Re: [Leica] [Partly Relevant] The Economical Leica and Future Film

["Patrick G. Sobalvarro" <pgs@sobalvarro.org>]

Bryan Willman wrote:
> 
> First, my few contacts in the movie industry tell
> me that modern 35mm commercial stuff isn't
> 1/2 frame, but some sideways panamorphic thing.

That's interesting; by contrast, my contacts in the movie industry tell
me that very few feature films are shot on anything but 35mm -- E.T. and
Lawrence of Arabia, for example, were shot in something approaching
70mm, but 70mm is mostly a _release_ format that allows you to get more
light through the film for projection in a bigger theater that has the
right equipment.  But it doesn't matter -- if you quadruple the film
size, the cost of the storage is still small compared to other
production costs.

> Second, MPEG works fine for things like TV sets,
> you couldn't possibly tolerate it for showing on
> a commerical movie screen.  People shoot or
> print up to 70mm (or even bigger for IMAX) because
> *uncompressed* 35mm isn't *good enough* for
> some big screen movies.  MPEG only helps for
> small screen things.

Your statements about compression here don't make any sense.  Why do you
think that saving storage space through compression is useful for small
formats and not for large formats?  First, MPEG II is not dependent on
on format -- the maximum dimensions in the standard are 16K x 16K pixels
per frame.  Second, you wouldn't be able to see the compression
artifacts from compressing a movie by a factor of 10.  If you're
thinking it's like compressing a still image by a factor of 10, think
again -- the redundancy comes from the fact that little changes between
frames.  Third, as I said above, most feature films are shot on 35mm,
not IMAX or 70mm.  And just printing in 70mm doesn't improve resolution,
of course.

> Third, 7,776,000 megabytes is 7.7 terabytes,
> which would be larger than, say, the Sabre airline
> reservation system.  About 6 years ago, there were
> thought to be at most 1/2 dozen database > 2TB in
> the world.  I'm sure there are more now, disk grows
> fast,  but make no mistake 7.7TB is a *BIG* system.

I'm not sure what relevance these statements have to the topic at hand.

> To pump movie data in and out of the machine fast
> enough to allow real time editing, you need fast disks,
> and lots of them.  So IDE drives don't cut it.

I think your information about this area may be somewhat out of date. 
First, nobody makes plain old IDE hard disks anymore.  They're all
ATA-3, which has a maximum transfer rate of 33 MB/s and sustained
transfer rates of 5-10 MB/s.  Second, the bandwidth requirements for
this problem aren't large.  You need about 86.4 MB/s, which you could
easily do under the sustained transfer rate by striping the data across
sixteen disks.

> Effective price is about $100/GB.  7700 * $100 = about
> 3/4 million $ in disks.

This figure is wrong, being based on your contentions that (a)
compression can't be used for movies, and (b) disks cost more than they
do.  Incidentally, I was basing my calculations on consumer-priced disks
- -- if you want to buy 500 to put in a system, you can expect to pay much
less.

>  But to keep that much disk
> running, you need hot backups (say, 1 for 3) and a
> staff of several humans to run around and replace them.
> Big bucks, even for a movie.)

You need about 80 10GB disks for the whole movie; you'd probably want
ten times that many available while shooting.  MTTF for modern disks is
about one or two years; you might expect two or three in 800 to fail
each day.  With modern distributed disk technology, you would keep about
15% extra storage in the system and replace bad disks once a week or
so.  Of course you need a couple of operators to maintain a system this
big, but -- do you have any idea how many people it takes to shoot a
feature film?  In any case, 32GB disks will be out next year, reducing
the part count and the failure rate for the system.

> BTW - it's very difficult to effectively back up or reload
> a multi-terabyte database.  Archiving the stuff would
> be quite hard.

That's certainly true if you're using antiquated software.  You should
read up on distributed disks.

> And finally, if we concede 36MB per frame,
> the 24fps data rate => 864mb/second, which
> is faster than any PCI bus.  In fact, it's faster
> than all but the largest commercial system buses.
> (It's faster then the main memory bandwidth of
>  almost all current computers.)

You are once again supposing that compression can't be used.  If we
discard this superstitious notion of yours, we get 86.4 MB/s, which is
even below the old 32-bit 33MHz PCI bus bandwidth.  Even if we did
lossless compression, getting rates of about 4 to 1, a 64-bit 66MHz PCI
bus could easily handle the bandwidth.

> So, in practice, there are no commercially viable
> current computers that can edit *all* of the data
> for a movie *on-line*.  Of course, it's easy to store
> and edit this data as film.  (Toy-story was made
> with a whole farm of computers, and took a long
> time, and it's a *cartoon*, not a live action flick.)

You've made a bunch of incorrect statements and derived a bogus
conclusion.  I'll just repeat my earlier conclusion:
> >
> >So it would be possible to build the equipment to do this sort of thing
> >right now, except that development costs would be large and the
> >potential market small.  Still, the parts cost is small enough that I
> >would guess that within ten years most feature movies will be shot
> >digitally.
> >
> >-Patrick
> >