From: ejmartin on
On Jul 19, 2:21 am, ejmartin <ejm_60...(a)> wrote:

> So, binning FZ50 pixels to the size of 1D3 pixels gives a DR of
> 64000/12.1=5290=12.4 stops. On the other hand, the 1D3 *sensor* has a
> DR of 71000/4.0=17750=14.1 stops.  John only gets the FZ50 to be
> competitive because he uses the read noise that the 1D3 delivers at
> base ISO, around 20 electrons, rather than the sensor read noise of 4
> electrons.  Doing so, it looks like the FZ50 wins since its measured
> base ISO DR is about 11.7 stops.  But that is misleading, because the
> 1D3 read noise at that ISO has almost nothing to do with the
> properties of the sensor, but as you say it is coming from the
> downstream electronics.

Sorry, the antecendent of "its measured base ISO DR is about 11.7
stops" is unclear as written -- that's the 1D3's measured DR at ISO
100. So the FZ50 pixels scaled up to tile APS-H format would do
better (12.4 stops) than the current implementation of the 1D3 at low
ISO, but worse than a hypothetical 1D3 (14.1 stops) where one is able
to get around the DR limitations of the ISO amplifier/ADC downstream
of the sensor. The FZ50 gets around this issue by having pixels whose
DR fits within the DR of available ISO amplifier/ADC components. I
think there are simple schemes to get the full DR of the 1D3 sensor as
well, but for whatever reason Canon/Nikon have not implemented them.
Of course, this problem does not arise at high ISO, where the sensor
noise dominates the read noise; and there, even the current 1D3
implementation has higher DR than the hypothetical scaled up FZ50.
From: ejmartin on
On Jul 19, 2:25 am, "David J Taylor" <david-tay...(a)blueyonder.neither-> wrote:
> Roger N. Clark (change username to rnclark) wrote:
> []
> > Once again, let's work a simple example.
> > Sensor A has 100 pixels for every pixel in sensor B.
> > Both have pixels with the same read noise = 4 electrons.
> > Assume 100% fill factors and same QE.
> > Compute dynamic range for the area of pixel in sensor B.
> > sensor A:  X = Signal A = Signal B if we sum the signal from each
> > pixel in A (sum(100* signal in one pixel).
> > read noise:  sensor A = 4*sqrt(100) = 40  sensor B= 4 electrons.
> > Dynamic range, DR: Sensor A DR= X/40, Sensor B: X/4
> > The large pixel has higher DR.
> []
> > Roger
> .. but sensor A is delivering a much higher spatial resolution.  Depending
> how the image is displayed, and how the eye/brain interprets the signal
> and the noise, it is not impossible that the perceived image from the
> higher resolution sensor will be preferable to that from the lower
> resolution sensor ....
> Having said that, it strikes me that for a particular
> scene/light-level/lens/display/viewing-distance combination there will be
> an optimum sensor size, and the different sensor-size cameras we use
> simply cover a different subset of the total picture-taking conditions.
> Cheers,
> David

Well, that's the tradeoff isn't it? At least, that's the tradeoff for
current implementations at high ISO, where it's the sensor properties
that control DR -- big photosites for more sensitivity in shadows, or
small photosites for more resolution. At low ISO, due to the
limitations of current implementations in the ISO amplifier/ADC, one
actually does better per unit area with small photosites, because they
place less demands on the dynamic range of that downstream
electronics. Small photosites will be better at low ISO until camera
companies start delivering all the DR that the sensor is offering.
From: Steve on

On Sat, 19 Jul 2008 03:25:40 GMT, John Sheehy <JPS(a)no.komm> wrote:

>Steve <steve(a)> wrote in
>> You said you're trying to measure the noise and DR of the sensor as
>> pixel density changes.
>This is not a measurement, per se. It is a real-image demonstration. I
>think measurement, as it is usually done, is a bit flawed. SNR of pixel
>statistics do no equal SNR of an image, and even if you take viewed pixel
>size into account, which, when factored with pixel noise, gives a good
>representation of visible noise, any image that is upsampled from the
>original capture resolution needs to use a scaling factor based on the
>*original* pixel's dispalyed size, not the individual pixel of the
>upsized image. Statistical noise measurement has a serious hole in it,
>where symmetry about the 1:1 scaling factor of pixels does not exist;
>standard deviation of noise drops at the new pixel level when you
>downsample, but it does not increase in an upsample, statistically, even
>though it does increase visually, because standard deviations are totally
>blind to the details of noise frequency.
>> But the only thing you're measuring above is
>> resolution of a given area of the focal plane.
>That is all I claim to be measuring. Why do you think I am or should be

If all you claim to be measuring is resolution at the focal plane then
why are you even showing pictures of anything at all? It's pretty
obvious that higher pixel density has more resolution at the focal
plane. Of course, that measure is completely meaningless to any real
world normal photographic application. It only is meaningful for
specialized applications where one is purposefully trying to image
something projected onto a very small focal plane. For real world
photographic applications, you have to take the entire sensor size
into consideration also, since that's what the camera is imaging onto.

