If this is not false advertisement, what is?

[Just Looking]

Pixels do not equal sensors. A pixel is the smallest unit of a
final image, regardless of how it was made. My monitor has 2 Mega
pixels, not 6 (2M red + 2M green + 2M blue). The Foveon puts out
3.4 Mega pixels, period...

And the Fuji S2 puts out 12 MP, so why do some people insist on calling it 6 MP?

j

 

[Champa]

I think Sigma is pathetic. If this is not deceiving, what is? They
make 3.3 MP sensor to look like 10 MP by counting 3 pixels (red,
green, blue) for each final image's pixel. Since Canon uses 4
elements for each pixel (2 greens, red, blue), here are the true
pixel counts: 10D = 24 MP (6x4), 1D = 16MP, 1Ds = 44 MP. Is'n that
sweet.
The next step will be counting hard drive capacities in bits, not
bytes. It's 8 times more space!
Yes, you can call this post wahtever you want. I'm just sick and
tired of this 'marketing'. I just baught a DVD, capable of
wrighting 4.7G of data. And guess what, I can only take 4.5
Gigabytes of data. Oh, I didn't know that 1G is not 1 Gigabyte,
it's just G, which means nothing.
If I had more time and cared more I'd sue Sigma. It plain SUX.
People will turn away from very good cameras only to find out that
they get only a third of image data in the end.
People!!! Beware of Sigma!!! It's a scam!!!

--

Now you know that's what we are here for...the place where those biggg guys come to take a dump when they need to feel some sense of security.
ca
http://www.pbase.com/champa
http://www.pbase.com/sigmasd9
http://www.pbase.com/cameras/sigma/sd9

 

[Just Looking]

You mean THIS language?
http://www.x3f.info/technotes/x3pixel/pixelpage.html

j

Peter and others,

I fail to see 10.2 Megapixels there. 10.2 million pixels are a
different thing, although we have been down this road a few times.

The 1Ds is advertised as a xx.x megapixel camera, the D60 as a x.x
megapixel camera, etc. Foveon and Sigma are trying to avoid this
because they are not comparable in this sense. Look at Phil's spec
charts to see how he handles this.

The only real point to all of this is to try and make some people
aware of the contortions that Foveon and Sigma are going through to
try to 1) get their point across while 2) not misleading people.
Obviously, within these organizations there are different voices as
well. However, if you read http://www.x3f.info carefully, you will find
what I regard as an honest effort to get this right. Of course, if
you have your bias filter turned off, it is easier.

None of us are naive enough to believe all marketing copy. However,
at times, people do try to state it as close to the truth as the
market allows.

And frankly, I agree, it might have been better to leave it at 10M
and 3.34 x3 sort of things.

How's that for forked tongue?

--
Laurence Φ€ 08 LL

http://www.pbase.com/lmatson/sd9_images
http://www.pbase.com/sigmasd9/root
http://www.pbase.com/cameras/sigma/sd9
http://www.beachbriss.com (eternal test site)

 

[Joseph S Wisniewski]

dropped by from the Canon forum just to take a dump and hasn't been
back to clean it up.

That's the sign of a master troll. The ability to craft an inflamatory post that will take on a life of its own and grow a nice, ling thread, without the troll having to come back and "fan the flames" or feed it more fuel.

Compare and contrast this to many of the folks who are often called trolls, suce as SG10 or Ger Bee. They are incapable of really good trolling. THeir threads only run long when they stay with the thread: arguing, elaborating, expounding point after point.

Kind of nice to see trolling done right, for a change.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[tammons]

It happens a lot on these forums. All these camera sensitive people from the other, ummm, you know forums, have to come by and put us in our place from time to time.

Personally I thought it was all about what was on the print.

Ansel Adams and Man Ray would probably roll over in their graves, or either die again laughing.

