Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!uunet!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!husc6!frooz!cfa250!john
From: john@cfa250.harvard.edu (John Roll)
Newsgroups: comp.graphics
Subject: Re: How many bits are needed for NTSC? [was Re: Workstations that can record/play]
Summary: NTSC Bits pre pixel
Message-ID: <1781@cfa267.cfa250.harvard.edu>
Date: 20 Nov 89 12:56:01 GMT
References: <1593@ndmath.UUCP> <46900048@uxe.cso.uiuc.edu> <630@alias.UUCP>
Organization: Harvard/Smithsonian Center for Astrophysics
Lines: 52


The problem with trying to figure out the # of bits per pixel that are needed
to reproduce NTSC quality video on a workstation is that the NTSC picture
is encoded in a different color space.  The usual RGB color repersentation
that we use for computer graphics display is over done.   When the NTSC spec
was created the television (analog hackers) engineers shoehorned the color
broadcast singal into the already defined black and white broadcast system.
They did this by limiting the sugnal to those colors and bandwidths that 
the eye could percieve.  The signal is not an RGB signal because the eye
descriminates the changes in colors at different bandwidths.  Black and White
is the most sensative.  (~8MHz).  The color portion of the picture is sent
along the IQ color axis.  I is the purple color at the bottom of an NTSC 
test pattern and Q is the more anber white color.  These signals make up the 
color picture and are sent as "sub-carrier" modulations of the BW signal.
Any way these signals are sent at much lower frequencies.

To obtain NTSC quality from RGB 24 bits are indeed needed. (fact)
To obtain NTSC quality from another scheme MUCH less bandwidth is needed.

Example: YRB color syncronized coding used in the New Media Graphics
	 VideoWindows digital video board.

	12 bits per pixel at 8MHz
	Endcode The BW as 7 bits lumanance at 8MHz
	Multiplex the two chroma channels to obtain two 4MHz streams.
		8 bits R-Y at 4MHz
		8 bits B-Y at 4MHz

This will give FULL NTSC color and spatial resution in a more compact form 
than RGB.


Other schemes are even better but more difficult to handle in a DSP (for
post processing of the video signal).  If the signal were encoded in its
native braodcast format YIQ

	10 bits at 8 MHZ
	7 bits Luma
	Multiplex I and Q to obtain 8 bits at 1.6 and 3.9 MHz respectivly

DISCALIMER: I am a past employee of New Media Graphics.

This answer is off the top of my head all flames ignored.  For a serious 
explaination see "The Television Engineers Handbook"  1000 pages of fine
print but really covers the subjects of color space representaion and 
signal encoding (recomended reading for all CG Engineers).

John@cfa

		
needs of the human 
~v