[filmscanners] Re: Digi, film and scanning in movies

["David J. Littleboy" <davidjl@gol.com>]

"Brentley Beerline" <yeltnerb@pacbell.net>
>>>>>>>>

The Foveon chip has some technical shortcomings right
now which is a possible reason that it is not widely
adopted.

1.  Poor ISO performance right now it is 100-400 ISO.
<<<<<<<<<<<

It's more like ISO 100 _only_. Above that it's pretty bad. The noise levels
at ISO 100 are quite good, though.

>>>>>>>>>>>>
2.  Bad multiplier 1.7x.  Which means that your widest
lens is like a 30.
3.  Image aberations, artifacts etc. that increase
toward the edge of the frame.  Their software does a
decent job at correcting this, but it is easy to
demonstrate.
<<<<<<<<<<<<<

4. Serious color problems. The trick of different frequencies absorbed at
different layers has two problems: partial absorption of the signal for the
lower layers in the upper layers, and being stuck with the frequency
characteristics of silicon. Both of these make extracting RGB difficult and
the former means that the technology has worse noise characterisitcs than
Bayer in principle. (Apparently the chip also has a lower fill factor than
Bayer chips, making noise, in principle, worse.)

>>>>>>>
Foveon does have some things in their favor
1.  Decent funding
2.  Interesting technology (although the fact that
they have three layers of sensors will make true wide
angle very very hard).
<<<<<<<<<

Well, they seem to be using that funding for market hype, and "interesting"
is often a pejorative term, rarely rising above the level of damning with
faint praise.

--- Arthur Entlich <artistic-1@shaw.ca> wrote:
> One thing to remember is that the actual
> resolution of the chip in terms of the color, is
> considerably less than the the stated 4 MP.
> Every color digicam except the Sigma uses Bayer
> interpolation pattern, as I am sure you know.   That
> means the color is
> interpolated in literally every pixel, for two color
> elements based upon
> its closest neighbor.  There are 25% R and B and 50%
> G color separating
> CCD sensors in the chip.  So, although the
> luminosity is relatively
> accurate for the image, the color information is
> still a lot of guesswork.

Since your audience is human viewers, this is exactly the right thing: human
eyes are horrendously bad at resolving color detail and humongously good at
recognizing luminance detail. At any viewing distance such that luminance
detail is acceptable, color detail won't be a problem.

> I am a great supporter of the Foveon technology,
> although for some
> reason, and I believe it to be internal industrial
> politics, it has not
> caught on.

Technical inferiority and not solving a real problem are extremely good
reasons for it not to catch on.

>  Take a look at reviews and comparisons
> of the Foveon chip to
> standard equivalent CCDs using the Bayer
> interpolation grid.

The worst thing about the SD9 is its lack of an antialiasing filter. That
makes the "detail" that it renders largely bogus. The reviews rave at how it
resolves up to the Nyquist frequency, but they fail to notice the Moire
patterns between 70% and 100% of the Nyquist frequency and the false
response above the Nyquist frequency. Every sharp pixel reported by the SP9
has a positional inaccuracy of up to 1/2 the pixel pitch.

Of course, if they had included an antialiasing filter, it would have
resolved 70% of the Nyquist frequency like every other digital camera, and
would have been laughed out of the market.

David J. Littleboy
Tokyo, Japan


----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body