Pixel Density

DjBeau · Sep 3, 2009

Hi

I know lots of people are saying that you can't tell the difference between the quality of a full frame image and a crop sensor image. Well, those people should obviously buy some glasses.

One of the things that really stands out on fullframe is the increased sharpness (in the center, anyway). The reason for this is - as far as I understand - that the pixel density is lower on a full frame which means the pixels are farther apart and do not "bleed" into each other.

Now, here's the question:

Is a 12 mega pixel full frame sensor sharper than one of 24 mega pixels? I mean, since you're cramming more pixels into a sensor of the same size wouldn't that mean bigger density and therefore more bleeding?

Or do that extra amount of pixels to define the image simply balance out the bleeding problem?

Regards
Beau

Plato · Sep 3, 2009

I've never seen pixel blood.

A lower density does not mean that the pixels are further apart. It means that the pixels are larger. The issue is electrical noise (the simplest example is static), especially at high ISO values. The noise is always present but it impacts small pixels greater than it does large pixels.

However, this is a bit of over-analysis. Just as with large frame sizes with film, large sensors produce better results than small sensors. The decision that each photographer must make is whether or not the difference affects his/her own pictures. If all of your shots are in bright sunlight at ISO-100 and you'll never print anything larger than a 4X6, even the DX sensor is major overkill.

Moglex · Sep 3, 2009

The reason that FF is sharper with any given lens is that the lens can produce a certain number of LPPM (Line Pairs Per Millimetre).

Obviously there are less millimetres for it to produce line pairs over with a crop frame sensor than one that has the full monty.

It really is that simple.

As Plato says, though, there are tradeoffs to consider - it isn't all about the ultimate sharpnesss.

Village Idiot · Sep 3, 2009

Crop sensors should be sharper over all since they're using less of the edges, which aren't as sharp as the center of a piece of glass.

Maybe it's the shallower DOF you're noticing?

Moglex · Sep 3, 2009

Village Idiot said:
Crop sensors should be sharper over all since they're using less of the edges, which aren't as sharp as the center of a piece of glass.

He did specifically say: 'In the center, anyway'.

Gaerek · Sep 3, 2009

DjBeau said:
I know lots of people are saying that you can't tell the difference between the quality of a full frame image and a crop sensor image. Well, those people should obviously buy some glasses.

Really? I'd be willing to bet that even most pros would be hard pressed to tell the difference between a photo shot with a crop frame and full frame at lower ISOs. This is especially true at web resolutions and anything the size of an 8x10 or smaller. Is there a difference, most certainly. But the only people who will really be able to tell the difference are pixel peepers.

I guess I need glasses for my 20/20 vision.

DjBeau · Sep 4, 2009

I am of course aware that there are other things to consider and that I might be starting a **** storm. On certain pictures, though, especially close-up portraits, I think the sharpeness does matter quite a lot. I've been working with crop as well as full frame and it's really not that hard for me to see the difference.

@ Plato: But does that mean that the pixels on a 24mpx are smaller than those on a 12mpx and will thereby be more influenced by electrical noise?

@ Moglex: Can you elaborate a bit on your post? I don't think I quite fully understand it

JamesMason · Sep 4, 2009

DjBeau said:
@ Plato: But does that mean that the pixels on a 24mpx are smaller than those on a 12mpx and will thereby be more influenced by electrical noise?

yes, taking into account sensor size i.e. full frame, 1.6x, ect.

Garbz · Sep 4, 2009

Noise yes, Noise does not necessarily reduce sharpness though. Noise reduction reduces sharpness.

What you will see is the difference in the lens. Diffraction amongst other things determine a finite size that any given point will be projected by the lens. If your sensor has more megapixels in the same area, and the point is larger than any single megapixel what you end up with is reduced visible sharpness.

But is it less sharp? One could consider it like this. You have a printed photo that looks perfectly sharp to the naked eye. Now you look at it under a magnifying glass and find that it is not as sharp, But is the photo itself now less sharp, or are you just showing it's flaws?

The easiest way around the problem... Downsample. take your 24mpx image and convert it to 12mpx, and it will look just as sharp as it was previously.

musicaleCA · Sep 4, 2009

Garbz said:
The easiest way around the problem... Downsample. take your 24mpx image and convert it to 12mpx, and it will look just as sharp as it was previously.

Woohoo! My favourite "trick". :lol:

(That is, downsampling an image can definitely save it. There are plenty of times when I look at an image and think "aw nuts, they moved a few millimetres, it's not sharp" and just downsample.)

Moglex · Sep 4, 2009

DjBeau said:
@ Moglex: Can you elaborate a bit on your post? I don't think I quite fully understand it

The resolution of a lens is measured in LPPM or 'Line Pairs Per Millimetre'.

