Digital Camera Sensor Performane Summary updated Feb 3

On Thu, 07 Feb 2008 22:11:22 -0700, "Roger N. Clark (change username to
rnclark)" <username.TakeThisOut@qwest.net> wrote in <47ABE47A.4000609.TakeThisOut@qwest.net>:

>John Navas wrote:
>>
>> but I see quite a few unsupported assumptions in your work,
>>
>So what might be those "quite a few" be?

I don't have time to go through all of it, so here are a few things
I noticed on my initial reading:

The first paragraph is composed of unsupported claims. If it's intended
to be a summary of the rest of the document, that should be stated. If
instead you have citations, footnote numbers would be good. As it is,
it reads like a set of unsupported starting assumptions.

Photons aren't converted into electrons -- they stimulate the release of
electrons, called photoelectrons.

ALthough not currently used in digital cameras AFAIK, there are multiple
electron sensors -- 1:1 is not ideal.

You use your own definition of dynamic range.

Read noise is affected by readout speed, which you seem to ignore other
than for the A/D converter.

Etc.

The essence of scientific analysis is that we either make postulates
(like "It turns out that the noise in good modern digital cameras is
dominated by photon counting statistics, not other sources.") and then
prove them, or accumulate evidence and draw conclusions from that
evidence. Many or even most of things you say may be true, but that
doesn't follow from your presentation.

--
Best regards,
John Navas
Panasonic DMC-FZ8 (and several others)
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

John Navas wrote:
> On Thu, 07 Feb 2008 22:11:22 -0700, "Roger N. Clark (change username to
> rnclark)" <username RemoveThis @qwest.net> wrote in <47ABE47A.4000609 RemoveThis @qwest.net>:
>
>> John Navas wrote:
>>> but I see quite a few unsupported assumptions in your work,
>>>
>> So what might be those "quite a few" be?
>
> I don't have time to go through all of it, so here are a few things
> I noticed on my initial reading:
>
> The first paragraph is composed of unsupported claims. If it's intended
> to be a summary of the rest of the document, that should be stated. If
> instead you have citations, footnote numbers would be good. As it is,
> it reads like a set of unsupported starting assumptions.

There are over 35 refere3nces on the web page, many from sensor
manufacturers. All the statements in the intro are supported
by multiple references.

> Photons aren't converted into electrons -- they stimulate the release of
> electrons, called photoelectrons.

This is semantics. Photons are absorbed, and that energy goes into
exciting electrons so there is a 1:1 correspondence in
photons to electrons. Once the photoelectrons are produced,
they are simply called electrons. Read any of the manufacturers
data sheets (many of which I reference) and you'll see
full well capacities and read noise specified in electrons,
not photoelectrons.

> ALthough not currently used in digital cameras AFAIK, there are multiple
> electron sensors -- 1:1 is not ideal.

These are electron multiplying CCDs. One photon produces one
photoelectron and that electron then gets multiplied. There is no
change in quantum efficiency. It is a method used to effectively
make read noise negligible. That was important when read noise
was 15 or so electrons, but is less important now that other methods
result in much lower read noise. And what relevance is this
to the properties of current digital cameras, none of which
use the technology.
>
> You use your own definition of dynamic range.

No! That is the industry standard definition. I even give a reference
to Hamamatsu, a sesnor manufacturer who gives the definition.


> Read noise is affected by readout speed, which you seem to ignore other
> than for the A/D converter.

While you are correct, read noise is reduced by very slow and
multiple correlated reads. In scientific sensors, often a
many seconds are used to read out a 1-megapixel chip. But,
again, that is irrelevant for two reasons: modern digital
cameras (e.g. 2007 models) have already achieved extraordinarily
low read noise, as low as about 2 electrons. And again,
how is slow read relevant? The read noise is a measured
property, whatever methods are used. The values for read noise
on the web page are measured, not theoretical or modeled.
>
> Etc.
>
> The essence of scientific analysis is that we either make postulates
> (like "It turns out that the noise in good modern digital cameras is
> dominated by photon counting statistics, not other sources.") and then
> prove them, or accumulate evidence and draw conclusions from that
> evidence. Many or even most of things you say may be true, but that
> doesn't follow from your presentation.

