ACE input table bug in PS6 suspected
ACE input table bug in PS6 suspected
- Subject: ACE input table bug in PS6 suspected
- From: email@hidden
- Date: Tue, 11 Sep 2001 12:02:17 +0100
Hi list,
I just stumbled over what seems to be a bug in Photoshop 6.
Can anyone confirm the following, or is it even a known problem?
A2B table lookup with Adobe ACE gives wrong results
when 8-bit profiles with non-linear input tables are used.
How to see it:
1. Take an 8-bit CMYK profile with linear A2B input tables.
In the mft1 tag of A2B1 you'll find the 256 bytes 00 01 02 03 04 05 ... for
the first (Cyan) channel input table. Change the 01 to 00, so that
you get a minimally non-linear table just for Cyan, save the profile.
2. Set the rendering intent to absolute colorimetric and use the
color dialog to enter CMYK values. Leave MYK at zero. Change
C in 1% steps from 0 to 10 and observe the Lab values. Around 4%
it appears as if you entered a value near 100%.
3. Verify that for the unaltered profile the results are correct.
This also happens in a mode change to Lab.
Create a CMYK document with a gradient in the cyan channel.
Do a mode change to Lab. You'll see a almost solid cyan bar
appearing in the light area.
If your profile has non-linear input tables in every channel, this
will happen in every channel, and always in the tones around 3-6%.
I see this happen in PS 6.01 with the Adobe ACE engine on Mac and PC.
It does not happen with PS 5 or 5.5.
It also does not happen with other CMMs.
It seems to me that ACE creates a new interpolation table with a
reduced number of grid points (probably to treat 8-bit input tables
similar to 16-bit tables). Somehow the second table entry wrongly
gets an output value of 100% regardless of what the input table has.
Regards,
Johannes
-----
Ebner Ulm / Abt. Druckvorstufe Tel.: +49 731 2056-153, Fax -151
Dr. Johannes Hoffstadt email: email@hidden
Eberhard-Finckh-Strasse 61
D-89075 Ulm