From: Paul Furman on
John Sheehy wrote:
> Paul Furman wrote :
>> I'm saying more dense pixels might be useful for astronomy where you
>> can stack them to overcome the noise & DR limitations and take
>> advantage of the higher resolution/magnification through the same
>> telescope. I'm not even sure about this, so I said 'may be useful'.
> The higher pixel density does not have more noise and DR limitations to
> overcome; that is my main point. Higher pixel densities generally give
> higher DR at base ISOs.

Why does base ISO matter?

> Only a *very* small number of DSLRs have lower
> area-based read noise than P&S pixels like the FZ50's do, and not by
> much.
> The 400D here, clearly has more noise at the same level of illumination.
>> Yeah, OK but the 400D goes up to ISO 1600 in fine form so this is
>> pretty much irrelevant isn't it?
> Not at all. Low signals can be deep shadows at low ISOs; not just
> midtones at high ISOs.

Sure, but if you have bright highlights also, the small pixels are badly
crippled. The sample shot does have bright areas but as I understand, it
was severely underexposed. Can you explain a little more clearly about
these factors please:

Shot noise -
Read noise -
Full well capacity -

I sort of vaguely can imagine the possibility that the noise may be
better due to the math when it all works out in the real world (perhaps)
but a larger pixel should absolutely have more dynamic range: that is
real simple math.

And of course an 85MP AP-S sensor is going to cost a heck of a lot of
money, but yeah, lets just say we want the best possible image for this

Paul Furman

all google groups messages filtered due to spam
From: John Sheehy on
Bates <nw1008(a)> wrote in

> You're saying that you could not compare the 350D to the 450D because
> you might attribute the differences to things such as minor
> improvements in read noise and microlens efficiencies but you CAN
> compare two completely different cameras with completely different
> size sensors with completely different lenses and attribute ALL of the
> differences to the difference in pixel density.

No. The pixel density or pixel pitch is so different beteween the two
cameras that if there were any truth to the notion that increased pixel
density increases noise and decreases DR, that it would surely show here,
and it did not. The big-pixel result is junk, compared to the small
pixel result. The big pixels did nothing for us here. A 40D would have
faired a little better, but still inferior.

> I was on your side
> for the most part originally but sorry to say you've lost me. I just
> hope you are not in science because this is NO way to run an
> experiment.

This is the best possible way to run this experiment, as there are no
possible pairings out there of cameras with the same mount and the same
size sensors with vastly different pixel densities of the same era.
Someone suggested the 1D vs the 1Dmk3. The 1Dmk3 would slaughter the 1D,
and someone would say, "well, the 1D has an old Panasonic CCD sensor; no
wonder". The 1D was perhaps a better collector of photons than APS
sensors of its time, but it is a poor imager by current standards, per
unit of area. Its read noise, per pixel, is as high or higher than the

There are better ways to run *other* experiments, yes.

So, where are other people's better experiments with "better" sensor


<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: John Sheehy on
ASAAR <caught(a)> wrote in

> The best thing I can say for John is that he might actually believe
> his own nonsense.

And what exactly is this nonsense? I've said several things here in this

> It's ironic that he has accused others of
> "cognitive dissonance" because he seems to be the one suffering most
> from that affliction.

How so? What reality am I refusing to accept?

> When people here recently saw DPReview's article demonstrating
> sensor noise and detail at high ISOs, only one person disagreed with
> the obvious demonstration that the best performer by a wide margin
> (after Canon's DSLR ringer) was Fuji's F30. Panasonic's FZ-50 was
> included in the comparison and it didn't do well at all. This
> shouldn't have been surprising given that it stuffed nearly twice as
> many pixels into a sensor smaller than the one used by Fuji's F30.
> Sheehy insisted that the FZ-50 produced more detail even though it
> the F30 produced clearly legible text in its images where the FZ-50
> rendered the text as totally unreadable fuzzy blobs, and resorted to
> a grossly unscientific description of the F30's images as (IIRC)
> plastic and cartoonish.

That's very simple. Fuji's NR algorithms look for edges, Panasonic's do
not try as hard to preserve them. The JPEG output of both cameras are
sickly, IMO, just in different ways. Fuji RAWs from the 6MP 1/1.7"
sensor are nothing special; here's the 6500 at ISO 800 (100% crop; simple
RAW conversion with no NR):

> If the FZ-50 really has some kind of noise advantage at ISO 100
> over Canon's DLSR it would be useful if he could demonstrate it
> using real images, and to the extent possible, with both cameras
> filling their sensors with the same image.

That makes no sense at all. I am not comparing cameras; I am comparing
the effects of pixel density. I am fighting the myth that pixel noise
determines image noise, and that Pixel DR determines image DR, and
pointing out that with similar era-technology, the effecticve *image*
read noise can be significantly lower with higher pixel density, because
read noise does not increase at a rate equal to the reciprocal of pixel
pitch, when you miniaturize pixels.

> You might want to take
> portraits where a person's face fills the frame and then compare
> greatly enlarged crops.

Again, nothing to do with pixel density, per se.

