Resolution thoughts from the sideline

I've been following many of the MP discussions about the MIII and D800. The ones that interest me the most discuss the ideal number of pixles on a cmos in order to best resolve images from the best available glass without over-exposing the weakness of the glass. We can all do the file size/storage/processing math, but the question of glass-resolution seems to be theory when it comes to the question of "which comany got things right?" with the lastest camera, at least until others can shoot side-by-side with similar glass and look at some details on print.

One thought I had as I read the dpreview preview of the MIII, is that if the MIII is really a FF 7D then should it have been a equiped with a 28.8 MP Cmos (1.6x18MP) or does the 7D already over-resolve/expose the weakness of available glass? Did Canon find 22MP was a more suitable resolution given available glass?

Thoughts?

--
Summit County Photog
5D & Glass

This whole high res/outsripping the lens debate heats up during releases. At other times, we just refer to lenses that fall off too much in the corners as crap lenses.

summitphotographer said:

I've been following many of the MP discussions about the MIII and D800. The ones that interest me the most discuss the ideal number of pixles on a cmos in order to best resolve images from the best available glass without over-exposing the weakness of the glass. We can all do the file size/storage/processing math, but the question of glass-resolution seems to be theory when it comes to the question of "which comany got things right?" with the lastest camera, at least until others can shoot side-by-side with similar glass and look at some details on print.

One thought I had as I read the dpreview preview of the MIII, is that if the MIII is really a FF 7D then should it have been a equiped with a 28.8 MP Cmos (1.6x18MP) or does the 7D already over-resolve/expose the weakness of available glass? Did Canon find 22MP was a more suitable resolution given available glass?

Thoughts?

--
Summit County Photog
5D & Glass

Click to expand...

--

Rick Knepper, photographer, photography never for sale, check my profile for gear list and philosophy.

If a 7D is extrapolated to full frame, it becomes approximately 43mp. The criticality of this kind of pixel density brings out all sorts of flaws not only in lenses but also focus and steadiness technique, which is why so many people had issues when viewing their work at 100% on their computer screens.

Frankly for most real photographers that have work to do, full frame and 16-20 megapixels is more than enough, and most lenses and photographers are up to the task of taking advantage of that. (Heck, I still can't believe how good my old 20D's photos are when printed at 20x24!)

The appeal of higher megapixel photography is very real, though, when your target audience is browsing photos in gigapixel-like fashion on a computer monitor, which is an increasing occurrence these days in some circles. It's undeniably cool, but not every photographer's goal.

Drew

summitphotographer said:

One thought I had as I read the dpreview preview of the MIII, is that if the MIII is really a FF 7D then should it have been a equiped with a 28.8 MP Cmos (1.6x18MP) or does the 7D already over-resolve/expose the weakness of available glass? Did Canon find 22MP was a more suitable resolution given available glass?

Click to expand...

First off, you can't out resolve glass. Higher MP will always provide more information/ detail. That goes even if there are only large details in the picture as the edges will be better defined, or if there are no edges, as gradual tonal changes will be better defined. (Obvious exceptions exist; take a picture of a black wall and you could do perfectly well with one pixel. Likewise differences will be smaller if all edges are perfectly vertical and horizontal).

If the 5D III had the same pixel density as the 7D it would have some 45 MP. (You have to go by differences in surface area, not diagonal).

The focus and motion problems doesn't make any difference with regards MP. More MP will still yield more information, and if anything make corrections easier.

It depends on what you want. If you want great pixels you need few and large ones, but if you want a better photo as such more MP will be key - even if every pixel is a bit worse off. I take photos, some prefer pixels.

(We still have the anti-aliasing filter for crying loud. It is there because the resolution is too low for fine details and edges at an angle. We pay huge amounts for better lenses, and then have a blur filter on the sensor...).

So 1.6x18MP does not yield the correct conversion for the MP's.

Another question, is the image quality of the 7D sensor much improved over let's say a D60 or is it marginal, for that matter is the 7D image quality much improved over a 5D when you take the same crop- surely there is a way to see if the increased MP's count.

You are correct if you are printing, but if your goal is gigapixel-type viewing on a computer monitor at 100%, then you are not going to like what you see unless your lenses, focus, and camera shake are all under control.

Drew

summitphotographer said:

Another question, is the image quality of the 7D sensor much improved over let's say a D60 or is it marginal, for that matter is the 7D image quality much improved over a 5D when you take the same crop- surely there is a way to see if the increased MP's count.

Click to expand...

I've never had a either a D60 or 7D so this is hard to comment. I have borrowed a 7D once though, and if I remember correctly the 5Dc batters it on pixel level and hard to tell on picture level. This is not a valid test though:

As these as are different sensor sizes you are not actually comparing different pixel densities in isolation, and the 5D collects a lot more light overall. To compare the effects of pixel size you should really compare just the center picture bit from the 5D that comes from an equal area on the sensor as the sensor size of the 7D. The 7D would win that easy with some 18MP versus about 4-5 MP or so. (Each pixel would still be worse though).

