[filmscanners] Re: scanner dmax discussion

["Karl Schulmeisters" <karlsch@earthlink.net>]

Austin
You are right about the difference between dynamic range and dMax.

But on A/D converters, the simplest way of doing it is very much along the
lines of what you are describing - 1bit/step, but there is no inherent
requirement to do it that way.  As for the linearity of sensors, from the
plots I've seen (and obviously Nikon, Canon et al aren't publishing their
plots for competitive reasons), the devices are linear over a particular
range, and while that is where they try to map the sensor into, at the ends
you get significant non-linearity - noise on the low end and
cascading/swamping/blooming on the high end.

As for driving the scanners into different parts of the curve, I haven't
seen any, and I've punted the question back to the A/D expert that claimed
this to me.  Will have answer soon.


Your point about the minimum step increment being the noise function is
correct.  The hard part is that I don't think the noise function is linear
over the full dynamic range of the sensor so theoretically the step
increments aren't linear either.

Karl

----- Original Message -----
From: "Austin Franklin" <austin@darkroom.com>
To: <karlsch@earthlink.net>
Sent: Wednesday, July 09, 2003 6:56 AM
Subject: [filmscanners] RE: scanner dmax discussion


Karl,

> So in part, the 'wizardry' of the electronics inside the scanner A/D
> converters is to try and compensate for the non-linearities of
> the detectors
> themselves.

That has nothing to do with the A/D converter, it simply divides it's input
voltage range by N steps, and records 1 bit per step.  The CCD sensors used
in scanners are relatively linear, but there is a calibration process that
does give a "table" of values, one per sensor element, that is applied to
the DIGITAL data after the A/D process.

> Note also, most of the detectors have photo response curves not that
> dissimilar from film, so by changing their detection threshold, you can
> drive them into different parts of the response curve.  Not all scanners
> allow you to do this, but in theory it is possible.

Actually, I haven't heard of any that do it that way.  All you have to do is
change the exposure time to do this.  I do believe that one claims to have
"analog gain", but that is not really analog gain as far as any electronics
between the CCD and the A/D are concerned, I believe it changes either
intensity of the illumination, or simply exposure time...I don't remember
which.

> There are scanners
> that allow
> you to scale (or autoscale) the image range to maximize the data
> extracted -
> but they are pretty pricey.

What scanners do you believe do that?  Typically all they do is apply
setpoints and expand the data so it occupies the full range of 8 or 16 bits,
but this is simply a conversion process, not an analog offset/gain function.
That COULD be done as an analog function, I've done it, but not on
commercial scanners, and there really is no need to do it anyway, as all you
are doing is expanding noise, so you really get no additional usable
information by doing it in the analog domain.

Austin

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

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