Posted by Howard Cihak on November 21, 2012 at 4:55 AM
I'm hoping someone can answer a question put to me for which I really don't know the answer myself.
If an APS-C sensor and a full frame sensor have the approximately the same total number of pixels, then the full-frame sensor's individual pixels would be about twice as large (all other things being equal).
The question is ... will the larger pixel yield better resolution than the smaller pixel? My guess, and that's all it is, is that it will not.
(Forget the noise issue; I already know the answer to that one.)
Posted by Howard Cheek on November 21, 2012 at 12:23 PM
Resolution has multiple factors other than total number of pixels as you know.The full frame will produce better "resolution" using the same sharp lens (all other things being equal), simply because it's sharper with less noise.
I don't see how one can separate or 'forget' the noise issue here..I guess it's all in how you define 'resolution' Howard.
Just one more thing...Your name sounds oddly familiar to me for some strange reason
Posted by Larry Fischer on November 21, 2012 at 12:41 PM
I'm not much into the technical.
I have read with the advent of the Nikon D800, 800E and D8 that at least one thing was noted.
Old lenses got very much better.
From that statement, if true, we can say perhaps that lenses have been outperforming digital 'resolution', whatever that is, for some time.
Optical resolution is being able to discern 2 distinct point as 2 points. I think that has changed and with light sensors [pixels]
the definition is skewed and somehow not accurate.
Each pixel records its wavelength so it is the space around the pixels that apparently contributes
to noise.
Pixels are round, are the not? I wonder if the could be made square? Would that eliminate 'noise'?
I'm looking forward to the fine minds who "know" contributing to this thread ... in succinct form.... [grins]
the fisch
Posted by David Bostedo on November 21, 2012 at 4:48 PM
All else being equal, smaller pixels will always have better resolution than larger pixels.
If you had a lens that could resolve more than the sensor in both cases, and the noise levels of the pixels were the same, then the smaller pixels would definitely out-resolve the larger ones. That's why cameras with the same size sensor still get better resolution going to higher pixel counts.
For example, a D800 and a D700 - the D800 has smaller pixels, but clearly out-resolves a D700. If you were to crop the center of a D800 sensor to APS-C size, you'd still get better resolution than a D700.
What makes it difficult is that "all else" is frequently not anywhere near equal. The technology used in the sensor plays a large part, as does the images you take based on the sensor. Noise is one big factor that will affect resolution. And if you're comparing the same shot on two different sensors, you've either switched lenses or moved where you take the picture from; That complicates comparison, but doesn't change the underlying facts.
The reason most cameras don't go to a crazy number of pixels is mainly because as you start to increase the pixels, you increase the complexity of the design, and the amount of support circuitry you need, and the overall quality of the image. So if you're designing a sensor, and you prioritize dynamic range/exposure range, and noise performance, you lean toward larger pixels. If you prioritize resolution, you shrink the pixels until the noise performance starts to get below your acceptable design level. (Note that for most cameras, the overall noise level of the image is dependent more on the size of the sensor than the individual pixel noise, given the same size output between both.)
And as far as the "shape" of a pixel, it's not standard or well defined on the sensor itself. It really could vary between various square or rectangular shapes, and octagons or more irregular shapes depending on the physical design of the pixel, and how the surrounding circuitry is fit in. (Here's an example layout of a pixel. The actual light gathering portion is the large irregular cross-hatched shape. The rest is support circuitry and metal connections.) It's then usually covered by microlenses and color filters that are more regular in shape. (Like this.)
Posted by Karen Kane on November 22, 2012 at 3:20 AM
Howard,
Afraid the answer is more complicated than a simple yes-no.
Bigger isn't always better, but for pixel size, it does have some advantages (greater dynamic range and signal-to-noise ratio). 18MP on a full-frame sensor should out-perform 18MP on an APS-C sensor if the output will be the same size. However, as David mentioned above, making the pixels smaller so there are more of them comes with a different set of advantages (clarity and sharpness). 22MP on an APS-C sensor should out-perform 18MP on another APS-C sensor. Camera manufacturers try to balance these and get the best of both worlds.