This is exactly what it says: the number of pairs of lines (one black, one white) that can be differentiated per millimetre at the focal plane.

The resolution of the image is determined by the resolution of the sensor (or film) in conjunction with the resolution of the lens.

Let:
Rs = resolution of the sensor
Rl = resolution of the lens
Ri = resolution of the image

Then Ri = 1 / ( 1/Rs + 1/Rl )

Caveat: This is simplifying considerably by ignoring, for example, edge effects and the Bayer filters on digital sensors. The numerical answer should not be taken as an accurate figure but the working shows the general way the figures will tend to pan out.

If your sensor has a horizontal resolution of 5000 pixels this is equivalent to 2500 lp (since it takes two pixels to represent a pair of lines).

Thus a full frame sensor has a resolution of 2500/36 = 69 lppm
And a Canon crop frame a resolution of 2500/22.7 = 110 lppm

Say your lens has a resolution of 100 lppm

Then the effective lppm for the full frame lens = 1/(1/69+1/100) = 40 lppm
and the crop frame = 1/(1/110+1/100) = 52 lppm

And the effective lp's are 40*36 = 1440 and 52 *22.7 = 1271 for the full and crop frame respectively.

Thus the FF will be approximately 13% sharper.

DjBeau · Sep 4, 2009

Thank you so much for all your answers, especially you Moglex.

I'm of course not trying to start another tedious debate on FF vs crop - just searching for the scientific explanation to what I see with my eyes.

This raises a few more questions.

I tried to use your formula and double the amount of pixels to 10.000 = 5000 lp. On a FF this adds up to about 2093 effective lp's which is 45% more than on a FF sensor with only 5000 pixels. Does this mean that a 24mpx FF sensor is 45% sharper than a 12mpx FF sensor or are there other factors to take into consideration? It doesn't really seem likely to me.

Derrel · Sep 4, 2009

The resolution of a lens may be measured in LPPM or line pairs per millimeter, but there's a new way of measuring the resolution of an imaging system, called line pairs per picture height, which is an effort to take into account how well the lens AND the sensor work together. Now that we have gotten into very high MP count sensors territory, many testers are reporting results using the Line Pairs Per Picture Height method. One problem now with 21 to 24 MP sensors is: Just how good is the lens????

Lower MP sensors, like the old 2.7 and 4.2 megapixel Nikon sensors in the D1 series and the D2h, did not have sufficient resolution abilities to show much of a difference between an average lens, a good lens, and a superb lens; the "system" resolution,meaning the lens and the sensor, together did not produce enough line pairs per picture height to make cheap glass look bad. And good lenses didn't resolve that much higher than superb lenses because until we got to about 12 MP on APS-C sensors.

Once we hit about 12 megapixels on Nikon's APS-C sized sensor, the sensor abilities and its demands upon lenses became a very significant,real issue. On the Nikon D2x for example, roughly a 12.2 MP camera, the SAME,exact Nikon lens shot on that camera could produce HIGHER resolution figures than that same,exact lens mounted via a lens adapter on a Canon EOS 1Ds Mark II, a 16.7 MP camera, provided the lens was used at wider-open apertures. Why? Pixel density on the D2x's sensor was very,very high--higher than that on the FF 16.7 MP body. With more pixels per millimeter, a high-resolving lens will record, well, more information on more, closely-spaced pixels. At least at wider lens apertures; stopping down to even f/8 can cut resolution and contrast, due to the effects of diffraction.

Theory and practice and physics all interact, and a lot of what one has read before about line pairs per millimeter in APS-C versus FF is influenced by things like how good the lens actually is, and how well the images are processed in software, and also how soon pixel density causes optical diffraction to cause loss of resolution; on a 1.5x camera with 12.2 MP, the Nikon D2x specifically, diffraction starts lowering image sharpness at f/5.6, and once the aperture becomes smaller, like say f/7.1, there is a *marked* and easily-seen lowering of overall resolution numbers.

On really high-resolution sensors like the D3x's 24 MP, diffraction becomes a big problem,as does the *quality of your lenses* and your camera technique--focusing and camera support become critical the higher the megapixel counts go. If one reads this review Nikon D3x Review by Thom Hogan

One can see that an up-sampled 12 MP image from a Nikon D3 can look almost as good as a D3x's 24 megapixel image, as long as it was shot with a really good lens like a 200 f/2, 50mm f/1.4 AFS-G, or 24mm PCE Nikkor, to name three specific lenses Thom tested.

