Archived posting to the Leica Users Group, 1997/11/28

[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]

Subject: Re: Digital storage???
From: Jim Brick <jim@brick.org>
Date: Fri, 28 Nov 1997 23:34:19 -0800

At 05:32 PM 11/28/97 +0200, you wrote:
>Excellent piece, Jim. Just one further question. You say
>
>>There is still one flaw. To produce a very high resolution digital  
>>image (400+ megabytes) takes time just to move that amount of  
>data.
>
>I don't know much about digital imaging but I've seen the following  
>estimate a few times: a 35mm frame of Velvia has about 64 megabytes  
>of useful information. "Useful" I think referring to how close  
>shades human eye can distinguish. Scanning resolution set somehow  
>according to the size of the grain. So is your figure referring to a  
>bigger format or is this just simply a hard thing to judge?
>
>I guess the bottom line is: Are there any standards/agreements  
>currently of what constitutes a digital equivalent of an image on  
>film?
>
>Kari

Kari,

*DISCLAIMER* I once new, but cannot remember the actual amount of data in a
Kodachrome slide in terms of pixels. The following is just an off the top
of my head discussion. No math was checked. But it should be fairly accurate.

The high end drum scanners can scan at 8000 DPI. The resolution at which
you scan is determined by what size and for what purpose the output is
destined.

If the output is to be printed via press (halftone) the scan dpi has to be
twice the screen frequency. If the output is going to a computer printer
(dye sub, ink jet, etc.) the scan dpi should equal the printer dpi. A 150
line screen will need a 300 dpi scan. A 600 dpi printer needs a 600 dpi
scan. The size of the output will then determine the size of the scanned
digital file. The formula is ScanSize=((height of image)x(2 times screen
freq) x (length of image) x (2 times screen freq)) x 3(for RGB). Or x 4(for
CMYK).

If your output is going to be film (quality hopefully equaling the
original) for optical enlarging on perhaps Ilfochrome, you want to capture
(scan) as many pixels as possible. But you should only scan at the dpi of
the film imagesetter.

To make things simple, let's assume that 35mm is 1"x1". A good drum scanner
can record at 8000 dpi. 8000x8000=64 megabytes. But there are three colors.
R, G, & B. So each pixel is recorded three times to produce a 3x64MB file
(192 megabytes.) Or times 4 for CMYK (256 megabytes). This is not all of
the information available on 35mm. It's all that an 8000 dpi scanner can
produce. Now take this out to 4x5 (my interest) 32,000 x 40,000 x 3 = 3.84
gigabytes and you still don't have *all* of the available information. You
only have what the scanner is capable of delivering. But for the printing
industry, you don't usually need an 8000 dpi scan. An 8000 dpi scan of a
35mm slide can produce a file capable of being printed at about 30"x36" on
a 400 dpi printer.

Everbest produces files somewhere between 400 & 500 MB, from a 4x5
transparency, to produce a 20x24 print. So the size of the final print or
page will determine how much data you need to capture. A photo CD image has
enough pixels to make a quality 8x10, if everything goes well. The
photographs in Fred Wards books were mostly produced from Photo CD images.

My original point was that there is more information available on film than
we can normally scan & record to a digital file. We are limited by scanner
hardware. However, in reality, we don't need *all* of the film's
information for most work. But if you wanted to archive that 4x5
transparency and could record *all* of the available information, we're
probably approaching a terabyte.

Film is still the best way to record images. Astronomical calculations,
weather predicting, and photographic image manipulation, still give the
best computers a giant headache.

For those who again want to know the nomenclature... here it is:


S P A C E - Bits or bytes
                Decimal Power of 10
Kilo	(K)	Thousand	3
Mega	(M)	Million	6
Giga	(G)	Billion	9
Tera	(T)	Trillion	12
Peta	(P)	Quadrillion	15
Exa	(E)	Quintillion	18
Zetta	(Z)	Sextillion	21
Yotta	(Y)	Septillion	24


T I M E -     Fraction of second
                       Decimal Power of 10
Millisecond	(ms)	Thousandth	-3
Microsecond	(aes)	Millionth	-6
Nanosecond	(ns)	Billionth	-9
Picosecond	(ps)	Trillionth	-12
Femtosecond	(fs)	Quadrillionth	-15
Attosecond	(as)	Quintillionth	-18
Zeposecond	(zs)	Sextillionth	-21
Yoctosecond	(ys)	Septillionth	-24


Storage/channel capacity measured in:

CPU word size	Bits
Bus size	Bits
Disk, tape	Bytes

MEMORY
overall capacity	Bytes
SIMM or SIP module	Bytes
Individual chip	Bits