Re: What's an "effective pixel"?

> Roy Smith wrote:
>
>> After 25 years with my OM-2, a couple of years ago I broke down and got a
>> Cannon PowerShot. Great little camera, but I'm ready to move up. One
>> thing I'm thinking of is the Olympus E-500, but I'm a long way from
>> plunking down any money yet.
>>
>> I'm reading a review that says its has "8.0 million effective pixels".
>> What the heck is an "effective pixel"?

As others have pointed out and for the purposes of this question the
effective pixels are those that contribute the final viewable downloaded
picture. There are more pixels on the sensor chip but remember the chip is
square but the lense projects a circular image. There are also parts of the
chip that are used for colour balance, light levels etc. The exact whys and
wherefores are irrelevant for the moment.

So when comparing manufacturers claims of megapixels its simpler when
comparing to count the effective pixels rather than the total. That is
all but sigma and fuji. Leaving these out for the moment will save the
arguments over interpolation, manufacturers wild claims etc.

Also because some manufacturers were quoting all the pixels on the sensor,
reviewers favour the 'effective' pixels when quoting numbers. these days
its not so much an issue when you have 6 or 8 megapixel cameras rather than
1 to and 3. Going from 1 to 1.3 was significant. Nowadays you woudn't
notice the difference between 6 and 6.3 without using specialist equipment
and looking at graphs and charts and things. Your own eyeballs won't be
able to detect them. When you get up to 8 then the difference between 8.1
and 8.2 or more worrying about it is pointless.

So what about fuji? Well keeping it simple fuji have two technologies that
vary from a standard sensor. The first is pairing sensors. That is 1 large
and 1 small sensor together in the one individual spot where a normal sensor
would only have 1. This gives fuji a unique advantage over normal sensors
in the way of dynamic range. Basically detail in very bright areas and very
dark areas can be captured using a single exposure.

The big argument is whether you count each sensor as a pixel or whether a
pair of sensors equals 1 pixel. My advice is to count in pairs but consider
the extended dynamic range when comparing the cost. To quote several
reviewers "i was surprised to find this isn't yet another gimmick this
technology works"

The other technology fuji employs are hexagonal shaped sensors instead
square ones. For clarity i must apologies that when i sensor in this case i
mean the individual photosensitive receptors that make up a sensor. This
leads me to another more confusing and more overgeneralised point which is:-

'what is a pixel really?'

For this conversation a 'normal sensor and a normal' pixel is considered to
be an individual receptor. Each one has a coloured filter over it so that
it only receives one colour. It registers the intensity of this colour. So
at this point the recorded info is greyscale. The camera must take the info
each receptor and then guess what the original colour was. Don't forget
there isn't a filter for every colour and in every conceivable shade. So
out of a grid of two green a red a blue and in some cases yellow to produce
and inbetween colour such as purple. Using program such as photoshop you
can look at each colour channel. When you do you will see that each
individual channel is black and white. What black and white image is really
how bright that particular ray of light was when it hit the sensor after
passing through the filter. Of course the complicate things further the
filters aren't 100% efficient. You would expect mauve and purple to get
through a blue filter to an extent more than a green one. This is also taken
into account.

Now fuji use a hexagonal array. So they can fit more similar sized
receptors into the same space. They also claim a hexagonal arrangement
makes for a more efficient grid for calculating a colour. A receptor might
be part of more than one group i shall now call a cluster in order to
provide info for an adjacent cluster.

I shall come back to this 'cluster' after explaining sigma technology.

Sigma and foveon.
A very clever man found that different wavelengths of light penetrate
silicon wafers to different depths. So place a receptor at each depth and
you can have RGB in this fashion. But why do this? Well with a normal
sensor everything is fine if a green ray of light reaches a green sensor.
But what if it hits the next one? Whats worse is 'what if the ray is
purple?', then you need several receptors to register the single point of
coloured light. With a foveon you still need more than one receptor '
cluster' but as they are behind each other rather than side by side means a
single coloured ray the width of one receptor and the height of one receptor
can be captured. Which means much finer detail.
So why aren't sigma outselling everyone? Well the years of research of
several companies and scientists have gone into refining the maths that will
accurately combine a normal grid of receptors into an accurate colour
whereas the foveon idea hasn't been around for so long. Also it seems
onlysigma and a few academic bods are working on the problem.