Read the references. The fact that CCDs and CMOS sensors have noise that
is dominated by photon noise is well established by multiple
sources, from the sensor manufacturers themselves, to scientists
working with the devices, to the amateur astronomers testing
the devices.

The title of the web page is "Digital Camera Sensor Performance Summary"
not "This is why...."

For the evidence of how this is determined, see references on the
web page in question,
http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary
and

Procedures for Evaluating Digital Camera
Sensor Noise, Dynamic Range, and Full Well Capacities;
Canon 1D Mark II Analysis
http://www.clarkvision.com/imagedetail/evaluation-1d2

Digital Cameras:
Counting Photons, Photometry, and Quantum Efficiency
http://www.clarkvision.com/imagedetail/digital.photons.and.qe

and references therein.

Roger
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

On Fri, 08 Feb 2008 07:53:00 -0700, "Roger N. Clark (change username to
rnclark)" <username.DeleteThis@qwest.net> wrote in <47AC6CCC.9020504.DeleteThis@qwest.net>:

>John Navas wrote:
>> On Thu, 07 Feb 2008 22:11:22 -0700, "Roger N. Clark (change username to
>> rnclark)" <username.DeleteThis@qwest.net> wrote in <47ABE47A.4000609.DeleteThis@qwest.net>:
>>
>>> John Navas wrote:
>>>> but I see quite a few unsupported assumptions in your work,
>>>>
>>> So what might be those "quite a few" be?
>>
>> I don't have time to go through all of it, so here are a few things
>> I noticed on my initial reading:
>>
>> The first paragraph is composed of unsupported claims. If it's intended
>> to be a summary of the rest of the document, that should be stated. If
>> instead you have citations, footnote numbers would be good. As it is,
>> it reads like a set of unsupported starting assumptions.
>
>There are over 35 refere3nces on the web page, many from sensor
>manufacturers. All the statements in the intro are supported
>by multiple references.

That may be true, but there's no way to tell, and I doubt many are going
to try to figure out which references are supposed to support which
claim. As I wrote, footnotes would be good. Likewise a clear chain of
reasoning, instead of sweeping summary statements.

>> Photons aren't converted into electrons -- they stimulate the release of
>> electrons, called photoelectrons.
>
>This is semantics. ...

I respectfully disagree.

>> ALthough not currently used in digital cameras AFAIK, there are multiple
>> electron sensors -- 1:1 is not ideal.
>
>These are electron multiplying CCDs. One photon produces one
>photoelectron and that electron then gets multiplied. There is no
>change in quantum efficiency. It is a method used to effectively
>make read noise negligible. That was important when read noise
>was 15 or so electrons, but is less important now that other methods
>result in much lower read noise. And what relevance is this
>to the properties of current digital cameras, none of which
>use the technology.

The issue comes from your use of the term "ideal".

>> Read noise is affected by readout speed, which you seem to ignore other
>> than for the A/D converter.
>
>While you are correct, read noise is reduced by very slow and
>multiple correlated reads. In scientific sensors, often a
>many seconds are used to read out a 1-megapixel chip. But,
>again, that is irrelevant for two reasons: modern digital
>cameras (e.g. 2007 models) have already achieved extraordinarily
>low read noise, as low as about 2 electrons. And again,
>how is slow read relevant? The read noise is a measured
>property, whatever methods are used. The values for read noise
>on the web page are measured, not theoretical or modeled.

Again, you're making claims without support.

>Read the references.

I've read many of them. The problem is that it's not at all clear which
references are supposed to support which claim, if any references at
all. Otherwise there's no good way to figure out whether your
interpretation of those references is correct or not, an issue
compounded by statements like photons being converted into electrons.

>The fact that CCDs and CMOS sensors have noise that
>is dominated by photon noise is well established by multiple
>sources, from the sensor manufacturers themselves, to scientists
>working with the devices, to the amateur astronomers testing
>the devices.

