Projecting Digital 'Slide' Images

This is an enhanced version of an earlier posting to the Electronic Media Group, American Institute for Conservation of Historic and Artistic Works EMG-membership list. Thanks to Mike Collette and Robin Myers of BetterLight; Peter Kraus (Agfa, retired); Ron Miller of LaserGraphics; and Dave Dicklich, President, Projector Central for their information and assistance.

Overview

Now that the some of the best slide projectors are being discontinued (Kodak Ektagraphic III), it is time to get serious about the viability of using video projectors for the projection of film-positive, single still image Artworks, i.e., digitized slide(s).

While images can be readily captured at the resolution of film (maximum of 4000 ppi) and beyond, projecting at that resolution is not possible. In JAIC, Fall/Winter 2001, there is a discussion of this problem, see Vitale, "TechArchaeology: Works by James Coleman and Vito Acconci," JAIC, F/W 2001, Vol. 40, No. 3, pp.233-258. The highest resolution projector (QXGA format) from the 2001 period, the LaserGraphics LG2001 projector with 2000 lumens output, 300:1 contrast range at $13,000, is still a valid recommendation based on the limitations of digital projection. It has been recently learned that the LG2001 is now a custom order from LaserGraphics and may not be manufactured much longer at the present one-at-a-time basis.

More recently, JVC has produced a "best in class LCD" QXGA (2048x1536) video projector, the JVC DLP QX1G outputs 7000 lumens at 1000:1 contrast, for $225,000 <http://www.projectorcentral.com/JVC-DLA-QX1G.htm>. It is intended for use in movie theaters. Sony makes the next closest resolution projector; the just announced Qualia-004, has HD (1920 x 1080) resolution, with undoubted high lumen output and possible 6000:1 contrast (if the announcement can be believed) at $25,000 <http://www.projectorcentral.com/Sony-Qualia-004.htm> and <http://www.image-acquire.com/sony/sony_qualia_004/sony_qualia_004.html>.

Projecting slides in a slide projector (the equivalent of 32-72 MB digital information) still yields more cost-effective and higher quality images than today's projected digital images, with their 9.4 MB of information, at the highest possible, QXGA, projected resolution of 2048x1436 pixels.

When digital projection in movie theaters becomes more common, projector resolution should increase to a scale that is "similar to film". However, today it is still below the resolution of film, by about half an order of magnitude.

Discussion:

Resolution of Slide Film

The most direct method of determining resolution would be to measure individual grain in slide film. However, actual grain size is almost impossible to evaluate in color reversal film for numerous reasons such shape of dye cloud, randomness of image cloud distribution with an emulsion layer, thickness of a layer and that fact that there are three dye layers.

Two values are routinely used to predict resolution. RMS Granularity shows the degree of noise (unevenness) in a continuous tone, it is a measure of irregularity in a virtually grainless substrate. Spatial resolution predicts how much detail can be resolved in an ideally processed film.

The value of 80-line-pairs/mm is the commonly quoted value for slide film. Fuji Velvia can actually deliver 80-lp/mm at 30% contrast, see below, but this is the only slide film capable of this performance. Most slide film has an average resolution of 50-lp/mm.

The often quoted resolving value, 80 lp/mm, is determined using high-resolution densitometer at very small contrast differences (10%), on film that was contact printed without the use of a lens. A 10% contrast difference is well below what a human could observe even if the film was enlarged substantially making the line-pairs human readable. As if to confirm this, Fuji doesn't report MTF values below 20-25% contrast difference in its more recent Film Data Sheets.

If the contrast difference between line-pairs is evaluated at a 30%contrast difference, something an average human could observe, the spatial resolution of slide film is about 35-60 lp/mm, using the MTF curves in (Kodak, 1986) & (FujiFilm, 2000-2003); excluding Velvia. Using the more realistic resolution of 50 lp/mm, and exposing the film through a very high resolution 200 lp/mm lens (most lenses actually have a 30-50 lp/mm resolution, lowering the final system resolution even more) the resolution of the system will be about 40 lp/mm (system resolving power equation, EQ1, from FujiFilm Professional Data Guide, AF3-141E, 2002, p 129).

