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


>Ken Robinson asks:- 

>With a CRT scanner why not keep a scan patch large and "lock >it at one

I appreciate his reasoning and the perceived benefits however real life
makes this impractical.

A fixed width patch would be no problem if we all accepted that pan moves,
zooms and all 35mm image widths were DVE'd downstream and out of a S35mm
width, this would mean resolution below broadcast standard except at full
S35mm camera apertures. BTW rotates would also be a DVE requirement.

A fixed height scan would need to be fixed at a still frame height covering
the highest film format, which happens to be 35mm anamorphic cinemascope.
All lesser heights even in still frame would then need DVEing with a loss
of vertical resolution.

The biggest drawback, as Ken alludes to is when the film is run.

Once the film starts to move an amount of the vertical scan is generated by
the films movement across the scanning aperture, to the extent that at
certain speeds, as we all know, no vertical raster scan is required, under
this condition we would be scanning in the same mode as a CCD line array
telecine .  If CRT telecines scanned at regular TV rates this zero vertical
scan would occur at the TV standard rate and there would be no vertical
scan at all. Cintel avoid this happening by deliberately scanning at faster
speeds than the TV standard and one of the reasons for this is to avoid
single line rasters at normal tv rates.

Now assume we retained a "locked still frame height scan" and then ran the
film. At 25 FPS PAL for instance the film would still be scanned with
approximately 600 lines however as both the  film is moving across the scan
aperture and the CRT raster is scanning "against the flow" of the film
across the aperture we would get a film image which had been completely
scanned during the first half of the raster scan, as Ken observed. The
second half of the scan would then scan the next frame. This results in two
film frames being displayed in one video frame, one above the other.

How about inverting the scan such that the vertical scan moved with the
film flow. In this instance each scan line would scan the same part of the
film and the resulting video image would be a vertical streak representing
a single line across the film image.  

Also consider variable speeds, the video image would vary dramatically as
speed changed, indeed at certain speeds the image would invert.

To avoid this stretch problem the vertical scan is scaled to the running
speed and overall zoomed raster size, such that the scanned height and film
motion across the aperture exactly combine to produce a correct height

The combination of film movement and CRT vertical scan rate (height) can be
equated to the time honoured description of an observer on one train
viewing an observer on another train on the adjacent track. The scanning of
film to obtain a correct image height is to always keep the relative speed
of the two trains the same.

Let's for arguments sake say 25 FPS=25MPH.

Train A=Film frame
Train B=CRT Vertical scan

A=still frame 0MPH
B needs to be doing 25MPH

A=PAL Rate 25MPH
B needs to stand still

A= silent speed 18MPH
B needs to be doing 7MPH

A= 30MPH
B needs to be doing -5MPH (ie the CRT scan reversed)

I won't even attempt to equate Zooms to this simile, suffice it to say the
observer's apparent relative speeds will change as the trains get closer
together or move further apart. As I seem to be going off the rails and
relativity is not my subject I will revert to the real world.

We do, with Ursa Gold, offer NewsFlip which when used in conjunction with
our scan rotate does rely upon scan vs film motion effect in the following

At scan sizes close to no vertical scan, rotating the scan though 360
degrees now causes the scan to need to keep up with the moving film, it has
to "run after" the film as it transits the aperture. This "running after"
the film causes the patch to transit the CRT giving a large patch. The
image due to scan rotation is now inverted and reversed, it is NewsFlip
that flips the image in the store right way up and flops the image

Given time and lots of coffee the above principle can be used to understand
why a 90 degree rotate gives a diagonal CRT scan.

I have no idea if this helps explain the reasons our telecinesí create such
contorted scans or whether it simply confuses the issue.

Ken's fixed scan solution would be tenable if the initial telecine scan
resolution was about 4K x 9K as this would allow limitless zooms, pans,
rotates and a full range of film frame speeds. If someone out there has a
DVE that can handle about 230 Mbytes of data in 30 milliseconds (6.9
Gbytes/second) sustained then I reckon I need to see what we could do. ! 

Meanwhile I guess we will retain the variable scan patch abilities that
give CRT telecines their flexibility.  

Peter R Swinson
16 October 1996