Printable View
I was recently in a forum and a member there mentioned that an aperature of a given lens behaves differently on different sensor sizes.
His example was based on the 85L f/1.4. The moment you putthis lenson an APS-C body the lens behaves like a 136mm f/2 lens. Is this correct (I understand the difference between FF and APS-C, it's the f/xx value that's got me wondering between the two)?
So confused!
The crop factor applies to aperture in terms of depth of field. So, an 85mm f/1.4 lens when used on APS-C has the equivaslent focal length as 136mm on FF, and has the equivalent depth of field as f/2 on FF. That applies for the same framing as FF, and the reason for the apparent effect on DoF is that to get the same framing on APS-C, you're further from the subject, meaning a deeper DoF. Sensor size does not affect exposure, so your f/1.4 lens on APS-C is giving your the shutter speed of f/1.4, not f/2.
Quote:
Originally Posted by neuroanatomist
Let me clarify this.
This statement is true in some sense but false in another. It is true in the sense that , if you use, eg, iso100 with both sensors and the same f number (ie, set your camera to f/1.4 for both cameras), the shutter speed will be the same.
However, it isn't the "iso number" that we care about, but rather how much noise we will get in our pictures. In the above scenario, the ff camera will have less photon noise. Alternatively, the ff camera could have used a higher shutter speed and gotten the *same* amount of photon noise as the crop camera. In this sense, f/1.4 exposes more quickly on larger sensors. Put another way, if you put the 85 f/1.2 on a crop camera and put the 135 f/2 on your full frame camera and set the camera isos so that photon noise is the same for both cameras, you'll get the same shutter speed (not exactly- 1.4 * 1.6 is not exactly 2, but close enough). The iso setting will be higher on the ff camera, but the photon noise will be the same.
Thus I think it is correct to think of cropping as changing the "effective f number" the same way it changes "effective focal length" (ie, multiply f number by crop factor to get "effective f number"), because cropping changes both DOF and exposure speed in exactly the same way as multiplying the f number by the crop factor.
John knows all this (after all, it's been hashed over dozens of times on this forum), so it is not my intention to correct him, but I think his statement is easy to misinterpret.
Jon, I think the noise issue is a separate argument. Sure FF tends to get less noise, but that
Quote:
Originally Posted by DavidEccleston
Same pixel size but different noise.
Am I missing something in your statement?
Mark
[img]/resized-image.ashx/__size/600x800/__key/CommunityServer-Discussions-Components-Files/24/3364.Untitled.jpg[/img]
Quote:
Originally Posted by DavidEccleston
Nope, you are not right. Here was discussed several times about sensor size and what depends on it.
And clemmb shows for you good camera comparison.
FullFrame sensor will have less noise not due to larger pixels, but due to larger surface of sensor = larger sensor, which can collect larger amount of light.
Imagine, you have:
- 10 buckets with 10 <span class="HW"]litre capacity each<span class="HW"]
- <span class="HW"]25 buckets with 5 <span class="HW"]litre capacity each.
And you place each type of bukets one nearby others. You have 2 areas with different size of buckets and of course different size of covered area. When the rain is starting, think, which area of buckets can produce larger amount of watter?
I think you will do the math [:)]
<span style="color: #ff0000;"]Edit: <span style="color: #ff0000;"]and it is due to overal larger capacity of buckets (pixels) = larger sensor.
Quote:
Originally Posted by DavidEccleston
So tell me, when here http://www.dofmaster.com/dofjs.html i select:
- 5D, 135 mm, F/2 and distance 5 meters, and i get 16 cm of DoF
- and with 7D, i select 85 mm (for same framing), F/2 and 5 meters distance, i get 26 cm of DoF.
Which number (Depth of Field) is smaller (more narrow): 16 cm or 26 cm ? [:)]
I
DoF is determined by four factors: aperture, focal length, subject distance, and circle of confusion. Only the last one is different with sensor size, but I
I'm adding more just to get back to the original argument. :)
As I mentioned, the noise should be identical in the middle of the frame, at a pixel level. Therefore nothing to do with effective apertures, a seperate argument. The extra size gives you less noise in prints to due extra pixels, shrinking the noise, or letting you downsample it away. I'm not saying there is no advantage to FF, just that the noise advantage has nothing to do with effective apertures.
