cause some blur will fudge away any difference between your 16 megapixel and 24 megapixel APS-C sensor. Still, you can probably see the difference in low iso in a studio.

[bkpix]

There could be something to the 12 meg limit, as you mentioned with the D700. I more easily get clear, sharp images with my 12-meg 1Ds than I do with my 16-meg K-5, though they're in the same ballpark. This has always seemed a bit curious, though I've generally blamed the fact that it takes a sharper lens to make an image of the same apparent sharpness and resolution on crop than on full-frame.

About the bigger question, I don't think that optical resolution is the sole defining characteristic of a lens and body; in other words, a sensor than can "out-resolve" a lens might still show elements of that lens' personality, for lack of a better term, that a sensor with less resolution might miss.

[Clicker]

You have to look no further than my Canon 7D for pixel density, Canon crammed it full of pixels and if your technique (shooting and PP) isn't polished you're in for a hell of a ride.
Over at POTN there's always someone (me included) who had initial problems with working an image with the 7D and those who just wouldn't listen and iron out the images just packed up and left never to look at another Canon i'm sure ...i'm still working the issue

[rawr]

Printing is also another element in considering the pros and cons of camera pixel count and lens resolution. Although there too the issues are complex - more megapixels (full-frame or otherwise) don't automatically mean bigger and better prints.

I also like to think that my K-5 isn't really limited to 16MP at all.

Thinking in 2D terms, panorama software (hugin, MS ICE etc) is so powerful and easy to use that I don't think of the K-5 as a 4964x3268 pixel camera anymore. I can easily shoot and print 15000x 3000 pixels if I want to without even busting a sweat. Thinking in 3D terms, stacking software lets me pile up 16MP shots into something equivalent to many more megapixels than are available in the camera hardware.

[normhead]

Again from luminous landscape from their notes on Photokina 2012.

The D800 is what analysts refer to as a “disruptive technology”. It shook the bedrock of the industry and changed the realities of what-was-what. But even before the D800 blew the roof off, the latest range of 24MP DSLRs were already starting to stretch the seems of the lens world. Simply put, most of the glass out there just isn’t good enough to take full advantage of 24, much less 36 megapixels much of the time.
And so the world turned.

Also on the ‘start saving’ front…Zeiss has introduced the most unusual and interest lens of Photokina – a 55mm f1.4 the size of large squash. Seriously, this thing is the size of an average “H” mount lens from ‘Blad. Think Nikon 14-24, just bigger.

And what’s all that size about? Well it’s about $3,000 or so. For that, however, Zeiss is promising that the heavens will part and pixelly goodness will rain from the heavens for frustrated high-end 35mm DSLR users who feel they just aren’t getting the most our of their camera with conventional glass.

I remember years ago a question in the general photography class, where a student asked an instructor what the difference between an expensive camera and a cheap camera was. The answer, "It costs a lot more to repair an expensive one." At this point it's starting to look like having a higher megapixel camera means you have to buy really expensive glass to get the most out of it.

[Nesster]

At this point it's starting to look like having a higher megapixel camera means you have to buy really expensive glass to get the most out of it.

Not to mention a ton of disk storage.

Excellent thread

[JinDesu]

I don't know about "most glass" - Again, it's the same pixel pitch as the k-5. Unless the k-5 has been outresolving "most glass"....

[jsherman999]

I remember years ago a question in the general photography class, where a student asked an instructor what the difference between an expensive camera and a cheap camera was. The answer, "It costs a lot more to repair an expensive one."

I've always felt that was true about almost all 35mm film cameras. DSLRs can contain parts that literally cost 10x as much as lower-end ones though, although there are bodies that land way outside of the value curve. Here's what Mike Johnston wrote about DSLRs and Leica and Veblen Goods:

"
As I've mulled over the worlds of high-end audio and photography, it increasingly seems curious to me that almost all products in high-end audio are Veblen goods, whereas almost none of the products in photography are. Top-end Canons and Nikons, while very expensive for cameras, are really not unreasonably priced given their technology and economics of production. Leica is essentially the only true Veblen good in photography.
"

At this point it's starting to look like having a higher megapixel camera means you have to buy really expensive glass to get the most out of it.

This isn't really true - what's true is that you have to buy high-resolving glass (which isn't always 'really expensive'.) Also, it's very important to know this: To maximize all those MP, you have to have high-resolving glass and use mirror lock-up, heavy tripod, etc - but to just get more resolution than you even have before, even using the same lenses, you only need to practice good technique - ie, don't do things that introduce motion blur, like use bad handholding technique or too-low shutter speeds. (and if you do use bad technique, you don't get 'worse results', you just don't get any better results than the lower-MP sensor would have given you using the same technique.)

These discussions got pretty involved in the Nikon forums right after it was announced that the D800 would have 36 MP. Some folks were worried that there would be no benefit to it unless you only owned Zeiss and only used tripods, but people like Luke Kaven and the creator of the sensorsgen site would often appear to set the record straight - that basically all your lenses perform better on a higher-resolution sensor under the same shooting conditions - even if you don't do things that 'max out' that sensor every time.

