EBU Technical Recommendation R92-1998
Charlie Pearce
confuse me more than I was before about 625-line TV/DVB broadcasting:
The ITU BT.601 standard (digital TV) specifies a line length of 720
pixels, whereas the BT.470 standard specifies that only the central
702 pixels be used to carry active picture information. (my
paraphrasing)
My questions:
1. What are the other 18 pixels used for? Horizontal retrace, in the
same way that 49 of the 625 lines are used for vertical retrace?
2. Are DVB broadcasts/European DVDs really 720x576 or actually
702x576?
3. Does the 4x3 aspect ratio currently used refer to 720x576 pixels or
702x576 pixels?
4. What about NTSC/US DTV? Are there 720 active picture information
pixels, or less?
5. Where can I get hold of ITU documentation? The EBU website
contains plenty of papers, but the ITU site seems to imply you need to
be a paying subscriber :-(
Thanks for any information
Charlie
Stephen Neal
Charlie Pearce wrote in message <373331b1...@news.globalnet.co.uk>...
>I recently found and read the above document, which has helped to
>confuse me more than I was before about 625-line TV/DVB broadcasting:
>
>The ITU BT.601 standard (digital TV) specifies a line length of 720
>pixels, whereas the BT.470 standard specifies that only the central
>702 pixels be used to carry active picture information. (my
>paraphrasing)
>
>My questions:
>
>1. What are the other 18 pixels used for? Horizontal retrace, in the
>same way that 49 of the 625 lines are used for vertical retrace?
Hmmm... Been out of technical stuff for a while - but seem to recall that
there is a different analogue blanking width for 525 and 625 systems - could
this be a legacy issue? i.e. 18 pixels more blanking in 625 than 525 -
therefore if picture content were contained in these 18 pixels it would be
blanked when converted back to analogue? Only a hazy recollection mind...
Peter Harris
Charlie Pearce <pea...@globalnet.NO-SPOO-PLEASE.co.uk> wrote in article
<373331b1...@news.globalnet.co.uk>...
> I recently found and read the above document, which has helped to
> confuse me more than I was before about 625-line TV/DVB broadcasting:
>
> The ITU BT.601 standard (digital TV) specifies a line length of 720
> pixels, whereas the BT.470 standard specifies that only the central
> 702 pixels be used to carry active picture information. (my
> paraphrasing)
720 luminance pixels (digital active line) at 13.5MHz pixel-rate is a time
of 53.3 microseconds.
Analogue video from a 625/50 blanked camera Y output will be 52
microseconds in width between half-amplitude points,
so the analogue active-line period of a picture source is nominally 702
pixels. (702 x 1/13.5MHz = 52 micosecs).
> My questions:
>
> 1. What are the other 18 pixels used for? Horizontal retrace, in the
> same way that 49 of the 625 lines are used for vertical retrace?
The 18 other pixels allow for the rise and fall time of analogue blanking
either side of the 52 microseconds nominal,
and for any positional lee-way (mis-timing!) of the analogue video relative
to digital syncs.
There are 144 Y "pixel periods" of digital blanking period for H retrace,
etc. (625/50)
>
> 2. Are DVB broadcasts/European DVDs really 720x576 or actually
> 702x576?
Most MPEG video encoders work with a signal input that was originated
according to analogue standards.
So the occupied pixels of the digital active line are a nominal 702.
Some encoders pre-filter and downsample to fewer pixels. This softens the
picture, but reduces the input data per frame
to the compressor.
>
> 3. Does the 4x3 aspect ratio currently used refer to 720x576 pixels or
> 702x576 pixels?
Aspect ratio calculations are made relative to the actual H width of the
analogue source picture, ie 702 pixels.
With 16:9 pictures, the 702 pixels still make 52 microseconds of analogue
waveform, but it is spread over 1/3 more
screen width for the same image height.
> 4. What about NTSC/US DTV? Are there 720 active picture information
> pixels, or less?
Try www.atsc.org?
> 5. Where can I get hold of ITU documentation? The EBU website
> contains plenty of papers, but the ITU site seems to imply you need to
> be a paying subscriber :-
Regrettable but true! Try www.etsi.org/BROADCAST/
> Thanks for any information
>
> Charlie
>
Could I suggest a Digital Video Standards and Measurements course at BBC
Training, Wood Norton?!
Peter - my thoughts not necessarily those of......
--
Charlie Pearce
getting the idea!
It would have helped if I'd thought things through to start with...
A pair of "frames" (is there a correct term?) takes 1/25 sec = 0.04
sec.
In this time there are 625 horizontal lines, and I'm assuming that
it's much easier to keep the horizontal scan regular than stop/start
etc., so a horizontal scan will take 0.04/625 = 64us, 5/6 of which
(53.33us) is "forwards" and 1/6 (10.67us) "backwards". To avoid
strange effects when the raster is beginning or ending a forwards
sweep, the first and last 9 pixels are blanked. In digital TV,
because MPEG-2 likes blocks of 16, this becomes the first/last 8.
How am I doing so far?
Vertical movement seems a bit more complicated. Presumably there's no
Y movement for 53.33us, then the raster moves down a line during the
10.67us retrace. But when and how does it move back to the top
(rather line 2 to handle interlacing)?
Presumably NTSC is the same, but with 63.56us scans (1/(29.97*525))?
How about PAL60 - presumably this has 53.33us scans, so what is the
active line length (it can't be the same as PAL/NTSC - does this cause
problems with format conversion?)
Thanks for any help (though I was probably better off in blissful
ignorance ;-)
Charlie
Charlie Pearce
pea...@globalnet.NO-SPOO-PLEASE.co.uk (Charlie Pearce) wrote:
>Vertical movement seems a bit more complicated. Presumably there's no
>Y movement for 53.33us, then the raster moves down a line during the
>10.67us retrace. But when and how does it move back to the top
>(rather line 2 to handle interlacing)?
I had a good long think about this at work today and think I've worked
it out. I imagine there are three regular mechanisms at work - one
that moves the raster from left to right for 53.33us then right to
left for 10.67us, one that moves the raster down 2 lines during the
10.67us retrace then waits for 53.33us, and a third which moves it up
625 lines every 0.02 sec. I suddenly realised why there is an odd
number of lines - line 313 traces an odd path along the bottom of the
screen then diagonally across the middle and ends up above line 1, so
that the next line fits neatly in place. After line 625 there's a
Y+625 and a Y-2 translation at the same time, so the raster returns to
square one. Very clever stuff!
Now I know this is probably very old knowledge to a lot of people, but
it was quite a revelation when all the pieces slotted into place ;-)
Of course, it's taken me until the advent of digital TV to work out
how the old system works!
Charlie