[Tig] Oxberry scanners
jasoncrump at gmail.com
Tue Nov 28 21:23:00 GMT 2006
I am a freelance DI Colorist working for a post house in Mumbai. I have no
affiliations with any manufacturer of any sorts. I am not an engineer and
these are my opinions and are always subject to being wrong, but hopefully
have constructive value. :) The kit i am grading on is a Lustre, Truelight,
and JVC 10K.
In an effort to analyze the scanning options in this city, we sent the exact
same shot/same negative (~650 frames) to every facility here. This included
three Spirit 4K's, one Sprit 2K, one Imagica Imager XE, one Oxberry Cinescan
(w/ wet-gate), one C-Reality, and one Cintel diTTo (demo machine). Final
delivered scan format for all scans were 2k 10 bit dpx. Of course, each
machine had a diff. operators, different calib. setups, and each Spirit had
different noise/grain reduction settings, so the results were widely varied,
even among the 4K's. This only proves how much the performance of a given
machine greatly depends on the competence of its operator. However this does
not rule out any analysis that can sussed out.
re: calibration/setup and density/code value relationship:
By looking at the scans, I think i can assume that most of the scanners were
"setup" by sampling base and then incrementally assigning code values based
on the densities of the layers as per cineon. This effectively meant that
all of the code values were between 95 and 685 with some specular highlights
jumping above 685. This was true for all expect the Oxberry. Oxberry scans
spread the same densities among a wider range of code values. So base was
around 50 and the specular highlights were in the 900's and the mid-tones
were much more spread out in between.
I can only assume that the scans whose visible black of base showed code
values roughly around 95, were as per cineon. However, since we did not scan
density steps from d-min to d-max on an OCN, I don't know whether the
density values from d-min to d-max were properly assigned their
corresponding code values as per cineon.
segue question: should scanners have different setups (scan LUT's) for each
OCN stock, so as to capture the different spectral responses from d-min to
d-max that each stock has?
Some raw scans were warmer, cooler, or more green. While grading, i found
that most color/code value discrepancies among the different scans could be
matched by independently shifting each RGB channels brightness while in log
mode on the Lustre. This allowed me to match the color of the scans without
disrupting the channels integrity. These differences were probably due to
the fact that each machine did not sample the exact same base on the neg for
Although the oxberry scan looked brighter than all the other scans, I found
that because the code values were more spread out, I had much more color to
play with and much more flexibility to push and pull these colors. I don't
know the exact sensor specifications of these machines, but it seemed that
from photon to code value, I liked the oxberry the best. Is it as per
cineon?... I don't think so. Is that a bad thing, if the results offer more
flexibility and dynamic range?... nope. Is this the result of a custom setup
by the facility?... probably. Can it be achieved on the other scanners?... i
My ranking in terms of color flexibility, dynamic range, and which ones I
could push and pull, and which were the easiest to get a pretty picture were
oxberry, imagica/diTTo close tie, Spirt 4k, Spirit 2K, c-reality.
The telecine scanners seemed to compress the information more than the
others. The blacks were more lifted and the whites pulled down. It was like
it was under-utilizing the available color resolution of 10-bit log. I had
to stretch the channels more to get to a pretty picture and this effected
the over-all quality of the picture (more noise). It was almost as if i was
killing the image because I was asking too many pixels to be so far away
from their original scanned value. make sense?
re: perceived sharpness
Oxberry, Imagica, and diTTo were all very close to each other. When turning
on any of the dust/scratch/grain tools on the diTTo, the sharpness of the
scan is effected. Sharpness among the Spirit scans varied as it seemed some
operators were more aggressive with the dust/scratch/grain tools. I also
found that the scans that had aggressive dust/scratch/grain concealment also
showed a drop in the midtones code values...similarly with the diTTo. (i can
safely say that dirt is more of an issue here than other places around the
world) I would say that the oxberry topped them all in perceived sharpness.
I think it has alot to do with the wet-gate. The scan was absolutely clean.
i think the proper technical way to say it is...it just popped more. ;) I
can remember another technical observation I had a few years ago when
putting negative up on a steadi-gate and then loading the same neg on a
Peterson wet-gate...wow! I know Imagica now offers a wet-gate option on
As I do not know the intricacies of a Spirit 4K scanner, I have some doubt
as to whether all scans are scanned at 4K and then down-sampled to 2K by
default or is this an option that one has to select while operating?
The difference between pin-reg and capstan were made clear in this
oxberry was the clear winner here. diTTo did a very nice job with minimal
effect on the sharpness and some effect on the mid-tone code values. Nice
results without any extra post-scan procedures. I would like to see these
compared to other concealment methods....IR, digital ice. The spirits also
made clean scans. From reading the manufacturers literature, i believe that
cintel and thomson use similar optical technology to conceal dirt and
scratches...a diffuse light source. i believe that on the diTTo, one can
vary the amount of diffusion. The imagica scans were filthy, even though it
was the second machine to get the neg. I think this model had no concealment
tools. Nice color, nice sharpness, but dirty.
obviously the 4K tops here. However, in my opinion, if you are working
reel-wise and scanning concurrently while grading, the scanner needs to have
the next reel scanned and ready for grading before you finish the one you
are working on. So ...say you are spending 2-3 days grading per reel, then
your scanner needs to be able to scan 20 minutes of material in 2-3 days.
With 2 shifts a day, an Oxberry can do this. Is it a hard sell to add 2-3
days at the beginning of a DI schedule to get pin-reg scans?... maybe not.
Otherwise, if you are grading one reel per day or speed is an issue, then
scan on a faster machine.
Regretfully, there are a few major players with the "latest" technology in
scanning (northlight/arriscan) that do not have a machine in mumbai when we
ran this test.
The oxberry here in Mumbai is sucessfully being used for production. I have
done work on oxberry scans, DSX scans, and spirit scans. I liked the oxberry
the best. Each scanner has its own advantages and disadvantages and there
are many different reasons to chose one scanner over another as either a
client or as an owner... ultimately its ones priorities that should guide
Andreas Wideroe Andersen <andreas-ml at wideroe.net> wrote:
> TIG is supported by subscribers.
> see http://tig.colorist.org for
> donation button interface. Bloggging now available
> at http://www.colorist.org/wiki3
> I was wondering if anyone use the Oxberry
> scanners for scanning new productions in 16mm and
> 35mm to high resolution (3-4k) digital files or
> are these machines mainly used for archival footage?
> Could they be used for new productions? Why/why not?
> I'm also interested in hearing user experiences with these machines.
> Norsk Smalfilm AS
> Andreas Widerøe Andersen <andreas at smalfilm.no>
> Filmshooting | Com - http://www.filmshooting.com
> Tel: (+47) 38 17 99 16
> Fax: (+47) 38 02 33 84
> No personal abuse; absolutely no advertising or marketing on the main TIG
> mailinglist. Help to maintain a high signal-to-noise ratio on the TIG.
> Tig mailing list
> Tig at tig.colorist.org
More information about the Tig