If so, then it should be straightforward to add citations/footnotes. As
it is, it reads like unsupported claims. There's no way to tell what
support exists for which claim. That you provide references is a very
good start, but expecting your audience to either read all the
references or just take you at face value greatly diminishes the
potential weight and value of your work.

>The title of the web page is "Digital Camera Sensor Performance Summary"
>not "This is why...."
>
>For the evidence of how this is determined, see references on the
>web page in question,
>http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary
>and

Again, your audience shouldn't be left to figure out on their own which
of your other pages is supposed to support which item.

It's a pain to properly structure a chain of reasoning with supporting
citations, but it is needed for any work that wants to be taken
seriously.

--
Best regards,
John Navas
Panasonic DMC-FZ8 (and several others)
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

[A complimentary Cc of this posting was sent to
Roger N. Clark (change username to rnclark)
<username.RemoveThis@qwest.net>], who wrote in article <47ABBB81.4050806.RemoveThis@qwest.net>:
> > Photons do not come into the context of this table at all. Full well
> > is measured in electrons; likewise for the noise. An electron
> > added to a smaller sensel will generate higher voltage than when added
> > to a larger sensel.

> >> Look at higher ISOs, and the 1DIII read noise drops to
> >> 2.1 electrons compared to the 1DII 3.9 electrons, about
> >> an 86% improvement, which helps high ISO imaging.

> > You know that THIS has nothing to do with high bit count. In the
> > regime of the full well of about 4K electrons, 2 electron noise is
> > "approximately 9-bits".

> BS


Com'on, Roger, you are an intelligent person. There is no reason to become
desperate and go to kindergarden-level defences. I'm sure you ARE able to
understand this!

Hope this helps,
Ilya

P.S. Here are the arguments again:

With capacitance, just go for the following thought experiment:

take a large flat capacitor; charge it by 2 electrons (I mean move
2e charge from one plate to another); measure the induced voltage.

Now cut the capacitor into two (cut perpendicularly to the plates).
You get two capacitors. Now charge each by 1 electron. Measure the
induced voltage on each.

Do you still think the voltages are different from those of the first
case? Then connect two halves together.

---

With noise of 9-bit ADC: mark 512 possible measurements on the interval
[0 .. 4096]; the step is about 8, right? So you are half-way there!

A good ADC would just, essentially, "snap" each point of the interval
[0..4096] to the closest marked point. So each marked point is
surrounded by an inteval of length 8 of values which would be discretized
to this output value. So the error is equally distributed in [-4 .. 4].

What remains is to calculate the RMS of a value equally distributed in
[-1 .. 1]; it is 1/sqrt(3). So the RMS of the error of discretization
is 2.3 (which is very close to the measured value of noise 2.1 from your
tables).
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

"Roger N. Clark (change username to rnclark)" <username.TakeThisOut@qwest.net> wrote:
>Now is when Floyd will probably jump in and the name calling
>will start. We last left Floyd when I pointed out there was a
>temperature gradient in cold weather gear when he said
>there was none, and then corrected him when he said there was
>no moisture gradient, I left while he was ranting and raving
>about how he knew how to survive in the cold (which in his rage
>he never realized I didn't question). Floyd's view was
>so focused on a specific thing he couldn't see the bigger
>picture.

Well you did get two things right. When I step in the name
calling starts... because *you* start with personal attacks as
soon as someone who focuses on a specific thing and won't let
go and can't be side tracked by your insults and diversions.

You could have argued technical points with Ilya (becaue you
are right and he is missing the point), but instead you have
to try to lead the discussion away from that, to anything and
everything else, including people not even involved in the
the thread. Then you bitch because other do the same thing!

--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd.TakeThisOut@apaflo.com
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

Ilya Zakharevich wrote:
>
> Com'on, Roger, you are an intelligent person. There is no reason to become
> desperate and go to kindergarden-level defences. I'm sure you ARE able to
> understand this!