With these same two cameras, though, adjusting your distance, or your lens, to produce the same framing, will have different DOF. This is an effect which is visible at the pixel level. This effective aperture change affects the ENTIRE FRAME. This is a FF advantage due to effective apertures.
edit: slight rewording of noise summary.
RE: I
Quote:
Originally Posted by elmo_2006
It seems every time crop factor comes it gets more confusing by the time you get to the end of the thread than at the start.
But.....Check this out for yourself.
Go to the DOF calculator. Enter 100mm at F1.0 for both camera at 10 feet.
Note the DOF of both lenses. Then if change the F of the 7D to F1.6. At that point the DOF of both lenses at the same distance will be identical.
I think this was what that individual was referring to.
In that respect they will carry the same identical 1.6 to 1 ratio.
If you try and prove that ratio in other ways, like comparing a 160mm on a full frame to a 100mm on the crop, the DOF ratio is not there. Nor if you try 100' compared to 160', the results will not hold the ratio.
So the crop ratio is only true to the F stop and the field of view. Other factors will come in to play in the other situations and I would think those would be the dynamics of how the lens passes the light and the width of viewable light on the sensor.
Now if you put a 100mm lens that is a F1.0 lens (if there were such an animal) on your crop camera, it does not change the fact that it is a F1.0 lens.
[View:http://www.robgalbraith.com/public_files/Canon_Full-Frame_CMOS_White_Paper.pdf]
"With all these benefits, it’s only natural to wonder why all DSLR cameras aren’t full-frame. Ultimately, the issue is money. Research, development, manufacturing and distribution costs are all independent of camera size, so a smaller camera will not cost appreciably less than a larger one for any of these reasons. The end cost difference between small mirrors, mirror boxes, chassis and so forth, and larger ones is not that great. The difference is the sensor."
Interesting this is Canon's thinking. If it weren't for the cost difference all cameras might be full frame.
And I think this paper addresses the "Noise" argument as well.
Quote:
Originally Posted by DavidEccleston
Agree. I was not understanding your first post on this. We are in agreement.
Mark
Quote:
Originally Posted by DavidEccleston
Yeah, maybe i wasn't reading your post carefuly. Maybe you were saying some truth [;)]
I know that Daniel's thread about all pixel size and dependency of that. In previous post forgot to mention the key think. I already edited it.
And in my opinion, the pixel size doesn't matter if the overal pixel surface is larger.
It does'nt matter if you have 1,6 square metre and you will split it to the 10 smaller squares or to the 20 squares . It still will be the same 1,6 square metre surface. And if you will compare it to the only 1 square metre surface, does not matter, how large the squares on that 1 square metter will be. The overal surface capacity will be smaller. Bigger surface can collect more light, and the noise is smaller. That's what i wanted to express [:)]
Quote:
Originally Posted by DavidEccleston
Yes, i agree. And want to add.. like were said in that big thread, that you can not compare pixels one by one (if this
is larger - good, smaller - worse). One pixel does not do the job. The
picture comes from collection of pixels.
Yes, I muddied things up by bringing up pixel level arguments without saying why. By showing the noise is equal at the pixel level, I show that the noise advantage of FF is due to pixel size and/or the bonus pixels at the edge. Is has nothing to due with effective apertures, which was the whole point of my original post. The noise advantage is a separate argument. The effective aperture argument is all about DOF changes. Jon pointed out a nice benefit of FF, which had nothing to do with the original question, and I just wanted to point that out (and then defend my argument when everyone said I was wrong :) ).
Quote:
Originally Posted by Kamelot
I am not sure that is correct. Sorry, I may be getting a little off topic here, but there has been a lot of bucket analogies going on where just the surface area is considered. However, (keeping the analogy going) there are losses that happen when you start breaking up the information into successively smaller buckets. Lets assume your buckets are square and fit together tightly into a rectangular array... you still have all the added edges that exist and that goes up as you go from 10 buckets to 20 buckets in the same area. There are losses at the edges. So unless you improve the technology to reduce the losses around the edges, then the "noise" is going to go up.