Either way, think of this analogy - pixels on a camera sensor are kind of like the spaces of an ice-cube tray (only they trap light, not water). If you have the same number of ice-cube spaces in two different trays, one small tray and one larger tray, then there will be more water (larger ice-cubes) in the larger tray. Similarly, a larger sensor with a certain # of pixels will trap more light over the same length of exposure time than a smaller sensor with the same # of pixels, and this provides the advantages of dynamic range and signal-to-noise ratio.
Discussion
4 repliesIf an APS-C sensor and a full frame sensor have the approximately the same total number of pixels, then the full-frame sensor's individual pixels would be about twice as large (all other things being equal).
The question is ... will the larger pixel yield better resolution than the smaller pixel? My guess, and that's all it is, is that it will not.
(Forget the noise issue; I already know the answer to that one.)
I don't see how one can separate or 'forget' the noise issue here..I guess it's all in how you define 'resolution' Howard.
Just one more thing...Your name sounds oddly familiar to me for some strange reason
I have read with the advent of the Nikon D800, 800E and D8 that at least one thing was noted.
Old lenses got very much better.
From that statement, if true, we can say perhaps that lenses have been outperforming digital 'resolution', whatever that is, for some time.
Optical resolution is being able to discern 2 distinct point as 2 points. I think that has changed and with light sensors [pixels]
the definition is skewed and somehow not accurate.
Each pixel records its wavelength so it is the space around the pixels that apparently contributes
to noise.
Pixels are round, are the not? I wonder if the could be made square? Would that eliminate 'noise'?
I'm looking forward to the fine minds who "know" contributing to this thread ... in succinct form.... [grins]
the fisch
If you had a lens that could resolve more than the sensor in both cases, and the noise levels of the pixels were the same, then the smaller pixels would definitely out-resolve the larger ones. That's why cameras with the same size sensor still get better resolution going to higher pixel counts.
For example, a D800 and a D700 - the D800 has smaller pixels, but clearly out-resolves a D700. If you were to crop the center of a D800 sensor to APS-C size, you'd still get better resolution than a D700.
What makes it difficult is that "all else" is frequently not anywhere near equal. The technology used in the sensor plays a large part, as does the images you take based on the sensor. Noise is one big factor that will affect resolution. And if you're comparing the same shot on two different sensors, you've either switched lenses or moved where you take the picture from; That complicates comparison, but doesn't change the underlying facts.
The reason most cameras don't go to a crazy number of pixels is mainly because as you start to increase the pixels, you increase the complexity of the design, and the amount of support circuitry you need, and the overall quality of the image. So if you're designing a sensor, and you prioritize dynamic range/exposure range, and noise performance, you lean toward larger pixels. If you prioritize resolution, you shrink the pixels until the noise performance starts to get below your acceptable design level. (Note that for most cameras, the overall noise level of the image is dependent more on the size of the sensor than the individual pixel noise, given the same size output between both.)
And as far as the "shape" of a pixel, it's not standard or well defined on the sensor itself. It really could vary between various square or rectangular shapes, and octagons or more irregular shapes depending on the physical design of the pixel, and how the surrounding circuitry is fit in. (Here's an example layout of a pixel. The actual light gathering portion is the large irregular cross-hatched shape. The rest is support circuitry and metal connections.) It's then usually covered by microlenses and color filters that are more regular in shape. (Like this.)
Afraid the answer is more complicated than a simple yes-no.
Bigger isn't always better, but for pixel size, it does have some advantages (greater dynamic range and signal-to-noise ratio). 18MP on a full-frame sensor should out-perform 18MP on an APS-C sensor if the output will be the same size. However, as David mentioned above, making the pixels smaller so there are more of them comes with a different set of advantages (clarity and sharpness). 22MP on an APS-C sensor should out-perform 18MP on another APS-C sensor. Camera manufacturers try to balance these and get the best of both worlds.
Either way, think of this analogy - pixels on a camera sensor are kind of like the spaces of an ice-cube tray (only they trap light, not water). If you have the same number of ice-cube spaces in two different trays, one small tray and one larger tray, then there will be more water (larger ice-cubes) in the larger tray. Similarly, a larger sensor with a certain # of pixels will trap more light over the same length of exposure time than a smaller sensor with the same # of pixels, and this provides the advantages of dynamic range and signal-to-noise ratio.
Here's a link with a more detailed discussion: http://www.cambridgeincolour.com/tutorials/digital-camera-sensor-size.htm