I think Sigma is pathetic. If this is not deceiving, what is? They
make 3.3 MP sensor to look like 10 MP by counting 3 pixels (red,
green, blue) for each final image's pixel. Since Canon uses 4
elements for each pixel (2 greens, red, blue), here are the true
pixel counts: 10D = 24 MP (6x4), 1D = 16MP, 1Ds = 44 MP. Is'n that
sweet.
The next step will be counting hard drive capacities in bits, not
bytes. It's 8 times more space!
Yes, you can call this post wahtever you want. I'm just sick and
tired of this 'marketing'. I just baught a DVD, capable of
wrighting 4.7G of data. And guess what, I can only take 4.5
Gigabytes of data. Oh, I didn't know that 1G is not 1 Gigabyte,
it's just G, which means nothing.
If I had more time and cared more I'd sue Sigma. It plain SUX.
People will turn away from very good cameras only to find out that
they get only a third of image data in the end.
People!!! Beware of Sigma!!! It's a scam!!!

--
Now you know that's what we are here for...the place where those
biggg guys come to take a dump when they need to feel some sense of
security.
ca
http://www.pbase.com/champa
http://www.pbase.com/sigmasd9
http://www.pbase.com/cameras/sigma/sd9

 

[tammons]

I will admit that even though SG10 was way over the top with his argumentative rhetoric, it was funny to watch him dig into some of these people pounding them constantly with specs and numbers, until they were so confused they gave up.

Maybe we should unleash him occasionally just to badger these knuckeheads to death.

dropped by from the Canon forum just to take a dump and hasn't been
back to clean it up.

That's the sign of a master troll. The ability to craft an
inflamatory post that will take on a life of its own and grow a
nice, ling thread, without the troll having to come back and "fan
the flames" or feed it more fuel.

Compare and contrast this to many of the folks who are often called
trolls, suce as SG10 or Ger Bee. They are incapable of really good
trolling. THeir threads only run long when they stay with the
thread: arguing, elaborating, expounding point after point.

Kind of nice to see trolling done right, for a change.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[Brian Slater]

dropped by from the Canon forum just to take a dump and hasn't been
back to clean it up.

That's the sign of a master troll. The ability to craft an
inflamatory post that will take on a life of its own and grow a
nice, ling thread, without the troll having to come back and "fan
the flames" or feed it more fuel.

Master troll? Give me a break. Nominated for Oxymoron of the Week. And of course you are a fan, as I am.

 

[Joseph S Wisniewski]

Joe,

I hope some day I can get to be as smart as you are,

Strange. All I hope for is to some day be happy...

but it's hard
when you don't reveal your sources of learning.

BS electrical engineering, Lawrence Technological University.

MS systems, Oakland University, emphasis DSP. My prof for image processsing and DSP was Manhauer Das (and there's a pretty good change I misspelled his first name), for pattern matching Hansen.

Postgrad work at Oakland in DSP, Bio-electronics, psychophysics, optics, and musical acoustics (never could focus on one field long enough to get a PhD).

BFA, Oakland. Been trying to recover from that and recapture my art for 14 years now.

Studied glassblowing for years under John Fitzpatrick.

Maybe you can
throw me a bone here...

What's the definition of "pixel" that you're using and arguing
about, and where can I find it in words others than yours?

I got mine from Jain, Anil "Fundamentals of Digital Signal Processing". These days, Gonzalez and Woods, "Digital Image Processing" is the standard. I'll pull a definition out of there for you tomorrow.

http://en.wikipedia.org/wiki/Pixel

And where is Foveon's admission that the X3 sensor needs an AA
filter that you're referring to here? (or did I mis-interpret your
dangling "this" in "And it's nice to see Foveon finally admitting
this."?)

Perhaps an overstatement on my part. The SD-10 does not alias as the SD-9 did. Therefore, it has an anti-aliasing filter. Either Foveon or Sigma put one in the camera. So, someone is admitting that it needs an AA filter. But that someone might not be Foveon.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[Joseph S Wisniewski]

dropped by from the Canon forum just to take a dump and hasn't been
back to clean it up.

That's the sign of a master troll. The ability to craft an
inflamatory post that will take on a life of its own and grow a
nice, ling thread, without the troll having to come back and "fan
the flames" or feed it more fuel.

Master troll? Give me a break. Nominated for Oxymoron of the Week.
And of course you are a fan, as I am.

I'm a "troll watcher" from way back. We've had trolls as long as we've had public forums. I've been watching them for 25 years or more on USENET.