> What Sheehy has done instead is to compare portions of each sensor
> having the same actual area.

That is the only way to compare pixel density, which is the only thing I
wish to demonstrate here.

> The 5D stores 12.8mp in a sensor that
> is 22.4 times larger than the FZ-50's 10.1mp sensor. So the largest
> part of the 5D's sensor that could be compared would contain only
> 0.571mp (if compared against the FZ-50's entire sensor). It really
> is ludicrous to design a test where more than 95% of the light
> reaching the DSLR's sensor is discarded, while proclaiming that this
> is a fair test designed to keep things equal.

You've got to be kidding me. The rest of the sensor is totally
irrelevent to the issues of the effects of PIXEL DENSITY or PIXEL PITCH.
Don't tell me to test/demonstrate something else, which has nothing to do
with my area of interest.

I don't start threads to state the obvious. My interest is areas of
false perception in "common wisdom". The big myth going around today is
that "cramming" more pixels into a given sensor size deteriorates image
quality, and that is generally nonsense. It seems to be that way

1) People are ignorant and judge cameras by 100% pixel views.

2) Downsizing software (like Photoshop's onscreen display at <100%) is
stupid, and drops pixels instead of weighing them evenly, increasing
visible noise more for higher-MP images; this is due in part because:

3) Our display devices have extremely coarse resolution. We should be
upsizing our entire images to fit them on the screen; not downsizing

4) Because people are viewing images at 100%, or with 100% pixel noise in
poor downsizing software, manufacturers are meeting the bad reactions
with ridiculous NR, destroying image detail.

> But then, Sheehy owns
> an FZ-50 which is known to have very poor high ISO performance.

Well, you show me an ISO 13,500 push from another small-sensor camera
that beats the one I have shown you, scaled to my 100% crop, that still
maintains some detail.

> And
> despite his claim that implies otherwise :

>> People actually believe that the noise of a pixel is the noise of an
>> image; that the DR of a pixel is the DR of an image.

> he seems to want to show that his FZ-50 is desirable by examining
> the noise and DR of a pixel,

No, I gave a visual comparison. When I look at pixel noise statistics, I
am fully aware that they are totally meaningless without a context.
Visual noise is pixel noise times displayed (original, if an upsized
image) pixel width.

My appraisal of the FZ50 is fairly objective. I don't have any problems
finding problems with things I have paid money for. You act as if I
spent all my hard-earned money on an FZ50 and now have to bolster it. I
could buy every DSLR, and every P&S in production, and every sub-$10,000
DSLR lens, without taking out a loan, and my FZ50 investment is only a
tiny percentage of what I have already spent on digital photography. Pay
attention now - It has an excellent lens, does RAW, and I own it, so that
is why I make reference to it; it is a convenient model of the small
pixel. It is not very good at JPEGs, and I would never recommend one to
a JPEG-only shooter unless they have looked at the samples and are happy
with them.

It *does* have the lowest standard deviation of read noise of any P&S
I've tested. It also has the highest quantum efficiency. There may be
others lower, but there is no way for me to know without buying every P&S
that does RAW, and testing myself, since this information is unavailable.
The read noise is a little worse than the standard deviation would
suggest, however, since it has some short line noises. Spending a little
bit more money on the readout circuitry would probably make the line or
dash noises less severe.

> while the rest of the world judges the
> worth of their cameras by the noise and DR of the images they
> produce.

People have proven themselves to be prone to illusion. Most people think
an image downsized to 10% with nearest neighbor is sharper and more
detailed than the original.


<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: John Sheehy on
Steve <steve(a)> wrote in news:hhdu741e94q168kgsfi3m8e79prn535eid@

> That may be what you did, but it doesn't prove what you're trying to
> prove, i.e., stuffing more pixels into a given sensor size doesn't
> degrade noise or DR. In doing what you did above, the sensor from the
> 400D would be capturing an image that is overall several times the
> area of the FZ50. No one in their right mind would consider that a
> fair comparison of different sensor sizes when what you're trying to
> prove is whether stuffing in more pixels does not degrade noise and DR
> for a "given sensor size."

I don't know what to say to you. You have a very warped view of reality;
you are just totally and completely wrong. The proper way to compare pixel
densities is to compare the same magnification of the same sensor area.
Anything else is a measure of something else.


<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: John Sheehy on
Scott W <biphoto(a)> wrote in

> It seems to me if you are shooting in low light conditions you would
> want to have both cameras at their best setting for that. You test is
> kind of like racing two cars but limited them both to third gear, even
> though one car has a forth gear.

Are you telling me that there is no need for good shadows at low ISOs? I
could have shot both of these at ISO 1600, and the 400D would have done a
tad better with the noise, but the resolution would still be lacking.

> If you keep the gain turn down so
> low, i.e. ISO 100, you are not seeing the read noise in the 400D, you
> will only see the read noise at higher ISO settings.

There is more read noise relative to an absolute exposure at ISO 100 than
at 1600 in the 400D. In the FZ50, they are nearly the same, regardless of
ISO (it seems to be lowest at 200 and highest at 800 in this camera, but
not by a lot).


<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><