'so why can't they use the same maths (algorhythm).' Well as a ray of
light enters passes through the silicon it loses energy. If it has to
travel to the rear of the sensor its signal will be much lower. So the
problem is two fold, not only has the actual intensity dropped but its so
low that you will probably have to amplify the signal which means the
possibiliy of noise.

'so wot? They tweak the algorhythm and everything is hunky dorey right?'
Erm well, yes and no. The thing is they haven't cracked it yet. Foveon are
notorius for getting innaccurate skin tones.

'phew thats alot to take in but why all the arguements ove effective
resolution?'

Well remember i mentioned clusters? Shouldn't that be a single pixel or
should you count the individual receptors?

So has the fuji got 6 or twelve. They claim that their more efficient
system should count as more pixels. Then other people say hang on mulitply
by 2 cos they have pairs. Others say hang on they don't count cos they are
2 receptors to one pixel. So the brass tacks is that 'effective megapixels
is 6.

Then the foveon crowd say well we have 3 million groups of 3. So that is
3x3 receptors so that is 9 million pixels. (n:b i know its ten cos it is two
green so its rggb not rgb). Others say no way its only 3 million.

Well i say it depends on how you define the pixel in the standard bayer
sensor. Are you counting each individual receptor as a pixel or are you
counting rggb as a pixel. To be perfectly honest are you counting the next
row down on the grid as included in that pixel.

should it not be
rggb
rgbg
rbgg
grgb
which means 10 million divided by 4 horizontal and 4 vertical = one pixel.
I believe only the one who wrote the algorthym can truly answer. That. but
there is a simple solution

1.Using their own eyeballs reviewers say fuji are being optmistic saying
their 6megapixel sensor is equivalent to 12. Its more like 9 or ten.
b.The extra dynamic range is worth considering in combination with the
supposed increase in pixels.
2.Regardless of how you rank the foveon in megapixels the colour balance
sucks so rule it out.
3.When buying into olypus system you are buying into the 4/3rds system.
This means a very small sensor which means you need to understand the myth
of focal length multiplication v field of view cropping and Depth of field
increase which isn't really increased cos you are actually using a wider
lens that does have lots of dof but you are selectively enlarging. Then
there is the issue of noise. then there is the issue that no matter what
further advances with other brands you can buy a new camera but you don't
need to replace your lenses. Whereas with the 4/3rds system the lenses are
solely designed for the current half 35mm sensor. Which happens to be the
smallest dslr sensor in the market.

All these i will leave for another time.



.

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: The Joy of Pixel Density
    ... critics here have no viable working model in their heads of what a sensor ... That is the only way to compare pixel density, ... that higher pixel density lowers DR or increases noise. ... Who said I was trying to tell people what camera to buy? ...
    (rec.photo.digital)
  • Kodak working to extend bounds of mediocrity
    ... technology with a new CMOS pixel to create the KODAK KAC-05020 Image ... Sensor, the world's first 1.4 micron, 5 megapixel device. ... "Camera phones and other small-pixel consumer imaging devices often ...
    (rec.photo.digital)
  • Re: Measure distances w 1mm precision
    ... For a good camera system you need at least 3000-5000 Euro (camera and lenses). ... With a 20 mm lenses and a sensor with 1280 x 1024 pixel you will have an field of view of 850 mm at a distance of 2 m. Lenses with a smaller focal length are also really expensive and not accurate enough for measurement systems. ...
    (sci.image.processing)
  • Re: Low cost image sensor
    ... >>>I have been searching the web for a low cost image sensor. ... >> read to 0.1mm resolution which actually means a minimum of 200 pixels ... resolution of the object and to get that from a 200 pixel device. ... Ideally I would like no moving parts, and no lens. ...
    (comp.arch.embedded)
Loading