There is such thing as a "master troll". By the normal USENET rules of troll watching, you have to make two posts that start 500 response threads. Since this forum has a 150 post limit, I guess that's the standard for trolling on dpReview.

We actually have a fellow here on dpReview, Gene Windell, who is a USENET master with several 500 post threads to his credit. He's very good, can make you think he's perfectly serious for months (even years) before pulling something.

It's a pleasure watching a real pro work...

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[Joseph S Wisniewski]

I will admit that even though SG10 was way over the top with his
argumentative rhetoric, it was funny to watch him dig into some of
these people pounding them constantly with specs and numbers, until
they were so confused they gave up.

Maybe we should unleash him occasionally just to badger

We don't need no steenkin' badgers...

these knuckeheads to death.

Well, that might be fun for a while, especially since this is autum and the weather is getting cool. But eventually the weather will warm again and the dead knuckleheads all over the floor will start to smelll...

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[Just Looking]

What's the definition of "pixel" that you're using and arguing
about, and where can I find it in words others than yours?

I got mine from Jain, Anil "Fundamentals of Digital Signal
Processing". These days, Gonzalez and Woods, "Digital Image
Processing" is the standard. I'll pull a definition out of there
for you tomorrow.

http://en.wikipedia.org/wiki/Pixel

That can't be the one you're using--you already said he Fuji S2 is not 12 MP. I look forward to you finding one that applies in the way you say.

And where is Foveon's admission that the X3 sensor needs an AA
filter that you're referring to here? (or did I mis-interpret your
dangling "this" in "And it's nice to see Foveon finally admitting
this."?)

Perhaps an overstatement on my part. The SD-10 does not alias as
the SD-9 did. Therefore, it has an anti-aliasing filter. Either
Foveon or Sigma put one in the camera. So, someone is admitting
that it needs an AA filter. But that someone might not be Foveon.

Doesn't alias like the SD9? Pretty nearly like it, according to the "Dave Box" test that I pointed out in a thread on "Just-right anti-aliasing":
http://forums.dpreview.com/forums/read.asp?forum=1027&message=6489299

j

 

[jeffkrol]

Sorry, I just think this should be repeated:

http://www.x3f.info/technotes/x3pixel/pixelpage.html

 

[Roland Karlsson]

Peter and others,

I fail to see 10.2 Megapixels there. 10.2 million pixels are a
different thing, although we have been down this road a few times.

And my mind just went blank .......

What are you trying to say?

Roland

 

[Roland Karlsson]

You mean THIS language?
http://www.x3f.info/technotes/x3pixel/pixelpage.html

This is a rather sad story. A new technology that does
it the "right way" have to fight the old technology that
does it "the wrong" way. Unfortunately, customers do
only understand numbers.

So, what do the new technology do? They invent new
ways ouf counting. They even write technically complicated
papers about it to make it more legitimate. They start to
talk about pixel sensors and then (without any problem),
calls the pixel sensors for pixels.

Clever. The only problem is that intelligent people that
understands what it is all about knows that this is all
the emperors new clothes. And the child said - but look
mom, he is all naked. And all understood - they had
been cheated.

Roland

 

[Joseph S Wisniewski]

Sorry, I just think this should be repeated:

http://www.x3f.info/technotes/x3pixel/pixelpage.html

You're quite right, it should be repeated. And disected. I especially enjoy this section, where the JCIA talks about pixel counts in a multiple sensor or multiple color plane camera.

b) For DSCs that use a plurality of image sensors, the number of
image sensors and the number of effective pixels corresponding
to each image sensor shall be noted. If the total number of
spatially sampled pixels is also noted, the description should
clearly state that the noted number is the total number of
effective pixels corresponding to all of the image sensors, to
avoid consumer misunderstanding.

This is relatively clear. If you've got multiple planes, you tell the consumer how many planes you've got, and how many pixels are on one plane.

The Foveon adds their interpretation of this:

This rule for three-chip cameras suggests that only
pixel sensors on a single chip should be counted, so the
"total number of spatially sampled pixels," or total
number of photodetectors on an X3 sensor,
could be used as the number of effective pixels."

The only problem is that the layers of sensors on the Foveon are spatially aligned, so that there are indeed just 3.4 million "spatially sampled" pixels, with three photodetectors per pixel.