[EQ1: 1/r=1/r (film) + 1/r (lens);
where r=resolving power and r=resolving power on each component]

Using the broadest range of 35-80 lp/mm resolutions, the MTF values are translated to about 1800-3000 ppi, an average of 2900 ppi resolution, before the effect of the lens resolution is applied. The same MTF range, excluding Fuji Velvia, would have an average resolution of 47.5 lp/mm or 2400 ppi (before the effect of the lens resolution is applied).

Comparison to Printed Images

In the field of the printed still image, critical observers are beginning to think that the resolution of color reversal (slide) film is equivalent to 2700 ppi, or 53 lp/mm. This is excellent agreement with the MTF data above. Mike Collette's clients tell him that his scanning backs deliver the resolution of 8 x 10 transparencies.

The new Canon EOS 1Ds has a 2704 x 4064 pixel, full frame CMOS sensor. Many photographers are saying its digital image is the equivalent of high-resolution slide film.

RMS Granularity of Slide Film

Kodak slide films have a RMS Granularity of between 8-13 and Fuji reversal film have values between 7-10. Some negative films have RMS Granularity rating of 5, but the negatives will be printed increasing its final "system" Granularity, markedly. The lower the Granularity, the lower the noise and the lower the perceived grain.

Granularity is a root-mean-square of "density differences," from the mean (average density), in a film of continuous tone (1.0 D). It has no units, and unfortunately cannot be translated into grain size in any way. However, Mike Collette is experimenting with the value using it to predict the unevenness of 12-micron square pixels. A 12-micron pixel is the size in his lower resolution BetterLight scanning backs. Preliminary equations suggest that the unevenness for the smaller area is 3.5 times greater, than that for the 48-micron test aperture. This would make the effective granularity for of 12x12 micron bit of film, an average of 35. This is a very grainy/noisy bit of film, indeed. The difference between neighboring bits of 12-micron film squares would be 35. The best Fuji film has a value of 7, and the worst Kodak slide film has a value of 13, using the standard test aperture.

RMS Granularity measures "perceived grain" because it measures the unevenness of a virtually grainless emulsion composite. The complexity of the silver-to-dye transition, filamentation of dye cloud and the multiple emulsion layers mean that only "the odd" single dye clouds can be observed at the edges of transitions. In the RMS Granularity test, the measured emulsion is composed of numerous (hundreds) translucent grain-clouds making up the 3-layer-emulsion (actually 9 layers) that the standardized, 48-micron diameter, test aperture evaluates. There is no chance of resolving a single grain in that structure, only at the edges between transitions. Note that Peter Kraus projects that the average dye cloud probably has a size of about 25 microns, and that it started from a silver grain that was about 1 micron or less.

However, a type of graininess can be observed in slide film. That graininess is the unevenness of the complex composite structure, its "noise." The degree of unevenness varies based on chemistry, physical and manufacturer variables.

Tonal Range of Slide Film

Ideally exposed and processed slide film has a Dmax between 3.0D and 3.9 D; see "Kodak Films and Papers for Professionals" (1986); the Kodak Professional Products website http://www.kodak.com/global/en/professional/products/colorReversalIndex.jhtml?id=0.3.10.8&lc=en; and the Fuji Professional Product website <http://home.fujifilm.com/products/datasheet/>.

The following data was calculated from Density vs Exposure (lux-seconds) curves (Characteristic Curve) supplied by the manufactures. Some will argue that humans can't see detail much above a density of 3.4 D in a transparency. However, I believe that under strong light conditions (in slide projectors) the highest densities are valid, but this point is under contention.

Video Projector Contrast

If an artist makes slides copies of the Kodak Q60 R1 color target (print target), and is having problems differentiating last few black patches, the "system" is not delivering the full tonal range of the chosen film.

Many of us have been saying for years that slide film lies about color. The first lie is that film increases the contrast of the subject by a factor of about 1.5 to 1.9, an average of 1.7. I calculated a 1.8 contrast enhancement for Velvia 50/100. That is, for an input of 0.1-2.5 density, it outputs 0.2 - 3.7 density, almost its full tonal range. The center slope of the "density vs exposure" (Characteristic Curve) is 1.8, while the overall contrast factor is 1.34; the long shoulder and toe of the curve are not linear, which accounts for the disparity.