Here's an example - a 50mm f/1.8 on a 12MP FF sensor vs a 24MP FF sensor - same lens:



Now, if you were going to try to 'max out' that sensor by using a $1300 Zeiss 50 f/2 macro planar on that 24mp d3x, you might bump up a bit over that red line in places - but if you were to use that same Zeiss on the 12MP d3, it's MTF curve would stil be under that red 24MP curve at probably every point. In other words, the $200 lens + 24MP combo would outresolve the $1300 lens on 12MP. (But just not any $200 lens - it would have to be a fairly high-resolving lens - like that $200 50 1.8 happens to be.)


.

[normhead]

Here's an example - a 50mm f/1.8 on a 12MP FF sensor vs a 24MP FF sensor - same lens:

I guess the next question would be would you get the same increase in resolution going from a 24 MP sensor to a D800 @ 36 MP. Those are the MTF numbers I'm waiting for. Saying that theoretically, the better sensor out performs the weaker sensor every time ignores the concept of a theoretical max. The whole jist of the article quotes would be that there is. Luke Kaven would seem to be saying there isn't? I't would almost seem to be a science versus faith argument. I would find it odd that someone would post response to a scientific theory, with data that didn't approach the limits discussed. My personal assumtion would be that the limit is somewhere between 20-36 for an FF camera. SO I'd like to se the MTF numbers for the same lens at 20, 24 and 36 Mp, since the FF cameras that interest me at the moment. Although really because I shoot a lot of sunsets, I probably want the FF with the best shadow detail and dynamic range. That is where the K-5 shines over the K20-d and I'm betting that will be where the K-5 II will shine over the K-5, and possibly where the 6D will shine over the D600 and D800.

[JinDesu]

Again - unless the k-5/D7000 has been outresolving FF lenses, the D800 isn't going to be any different. It has the same pixel pitch. And if you take the k-5 sensor and make it into FF, it matches all the properties of the D800.

I don't think the 6D is going to beat the D600 and D800 in DR. Canon hasn't been exceptional in their sensors with regards to noise and DR (their best DR sensor is in the 1Ds Mk3 ranked 47th on DXOmark at 12Ev...)

[Verglace]

This article assumes lenses don't improve over time. Sensors may resolve higher then current or old lenses, but that doesn't mean future lenses would be out resolved by it. Like sensor technology, the technology that goes into lenses also improve over time. while not necessarily always true, it is generally true.

There's this myth that lower resolution = better (which runs opposite to more = better). But that's not necessarily true, the sensor on the k-r(12mp) for example is clearly inferior to that of the k-5's (16), if the lower = better argument holds true this would not be the case. The fact is there is more to sensors than just the mega pixel count.

As far as I'm concerned I don't care if they increase the megapixel count so long as the overall performance is increase. Hard drives are extremely cheap, and computers right now are ridiculously fast, if people have trouble reading big files with their computer then maybe they should consider diverting lba funds to upgrade their computer. I also wont mind a low pixel count so long as the overall performance is increased (as in the case of the current flagships).

So what would I do if I were a manufacturer? I would go with whatever the customers actually want rather than what they should want.

[jsherman999]

There's this myth that lower resolution = better (which runs opposite to more = better). But that's not necessarily true,

It is a myth, but people can be forgiven if they still believe it, because for a while sites even like dpreview were sticking to the 'larger pixels = less noise, always' claim. Later on evidence started to appear that at similar QE, more pixels over the same area weren't necessarily detrimental to overall image noise and in some cases better, while giving you the other benefits higher-res brings.

...Hard drives are extremely cheap, and computers right now are ridiculously fast, if people have trouble reading big files with their computer then maybe they should consider diverting lba funds to upgrade their computer.

Going from 12 & 14MP to 36MP with my current computer was kind of a pain. Everything just slows down, every little task that was instant before now maybe takes a few seconds. It's nothing terrible, but add all the little slowdowns together through the workflow and it bumps up the annoyance factor a bit. But I'm doing my work on a 1st gen i3 with 4GB ram, no ssd or anything, so it's not going to be as noticeable if your workstation is up to speed.

So what would I do if I were a manufacturer? I would go with whatever the customers actually want rather than what they should want.

That's the conventional wisdom, but I wonder if that's the wrong approach with technology.

Sometimes folks ask for the wrong things, because they think those things are better.

Steve Jobs used to say that Apple customers would know what they wanted when he gave it to them, and that sounds flippant, but I think he was right. The Nikon fora was thick with folks wanting 'no more than 16MP' for the D700 replacement, absolute handwringing over anything more than that being 'noisy', and then when the D800 appeared everyone saw the real benefits and saw that their concerns about noise were based on experiences with older tech, incorrrect assumptions and some myths.


.