This is the authorative argument which should end this issue once and for all. Foveon says their sensor is 3.4 megapixels.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[Just Looking]

We can only hope that your quotes out of context won't be accepted by everyone as ending the discussion...

j

Sorry, I just think this should be repeated:

http://www.x3f.info/technotes/x3pixel/pixelpage.html

You're quite right, it should be repeated. And disected. I
especially enjoy this section, where the JCIA talks about pixel
counts in a multiple sensor or multiple color plane camera.

b) For DSCs that use a plurality of image sensors, the number of
image sensors and the number of effective pixels corresponding
to each image sensor shall be noted. If the total number of
spatially sampled pixels is also noted, the description should
clearly state that the noted number is the total number of
effective pixels corresponding to all of the image sensors, to
avoid consumer misunderstanding.

This is relatively clear. If you've got multiple planes, you tell
the consumer how many planes you've got, and how many pixels are on
one plane.

The Foveon adds their interpretation of this:

This rule for three-chip cameras suggests that only
pixel sensors on a single chip should be counted, so the
"total number of spatially sampled pixels," or total
number of photodetectors on an X3 sensor,
could be used as the number of effective pixels."

The only problem is that the layers of sensors on the Foveon are
spatially aligned, so that there are indeed just 3.4 million
"spatially sampled" pixels, with three photodetectors per pixel.

This is the authorative argument which should end this issue once
and for all. Foveon says their sensor is 3.4 megapixels.

--
Ciao!

Joe

http://www.swissarmyfork.com

 

[Just Looking]

What's the definition of "pixel" that you're using and arguing
about, and where can I find it in words others than yours?

I got mine from Jain, Anil "Fundamentals of Digital Signal
Processing". These days, Gonzalez and Woods, "Digital Image
Processing" is the standard. I'll pull a definition out of there
for you tomorrow.

I could buy them both today on Amazon for $230, but I can't figure out how to get the new full-text search to find me the definition of pixel for free.

The Jain book title actually is "Fundamentals of Digital Image Processing", btw.

My old Gonzalez and Wintz only talks about gray-scale pixels, and nothing about how to apply the term to sensors or cameras. I look forward to learning more about modern usage of this term.

j

 

[Just Looking]

... Plain and simple. Luminance is an aspect of a pixel.

That's a new piece of definition I hadn't seen yet.

Are you talking pixels in files? Or on sensors? On a Bayer sensor, doesn't this definition then only have half the count that's usually given, since a Bayer filer array by definition (from the original Bayer patent) has luminance sensors at only 50% of the locations?

Weather it's a single scalar value (the luminance of a monochrome
pixel), a three dimensional vector of the three Foveon layers, or a
61 dimensional vector of luminance produced by a scientific camera
with a rotating wheel of 61 narrowband 5nm filters).

It doesn't matter if the single luminance value per pixel is the
result of a pure monochrome sensor, or weather there is a CFA
(color filter array) in front of the sensor. Nor does it matter if
said CFA is a 3 color Bayer pattern, the 4 color Sony pattern, the
6 color Kodak pattern, a 3 color pseudorandom pattern. A pixel is a
pixel.

Fascinating. I'm sorry I'm impatient and keep bugging you, but I'm looking forward to that definition of pixel in which each one has a luminance and all these things don't matter. My mind is too feeble to conceive it.

Pixels have 2 dimensional spatial properties: frequency, bandwidth,
etc. They are not voxels (three dimensional pieces of a solid
object) and cannot be treated mathematically as such.

Right. Unless you use a different definition. For camera sensors, the thing counted as a "pixel" traditionally is a thing that Foveon has re-arranged into the third dimension. Perhaps some definitions say it can still be counted as a pixel? I haven't finished trying to survey the definitions, but people seem to have a lot of them.

And yes, a Fuji S2 sensor has 6 million of them. Not 12 million.

6 million of what? Are half of its output pixels (or whatever you call them) missing their luminances? Or are you switching definitions again?

j

 

[Mike Fried]

I think Sigma is pathetic. If this is not deceiving, what is? They
make 3.3 MP sensor to look like 10 MP by counting 3 pixels (red,
green, blue) for each final image's pixel.