Probably more important is the losses you get when converting light energy to electrical energy at the photo site. I do not claim an relevant expertise here, but it would seem reasonable to expect that as you reduce the size of the pixel, you also reduce the the light gathered at that pixel, and therefore reduce the electrical signal that you are trying to produce to transfer that information. I guess my point is, as you reduce the pixel size (and increase the number of pixels for the same area) you are reducing the electrical signal at each pixel- require more amplification and processing, and you would have more efficiency losses... so unless you have an increase in technology (pixel, electrically, and processing), then you will get more noise and losses.
So, to give you a little different perspective on the analogy: Let's say you have two bucket arrays, one with 10 buckets and one with 20 smaller buckets covering the same area. Now put a whole in the bottom of each, and add a tube to take that water to a point that will monitor the the water that drains from each bucket. Now the set of 10 buckets has less connections, less tubing, less losses due to leaks at the connections, and less friction losses in the tubing.Also, because the 20 bucket set has less water to look at, your monitoring equipment needs to get better to be able to measure the differences and account for the added losses.
There are many reasons for more pixels (e.g., cropping), butmore may not always be a good thing unless there is an upgrade in the technology thatis used to handle the detrimental effects.
Sorry this got a lot longer than I expected.
Pat
I see and of course understand your point conropl. And you are correct. But if we have 1 biger pixel and 4 smaller pixel, you can not say, that this one bigger will produce stronger signal with less noise. Because maybe that 4 smaller pixels together would produce 2 time stronger signal comparing with that 1 bigger pixel (if the smaller pixel number would be enough bigger).
The main misunderstanding comes from that we are trying to meausre noise level, but do not compare things equaly.
Because there is 3 varying things:
- pixel size
- sensor size
- technology
If we would compare only one thing, leaving two others as constant, there will be less misunderstanding [:)]
Quote:
Originally Posted by Kamelot
When the Canon White Paper was written, the 1Ds II was canon's flagship with 16.7 megapixels. Now we have the 5D II at 21 megapixel meaning smaller denser pixels.
#3 matters more than 1 and 2. We know that 5D II will handle noise better.
#1 and #2 are only valid arguments if, you are applying the same technology to both.
Quote:
Originally Posted by Kamelot
Given the same technology, each area is receiving the same light. So four pixels would have more edges which do not gather light the same, and there has to be some finite space between pixels that do not gather information. This all causes the need to make processing assumptions about what is happening at the edges. Therefore, given the same light energy, the larger pixel gathers more information with less processor interpolation required. And I also contend that when youconvert light energy into some signal, there are losses every time you have to do that. With 4 pixels compared one large one, it would seem you have four times the likelihood of having conversion and processing errors/losses.
I agree with HDNitehawk... technology change is the big equalizer. Comparing the 5DII and the 7D. The pixel density on the 5DII is less than the 7D. The 5DII may have more pixels, but they are spread out over an area 1.6 times bigger. So if you cut down the 5DII sensor to the same size as the 7D, you would only have 13.1 mega pixels compared to the 7D's 18 mega pixels. The 5DII has bigger pixels but less of them over the same area. The end result is a sensor that handles noise better IMO.
Quote:
Originally Posted by HDNitehawk
HDNitehawk:
Nice paper... it explains a lot. Thank's for sharing.
Beware that the Canon CMOS "White Paper" is hardly a white paper at all -- more like a marketing fluff piece. While it does have quite a bit of good information, Canon also mixes in enough misinformation and bald-faced lies (oops, I mean "marketing") that I can
Quote:
Originally Posted by Daniel Browning
I would think it is obvious it is marketing. It is an old document, and was produced to promote the CMOS sensors.
I guess I would have to ask, where are we being mislead?
Quote:
Originally Posted by HDNitehawk
I agree that #1 is irrelevant to the discussion. Talk of pixel density only adds to the confusion.
Tecnhology, however, can only take you so far. Sensors are less than 100% sensitive (ie, they register fewer than 100% of all photons), but the day isn't far off when sensors will be close enough to 100% sensitive that the only real parameter of the three will be sensor size. In the end, it is really effective f number that determines how fast one can expose (or, with a given f number, sensor size is what matters).
