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

Coments re: Film Reg. (combined responses)

Although I have been out of town for a week, I still wish to
respond to the various discussions regarding film registration.
I have had some experience in this area.


> Brad Hunt wrote: 

> I attended the demonstration at the ITS show of the new Sony HD 
> telecine prototype. For those of you that couldn't make this 
> show, I thought I would share my views of this demonstration.

> First off, Sony's engineers should be commended on the brilliant
> approach they came up with for correcting image steadiness. They 
> use several small capacitance sensors in the perforation area to 
> measure the vertical and horizontal position of the perforation.
> Since film base acts like a dielectric, the capacitance value 
> varies across the sensors as the perforations go by the sensors.

I would question a few things such as, what the density of sensors 
is, what the speed and accuracy of the capacitance sensing is and,
lastly how often do the horizontal and vertical errors of the film
get sampled?

> They use several small capacitance sensors within the perforation
>  area.

How many is several?  And in what direction?

> By tracking the rise and fall of capacitance across these multiple 
> sensors, a horizontal- and vertical-position correction signal can
> be produced. The X- and Y-correction signals drive servos controlling 
> two optical glass plates which can shift the optical film image 
> falling on the CCD area array imaging block.

This was the same method that I initially worked on with BTS/Philips
for their FDL-90/Quadra CCD telecine and quickly abandoned 
because of the realization that this method would have resulted in 
a very limited response time simply too inadequate and impractical
for the required number of corrections necessary to correct the 
positional errors within the frame.  

Realize that even though the size of the correction glass is 
physically small, the mass of the glass is enormous relative 
to the acceleration forces required to correct (or reposition)
the glass at the speeds necessary.

There are not many adequate methods available in which to 
cause a correction of the image to occur at the speed necessary.
Presently, the only two electro-mechanical methods available 
(that I am aware of) are piezo electric transducers and high 
speed electromagnetic coils.

1. Piezo cannot provide enough physical range which 
   therefore requires thicker glass and consequently 
   increased mass slowing it even further.

2. Coils can provide the physical range of movement 
   but are way too slow to react quick enough for 
   any reasonable correction time.
> To eliminate problems with dirt or damage in a particular 
> perforation, they measure 8 different perforations.

First of all, in order to provide corrections as the film is 
being scanned in continuous motion, the relative position of
the film (or perfs) must be measured at a rate very close to
that of the horizontal scanned image.  It must also be measured
as continuous as possible in the longitudinal direction for 
vertical correction.  The more H & V samples, the better. 

Second, if an error for a particular H sample exceeds the 
limits of a predetermined window, as in the case of a damaged 
perf, it is easy enough to hold over the last correction until
the next valid correction occurs.  

If they measure 8 different perforations then likely there 
does not exist error correction sampling at anywhere near a 
high enough rate to be meaningful to correct for continuos 
motion H and V errors on a film transport.

> The steadiness correction system worked very well. They 
> shifted the image scanning area to show the perforations.
> They turned the steadiness corrector on and off. You could
> watch the perf edge locked steady and then watch it float 
> around. Very impressive. 

Very deceiving.  I have seen this demonstration once before 
(by someone else) and it meant absolutely nothing when 
compared to the inability for it to accurately correct the
normal film errors across the entire picture frame.

Remember, when you are viewing an expanded perf you are looking 
at that small area of the film.  Also, the perf is seen as a 
large white area surrounded by dark area.  You have little 
reference and indication of the movements elseware at various
points within the frame.  

Also, the mere fact that you saw an image of a perf while
the film was running in real time highly suggests that 
the system is definitely not scanning at line rate.  Otherwise,
you would not see the shape of the perf, instead you would 
see a straight line top to bottom (the perfs edge scanned 
in the same place for every line).

> It will be interesting to see how the correction system works with 
> real world film with dirt in the perforations and with perf damage.

It would definitely be interesting to see the results with 
real world film but, please use at the very least a 12 field 
registration chart and of course real film.