Ok, here's a subtle distinction that Andy doesn't get:
Number of Sensors is NOT the number of Pixels.
Let's define:
MS = Million Sensors

MSS = Million Sensor Sites (actual number of sensor locations -- Sensors like the Foveon and the Fuji SR have multiple sensors in one site. The Fuji SR does it for dynamic range, whereas the Foveon uses it for color)
MP = Million Output Pixels

I'm going to make a table here:
Camera Model MS MSS MP
SD9/SD10: 10.5 3.5 3.5
1Ds 11 11 11
1D 4 4 4
D60/10D/300D 6 6 6
S2Pro 6 6 12
Fuji SR Sensor 6 3 6

You decide how much interpolation goes on. Who is advertising falsely? Fuji SuperCCD? Canon? Sigma? Obviously, some people don't get the distinction between sensors, sensor sites, and pixels. Phil Askey does, so you should read his reviews. Fuji makes great sensors, so do Canon and Foveon.
-Mike

 

[Joseph S Wisniewski]

... Plain and simple. Luminance is an aspect of a pixel.

That's a new piece of definition I hadn't seen yet.

Are you talking pixels in files? Or on sensors? On a Bayer
sensor, doesn't this definition then only have half the count
that's usually given, since a Bayer filer array by definition (from
the original Bayer patent) has luminance sensors at only 50% of the
locations?

Not exactly. Since pure monochromatic light sources (or objects that reflect monochromatic light) are pretty rare, just about any color will excite all the cells of a Bayer array to some extent, so you can theoretically derive luminance from any channel or combination of channels. Bayer preferred to derive it from green because tht's the strongest of the colors (highest amplitude reflections and therefore the best SNR of all the colors) and being in the middle of the spectrum, generally the sharpest for most modern lenses. Since a uniformly spaced rectangular array of three colors means you're going to have twice as many of one color as the others, why not use green?

The Foveon sensor is actually similar. Red is noisy and blurry. Red has 1/2 the sensitivity of the other channels (50% of visible light penetrates deeper than the Foveon cells) and the sensitivity curve means you need to use a differential gain of about four between the red and green layers to get a true "red" signal. This means red ends up 8 times as noisy as green, so it takes a lot of work using the green and blue signals to fix red noise.

The blur comes from diffusion. Because red photons have to penetrate deeply into the sensor (about 5um, if memory serves) they have plenty of room to "drift" across the cell borders and leak into sdjoining pixels.

Weather it's a single scalar value (the luminance of a monochrome
pixel), a three dimensional vector of the three Foveon layers, or a
61 dimensional vector of luminance produced by a scientific camera
with a rotating wheel of 61 narrowband 5nm filters).

It doesn't matter if the single luminance value per pixel is the
result of a pure monochrome sensor, or weather there is a CFA
(color filter array) in front of the sensor. Nor does it matter if
said CFA is a 3 color Bayer pattern, the 4 color Sony pattern, the
6 color Kodak pattern, a 3 color pseudorandom pattern. A pixel is a
pixel.

Fascinating. I'm sorry I'm impatient and keep bugging you, but I'm
looking forward to that definition of pixel in which each one has a
luminance and all these things don't matter. My mind is too feeble
to conceive it.

Pixels have 2 dimensional spatial properties: frequency, bandwidth,
etc. They are not voxels (three dimensional pieces of a solid
object) and cannot be treated mathematically as such.

Right. Unless you use a different definition. For camera sensors,
the thing counted as a "pixel" traditionally is a thing that Foveon
has re-arranged into the third dimension. Perhaps some definitions
say it can still be counted as a pixel? I haven't finished trying
to survey the definitions, but people seem to have a lot of them.

I wouldn't go with that definition. It will drive the folks who buld displays crazy, and they predate camera builders.

And yes, a Fuji S2 sensor has 6 million of them. Not 12 million.

6 million of what? Are half of its output pixels (or whatever you
call them) missing their luminances? Or are you switching
definitions again?

6 million photodiodes at 6 million discrete spatial locations. The raw output is also 6 million cells, just skewed at a 45 degree angle relative to a more conventional sensor.

--
Ciao!

Joe

http://www.swissarmyfork.com