Yes, of course I understand capacitance and charge.
I told you pixel size is is not a factor in the read noise
in digital cameras. Again, look at Figure 3 at:
http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary
There is no trend with pixel size. Other effects swamp any
such pixel size trend. For example, look at the 2007 era cameras:
the 1D Mark III versus the 40D. The 40D has almost double the
read noise despite smaller pixels.

You might actually look at Figure 3, and seeing there is no
trend, try to do some research as to why that might be
(hint: technologies like double correlated read, or even cheaper
cameras use cheaper and noisier electronics).

But in your typical response, you'll explain in some bizarre
way why I'm wrong and don't have a clue, as you've done to every
answer I've ever given and similarly the way you've responded to
many others who have tried to help you over the years.

I originally responded to your post explaining the larger picture,
but you responded:
"It would help if you FINALLY get some clue about..."
And then you complain about how someone responds to YOU.

So it seemed that a simple response of BS is much quicker
and to the point. But I'll make it even faster: I'll no
longer respond.

Now is when Floyd will probably jump in and the name calling
will start. We last left Floyd when I pointed out there was a
temperature gradient in cold weather gear when he said
there was none, and then corrected him when he said there was
no moisture gradient, I left while he was ranting and raving
about how he knew how to survive in the cold (which in his rage
he never realized I didn't question). Floyd's view was
so focused on a specific thing he couldn't see the bigger
picture.

Similarly, you are focused of the size of the capacitor and
not the larger picture.

I think this newsgroup has degraded into argument for the
sake of argument, and little to do with real intelligent
discussion. Look at what happened in this thread in
rec.photo.digital.slr-systems. It degraded into bitching
without any discussion of the content of the original
post or web site. This thread isn't much better.
And most threads quickly degrade into name calling, and very
very rarely an apology.

Now several people will jump in and continue name calling.
It's so predictable and is the reason many have left or
rarely participate. That and the constant Nikon/Canon bitching.
No one can ask a simple question because someone will jump in
about that manufacturer sucks. This used to be a great
newsgroup, but many knowledgeable have people, but few of them
are left. And that is why I left for months. I was here to learn
from people who had knowledge I did not have, and in turn
I tried to help in areas where I had knowledge. I did not
come here to argue.

Count me as one who will rarely, if ever participate in the future.
And I'm done with this thread too.

Roger
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

Ilya Zakharevich wrote:
>
> Com'on, Roger, you are an intelligent person. There is no reason to become
> desperate and go to kindergarden-level defences. I'm sure you ARE able to
> understand this!

Yes, of course I understand capacitance and charge.
I told you pixel size is is not a factor in the read noise
in digital cameras. Again, look at Figure 3 at:
http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary
There is no trend with pixel size. Other effects swamp any
such pixel size trend. For example, look at the 2007 era cameras:
the 1D Mark III versus the 40D. The 40D has almost double the
read noise despite smaller pixels.

You might actually look at Figure 3, and seeing there is no
trend, try to do some research as to why that might be
(hint: technologies like double correlated read, or even cheaper
cameras use cheaper and noisier electronics).

But in your typical response, you'll explain in some bizarre
way why I'm wrong and don't have a clue, as you've done to every
answer I've ever given and similarly the way you've responded to
many others who have tried to help you over the years.

I originally responded to your post explaining the larger picture,
but you responded:
"It would help if you FINALLY get some clue about..."
And then you complain about how someone responds to YOU.

So it seemed that a simple response of BS is much quicker
and to the point. But I'll make it even faster: I'll no
longer respond.

Now is when Floyd will probably jump in and the name calling
will start. We last left Floyd when I pointed out there was a
temperature gradient in cold weather gear when he said
there was none, and then corrected him when he said there was
no moisture gradient, I left while he was ranting and raving
about how he knew how to survive in the cold (which in his rage
he never realized I didn't question). Floyd's view was
so focused on a specific thing he couldn't see the bigger
picture.

Similarly, you are focused of the size of the capacitor and
not the larger picture.

