Too Many MP Part II

[strawman_John]

Having too much fun on this one and still I guess I feel emil has not grasped the detail of the conversation.

So what happens as you add more pixels for a given sensor size. There
is strong evidence that dynamic range is reduced. Take the 50D,
despite it having a later sensor design and better micro lenses, plus
a later generation of processing software and silicon, it has less
dynamic range than the 40D. (For a fairer comparison take some 500D
images and compare to the 40D, without the improved micro lenses it
makes a more comparable baseline).

What is this strong evidence? News for you: DR is about noise. Noise is scale > dependent. Pixel DR has little to do with image DR, for the reasons already > explained in this thread; DR is a scale dependent quantity, since noise is a scale > dependent quantity. The 50D and 40D have about the same DR at the same > image scale.

I fear you are letting the maths get in the way of comprehending the real world situation. The lower MP camera can cover a greater level of shades of light before you get into noise issues. It is something I have observed.

Then I reflect back on my own experience. There are lenses out there
that the 40D can get close to out-resolving, and there is reasonable
evidence of the 50D doing so on more lenses. So I see evidence of the
point of diminishing returns coming in crop sensors.

So using poor lenses and poor technique is a reason for those of us who don't > not to have the benefits of higher resolution? No thanks.

When did I say I was using poor lenses. You make assumptions you just cannot back how much else of your argument is constructed so poorly. Well if you count Canon's L series as poor lenses then so be it, but yes under certain conditions the 40D can be seen to be out resolving some of them and the 50D definitely can be seen to do so. And again we come to budget, if you buy a camera like the 40D or 50D how much will you spend on lenses.

All I dare do is challenge the more pixels in same sensor size is good regardless mentality. At least credit me with reviewing what I can and cannot achieve with my camera and deciding what I would like to see done next. Or is that not allowed.

It is sad you feel the need to play the person and imply they are doing something wrong rather than stick to the issue.

So to impress me, make an APS sensor of @ 12mp with noise performance
at ISO1600 of a 40D at ISO200. More pixels, of limited use, unless
the L series range are going to be rolled over, and even then, can I
use the extra resolution????

Again hint I was talking about L lenses.

The noise performance at ISO 1600 is actually better than at ISO 200 with > either the 40D or the 50D. It's just that for the light levels it is used for, the
signal is weaker, hence the S/N is lower.

You need to think of real photography and back out of abstract maths, or at least apply it to the real situation. When you get to low light levels you want higher ISO. If I were to take a shot at ISO 1600 and a shutter speed of 1/1000 for and 1/60 at ISO 100, same lens settings, which would have the higher noise? This is a real world situation that photographers can comprehend. For the landscape ISO100 fine, for the elusive flying fox wombat flying through the air the 1/1000 shutter speed might be essential.

The good camera developments will open up using the cameras under a greater range of conditions, I hope. Certainly the developments since the 10D have allowed good image capture at higher ISO/lower light it managed.

If you knew what you were talking about, you would want at ISO 200 the noise > performance of ISO 1600 of a Canon DSLR (my choice would be the 1D3, not > the 40D; less pattern noise)...

Now again the personal comment, is it pompous or patronising, or ignorant? Hard to tell, but it looks like you want to say that if you can patronise me it makes you think your point is OK. It undermines your argument as you are not apply decent honest debate to the conversation. So go on try harder to keep to the issues not the need to make petty comments.

By all means get the more expensive camera, I never contended that the 40D was perfect, quite frankly the AF performance and robust build lift the 1DMKIII well beyond the 40D/50D, and alas well past my budget.

 

[igb]

Having too much fun on this one >

I don't think that's possible

and still I guess I feel emil has not
grasped the detail of the conversation.

So what happens as you add more pixels for a given sensor size. There
is strong evidence that dynamic range is reduced. Take the 50D,
despite it having a later sensor design and better micro lenses, plus
a later generation of processing software and silicon, it has less
dynamic range than the 40D. (For a fairer comparison take some 500D
images and compare to the 40D, without the improved micro lenses it
makes a more comparable baseline).

What is this strong evidence? News for you: DR is about noise. Noise is scale > dependent. Pixel DR has little to do with image DR, for the reasons already > explained in this thread; DR is a scale dependent quantity, since noise is a scale > dependent quantity. The 50D and 40D have about the same DR at the same > image scale.

I fear you are letting the maths get in the way of comprehending the
real world situation.

What a pearl.

The lower MP camera can cover a greater level
of shades of light before

shades of light?? what is it? pompous, patronizing and/or ignorant?

you get into noise issues.

It is something

I have observed.

Ah. YOU have observed. Of course. That explains everything. Including the notion you have about what an observation is.

I have a request to all knowledgeable poster in the forum. The discussion about the effect of increased PD in IQ is very interesting and necessary. Therefore it is important to keep it out of the boundaries of imbecility. Please don't waste energy in replying to posters like the OP. TIA.

--
-------------------------------------------------------
My Galleries: http://webs.ono.com/igonzalezbordes/index.html

 

[ejmartin]

Having too much fun on this one and still I guess I feel emil has not
grasped the detail of the conversation.

Yeah, I can be a bit slow on the uptake at times.

So what happens as you add more pixels for a given sensor size. There
is strong evidence that dynamic range is reduced. Take the 50D,
despite it having a later sensor design and better micro lenses, plus
a later generation of processing software and silicon, it has less
dynamic range than the 40D. (For a fairer comparison take some 500D
images and compare to the 40D, without the improved micro lenses it
makes a more comparable baseline).

What is this strong evidence? News for you: DR is about noise. Noise is scale > dependent. Pixel DR has little to do with image DR, for the reasons already > explained in this thread; DR is a scale dependent quantity, since noise is a scale > dependent quantity. The 50D and 40D have about the same DR at the same > image scale.