>measuring something else. FOR THE SEEMINGLY MILLIONTH TIME: I AM NOT
>"PIXEL DENSITY". Comparing a 12MP 1/2.5" sensor's image to that of a
>12MP FF with the same field of view is *NOT* a comparison of pixel
>density. It is a comparison of full images from different sensor sizes,

Actually, that *IS* comparing pixel density. Because that's the only
thing different between the two sensors. The total number of pixels
is the same, the image falling on them is the same, etc. etc.... The
only thing different is the pixel density, and that is what you're
trying to compare. In your test, you are not keeping the total number
of pixels that capture the image the same. So you are comparing more
than just pixel density, you're also comparing resolution. And in
your resolution comparison, you are severly handicapping the 400D by
upscaling it 3 times more than the FZ50.

>with the same number of pixels, which happen also to have different
>> I don't think anyone
>> is going to argue that higher pixel density gives better resolution at
>> the focal plane. But that is absolutely meaningless if the size of
>> the sensor is different.
>Pixel density is discussed concerning sensors of the same size. There
>the same area from cameras with vastly different pixel densities, AND
>sensor size, to extrapolate what a large sensor with high pixel density
>might do, compared to the low pixel density that larger sensors now have.

There are plenty of cameras out there with the same size sensor and
pixel densities that are as different as what you're comparing. The
fact that doing so disagrees with your results shows that your
extrapolation is not a valid one.

>> Almost nobody except you and some people
>> with special applications that may require a small sensor size cares
>> about resolution at the focal plane.
>So why the hell are you and all the rest of the peanut gallery replying
>to my posts, if this doesn't interest you? Do you burst into churches on
>Sunday morning to anounce that their religion doesn't interest you? Do
>you go to sporting events to protest people's interest in sports? So
>why, then, do you fell a need to come here in this thread and tell us
>that the subject matter is irrelevant to you?

Maybe because I have a special application that requires greater pixel
density and a small sensor? But I know for absolute certainty that my
camera with higher pixel density is noiser with less DR than the one
with lower pixel density. However, the small sensor camera still
works better than the large one for afocal astronomical photography.
But not for anything else I've ever done.

>> The measure of resolution the
>> rest of us care about is the resolution of the image captured by the
>> entire sensor.
>Who said I don't care about that? What I am interested in, is what
>different pixel densities or pitches do in sensors of a given size. A

If that's what you were really interested in, that's what you would
measure. And in measuring that, you would scale the lower pixel
density to 100% and the higher one to something less than that rather
than scaling the high pixel density to 100% and the lower one to
something much greater than that. Also, you would look at the entire
DR, not just a tiny part of it.

>> Your test of DR and noise vs. pixel density is not valid because you
>> did not take the variable of resolution out of the equation by keeping
>> it constant for the images captured.
>Nonsense. It is valid for what *I* am testing. I don't care what you
>think I should be testing. I am not comparing sensors of different
>sizes, with similar MP counts. That is old hat; it is accepted and
>doesn't need to be proven that bigger sensors usually collect more
>photons and give lower image shot noise!

The problem is that your test is not testing what you claim it is. You
are purposefully handicapping the low pixel density sensor, both in
terms of resolution and in terms of not taking advantage of it's full
DR, just to prove a point. That's not a valid test. It's a
hypothesis where the test is skewed to come to the conclusion that the
tester wants. That happens all the time, usually by testers with an
agenda which you obviously have.

>> I.e., if you have two 10MP
>> sensors of different sizes, the only real world meaningful way to
>> measure noise and DR vs. pixel density is to allow both 10MP sensors
>> to capture a similar scene across the entire sensor and then compare
>> the results at 100%.
>Nonsense. That is a test of SENSOR SIZE, not PIXEL DENSITY.