I have no horse in this debate here, but I remember reading this from dpreview when the 50D came out (read below)... I remember thinking to myself that a camera like the 7D would never come out, because the pixel race was obviously over. Now, I don't think anyone would argue that the 7D not only resolves more detail than the 40D, but also has better high ISO performance as well (read 7D review following 50D). In conclusion, I agree with HDNitehawk. Technology can even the playing field even if pixel density is greater.
50D review: "Let's have a look at the really important stuff then: Image quality. Below ISO 1600 image output is clean with well balanced contrast and colors and as you would expect from a DSLR with a 15 megapixel sensor the 50D delivers a fair amount of detail. Having said that, in terms of per-pixel sharpness the 50D cannot quite keep up with the better 10 or 12 megapixel APS-C DSLRs in the market. At higher sensitivities the smaller photosites are clearly producing more noise (as shown from our RAW comparisons) and so Canon is having to apply more noise reduction to keep to acceptable noise levels, this of course means a loss of detail from ISO 1600 upwards.
It appears that Canon has reached the limit of what is sensible, in terms of megapixels on an APS-C sensor. At a pixel density of 4.5 MP/cm² (40D: 3.1 MP/cm², 1Ds MkIII: 2.4 MP/cm²) the lens becomes the limiting factor. Even the sharpest primes at optimal apertures cannot (at least away from the center of the frame) satisfy the 15.1 megapixel sensors hunger for resolution. Considering the disadvantages that come with higher pixel densities such as diffraction issues, increased sensitivity towards camera shake, reduced dynamic range, reduced high ISO performance and the need to store, move and process larger amounts of data, one could be forgiven for coming to the conclusion that at this point the megapixel race should probably stop. One consequence of this is that the 50% increase in pixel count over the 40D results in only a marginal amount of extra detail."
7D review: "Despite the highest nominal resolution of all APS-C DSLRs and therefore a very small pixel-pitch the EOS 7D performs very well in low light situations and manages to maintain a good balance between image detail and noise reduction up to very high sensitivities. It's visibly better than the EOS 50D and as good as it gets in the APS-C class (if you prefer the 7D or Nikon D300S in this respect is probably a matter of taste). If you require significantly better high ISO performance than the EOS 7D can provide, your only option is to move into the full-frame segment."
I don
dsiegel5151, quite a difference between their 50D and 7D reviews, huh? [:D] More reason to distrust what you read on DPR. Let me pick apart of a few of their summary points:
Quote:
Originally Posted by DPR
<meta content="text/html; charset=utf-8" http-equiv="content-type" />
This is doubly incorrect. First, their "RAW" comparisons are badly flawed and not even raw at all. Second, even from their flawed comparisons it's possible to see that for similar levels of detail the 50D has the same noise level as the 40D -- not worse.
Quote:
Originally Posted by DPR
Several years later, we now have Nikon's D7000, which is so far ahead of anything Canon has ever made that it's not even funny, especiallywhen it comes to base amplification read noise. And there are 2-micron digicam pixels that even better than that (but haven't been made into APC-S yet).
Quote:
Originally Posted by DPR
DPR's silly made-up terms and measurements annoy me a lot I prefer to use industry standard terms likes "pixel pitch" and measurements such as microns for pixel diameter. It's not very smart that they think it is a bad thing for the lens to be the limiting factor. Are lenses some sort of cheap throw-away accessory that is easy to improve, and that's why we never want our sensor to be limited by them? I take the opposite view: our lenses are the most expensive and important part of the system, and they are more expensive to improve than the sensor, so we never want our lenses to be limited by our sensor. So if your sensor has such poor resolution that it can't see the lens flaws, that is suboptimal.
Quote:
Originally Posted by DPR
Nonsense. The cheapest zoom at the worst aperture in the very corner of the frame still has more than 10% MTF necessary to achieve line separation on a 15.1 MP sensor. That means it's still necessary to have a contrast-reducing optical filter to limit aliasing artifacts.
Quote:
Originally Posted by dsiegel5151
What a load of BS (Bad Science). [;)]
<meta content="text/html; charset=utf-8" http-equiv="content-type" />
Yeah, it
Thats like saying you shouldn't scan a slide more that an XXXX about of resolution because your limited by your slide in the first place. What a load of Bad Science.[:D]
Cheers,
John.