[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: RE>Re: Thunder and Spirit
I have to disagree with my friend Paul Grace on the subject of data scanning
speeds on Spirit/Thunder. My understanding is that Philips claimed a maximum 2k
scanning speed of 23 and a bit frames per second. They cannot achieve the last
little bit and get to real time because of the physical construction of the CCD
block and the architecture of the electronics.
Without going into the depths of how it works (and I am typing this early on a
Saturday morning, when my brain is not at its sharpest, so I would probably get
it wrong), there are two CCD blocks in the Kodak scanning head used by Spirit
The first is for colour. This has two 480-photosite sensors for each colour
(red, green and blue), and they are self filtering (i.e. in effect they have the
appropriate colour filter glued to each line array). They are mounted one above
the other, to provide a 960 pixel per line signal in each of red, green and
The second block contains four identical but unfiltered 480-photosite line
arrays in two staggered rows of two, to give a 1,920 pixel detail signal. Note
that there is no colour information in this: it is in effect a luminance signal,
although it is not treated that way, it is added to each of the colour signals
as a detail channel.
This is an excellent analogue of the way the eye works: we see fine detail in
luminance only, and colour only in larger areas. Broadcast colour television
systems have always used this fact, transmitting a high resolution luminance
signal and adding a wash of colour. Using this, Philips has achieved excellent
visual image quality while minimising CCD noise.
Anyway, that was not the point I set out to make. What is important to
understand is that the line arrays are staggered. On the colour block, the
colours are scanned one after the other (to scan them simultaneously would
require a beam splitter which reduces light and therefore increases noise). On
the detail block, pixels 1 - 480 and 961 - 1,440 are scanned first on two of the
line arrays, then the second pair scans into the two gaps.
All these signals have to be timed together, without adding noise or shifts to
these delicate signals at the start of the passage through the telecine. My
understanding is that this is done in hardware, which was optimised for high
At Spirit/Thunder's 2k resolution there are about 1,400 lines on a 35mm film
frame; in 1125/59.98 HDTV there are 1,080. To get the increased line resolution
you have to reduce the running speed. 1080 divided by 1400 multiplied by 29.98
gives you about 23.1 frames per second as the maximum speed. All this is
described in horrific mathematical detail in a paper given by Mike Christmann of
Philips to SMPTE earlier this year.
On Paul's other point, I was told by Philips that the data pack would be on
display at IBC and deliverable immediately afterwards, but Paul is very close to
the business and his information may well be more up to date than mine.
Hope this helps.