This note is a follow up to the comments I made comparing Silverfast 5 AI to
Polacolor Insight on a SS4000. One of my comments was that Silverfast (IT8
calibrated) produced scans that are less saturated that Polacolor Insight.
I was contacted by a couple of folks off-list who said that for them getting
good color saturation is not a problem with Silverfast 5 AI.
I revisited my experiments to confirm my impressions. I was wrong. I retract
my comment that Silverfast 5 AI (IT8 calibrated) has limited saturation as
compared to Polacolor. I must have made a mistake while using these tools.
The Silverfast output is different than the Polacolor output, as I'll show
below, but lack of saturation is not one of the differences. I can't account
for this discrepancy.
Before I reveal the differences I've found, let me explain the process I
used to make my new comparison:
- I use a D65, gamma 2.2 display.
- The test image is a Kodak IT8 Q60-E3.
- Ensured appropriate general driver options (CMS settings, etc).
- Set the image type. For Silverfast I selected '<custom>', zeroed the
associated controls, and saved this as 'Motive_Raw'. For Polacolor I
selected 'Color Slide 2b.'
- Disabled (did not use) user-selectable forms of auto-exposure.
- Pre-scanned the test image.
- Set the crop area to be within the bounds of the film frame.
- Reset the levels, curves, color, and sharpening adjustments to their
positions of minimum effect.
- Used levels (Silverfast) or curves (Polacolor) with the histogram to set
the exposure. My goal was to capture as much range as possible with the
minimum influence on the software's alignment for color balance. I set
exposure by expanding the pre-scan data to fill the entire range of values
(0-255) in the histogram. I did this for the combination of all three
channels as a unit. In other words, I did not individually adjust the
channel exposures. For example, if the red highlight was the most extended,
I set all channels exposure equally so that the red highlight bumped into
level 255 (roughly), letting green and blue fall below where they may. I
used the same approach for shadow exposure.
[BTW--Wayne Fulton points out that Polacolor does have a histogram tool,
selectable under the Edit menu. I thought it didn't have this feature.]
- Finally, I scanned the IT8 at 1500 spi into Photoshop 6.01 and did not
color manage the inputs (no profiles).
I then compared data the resulting images.
Here are copies:
What I notice:
The Polacolor scan is neutral in the grey stepwedge, tone slightly dark and
color rendition a good match to the IT8 as I see it on a florescent D50
lightbox. Overall cast is a slight yellow/green compared to the film on the
The Silverfast scan looks darker. In the grey stepwedge, midtone values are
down about 15-20 units from the Polacolor scan. There is a slight yellow
cast to the midtones--more so than the Polacolor scan--a cyan cast to the
highlights, and a major red cast in the shadows.
To compare saturation, it's important for the images to have similar tone.
I normalized the midtones by using Photoshop to lift the Silverfast RGB
curve up at the midpoint to match what I feel to be the more accurate levels
of the Polacolor scan. The Silverfast scan shows more red saturation of the
face, especially the eyes and the shadow at the neck. Saturation of the
primary color stepwedges is slightly greater in the Silverfast scan.
My opinion is that Polacolor rendering is closer to the film as viewed on
the lightbox. But these results clearly contradict my earlier comments about