[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Jude Ng asked some questions about "what is film resolution?". My feeling
is that there is still a great deal of misconception here, and I offer
absolutely no guarantees that I understand it correctly. However, that has
never stopped me speaking out before.
Film manufacturers like Kodak publish measurements of what we could regard
as the limiting resolution of film: they are called modulation transfer
curves. These measure the number of cycles from black to white you can
resolve per millimetre. As the name modulation transfer curve suggests,
they are not absolute figures, but curves of line-pairs against resolution
(expressed as a percentage), for each colour dye layer. You need to pick a
suitable mid-point as the limit of what would be acceptable.
If you then convert this to real resolution across the film frame, and do
suitable Nyquist corrections to get to a digital sampling rate, you find
out that to make a true digital copy of a 35mm film frame, you need to
sample at 4k resolution. Note that this means 4,096 pixels across the width
of the film. Super 35mm is about 4:3 in aspect ratio, so that means around
3,000 lines, each of 4,096 pixels.
However, this results in such a massive data file (around 75 megabytes a
frame) that it has generally been regarded as unwieldy. Practical
experience with 2k resolution - resulting in a quarter of the data - have
shown that it is quite good enough for most applications. My understanding
is that most of the famous digital special effects movies have been made at
A subsidiary benefit of this is that 2,048 pixels on around 1,500 lines is
not that far off HDTV's 1,960 pixels on 1,250 lines, which is what made it
possible for Philips to migrate from the FLH-1000 to the Spirit.
Remember also that all these numbers are scaled down for 16mm.
I can bore you all rigid with the calculations necessary for digital film
resolution if you need me to, but they are not very interesting. Indeed, I
may already have bored the group on the topic, in which case you will be
able to find it in the web site.
With regard to colorimetry, again accepted wisdom is that to capture the
full range of colour depth you need either 14 bit linear or 12 bit
logarithmic quantisation in each colour channel. However, the pragmatic
view of some manufacturers is that you can get away with smaller colour
depths to reduce the size of the data files you need to move and store.
Jude's final question - do you need higher resolution film formats like
VistaVision - is an interesting one. Do you go to greater-than-35mm
resolution to compensate for the losses in the post production and
duplication chain, or is it for other creative reasons (like framing)? You
need to talk to someone who understands these things.
At IBC last year Philips was talking about a research project with Warner
Bros. for digital production of every stage of a movie, not just effects
work. The days negatives would be scanned into a vast digital store, the
non-linear edit could be then be on-lined, and a new film master could be
output for distribution. It would be interesting to hear from anyone at
Philips or Warner Bros. on how this is going.