If the 24-patch gray scale in the Kodak Q 60 R1 reflective color target is shot using Fuji Velvia, an ideally shot and processed 2.4 D tonal range (2.5 Dmax) will produce a transparency with a 3.7 Dmax, assuming a 0.5 Dmin. On the gray scale, the next to last patch, GS22, patch #23 would be 3.45 D; GS21 (patch #22) would be 3.25 D; GS20 (patch #21) would be 2.7 D; and GS19 (patch #20) would be 2.5 D.

If the duplication process produces a Velvia slide yielding only 21 discernable gray patches (of 24), the film could be delivering a tonal range of about 2.6 D, or a contrast ratio of 400:1. The Dmax would probably be at 3.1 D with a 0.5 Dmin, a 2.6 tonal range. If the projector's Dmin can be adjusted using the Brightness control, from 0.0 D to 0.5 D, the projector's full contrast will be usable, and could reproduce the contrast in the slide (at lower resolution).

Translating density into "contrast range" will help correlate video projector characteristics to film equivalents. Basic contrast ratios are:

1.0 d=10:1 contrast ratio
2.0 d=100:1 contrast ratio
3.0 d=1000:1 contrast ratio
4.0 d=10,000:1 contrast ratio;

where contrast=antilog of D, density. Typical projectors have contrast ranges of:

100:1=2.0 D
300:1=2.47 D
500:1=2.7 D
3000:1=3.48 D.

The highest contrast I've seen advertised for a LCD-based device is 3000:1. The average contrast range for common video projectors is 300:1; the lowest commonly quoted range is 100:1. In reality, a projector with a contrast range of 500:1 is very good, and 1000:1 contrast is excellent. Some consider 1000:1 contrast an optimistic specification for today's LCD projectors.

In conclusion, video projectors cannot match (1) the resolution or (2) the tonal range, of common slide film. The resolution of even today's best resolving devices is only about 13-29-50% of the perceived resolution of slide film. Some will argue that a projector delivering a contrast of 500:1 is as good as can be expected, but it is still short of the tonal range that native Kodachrome 25, Ektachrome 100GX and Fuji Velvia 50/100 can deliver, by a factor of about 10, (an order of magnitude). If however, the original slides have been duplicated onto Kodak 5071, 6121 or EDUPE slide dup film, their tonal range will have a maximum of about 3.0 D (1000:1) if perfectly duplicated.

Less than ideally processed slide dups may well be within the contrast range of selected video projectors, but never their resolution. Less than ideally shot and processed Velvia or Ektachrome 100GX could have a tonal range that is within the contrast range of selected video projectors, but never their resolution

For the Sake of Completeness: Printing & Video-Out Issues

Printed images are generally sent to an inkjet printer between 150 ppi and 360 ppi resolution. It is said that children with perfect 20-20 vision can resolve no more than 300 ppi at normal viewing distance. [Most digital professionals accept the figure as a given, but I question the value as too low.] The 150-360 ppi images are printed at 720, 1440 or 2880 dpi on the typical photographic quality inkjet printer.

The difference between output file and printer resolution, is that the printer lays down inkjet dot clumps of color inks at 1440 dpi resolution, to yield "360 ppi-pitch" pixels composed of about 16 dots. [For a 150 ppi-pitch image, it would print 92 dots per pseudo pixel at 1440 dpi.] The printed "1440 dpi-elements from 360-ppi original" don't have the possibility of representing 16.7M colors each; rather they have just 1-8 ink dots for each color/black ink. The "360 ppi-pitch" pseudo-pixels are made up of the printer driver's best approximation of the target color using its available inks, at the image resolution.

Laser photographic printers using color photographic paper (Fuji Crystal Archive), such as the Durst Lambda (200-400 dpi), Cymbolic Sciences' LightJet (200 ppi) and ZBE Chromira (300dpi/425ppi), blur the laser point source with multiple color laser dots, of varying intensity, producing pixels with 24-bit color depth. This continuous tone image appears significant when viewed under magnification but can only be printed on photographic papers (with their limited surface textures). However, when humans view without the aid of magnification, inkjet printers produce equivalent or better results, at much lower cost and on a wider variety of papers, films, cardboard's and plastic sheets.

Video projector produces pixels that have a minimum of 16.7M possible colors. This is quite different from an inkjet-printed image, but similar to your CRT monitor, laptop display or laser-photo printer.