[argyllcms] Re: Calculating film LUTs for a film recorder

  • From: Doug Private <doug17r@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Tue, 28 Aug 2012 15:28:26 +1000

On Tue, Aug 28, 2012 at 5:04 AM, Philip Pemberton <philpem@xxxxxxxxxxxxx>wrote:

> On 21/08/12 02:16, Graeme Gill wrote:
> >> I've actually got a Spyder2Print calibrator, though as I understand it
> >> that isn't supported by Argyll :(
> >
> > Sorry, it's a lot of work to RE something like that, since it's
> > quite different to other spectrometers.
> Yeah - I've been trying to reverse-engineer it off-and-on for a while.
> I'm not really getting anywhere, unfortunately.
> >> It's a linearisation LUT. From what I can tell, it's intended to counter
> >> the effects of the film's characteristic curve (H-D, or
> >> "Hurter-Driffield" curve). Although as you say, matching to sRGB may be
> >> easier and more productive.
> >
> > printcal wouldn't be too hard to modify to target a particular curve
> shape.
> Well based on what I've reverse engineered from the film profiles, I
> need to set the luminance of the CRT (this is different for each film)
> then build a LUT to linearise the response. I suspect the luminance is
> set to allow the film to be exposed in a reasonable time while also
> providing a usable dynamic range and avoiding issues with reciprocity
> failure in the film.

Another issue to consider is flare which seems to be a significant problem
in these CRT film recorders.  Perhaps using the minimum necessary exposure
to achieve a 'white' level will also help minimize the flare.

The required luminance may also be related to the scanning resolution and
the scanning speed if this can be varied.  Perhaps the film recorder needs
to slow down the scanning speed at higher resolutions in order to keep the
pipes full.  Seem to recall being able to program the target scanning speed
on some of these film recorders - these were developed in the days of 68k
Macs etc and the software limited the recording speed.

Positioning a digital camera to recorder the CRT can be useful for
experimenting with the software, and perhaps you could attempt this.

After looking into patching printcal it seems a good solution is to just
pre-scale the device values passed to printcal in the it3 file so that the
RGB white device levels range to 1.0 for white, and then scale the printcal
result when writing the LUT.

The calibration procedure would then be: 1. choose a starting LUT that will
adequately expose the film.  2. Generate calibration patches.  3. Scale the
ti3 device values with the LUT values and also scale them to a maximum of
1.0.  3. Generate the calibration table using printcal.  4. Scale the
calibration table to back to a LUT.

I'll hope to give this a try within a few weeks and will report back.

Other related posts: