aCuria
Offline
• • •
Upload & Sell: Off
|
 p.12 #20 · 61MP Sony A7R V specifications leaked | |
ruthenium wrote:
Some of the disagreement around your statements might be due to the categorical nature of your writing. For example, "edge to edge performance that really matters too" makes me wonder how much, on average, we are pixel-peeping at the edges? Arguably, most times the area outside the central APS-C zone does not need to be razor-sharp.
By "edge to edge" I mean that the "whole frame" has to be considered, not just the center.
ruthenium wrote:
I am also not sure about the truth of "100 MP FF aka 160lp/mm sensor from Sony is useless with the existing lens lineup alone." Here, I am thinking about those situations when an image is cropped. Intuitively, it seems that having more pixels in the cropped image should be better for the image quality. Is this wrong?
There are multiple layers of possible misconceptions in this kind of discussion, and its difficult to get people to understand whats really going on. You probably already know some of this explanation below:
The first and most common misconception is that people think that detail in the photo will increase linearly with pixels, and this is wrong because the "megapixel" tag is a deliberately misleading metric. For us to perceive 2x the detail in an image you would have to double the pixels on both axis (and double the sensor size on both axis too)... Meaning that going from a 15MP m4/3 sensor to a 60MP FF sensor is necessary to get 2x the detail, (assuming the lens quality is unchanged, which is another whole can of worms)
The second possible misconception is in just looking at the linear resolution differences, so 7008 (100lp/mm) px A7iv, and 9504px (130lp/mm) on the Riv, and expecting to get that 9504 / 7008 = 130% more detail. However, its necessary to use the same pixel pitch, and a larger sensor to truely get 130% more detail.
So then the question is what do we actually gain? This can be approximated with the transfer function:
image = image lp/mm, meaning what we get in the raw file, printed the same size as the sensor
sensor = sensor lp/mm, computed based on pixel pitch
lens = lens lp/mm, the resolution of the virtual image projected by the lens
x = experimentally determined (assume x=2)
So we have this relationship:
1/image^x = 1/sensor^x + 1/lens^x
So if we look at the 135GM f/2.8 MTF chart and plug in the MTF 50 at the 10mm mark (~70 lp/mm) (note that I am assuming lensrentals is measuring the resolution of the virtual image directly, which is what I think they are doing) we get the following values:
A7iv image resolution = 57 lp/mm
A7Riv image resolution = 61.6 lp/mm (7% increase from 7iv)
160 lp/mm sensor image resolution = 64.1 lp/mm (3.9% increase from Riv)
Note that if we were to use a lens that can do 79 lp/mm on the 7iv, vs a 70 lp/mm lens on the Riv, the A7iv will theoretically produce an image on par with the Riv shot!
Now, if we re-do the computations with a theoretical lens that does 160 lp/mm edge to edge:
A7iv image resolution = 84.7 lp/mm
A7Riv image resolution = 100.9 lp/mm (16% increase from 7iv)
160 lp/mm sensor image resolution = 113.1 lp/mm (10.8% increase from Riv)
As you can if the lens is better, we can get a more significant real resolution increase on the final raw file.
Note that using MTF50... just means we assume that 50% contrast is good enough for us. This means that a black hair will end up as middle grey rather than black, which may or may not be acceptable. Of course the calculation can be repeated reading off the graph at higher contrast levels
One thing you may notice from this exercise, is that the center of the image contains more information than the outside, due to the MTF curve being the strongest in the center. This also means that using apsc mode is less damaging than one might expect. Maybe the area under the mtf curve can be used somehow.
ruthenium wrote:
I am not arguing about the substance(!) of your claims. I just wanted to say that strong statements, expressed in terms "no ifs ands or buts about it" possibly make people uncomfortable. The writing style can be strongly influenced by the professional occupation of the person. E.g., I expect some technical professionals (scientists) who are experts in their field (or consider themselves to be experts) may use strong categorical statements when they write (I am guilty of this myself).
Well... thats actually an uncomfortably accurate guess 
|