I think this newsgroup has degraded into argument for the
sake of argument, and little to do with real intelligent
discussion. Look at what happened in this thread in
rec.photo.digital.slr-systems. It degraded into bitching
without any discussion of the content of the original
post or web site. This thread isn't much better.
And most threads quickly degrade into name calling, and very
very rarely an apology.

Now several people will jump in and continue name calling.
It's so predictable and is the reason many have left or
rarely participate. That and the constant Nikon/Canon bitching.
No one can ask a simple question because someone will jump in
about that manufacturer sucks. This used to be a great
newsgroup, but many knowledgeable have people, but few of them
are left. And that is why I left for months. I was here to learn
from people who had knowledge I did not have, and in turn
I tried to help in areas where I had knowledge. I did not
come here to argue.

Count me as one who will rarely, if ever participate in the future.
And I'm done with this thread too.

Roger
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

On Feb 10, 7:53 am, "Roger N. Clark (change username to rnclark)"
<usern... RemoveThis @qwest.net> wrote:


> I think this newsgroup has degraded into argument for the
> sake of argument, and little to do with real intelligent
> discussion.

Yes it has. And why? Not because of the people who are more interested
in arguing than exchanging information, but because some of those who
would like to actually have an intelligent discussion treat everybody
as if they're the same. Thus, if someone is simultaneously incapable
of admitting a mistake, is arrogant enough to think that he knows
better than everybody else and has the patience and time to jump into
every discussion and disagree, the he'll obviously just ruin the group
for everybody.

In short, the problem isn't that such people exist (there aren't many,
3-4 basically), it's that they are allowed to ruin everything by being
taken seriously.


>
> Now several people will jump in and continue name calling.
> It's so predictable and is the reason many have left or
> rarely participate. That and the constant Nikon/Canon bitching.
> No one can ask a simple question because someone will jump in
> about that manufacturer sucks.

Yes, it's useless to attempt most discussions now.

> Count me as one who will rarely, if ever participate in the future.

That's a shame, but understandable. Arguing just to argue is getting
tiring.
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

[A complimentary Cc of this posting was NOT [per weedlist] sent to
Floyd L. Davidson
<floyd.TakeThisOut@apaflo.com>], who wrote in article <8763wxuxvs.fld.TakeThisOut@apaflo.com>:
> Well you did get two things right. When I step in the name
> calling starts... because *you* start with personal attacks as
> soon as someone who focuses on a specific thing and won't let
> go and can't be side tracked by your insults and diversions.
>
> You could have argued technical points with Ilya (becaue you
> are right and he is missing the point)

May I somehow invert the charge, and note that your claim that "Ilya
is wrong" does not look very technical... Could you at least mention
which argument of mine do you consider wrong, and why?

Thanks,
Ilya
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

On Feb 10, 9:35 pm, Chris Malcolm <c... DeleteThis @holyrood.ed.ac.uk> wrote:
> acl <achilleaslazari... DeleteThis @yahoo.co.uk> wrote:
> > On Feb 10, 7:53 am, "Roger N. Clark (change username to rnclark)"
> > <usern... DeleteThis @qwest.net> wrote:
> >> I think this newsgroup has degraded into argument for the
> >> sake of argument, and little to do with real intelligent
> >> discussion.
> > Yes it has. And why? Not because of the people who are more interested
> > in arguing than exchanging information, but because some of those who
> > would like to actually have an intelligent discussion treat everybody
> > as if they're the same. Thus, if someone is simultaneously incapable
> > of admitting a mistake, is arrogant enough to think that he knows
> > better than everybody else and has the patience and time to jump into
> > every discussion and disagree, the he'll obviously just ruin the group
> > for everybody.
> > In short, the problem isn't that such people exist (there aren't many,
> > 3-4 basically), it's that they are allowed to ruin everything by being
> > taken seriously.
>
> If there's only a few then a good start to solving the problem is
> killfiling or ignoring them.
>

Yes, that was part of the point I was trying to make. The rest is that
this doesn't really work unless most people do it ("most" may not only
mean numerically, but also somehow weighed, eg if there are two people
who do nothing than flame each other and don't disturb the rest it's
fine).

