OT: Composite/Component [was Re: [mythtv-users] Re: Image quality, what effects it?]

[Cory Papenfuss]

> The knee for luminance bandwidth in a traditional NTSC signal is
> actually 4.2MHz, and, if your filtering wasn't good enough, you got
> less.  In *theory*, it's possible to move more through an S-video link,
> but you have enough bandwidth on the tape to store it.  S-VHS and Hi-8
> can make use of more of what's already there, but I don't know that you
> can usefully extend it any.
>
> In any event, Yves Faroudja has proved in his Super-NTSC labs that you
> can fit into an NTSC channel a picture that would knock most
> network-affil chief engineers *flat* on their ass; most NTSC gear is
> truly garbage.  But then they wouldn't be able to *sell* HDTV.
>
> Cf. Pogue Carburetor.  :-)

 	No big surprise there.  It's sorta a "digitally enhanced" analog signal 
then... some post processing to recover some information to 
deinterlace/interpolate.  I guess I recall the 4.2 MHz knee in the luminance. 
My main question was whether or not most consumer gear takes advantage of the 
possibility of more luminance bandwidth in S-Vid.  If your source had plenty 
(DVD, CG, etc), you could even get more out of composite since the audio 
subcarrier isn't necessary anymore either, right?

>>   It's almost exactly the same as the old "sync-on-green" three-coax
>> workstation monitors. Oversimplified, Y=R+B+G, Pb=B-Y, Pr=R-Y and
>> H/V sync is also added to Y. Realistically, they're weighted by
>> their perceptual luminance, but for illustration it's not important.
>
> Well... yeah.  And you're right; I and Q are *modulated* chroma, where
> Pr and Pb are luminance signals, so there's more difference than I let
> on.
 	Luminance signals... of a specific color, I guess.  True enough.... at 
least they're baseband video information, not modulated.

>
> Yeah; this is an issue I've always wondered about; those component
> cables can carry signals that are *not* RS-170A timed, then, right?

 	Sure... most of the transcoder boxes (A960, etc) wouldn't inherently 
have that limitation.  Garbage in = garbage out.  If TV's were "multisync" like 
computer monitors, they'd even work on it.  I suspect that some of them 
would... some would also probably blow up and die, though too.

>>   Interesting. What do you mean exactly? Did the U-matic remix a
>> "baseband" chrominance (demoded from the 3.58 subcarrier) onto 729kHz.
>> Then send a differential I and Q down 2 pairs?
>
> No, it was like S-video: the chrominance was quadrature modulated, but
> at 729kHz, instead of 3.58MHz; they shipped it out as Y/C; you can get
> frequency heterodyne boxes to convert U-Matic Dub Out to "proper"
> S-video.
>
> Remember, even on VHS, the NTSC color subcarrier is *not* recorded at
> 3.58MHz.
 	Gotcha... it's just that you said "4-wire" so I thought maybe they ran 
them separately.  What's the color subcarrier on VHS?

>
> I understand that the problem is specifically the Bt878 chip, which
> wants to bus-master its PCI transfers; the video4linux at redhat list
> archives for the last month might be informative on this.
>
 	?  I was basically referring to the quantity of data that had to be 
passed.  If it were true RGB (which it's not... probably a factor of 1/2 for 
4:2:2), 720x480 * 30fps * 3 colors * 8bits/color = 31MBps... that's a lot of 
data for a PCI bus.  Even with a 1/2 reduction for YUV and 1/2 D1 of 352x480, 
that's 7.5 MBps... still plenty.  Trying to do HDTV would require 
proportionately more.  That pretty much means it would have to encode to MPEG2 
(or MPEG4) on the fly to stream over the PCI bus.  Lots of embedded CPU power 
necessary for that... 4x-8x hauppauge I suppose.


>>> Ok, I think it's time for someone to snipe at me now for trying to
>>> be informative. ;-)
>>
>>   Me too... :)
>
> Cheers,

 	"Techo-goobers hijack mythtv thread... full story at 11!"  ;)

-Cory