I fear you are letting the maths get in the way of comprehending the
real world situation. The lower MP camera can cover a greater level
of shades of light before you get into noise issues. It is something
I have observed.

I am guessing it is something you have observed by pixel-peeping; am I right? If not, tell me how you arrived at this conclusion. If so, then you are comparing images at different scales. Again, noise is a function of spatial frequency; look at an image at two different scales, and different levels of noise are involved. The ability to resolve tonality depends on how large a region that tonality is being averaged over. A small patch of image has large fluctuations of tonality, and therefore poor tonal resolution; a large region has small fluctuations, and correspondingly good tonal resolution. The only thing that large pixels do is place a bound on how fine a region one is allowed to consider. Small pixels of the same light-gathering efficiency per area have the same tonal resolution at any scale that the large pixels can sample, in addition to having tonal resolution on scales where the large pixels cannot go.

The noise performance at ISO 1600 is actually better than at ISO 200 with

either the 40D or the 50D. It's just that for the light levels it is used for, the
signal is weaker, hence the S/N is lower.

You need to think of real photography and back out of abstract maths,
or at least apply it to the real situation. When you get to low light
levels you want higher ISO. If I were to take a shot at ISO 1600 and
a shutter speed of 1/1000 for and 1/60 at ISO 100, same lens
settings, which would have the higher noise? This is a real world
situation that photographers can comprehend. For the landscape ISO100
fine, for the elusive flying fox wombat flying through the air the
1/1000 shutter speed might be essential.

Again the straw man to be knocked down. Dullard that I am, even I understand that a higher exposure gathers more photons, resulting in higher S/N in the image (leading to what some people call higher IQ; something I aspire to ;-)

However, that was not under discussion. When asking which is better for noise relative to signal, high exposure or low exposure, sure enough high exposure wins every time. When asking which is better for noise relative to signal, ISO 200 or 1600, one should fix the exposure so that both camera settings are recieving the same signal, otherwise you are varying two variables (exposure and ISO) and attempting to infer that ISO A is better than ISO B, when what you actually determined was that exposure X was better than exposure Y.

At fixed exposure, ISO 1600 turns out to have less noise. One doesn't use high ISO in all circumstances, of course; when conditions allow an exposure beyond the raw saturation level of high ISO, then a longer exposure at lower ISO wins every time. One does use high ISO when it doesn't exceed that saturation of the ADC, precisely because it intrinsically has lower noise than low ISO at that same exposure. If it did not, there would be no reason to even have an ISO control in the camera -- exposure compensation in the RAW converter would be all that would be needed. In fact, some medium format backs work in precisely this way (they have a fixed ISO gain in the camera hardware); but I digress.

The good camera developments will open up using the cameras under a
greater range of conditions, I hope. Certainly the developments since
the 10D have allowed good image capture at higher ISO/lower light it
managed.

Yes, because read noise has been lowered substantially, in spite of the handicap those silly engineers have given themselves of shrinking the pixel size with each generation.

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/

 

[strawman_John]

I like how you guys are unable to carry the argument without a toys out of the pram moment.

Lets make it easy do the comparison on prints for example.

 

[strawman_John]

I apply the arguments I do as I compare prints as that is my desired output.

In the end it all comes down to what you can get out of the camera.

 

[S p i t z e r]

We need to make a few distinctions that are critical to the discussison.

1) We are talking about more pixels vs less pixels on sensors of the same size.
2) We are talking about the noise and DR of the image not the pixel.
3) We are talking about images at the same display size.

If you do not agree with any of the three statements above, then this is the reason for the discord. However, if you agree with the three statements above, the debate is about whether fewer pixels has less total image noise and more DR than more pixels created from the same size sensor and displayed at the same size.

In answering this question, we must first distinguish between the two types of noise: shot noise, which is a function of the total amount of light that the sensor captures, and read noise, which is a function of the sensor efficiency and supporting hardware.

For the same size and technology sensor, scene, and exposure, the shot noise will be the same, regardless of the pixel count. That is, the total light absorbed by all the pixels will be the same. The microlens covering for smaller pixels is exactly as efficient as it is for larger pixels.

We are now left with the read noise. It is my understanding that for sensors of the same design that the read noise from the pixel is proportional to its linear dimensions rather than its area. That means that there will be more total read noise with smaller pixels. For example, if a 2x2 pixel has a read noise of 2, a 1x1 pixel would have a read noise of 1. Thus, four 1x1 pixels (same area as one 2x2 pixel) would have a total read noise of sqrt (1^2 + 1^2 + 1^2 + 1^2) = 2, which is the same read noise as the 2x2 pixel. However, they would also record 4 times as much detail.

As I posted in the previous thread, the pics in this link demonstrate the point I am making:

http://forums.dpreview.com/forums/read.asp?forum=1018&message=28607494

Both pics use the same amount of sensor area with the same exposure, thus the shot noise is the same. However, the FZ50 pic is made with many more pixels giving it substantially more detail, and the same amount of noise.

Thus, more pixels do not result in more noise, but they do yield more detail.

 

[berleconi]

from micron sensor small sensor maker example
claim that more megabyte with about the same noise

 

[berleconi]

Is this a example ?
from micron sensor small sensor maker example
claim that more megabyte with about the same noise



Berl.

 

[bikinchris]

Some things that seem to be missed by people who need more pixels. It's probably intentional. As pixel density increases you not only have fewer photons in each well to work with and produce signal, but higher pixel density also reduces your ability to stop down and get more depth of field. Diffraction comes into play. When you shoot with a higher pixel density sensor, you wind up with a blurry image that lacks detail.
--
Chris, Broussard, LA