Obviously, one possible problem is that not everybody agrees who
should be ignored, but that's not the reason for the current mess: if
it was only that, then people could have just segmented into
orthogonal groups and there would be no problem. So the problem lies
elsewhere.
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

acl <achilleaslazarides RemoveThis @yahoo.co.uk> wrote:
> On Feb 10, 7:53 am, "Roger N. Clark (change username to rnclark)"
> <usern... RemoveThis @qwest.net> wrote:


>> I think this newsgroup has degraded into argument for the
>> sake of argument, and little to do with real intelligent
>> discussion.

> Yes it has. And why? Not because of the people who are more interested
> in arguing than exchanging information, but because some of those who
> would like to actually have an intelligent discussion treat everybody
> as if they're the same. Thus, if someone is simultaneously incapable
> of admitting a mistake, is arrogant enough to think that he knows
> better than everybody else and has the patience and time to jump into
> every discussion and disagree, the he'll obviously just ruin the group
> for everybody.

> In short, the problem isn't that such people exist (there aren't many,
> 3-4 basically), it's that they are allowed to ruin everything by being
> taken seriously.

If there's only a few then a good start to solving the problem is
killfiling or ignoring them.

--
Chris Malcolm cam RemoveThis @infirmatics.ed.ac.uk DoD #205
IPAB, Informatics, JCMB, King's Buildings, Edinburgh, EH9 3JZ, UK
[http://www.dai.ed.ac.uk/homes/cam/]
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

[A complimentary Cc of this posting was sent to
Roger N. Clark (change username to rnclark)
<username.DeleteThis@qwest.net>], who wrote in article <47AE8333.3070207.DeleteThis@qwest.net>:

> Yes, of course I understand capacitance and charge.

Good. Then you UNDERSTAND that using the same ADC, 25% smaler sensel
will produce 25% full well, and 25% smaller noise (in electrons).

> I told you pixel size is is not a factor in the read noise
> in digital cameras.

As you UNDERSTAND (see the preceding sentence), it is.

> You might actually look at Figure 3, and seeing there is no
> trend

??? I was not discussing any "trend". What I did was I compared 1DmII
with 1DmIII, and noticed that the performance at 50ISO is 90% compatible
with an assumption that they use the same ADC, and just the
capacitance changed.

> But in your typical response, you'll explain in some bizarre
> way why I'm wrong and don't have a clue, as you've done to every
> answer

I'm sorry if you got such an impression. IMO, about 2/3 of the
answers you give is correct, and is very useful to the discussion.

What I'm trying to correct is the remaining 30%. And, AFAIK, you
improve often (although very slowly); it takes you from half-a-year to
a year, but for many particular BSes you believed in, you stop to
believe in it.

[Of course, during these "half-a-year to a year", you will call the
people who try to teach you names. But nobody is perfect...]

Yours,
Ilya
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3 

On 10 feb., 22:08, Ilya Zakharevich <nospam-ab....TakeThisOut@ilyaz.org> wrote:
> [A complimentary Cc of this posting was sent to
> Roger N. Clark (change username to rnclark)
> <usern....TakeThisOut@qwest.net>], who wrote in article <47AE8333.3070....TakeThisOut@qwest.net>:
>
> > Yes, of course I understand capacitance and charge.
>
> Good. Then you UNDERSTAND that using the same ADC, 25% smaler sensel
> will produce 25% full well, and 25% smaller noise (in electrons).

Nice. 25% smaler (<- that's not phrased correctly) sensel also means
25% of actualy excited electrons at given light level.
"In the physics of photon counting, the noise in the signal is equal
to the square root of the number of photons counted"
That means 100% more noise at given light level.


> > But in your typical response, you'll explain in some bizarre
> > way why I'm wrong and don't have a clue, as you've done to every
> > answer
>
> I'm sorry if you got such an impression <snip>
>
> What I'm trying to correct... <snip>
>
> Yours,
> Ilya

I just wish you and your likes would shut up when Roger decides to
participate in this NG is all I'm saying.
 >> Stay informed about: Digital Camera Sensor Performane Summary updated Feb 3