IT8 calibration problem in Ai 6.0.2r31

[obese]

It seems the IT8 calibration is changed significantly in version 6.0.2r31. While the procedure is simplified, I am having problem with colors, especially in shadow area. Before calibration, patch GS0 reads about 250, GS11 about 110, and GS23 about 22, as suggested in Ian's tutorial. After the calibration, GS23 becomes 0, and there are a lot of noise in patch GS20 and above, which suggests early clipping. This behavior seems to be different from earlier versions (I was using 6.0.2r01 and it is good). As a result, the scans lose most of the shadow details. It's not a hardware problem because when I turn IT8 off, I get all the shadow details back. Is this a bug in 6.0.2r31?

I am using SS4000+ (firwire)/PS7.01/WinXP. The IT8 target is fairly new, not the old C-Roes. I used the same target to calibrate 6.0.2r01 and had good results. I followed the calibration procedure in Ian's tutorial.

[ilyons]

Check that the Limit Gamma Slope feature in the Options>Special dialog has not ben activated. What you describe is consistent with it being active.

[obese]

I just checked and the Limit Gamma Slope option is deactivated. If I activate it, then both patch GS20, GS21 and GS22 become black. The following two images are scans of an IT8 target, the first one with calibration on, the second one without calibration. The Limit Gamma Slope option is off for both.


BTW, Ian, thanks for the nice tutorial.

[cew3jsm]

Hi,
I have experienced exactly the same problem. With calibration, the scans are severly clipped in shadows and the shadows show a green caste. (The limit gamma slope is not checked) Switching off IT8 calibration, the scanner works fine, shadows are not clipped and have plenty of detail, but I have to slightly tweek skin tones (they are a bit too magenta/red). Seems like a bug to me.

Clay

[cew3jsm]

Hi,
I have experienced exactly the same problem. With calibration, the scans are severly clipped in shadows and the shadows show a green caste. (The limit gamma slope is not checked) Switching off IT8 calibration, the scanner works fine, shadows are not clipped and have plenty of detail, but I have to slightly tweek skin tones (they are a bit too magenta/red). Seems like a bug to me.

Clay

[harry shin]

1. This is somewhat off topic and thus I'll probably post this as a new topic elsewhere... However, I recently downloaded the current version of silverfast 6ai. What I've noticed is that when I use negafix (I just shoot negs), there is a definite mild to moderate reddish colorcast. I've reviewed how to use negafix, I've been using negafix for quite some time, AND I've been getting this reddish color casts with the same images (NOW) that I didn't have problems with before. Thus, I also think there is some type of bug in the current version--I hope someone out there at LSI will work on this.


Harry

[LSI_Belli]

Dear All,

we will check straight away, if anything is wrong.

To help us localizing an eventual problem, could you please send us all necessary information (target type, images, etc.) at support@silverfast.de.

Thanks,

Thomas

[obese]

Thanks for the quick response. Target info and images has been sent to support@silverfast.de.

[ilyons]

Thomas,

I think the above issues are only the tip of the iceberg. Please contact Martin M and KHZ they both details on my findings. You'll also need to look at HDR r32

[harry shin]

1. Just a follow up (tangent) to my problem; it may be somewhat relavent to this issue (ie bug with current version of silverfast). I was having a reddish color casts to my scans (negatives). After reading these posts, it finally dawned on me that it could be a bug in the software--> I got rid of the current version of silverfast, I then installed the prior version (2r20)--> scanned the SAME negative--> NO color cast problems. Thus, from another point of view / issue, it clearly appears that there is a bug with the current version of silverfast in regards to color balance / negafix etc...
2. With the bug related to SDR and now this, I haven't been able to scan for... 1.5-2 months now. I hope that the next update will be bug free.


Harry

[cew3jsm]

Hi,

Is there an ETA on a revised fixed version of SF6.0.2 for the Polaroid Sprintscan 120? Just curious and hopeful.
Thanks,
Clay

[obese]

Just tried Ai r37 today and I am still having problem with IT8 calibration. I removed the entire Silverfast directory before installing r37. And I followed the exact procedure (the newly updated version) described on Ian's website. The target I use is brand new just ordered from Wolf *****. The calibration was reported success. However, as soon as I turned calibration on, the clipping problem came back. Starting from GS17, all the grey patches have strong green cast. The green channel seems to have about the right values, but the red channel and blue channel are severely clipped. Basically the profile is wrong. By accident, I switch to a profile generated by 6.0.2r01 and the problem is gone. It looks like a bug in IT8 calibration routine to me. I don't know if the problem is Windows specific. Ian doesn't seem to have problem on Mac.

[ilyons]

Just tried Ai r37 today ........... I don't know if the problem is Windows specific. Ian doesn't seem to have problem on Mac.

"Ian" uses the Mac version and it was only updated today - 25 July (been lying in the rough of r31 since May). I just downloaded it and as you have noted the shadows are GREEN!



I've also scanned my targets as HDR color images and then opened them in HDR with IT8 module. I built profiles for each target (one scanned on Polaroid and the other on the Nikon). The Polaroid based profile which I created in HDR is GREEN and the Nikon based profile whcih I created in HDR is good!

Sadly as things stand R39 for the Polaroid is unusable with or without LaserSoft IT8 calibration. If I go back to r23 and the old version of IT8 all is fine.


Thank goodness I also have Nikon LS4000ED to work with or I'd be in deep smelly stuff. I'll also continue building my profiles using ProfilemakerPro until this issue is fixed.

[obese]

As a test, I did IT8 calibration first with r01 (which seems to be the last version that does IT8 correctly for my SS4000+ under Windows), then with r37. I then used Iphoto ICC Inspect to check the 2 profiles generated. Compare with the r01 profile, the r37 profile covers slightly more blue and red while significantly less green. That kind of explains the green cast and clipping problem.

I also did a raw scan of the target with all CMS settings off. The grey patchs are all quite neutral, maybe with just a little bit magenta cast. But no way this little magenta cast should cause SF to oversaturate green by that much. I don't have HDR with IT8 so I can't verify Ian's observation. But my HDR scan (no profile embedded) looks OK.

My current solution is to build the profiles with r01 and use them in r37, which seems to work fine for now. Hope this problem can be fixed soon.

[ilyons]

I also did a raw scan of the target with all CMS settings off. The grey patchs are all quite neutral, maybe with just a little bit magenta cast. But no way this little magenta cast should cause SF to oversaturate green by that much. I don't have HDR with IT8 so I can't verify Ian's observation. But my HDR scan (no profile embedded) looks OK.

Even the slightest magenta shift in the neutral patches of the target seem to cause the SilverFast IT8 module to mess up on the shadows and turn them green.

The EXACT same targets scanned on my Nikon and Epson do NOT show ANY magenta shift in patches G16 through GS22 and the final profiles are PERFECT. ONLY the images scanned from the Polaroid result in green shadows after the profile is built, but ONLY if the profile is built using the LaserSoft IT8 module.

I have used ProfileMaker 4.1.5 to build profiles from the magenta shifted polaroid scans and they are PEREFCT. The Polaroid version of Silverfast is doing something WEIRD to the neutral patches and the LaserSoft IT8 modules tries to compensate.

BTW: I ahve looked at the Polaroid profiles using anumber of apps designed to check their quality, gamut and a data - they are broken!