Fair enough, but then you are photographing for pixels. Which would also be better if you downsized a higher resolution image to whatever resolution you want to view in. (At the very least in downsizes in factors of 2).

way back when the discussion was about pixel size. Glass and CCD/CMOS aside, the discussion was about how two sensors of the the same MP's but different sizes yielded different photo qualities, the larger size being superior. So, has this thought that changed? What's the optimum pixel density for a FF sensor? Surely Dpreview guys could revisit this in a test.

summitphotographer said:

way back when the discussion was about pixel size. Glass and CCD/CMOS aside, the discussion was about how two sensors of the the same MP's but different sizes yielded different photo qualities, the larger size being superior. So, has this thought that changed?

Click to expand...

No, that is correct. The change is that for equal size sensors more MPs are better even if they are smaller. Slight exception for high ISO.

"What's the optimum pixel density for a FF sensor? Surely Dpreview guys could revisit this in a test."

More will always be better as long as each pixel still have a meaningful signal to noise ratio. Personally I think we will have full frame sensors with several hundred MPs within a few years from now.

BadBeta said:

More will always be better as long as each pixel still have a meaningful signal to noise ratio. Personally I think we will have full frame sensors with several hundred MPs within a few years from now.

Click to expand...

Is this universally accepted?

summitphotographer said:

summitphotographer said:

More will always be better as long as each pixel still have a meaningful signal to noise ratio. Personally I think we will have full frame sensors with several hundred MPs within a few years from now.

Click to expand...

Is this universally accepted?

Click to expand...

Looking at this forum I will have to say no. Technically speaking though it is hard to argue against, with the exception of low ISO. (For as long as there is an element of extra noise introduced per pixel).

Naturally I'm quite interested in a seeing the new Nokina cell phone with a 41MP camera in a sensor much smaller than both APS and full frame. Reports are good though.

No. At least not until you approach 100% quantum efficiency. But it's truer than it was some years ago. Each photo site needs a certain amount of space around it and it can't capture all the light that hits the sensor. The smaller the pixels, the more wasted space between photo sites. Microlenses and backlit sensors, etc. have helped improve this, but larger pixels will result in better DR and better low light. But as sensor design improves this will matter less and less.

Hundreds of megapixels is a bit outrageous. Even if the technology is there, lenses are diffraction limited at reasonable stops way before that. Large format landscapes are often shot at f32 or f64 to get enough depth of field (equivalent to f8 or f11 on APS-C, I think?) and by then you're diffraction limited somewhere around 20 or 30 megapixels I think (this is all fuzzy math due to tons of factors)... And you still get great wall-sized prints. So the world's best lenses being used at shallower stops (f4 or so) might benefit from 100 megapixel sensors, but past a certain point not so much... With great lenses the d800e might even approach 4x5 in terms of resolution at normal (equivalent) stops--and past that point it really will "out resolve" most lenses, since few lenses are diffraction limited below f8. Also, 4x5 turns out flawless prints up to 40'x50' even at deep stops. I've seen 80'x100' prints from 8x10 that benefitted from the additional image quality, but those I believe the prints cost thousands of dollars each to print. For FX and DX, at least, we're approaching a point where lenses will be "out resolved" in that we'll legitimately be at 0mtf... Tonality would still benefit a bit, though, but again...that would only matter for prints more expensive to purchase than the camera.

Matty W said:

No. At least not until you approach 100% quantum efficiency. But it's truer than it was some years ago. Each photo site needs a certain amount of space around it and it can't capture all the light that hits the sensor. The smaller the pixels, the more wasted space between photo sites. Microlenses and backlit sensors, etc. have helped improve this, but larger pixels will result in better DR and better low light. But as sensor design improves this will matter less and less.

Click to expand...

Assume same quantum efficiency, and assume that Canon is not lying about its gapless sensors.

As for the diffraction limit and all that I've written that here before so I wont repeat in every thread here. Take a look at http://luminous-landscape.com/tutorials/resolution.shtml , table 3, and then double that to ditch the anti-aliasing filter. Then you go multiply a few times still as the Airy disc isn't a flat plateau, and you could expect resolution improvements at least up to 9 or 16 pixels per Airy disc.

On your general assumptions...

On the subject of the relationships between sensor and lens resolution, there are three possibilities:

1. Lens outresolves sensor - In this case, we would perhaps still like to see better resolution from the sensor so that it could fully capture the resolution provided by out best lenses. So, having the lens out resolve the sensor is not where we want to be.
2. Lens and sensor resolution are equal - Unfortunately, this is an impossibility. Lens resolution varies across the frame, by aperture, and among lenses. If you were to try for the "equal" outcome, I suppose you would be looking for sensor resolution high enough to resolve the very best image that could be produced in the best part of the frame at the ideal aperture on the best lens available.
3. Sensor outresolves lens - This makes the most sense, and for a variety of reasons. Obviously, such a sensor would be "up to" any lens you might use. Additionally, the finer pixel pitch could contribute in some small way to smoother gradients.

