[filmscanners] RE: Density vs Dynamic range - was: RE: opinions? Reviews? of Primefilm 1800 ?

["Tony Terlecki" <ajt@mrps.demon.co.uk>]

On Sat, Jun 08, 2002 at 09:43:00AM +0100, dickbo wrote:
> Bits equals available grey levels per pixel
>
> ----- Original Message -----
> From: "Laurie Solomon" <laurie@advancenet.net>
> To: <dickbo@btopenworld.com>
> Sent: Friday, June 07, 2002 4:22 PM
> Subject: [filmscanners] RE: opinions? Reviews? of Primefilm 1800 ?
>
>
> >Another misconception...though equally as common...the number of bits the
> converter
> >has, have nothing to do with the scanners ability to capture any particular
> >density range.
>
> Just out of curiousity and in simple layman's terms, what do the number of
> bits that the converter has  have to do with if not the density range? How
> does it impact on what is captured?
>
> Austin, I am asking a serious question here out of my lack of knowledge and
> sure would appreciate a good discussion in layman's terms so that I can
> understand what is being said without having to hire an engineer to
> interpret. It has always been my understanding, rightly or wrongly, that the
> higher the number of bits the more detailed or refined the informational
> date captured from the original that is transmitted as data in the digital
> file with respect to highlight and/or shadow detail with the density range
> figure representing the range of contrast that can be captured.  In other
> words, "dynamic range" representing the contrast range of the capture's
> capabilities, while the bit depth represented the quality of the data
> captured within that range particularly the extremes.  If this is wrong,
> please explain where and how it is and provide me with a more accurrate
> description (but once again, I urge you to try and do it in non-engineering
> terms if possible).
>

The number of scanner bits is a necessary but not sufficient requirement for
seeing densities at a scanner's theoretical dmax (i.e. log 2^bit-depth).

Number of bits have two main roles. They do indeed represent the theoretical
maximum density that a scanner could have if the electrical components were
up to scratch and could use those bits to their full. More importantly
though the number of bits determine with what resolution the scanner can see
the density levels that it is capable of scanning. Reread that last
sentence again!

Lets take a sample scanner which can only record film density up to a 2.4
dmax. This should be within the realms of even the cheapest modern film
scanner hardware. This density limit is imposed by (let's say) the cheap
design of the scanner and the manufacturer's requirements to be cost
effective. Now lets take this scanner and give it an 8-bit CCD to start
with. In a minute we'll give it a 14-bit CCD and see the difference.

The number of bits represents the POTENTIAL density range that the scanner
is capable of. Scanner CCDs are linear devices - it is important you
understand this concept because it determines how well scanners can
potentially see into the shadows. With a linear device each doubling or
halving of a scanned linear data value represents a doubling or halving of
light, or 1 stop, or 0.3 density units.

Let us take our 8-bit scanner then - 8-bits gives us a total of 256 values
that the scanner can use to scan the image. Let us look at the numbers to
see how these data values can be spread out across a density range. I'm
going to assume that the operator has placed the scanner's white point
(value 255) exactly at the lightest portion of the film/paper.

128-255: 1 stop (0.3 density units)
64-127: 1 stop (0.3 density units)
32-63: 1 stop (0.3 density units)
16-31: 1 stop (0.3 density units)
8-15: 1 stop (0.3 density units)
4-7: 1 stop (0.3 density units)
2-3: 1 stop (0.3 density units)
0-1: 1 stop (0.3 density units)

Note that I cannot go any further than this because I need at least two
distinct values to represent a change in density.

The 8-bit scanner then can scan up to 8 potential stops. It is impossible to
see anything denser because it would need more bits. Even now you have so
few bits at the shadow end that you would almost certainly have
posturization.

Let us move now to the same scanner with 14-bits to play with. How do those
8-stops map to the scanner data values? Well 14-bits gives us 16,384 values:

8192-16383: 1 stop (0.3 density units)
4096-8191: 1 stop (0.3 density units)
2048-4095: 1 stop (0.3 density units)
1024-2047: 1 stop (0.3 density units)
512-1023: 1 stop (0.3 density units)
256-511: 1 stop (0.3 density units)
128-255: 1 stop (0.3 density units)
64-127: 1 stop (0.3 density units)

Heck we've reached 8-stops here and there's loads of bits to spare! Better
still that last stop which, on the 8-bit scanner, could only be represented
by either a 0 or 1 can now be represented by 64 distinct values. Undoubtedly
you could get better tonal resolution in that last stop of shadow area with
14-bits.

What about the values from 0 to 63 I hear you ask? Well as agreed at the
start our scanner is only capable of recording up to 2.4 dmax from the film.
This is a hardware limitation. What if they came along and created a better
scanner which did better at seeing into the shadows? As a 14-bit device, it
could use those extra bits to record information for densities greater than
2.4. For the 8-bit device there is nothing that can be done because you
cannot halve the number of bits any further - you cannot have fractions of a
bit.

One other issue here is that I have been dealing with linear data. The norm
is then to apply a power transform to this data (usually referred to as a
gamma curve) so that the data values are better distributed according to
human perception. We do after all interpret light logarithmically rather
than in a linear fashion but we are getting beyond ourselves here and should
perhaps save that discussion for another time!

It's difficult describing this purely in layman's terms. If you don't
understand anything here then please say so and I'm sure we can
elaborate/simplify further.


>
> -----Original Message-----
> From: filmscanners_owner@halftone.co.uk
> [mailto:filmscanners_owner@halftone.co.uk]On Behalf Of Austin Franklin
> Sent: Friday, June 07, 2002 7:13 AM
> To: laurie@advancenet.net
> Subject: [filmscanners] RE: opinions? Reviews? of Primefilm 1800 ?
>
>
>
> > However, it is rated with 3.2 dynamic range, which is a bit low for a 14
> > bit/channel.
>
> Even though they may call it "dynamic range", it is DENSITY range.  Another
> misconception...though equally as common...the number of bits the converter
> has, have nothing to do with the scanners ability to capture any particular
> density range.
>
> 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
>
>
>

--
Tony Terlecki
ajt@mrps.demon.co.uk
Running Debian/GNU 2.2 Linux

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