The problem with registering film in continuous motion is 
that the film has a multitude of various rates of movements 
in both H and V directions.  Also the more errors you correct,
the more subtleties of tiny intricate uncorrected movements
will visually surface otherwise masked by the weave.

While the film is in continuous motion across a gate it is 
typically bending, twisting and turning, not just moving left
to right.  Even though these errors are ever so slight they
are still enough to cause perceptible movements to a picture.
Although not impossible, it is not so simple to correct for
all of these errors.

> You should remember that steadiness correction systems that use 
> the perforation for reference are great for camera films (assuming 
> you use the correct perf for reference) but do not fully correct the 
> unsteadiness introduced in continuous contact printing used in the 
> production of most interpositives.

Very true, if registration error is printed into any film then 
the perfs are no longer the reference to restore steadiness. 


> Dick Hobbs wrote:

> Clearly Sony has made great strides with its HD telecine since
> the presentations they gave at NAB. The position sensor system
> is extremely ingenious - I bet there are a few telecine designers
> out there saying "I wish I had thought of that" 

Only if it works.

> but how do they adjust the position of the frame?
> Are there micro-adjustments to the sprocket drive to get each 
> frame accurately aligned? How does it work horizontally?

You ask some very excellent questions.  I wonder if someone 
out there from Sony would wish to comment.

> Apart from that, my other concern is with the zoom and rotate. Doing it
> mechanically of course eliminates digital artifacts, but I very much doubt
> if it can be achieved in a frame bar, which means that changing settings
> can only be achieved with assemble edits. Is that going to be practical in
> use?

All this would indicate clearly that this system is an adaptation of 
a camera to a telecine.  I for one, am disappointed yet not surprised.


              RE: RE>Re: Sony HD Telecine Review
              Thu, 17 Jul 1997 12:42:37 -0700 (PDT)

>> Clearly Sony has made great strides with its HD telecine since the
>> presentations they gave at NAB. The position sensor system is extremely
>> ingenious - I bet there are a few telecine designers out there saying "I
>> wish I had thought of that" - but how do they adjust the position of the
>> frame? Are there micro-adjustments to the sprocket drive to get each frame
>> accurately aligned? How does it work horizontally? <<

> Mike Most wrote: 

> My understanding is that the light source is mechanically deflected 
> (by a mirrorred surface mechanically driven by the correction signal) 
> to scan a slightly different part of the frame to account for the 
> registration drift.

The light source?  I think you mean the image.

> I might point out that it's actually quite similar conceptually to EPR 
> (in which the tube's scan patch is moved to stay registered with the film,
> rather than the film being mechanically pinned),

In fact, with EPR the scan is not really "moved to stay registered with
the Film" as you say, it is the mis-registration of the film against 
the reference scan that generates the error correction signals.  

> although, of course, a very different arrangement for both generating 
> the correction signal (variable capacitance) and deflecting the scanning 
> source. 

And Mike, would you press "Enter" once and a while?


              RE: RE>Re: RE>Re: Sony HD Telecine Review
              Mon, 21 Jul 1997 00:06:41 -0700 (PDT)

Paul Grace wrote:-
>> Projectors run film many times over and do not shed the sprockets 
>> or allow only one showing per print.<>

> Dominic Case wrote: 

> True . . . but print film has Eastman Kodak rectangular perforations
> which are optimised in shape for repeated runs without wear and tear.
> Camera negative stock has Bell & Howell perforations (barrel-shaped)
> which are designed for precise fitting to register pins.
> Try running camera negative through a projector back and forth a few
> times then look at the perforations.

Very well said.  

> Also, sit close to the screen at your local cinema and watch the
> steadiness of the image.  It's not really a fair comparison with the
> requirements of telecine transfer.

True but I do believe that most of the registration errors you see on 
theatre screen are actually printed in via contact printers rather than 
the fault of the projectors. 

As a matter of fact, I often notice titles dancing around all over the 
place against the backgrounds.


Michael Kaye

"Fighting for Goodness instead of Evil" - Maxwell Smart.

mailinglist digest available......posting guidelines on the webpage
the Telecine Internet Group  <http://www.alegria.com/telecinehome.html>