Re: The Joy of Pixel Density

ASAAR <caught@xxxxxx> wrote in
news:6rov74hjr208gua15e3bcv8c7hm4sq62c7@xxxxxxx:

On Thu, 17 Jul 2008 22:15:06 GMT, John Sheehy wrote:

That makes no sense at all. I am not comparing cameras;

Only because sense can't penetrate your thick skull. People don't
buy sensors.

I don't care what people do. I care about what gives or potentially
gives the best imaging.

They buy and use cameras. If you can't show that your
warped theory provides any real world benefit, you may want to
examine why that is so.

Why in the world would anyone with any amount of intelligence, with the
history of all that came before us, believe that the best things already
exist?

BTW, what you failed to understand is that
the direct comparison would be between the images, not the cameras.
See above . . .

The direct comparison of what? I am getting the sense that most of the
critics here have no viable working model in their heads of what a sensor
is, and what it does, yet they feel obliged to make all kinds of comments
assuming some kind of magic.

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.

I hate to break it to you, but you're demonstrating much more than
you realize.

I know; I clearly don't have patience for stupid people; that is obvious.

I don't start threads to state the obvious.

They are obvious because the threads are all rehashes of what
you've previously posted. Unconvincing before, unconvincing now.

And where is your counter-evidence? There is none. There is no evidence
that higher pixel density lowers DR or increases noise. Every shred of
so-called evidence assumes a 100% pixel view, uses JPEGs, or unequal
conversions (No commercial RAW converter converts different cameras the
same way; Brands that have a history of high RAW noise get extra NR by
default, just like the manufacturer's JPEGs do).

The big myth going around today is that "cramming" more pixels
into a given sensor size deteriorates image quality, and that is
generally nonsense.

Your nonsense continues. Consider DPReview's conclusions about
your pet FZ50. They admitted that it had "excellent resolution",
and "under the right conditions it produces superb output." It
seems that the myth creator is *you*.

What myth? I think the FZ50's JPEGs are miserable - that's the camera's
big downfall. It should have had two options; honest, simplehonest
conversions for those who know how to deal with small-sensor noise on
their own, and Fuji-like over-the top noise reduction. I don't know why
they make the choice that they do here; maybe something as simple as
someone well-liked by the CEO (his nephew) coming up with the NR that is
pleasing to no one.

DPReview recognizes what you
apparently can't, that cramming too many pixels into the sensor
increases that probability that more inferior images will be
produced because conditions aren't always optimal.

Guess you're out of the loop, but there's a big showdown going on at
DPReview right now, because the intelligesia there have challenged Phil
on his pixel-centric, noise-reduced, unequal conversion comparisons, and
it has become clear to him that his most knowledgeable forum members
almost unanimously believe that the conclusions are invalid. He was
engaged in such a discussion, and when asked for proof of his viewpoint,
he became mute. Hopefully, he's doing a lot of thinking about this.

Of course, for cameras that don't do RAW, he can't possibly compare equal
conversions, but for those, he still doesn't have to compare at the pixel
level. And guess what, his most recent comparison, written since his
silence, has had one image scaled to meet the other.

If everything you say about pixels evaluated for noise and DR at
low ISOs is true, it still doesn't matter if you can't show how that
will help people to select a camera that produces superior images.

Who said I was trying to tell people what camera to buy? Boy, do you
have a warped sense of reading. No matter how many times I tell you I
think the JPEG output of the FZ50 is not the best, and no matter how many
times I tell you I am only using it's pixels because of their RAW
qualities, you will keep on insisting in your warped mind that I am a
rabid fanboy, justifying his purchase.

This conclusion is far clearer and far more helpful for that purpose
than anything you've ever written.

Many things that people find clear are actually false, under the hood.
People like simplicity, more than they are interested in truth.

Producing a camera this size with 35-420mm (equiv.) F2.8-3.7
zoom lens means you have to use a small sensor, and small
sensors mean compromise because noise is always going to be
an issue.

True so far, except that at base ISO the RAW images are quite acceptable;
as good as a Canon DSLR at ISO 400 or 500.

I completely understand why Panasonic chose to jump
from 5 to 8 to 10 megapixels in three generations of its flagship
FZ camera - it's a lot easier to sell a camera on pixels than
picture quality, and the average consumer (and it must be said
the average camera store buyer) uses megapixels above all else
when sorting cameras into categories. Panasonic's marketeers
knew full well that with 10MP 'consumer' SLRs on the horizon
(and 10MP the new high end for compacts too) the FZ50 needed
a headline 'resolution' that kept it near the top. I do, however,
think it was a mistake to think that the FZ50 buyer is the 'average'
consumer, unable to base decisions on anything but megapixels...