Nonsense. That is a test of pixel density. Since the pixel density
is what determines the sensor size for a given number of pixels. If
you don't keep the number of pixels the same, you are testing
something more than pixel density. If you want to take SENSOR SIZE
out of the equation and truly measure only PIXEL DENSITY, then you
have to compare the same sensor size since it's obvious that your
extropolation doesn't work. And the main reason your extrapolation
doesn't work is that you are blowing up the low density camera to over
100% rather than shrinking the high density camera to less than 100%.
If you did your test that way, you'd have a better chance of your
extrapolation giving valid results. That still doesn't solve the
problem you're having with only using a small portion of the low
density sensor's available DR though.

>> When you do
>> that, you are no longer measuring the noise and DR of a sensor when
>> only the pixel density changes.
>> I'm pretty sure you understand this and are just trying to troll for
>> an argument.
>I've given far more than just a hint of what I am demonstrating here, yet
>almost no one who has responded seems to understand. My thoughts are
>going way over your head, apparently.

Not at all. I know exactly what you're trying to show. The problem
is that your test is flawed in several severe ways and is designed to
give the results you're after. That is not a valid scientific test.

From: Steve on

On Fri, 18 Jul 2008 21:55:18 -0600, "Roger N. Clark (change username
to rnclark)" <username(a)> wrote:

>It seems most people here realize that there is an optimum pixel
>density that trades focal plane resolution with S/N, dynamic range,
>and real resolution. That is also what the scientific models

Exactly. Taking the little table from your previous post, I'll add a
little more the the upper end to make a point:

APS-C DSLR with 1 pixel?
APS-C DSLR with 9 pixels?
APS-C DSLR with 0.1 megapixels?
APS-C DSLR with 1 megapixels?
>APS-C DSLR with 3 megapixels?
>APS-C DSLR with 6 megapixels?
>APS-C DSLR with 8 megapixels?
>APS-C DSLR with 10 megapixels?
>APS-C DSLR with 20 megapixels?
>APS-C DSLR with 40 megapixels?
>APS-C DSLR with 80 megapixels?
>APS-C DSLR with 120 megapixels?
>APS-C DSLR with 200 megapixels?
>APS-C DSLR with 500 megapixels?
>APS-C DSLR with 999 megapixels?

Somewhere in there is the sweet spot where the resolution is
sufficient to capture the spatial detail you want in an image and the
S/N and DR is sufficiently low to capture the light intensity detail
you want. If you look at graphs of S/N and DR vs. the resolution
figures above, the increase in S/N and decrease in DR is going to be
exponential. So stay somewhere where the slope isn't too bad yet and
the resolution is sufficient and you'll be happy.

It's pretty obvious that you don't want to be in the 1 or 9 pixels
area even though I'm sure each of those pixels would have tons of DR.
It's also pretty obvious you don't want to be in the 200-999 megapixel
area either because each pixel could probably barely manage a single
photon capture.

However, using John's test, you can show that 1000 megapixels for an
APS-C sized sensor has better S/N and DR than the 10 megapixel sensor.
Because he would only use the portion of the DR that the 1000 MP
sensor can manage when comparing it to the 10MP sensor, which would
have tons of wasted headroom in his test. And he would cut a tiny
area of the focal plane image out of the 10MP sensor and blow it up to
10,000% and compare that with a 100% crop of the 1000MP camera.


From: Steve on

On Sat, 19 Jul 2008 04:28:24 GMT, John Sheehy <JPS(a)no.komm> wrote:

>Steve <steve(a)> wrote in
>> Or Nikon's D3 vs. Canon's 1DsMkIII. Both have very similar sensor
>> sizes but very different pixel densities. Both are pretty much the
>> top of the heap as far as what either manufacturer is capable of in
>> every other respect so it's a fair comparison of what pixel density
>> does to noise and DR.
>> The problem there is that just like comparing almost identical low-end
>> cameras with different pixel densities, the results would not support
>> his hypothesis that increasing pixel density increases DR and
>> decreases noise for a given sensor.
>The quantum efficiencies and the per-pixel low ISO read noise are similar,
>though, so you can still see what the noise is like at the image level,
>crops blown up, and the same downsample ratios for both. I know the 1Dsmk3
>will do better, but you might need proof.

You'd be wrong if you are comparinig what you say you are trying to
compare, S/N and DR. In any valid test, the D3 would do better if you
look at a crop at 100% for the D3 and 50% for the 1DsMkIII to account
for the fact that the D3 has lower spatial resolution. The D3 would
do better specifically because it has larger pixels. Of course, you
could also look at 100% crops for both. But that would be incorrectly
handicapping the 1DsMkIII, as you have done to the 400D in your test
against the FZ50.

Now, if you wanted to measure spatial resolution, of course the
1DsMkIII would do better. But that's not what you say you're trying
to measure.