It seems right now, that with current, actual cameras, the high-MP cameras like Canon 5D Mark II, Sony a 900, and Nikon D3x there is a huge strain put on the real-world photographic chain,requiring a steady camera, absolutely dead-on focusing,and superior lens performance, all in the 21 to 24 Megapixel on FF range. Looking at the dPreview full-sized samples from the 18 MP Canon 7D,which would scale to 46 MP on a FF sized sensor with that same pixel density, I personally feel that the 17-55 f/2.8 EF-S f/2.8 lens does not have the resolving power and CA correction to make use of such a high pixel density--but the 70-400 f/4L and 85mm 1.2L do have the ablity to eliver a good enough image on that same sensor.

Like the Nikon D3x, the Canon 7D's sensor does NOT deliver images that look all that great above ISO 1600; it seems that at the really elevated ISO settings like 3200 and higher, a 12 Megapixel full-frame sensor actually does deliver a better photograph,overall, than the a higher-MP count FF sensor.

We have to realize that theory and mathematical performance does not always translate into actual resolution in-camera; Higher MP counts also bring big problems with insufficient lens quality to utilize the added MP counts, diffraction problems, camera shake, mirror slap,and the need for absolutely DEAD-ON focusing. There is a really large "Sweet Spot" that seems to be around 12 to 13 megapixels (Nikon D3, Nikon D700,Canon 5D original) that delivers high resolution while being accepting of a lot of lenses, and also delivering excellent image color, resolution, across a wide range of ISO settings, from basically 50 to 3200 ISO, and in the case of newer sensors, up to ISO 6,400 with pretty amazing results; the higher MP sensors like the Sony a900 start getting pretty noisy at ISO 800, while the Canon 5D Mark II looks quite excellent up to ISO 1600, and that's about the top end for the Nikon D3x.

Lens performance also is a bit different for APS-C and FF: a lens designed for DX needs to deliver a VERY high degree of resolving power across a relatively small image circle, like the Nikon 70-200VR was designed to do. That same lens is much, much weaker on FF, which demands high optical performance across a significantly larger area of 864 square millimeters, as opposed to the 370 square millimeters of a Nikon DX sized sensor. Canon's 70200 f/2.8L on the other hand, is an excellent lens on a full frame digital sensor. Now that we have arrived at cameras that have moved from the 6MP level of 1,700 LPPH to over 3,000 LPPH, we have a lot of issues with lenses and technique (focusing,vibration,subject motion) to deal with.

DjBeau · Sep 4, 2009

What an extremely interesting post! Now I can finally get to the concrete matter at hand - well maybe not the one I started out with but something related that has been bugging me:

There are rumors that Nikon will release the D700x, a 24mp camera (yes possibly just rumors but just go along with me here). It has been my plan to upgrade to FF when this is released. Why, you might ask, not the D700? Well, I have this DX wide angle lens that is performing incredibly well on my crop sensor (D80) and I would like to still be able to use it (mostly because I can't afford any of Nikons FF-alternatives if I buy a new camera). On a D700 in crop mode this would mean I'd end up with only a 5.1mp file whereas when used on the D700x I would still have a, say, 10mp file.

But as far as I can understand from Derrels post it wouldn't make much sense anyway since DX lenses are apparently not performing that well on FF. Did I understand this right?

And another thing: Will the new Nikon 70-200 f/2.8 be designed for FF sensors and therefore be visibly better than the old one?

Derrel · Sep 4, 2009

Nikon's DX lenses will mount and will shoot pictures on Full Frame or "FX" Nikon bodies,and can do so with complete coverage of the DX image circle of the three FX Nikon bodies (D3,D3x,D700).

What I meant was that a lens optimized for the smaller, DX image circle, needs to have very good optical performance across the entire width of the DX image sensor; that high resolution ability and excellent optical correction is needed for a lens to make a DX sensor really perform great. The small, high-pixel DX sensor Nikons with 12.2 MP sensors demand lenses that have *excellent* optical quality in order to leverage the small, very densely-packed sensors.

The FX sensor is much larger than DX, and the pixel density is lower,but the image capture area is significantly larger,and therefore a lens with somewhat lower optical quality can produce a truly excellent image. This is the same way medium format worked: a significantly larger capture area than 35mm, with negatives that would be enlarged fewer diameters than 35mm negs, allowed Medium Format film cameras to shoot with lenses that were much lower in resoultion than lenses designed for 35mm. For example, a lens for 120 rollfilm that resolved say,roughly 46 center/37 edge lines per millimeter in the old days of the 1970's would produce a comparable image to a 35mm film lens that turned in 68/55,and both lenses would be considered "Excellent". The old Modern Tests bore this out many times.

Yes, if a D700x at 24MP hits market, you'll be able to get a higher MP count DX-sized file by shooting the D700x in DX-crop mode than you get with your D80. You could also buy a used or new D300 and get the 12.2 MP in APS-C mode. And yes, the new 70-200VR has been re-designed to optimize it for FX sized sensor use.