[filmscanners] RE: Newish Digital Tech

["Austin Franklin" <austin@darkroom.com>]

Hi Karl,

> Yes you can distinguish without a filter.

I believe you misunderstood.  You can not distinguish without some kind of
"filter"ing happening, or you would not be being selective about what it is
you are sensing.  "Distinguishing" IS filtering.

> Clear silicon transmits different frequencies of light to a different
> extent.  Red has the least transmissivity, Blue the greatest.
> This is very
> similar to water - which is why everything below about 50' starts to look
> green and below that it rapidly shifts to blue.  So imagine if you put a
> full spectrum photocell at 15', 50' and 100',  you'd then be able to
> calculate the blue component from the 100' sensor multiplied by Blue
> Frequency attenuation.  Green would be calculated by subtracting the Blue
> Freq component from the50' sensor (again normalized for both blue
> and green
> attentuation), and Red would be calculated by subtracting Blue and Green
> from Red.

I have to think about that, and thanks, it sounds interesting.  But, my
question is, how much light actually gets through?  Certainly not %100...so
as the depth increases, the light decreases, and the upper sensing "area"
will have much more "light" to "work with", right, no matter what the color
is you are sensing?

> But the gotcha is that the sensor in the Foveon case is not at a single
> depth, its a 'region' of doped silicon.  And so you are going to get an
> averaged value, and you are also going to get some red propogating all the
> way to the Blue and Green sensors quantum mechanically - ie
> probablistically.

Again, I have to think about this.

But, what is the EXACT photon to electricity conversion?  Does the mechanism
that does the conversion block light, therefore decreasing the amount of the
other frequencies light to the lower sensing layers?  What is the light vs
area efficiency...in other words, if you have an area that has "M" photons
hitting it overall, how many of the photons "P" are converted to electrical
energy over that area, such that the efficiency = P/M?

> So you can probablistically compensate for it, but its going to take time
> for them to tune the compensations.

How predictable is it, and how consistent is it from cell to cell, and from
sensor to sensor?  That could work out to be a calibration nightmare, if
even possible at all.

Where are you getting your information from?  I have not read the patents,
but I probably should.  You've given some great "information" (don't know
how much of it is speculation or actual information, but it's quality is
high none the less, and your explanations have more probability/soundness
than others I've seen ;-), thanks.

Regards,

Austin

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