[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


Dear Paul
        As a manufacturer I generally stay quiet on the Tig as it is meant
to be impartial and I, obviously have a vested interest. But, when another
film scanning product manufacturer comments on one of my film scanning
products I feel I have to write.
        Negative aperture correction has been used by many URSA users for
years now to deal with something known in the television industry as
interlace flicker. This generally happens if you have information on one
line that is significantly different to the information on the adjacent
lines above and below. This is not telecine generated alias of the sort
SCAN'dAL was designed to fix, it is merely an artefact of the display only
showing the information once every 1/25th of a second (in our country).
This can be proved by viewing an image on a non-interlaced monitor and an
interlaced monitor simultaneously (I believe you may have access this sort
facility in the post house you also work for). If you really want to see
interlace flicker, try displaying something like the Flame desk top an
interlaced monitor. The Cintel aperture corrector (very good), does contain
a vertical filter, and the user has control of some of the coefficients and
these can be used to vertically soften the picture which will deal with
interlace flicker, but not vertical aliases.
        Once a signal contains true aliases there is nothing you can do
with any amount of post filtering, no matter how clever that filter is, to
remove it. Aliases in sampled systems often happen due to incorrect
sampling procedure. One mistake that can be made during the sampling
procedure in to not take all of the input signal into account. This is what
a standard URSA does. The scanning beam is focused finely enough to resolve
the film information to 2K no problem. Trouble is we are not doing 2K. In
the vertical direction we're doing 487 or 576 lines and basically the
vertical sample size on a well set-up URSA is incorrect and fails to 'scan
all' of the film. If vital information remains unscanned, or unsampled, it
will not appear in the video signal and cannot be guessed at by some sort
of post filter. SCAN'dAL corrects the vertical sample size automatically
for every instance of XY zoom, rotate, or for that matter any other wacky
scanning patch you can create with your scan effects, dynamically on a
sample by sample basis.
        This involves some interesting calculations, based on the distance
between the scan lines and the direction of vertical scan, in real time.
The results of these calculations are applied to three high speed
amplifiers, two of these drive a brand new astig' coil (fitted quite easily
behind the existing one, which we unplug), the other amplifier drives the
existing focus modulation coil but with enough 'go' to make it work
properly in all instances. We even take into account the lens magnification
when making the calculations, it makes a difference! This I think you will
agree is a little more involved than 'tweaking the astig' on the package.
You do not have to remove the tube to fit the new components. The upgrade
is a far less daunting task than our TWiGi upgrade, you need a screw
driver, but you don't need a soldering iron.
        In many instances the improvements to spot focus we have made,
improves picture sharpness. In all cases the pictures we now produce have a
much more naturally detailed look.
        As regards to controllability, we have given the colourist one
control, vertical sample size. If you do nothing with this control, and
leave it at the default setting the pictures look lovely, like film in
fact. If you turn it down to zero your back to normal horribly aliased
pictures. Oddly enough, if you turn it up you can deal with interlace
flicker. This makes the scan lines overlap slightly, but it still looks
better than negative aperture correction, because if film has the sort of
fine detail that produces interlace flicker it is bound to have the sort of
detail that would alias if sampled incorrectly.
        I could have explained all of this in the frequency domain but
unless you really know your sampling theory it's less obvious. However,
another definition of an alias is a low frequency signal that was not
present in the original signal information. Because it is a low frequency
component and therefore 'in-band' it cannot be distinguished from the
wanted 'in-band' information, not by electronic filters at least. We can
tell the difference with our eyes, but that's because we know that dreadful
pattern was not on that brick wall in real life!
        SCAN'dAL also produces still further improvements in noise, both
measurable and subjective. Firstly, because the scanning beam is now oval,
the electrons are falling on a larger area of CRT phosphor, thus the
phosphor is less close to saturation and more efficient, producing more
light. This can only enhance tube life still further! But the subjective
improvement in noise is really due to sampling film grain correctly.
        Anyway, perhaps I rambled enough now, so before I leave and for the
benefit of information, here's a quick summary of SCAN'dAL:-
1. Product consists of replacement scan output card, high speed amplifiers,
and a 38mm octopole stigmator coil! Keep your old output card as a spare.
2. Operator control of vertical sample size, fully programmable from the
desk, as a dynamic if you wish, now! 
3. Fully compatible with all other known telecine options.
4. Fixes picture geometry.
5. Accurate beam current clamp. This means the PEC auto align will now be
accurate and consistent.
6. Intelligent tube protection schemes that won't annoy!
7. Sensible price - No hidden cost of ownership, the tubes can only last
longer still!

Stu Hunt

thanks to Phil Kroll of Otti International for support of the TIG in 1997
      TIG subscriber count is 843 on Mon Aug 18 01:42:13 PDT 1997
   archives and much more at http://www.alegria.com/telecinehome.html
     mailinglist digest available.... unsubscribe via a message to
        'telecine-request at alegria.com' with Subject: unsubscribe