Also, understand that the way lens resolution is measured, it is not exactly a binary. In other words, it isn't that case that a lens "resolve" X lp/mm but doesn't resolve X=1 lp/mm. This is based on a somewhat arbitrary point at which the ability to maintain contrast between alternative light and dark bands degrades to a certain level of contrast. These things are a bit more complex and messy than some want to realize.

Dan

summitphotographer said:

I've been following many of the MP discussions about the MIII and D800. The ones that interest me the most discuss the ideal number of pixles on a cmos in order to best resolve images from the best available glass without over-exposing the weakness of the glass. We can all do the file size/storage/processing math, but the question of glass-resolution seems to be theory when it comes to the question of "which comany got things right?" with the lastest camera, at least until others can shoot side-by-side with similar glass and look at some details on print.

One thought I had as I read the dpreview preview of the MIII, is that if the MIII is really a FF 7D then should it have been a equiped with a 28.8 MP Cmos (1.6x18MP) or does the 7D already over-resolve/expose the weakness of available glass? Did Canon find 22MP was a more suitable resolution given available glass?

Thoughts?

--
Summit County Photog
5D & Glass

Click to expand...

--
---
G Dan Mitchell - SF Bay Area, California, USA
Blog & Gallery: http://www.gdanmitchell.com/
Google Plus: https://plus.google.com/u/0/102554407414282880001/
Facebook: http://www.facebook.com/gdanmitchellphotography
Flickr: http://www.flickr.com/photos/gdanmitchell/
Twitter: http://twitter.com/gdanmitchell

summitphotographer said:

summitphotographer said:

More will always be better as long as each pixel still have a meaningful signal to noise ratio. Personally I think we will have full frame sensors with several hundred MPs within a few years from now.

Click to expand...

Is this universally accepted?

Click to expand...

No it is not, because it is not generally true if you look at image quality and not resolution alone.

You can not add the signal of 2 small pixels and get the same signal to noise ratio as a single pixel with twice the size. This is because the read noise, which is per pixel and not per size, is added up and not averaged out. This will be visible in the darker areas of the pictures, mostly.

The question just is if you want "cleaner" or "sharper" pictures.

At ISO 100 you will always get clean pictures and the higher resultion sensor will win (if the lenses can keep up and the aperture is large enough)

At higher ISO the sensor with larger pixels and lower resolution will have the cleaner pictures.

It is not that the one is better than the other. It is just a matter of choice.

Even Nikon gives you this choice:

D4: 16.2 megapixels
D800: 36.3 megapixels

This conclusion does not depend much on the brand but is limited by physics. Photon noise is the limiting factor here.

This webpage gives a very good explanation:

http://www.clarkvision.com/articles/digital.sensor.performance.summary/index.html

Best regards

Some lenses are just amazing with my 7d. The 70-300L has crop detail I'm still getting used to.

In my opinion most high end Canon glass can support sharpness at a higher pixel density than the 7d. The 24-105 is at the low end, primes and some telephotos are at the high end.

I'd hoped the III would have enough pixels so I could crop to the 7d level. That way one camera could serve as a FF and crop, but alas it was not to be.

But your math is wrong. Should be 1.6^2 X 18 = 46 MP. Given that, the 800's 38 MP seems a pretty reasonable compromise.

summitphotographer said:

One thought I had as I read the dpreview preview of the MIII, is that if the MIII is really a FF 7D then should it have been a equiped with a 28.8 MP Cmos (1.6x18MP) or does the 7D already over-resolve/expose the weakness of available glass? Did Canon find 22MP was a more suitable resolution given available glass?

Click to expand...

tko said:

But your math is wrong. Should be 1.6^2 X 18 = 46 MP. Given that, the 800's 38 MP seems a pretty reasonable compromise.

Click to expand...

why 1.6^2 X 18 = 46 MP vs. 1.6x18 ?

schwoofi said:

You can not add the signal of 2 small pixels and get the same signal to noise ratio as a single pixel with twice the size. This is because the read noise, which is per pixel and not per size, is added up and not averaged out. This will be visible in the darker areas of the pictures, mostly.

Click to expand...

Which I wrote "slight exception for higher ISOs" in the previous reply.

schwoofi said:

The question just is if you want "cleaner" or "sharper" pictures.

Click to expand...

Higher resolution sensors will give you both better tonal graduations and more details.

schwoofi said:

At ISO 100 you will always get clean pictures and the higher resultion sensor will win (if the lenses can keep up and the aperture is large enough)

Click to expand...

Agreed, except for the lens part. Higher resolution will be better regardless of lens, and although diffraction will eat away at the relative advantage it will newer be worse than lower resolution sensors.

schwoofi said:

At higher ISO the sensor with larger pixels and lower resolution will have the cleaner pictures.

Click to expand...

Agreed.

schwoofi said:

It is not that the one is better than the other. It is just a matter of choice.

Click to expand...

Again agreed. The choice being between lower resolution for low light photography and high FPS (due to less processing), and higher resolutions for higher picture quality in normal light. Or to get more resolving power out of all your lenses. Or to use and crop a 200mm 2.8 instead of that expensive and heavy 500mm 5.6 - which should be point for quite a few even if they only need low resolution in their end process.