There are three layers to this onion. There is the hypothetical blind
consumer who thinks that more MPs are always better in every way. Then
below that there are people who think that more MPsruins images in a
given sensor size. Below that are the people who understand that there
really is some value in higher pixel density, an that most of the
perceived problems by the middle layer are due to NR and display
technology problems (100 PPI monitors are a cruel joke!; so is
Photoshop's noisy zoom at less than 100%), and that the underlying data
is almost always better at higher pixel densities, except when the MFR
starts taking shortcuts, such as the cheaper late stage/ADC circuits to
save money (ala Canon 450D and Pentax K20D, whose high-ISO noise improved
over their predecessors', while the low ISO noise increased)

And so what we have is a camera that stretches its sensor to
almost breaking point and compensates for the lack of sensitivity
in anything but the brightest conditions by using excessive noise
reduction. The FZ50 is an excellent 5 or 6MP camera, but a
rather less impressive 10MP camera. Is this a problem? Probably
not - by the time the huge files have been shrunk down for
printing or viewing on-screen they look fantastic,

Maybe that should have been a hint to himself and to you. All sensors
should be compared at the same size output; not zoomed in 100% to the
pixels. We take images, not pixages.

and the
handling and features are quite simply peerless. But do not, for
a minute, think that the 10 million pixels you're getting with the
FZ50 bear anything but a passing resemblance to the 10 million
pixel images you'll get from a good SLR once you get above
ISO 100, or once light levels start to drop.

I agree with that comment 100%. Are you surprised? If you are, perhaps
you have some major problems when reading my words. I have NEVER, EVER
claimed that a 1/1.8" sensor can outperform an APS-C sensor, except
perhaps at read noise; quite a few compacts have the same or less read
noise than some DSLRs. If you think I just contradicted myself; think
again. Read noise is not the only noise. Lower shot noise is had in the
bigger sensors (given equal exposure, the image shot noise is inversely
proportional to the square root of the sensor area, times the quantum
efficiency).

And so, to sum up; for the serious user the FZ50 is without doubt
the best equipped, best specified and best handling 'bridge camera'
on the market today, and under the right conditions it produces
superb output. It is a rewarding and enjoyable photographic tool
that - once you've learned its quirks - offers a compact 'all-in-one'
solution to anyone wanting a huge zoom range without all that lens
changing and all that bulk. Inevitably this involves a certain amount
of compromise; the smearing of fine, low contrast detail that is the
hallmark of the Venus III engine limits the FZ50 to low ISO settings
for any serious photography unless you're happy to accept that
you'll never be able to produce big enlargements. For me this is an
acceptable compromise, and - though I wish Panasonic would drop
the megapixel race and concentrate on picture quality - it does
produce excellent printed results.

That's because the nose isn't in the monitor at 100% zoom. You do
realize that 100% zoom on a monitor means a printed image 3 to 4 feet
wide, something you're not even likely to try, don't you?

Yet, people make judgements condemning tiny-sensor cameras for noise at
100% pixel view.

If this had been a mould-breaking
5 or 6 megapixel with excellent low noise performance throughout
the ISO range (and particularly up to ISO 800) it would no doubt
have performed considerably better and would have been an easy
choice for a Highly Recommended. As it is it just squeezes in thanks
to its many other outstanding qualities - and only for those users
who can live without anything over ISO 200.

What Phil does not understand is that the FZ50 has less RAW noise than
just about any 5 or 6 MP camera when downsampled to 5 or 6 MP, or any 5
or 6 MP camera upsampled to meet the 10.1MP at full resolution, based on
the RAW data. The problem with NR is that the manufacturers are
approaching it from a pixel-centric viewpoint, and as the per-pixel shot
noise rises as they increase the pixel density, they lose the ability to
easily assess what pixels need to be squashed, as individual pixel noise
starts to approach hot pixel levels, so, instead of that waxy, continuous
texture at 5MP, they have little pock-marks in the waxy texture at 10.1
MP, from the pixels that broke the threshold. If there were no NR at
all, except a small amount of high-frequency chroma filtering, this would
not even be an issue when viewing 5 and 10 MP images at the same size;
the higher individual pixel noise would dissolve more into the scene,
with potentially lower image-level read (but not shot) noise. The shot
noise would only improve if the QE improved. There is little room for QE
improvement in the FZ50's sensor; according to some sensor experts on
DPReview forums (industry insiders), the FZ50 microlens/nMOS sensor
design is top-notch.

--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@xxxxxxx>
<<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
.

Relevant Pages

  • Re: D3x great camera, but...
    ... decrease pixel size, its like cutting a pie into smaller pieces, so ... Bayer sensor design and the need for a blur filter. ... equal comparison I've seen, shows that P&S-sized sensors are as good or better than DSLR sensors at photon efficiency, and that base ISO read noise, integrated by pixel pitch, is also equal or lower ... But neither full well capacity nor electron densities are ...
    (rec.photo.digital.slr-systems)
  • Re: DPI
    ... The pixel originates in your camera. ... There are roughly eight million Pixels on the sensor in your ... resolution discernable by the human eye. ...
    (alt.photography)
  • Re: Nikon D80/200 - Canon 30d
    ... than an el cheapo Point&Shoot camera with a tiny teeny-weeny CCD ... from a manufaturer that its CMOS sensor has less noise than a CCD ... just look at the images (not the ...
    (rec.photo.digital.slr-systems)
  • Re: The Joy of Pixel Density
    ... the sensor DR you're actually interested in. ... amplification) and its noise is probably the irreducible minimum read ... How does pixel read noise scale with pixel pitch? ... programmable gain amplifier into the "middle read noises" Nm. ...
    (rec.photo.digital)
  • Re: Any thoughts /news on Foveon sensors?
    ... >the design of the sensor and the fact that the pixel count hasn't changed. ... The X3 sensor doesn't suffer from colour moire on B&W images, ... >anti-alias filter of a Bayer sensor camera. ...
    (rec.photo.digital)