| A DV(L)-FAQ [e] |  |
DVL-Digest 552 - Postings:
Index
'That' Panasonic booklet - (2)
EIDE tech questions
Filming sunset. How the pros d - (2)
PAL resolution
videoke chargen
'That' Panasonic booklet - "Perry" nospam-perry.mitchell@btinterne
Jan Crittenden posted:
> And yes while it is a free lunch, Perry, the reality is there, that
DV and MPEG2 Compression as it begins in both compression schemes is using
DCT. I do recognize that this may take a little synthesizing of information
to make the connection between them, but there is the similarity. Now
certainly the book was written in an effort to help people understand that
once you throw information away, it is gone, which is what happens in MPEG2.
However the initial discussion was how would DV look when recompressed into
MPEG. I said that if you brought it back to the DCT and then applied the
MPEG2 algorithm that it would look pretty good.
DCT is (at least theoretically) a completely transparent and reversible
transform to and from the normal raster scan. What differentiates the
compression schemes is what you do before DCT to filter color and then with
the data obtained after the DCT stage. MPEG is more efficient overall and
therefore for the same target data rate it needs less compression after the
DCT stage and more importantly it needs less color filtering (known as
'Raster Transformation'). What is important is that the data that is
'thrown away' is either redundant or irrelevant.
If we take for instance standard DV (or DVCPRO) and MPEG2 442@ML (or
BetacamSX) then they are both of similar data rate (DV is slightly higher).
If you took an individual frame then the DV version would have less color
information and it would usually be more highly compressed (DV has a more
dynamic variable rate compression dependent on the picture content).
All this is beautifully explained in the Panasonic booklet, and I can only
once again please urge Jan to see if it can be returned to availability off
the web site. I notice now that it has a few silly but obvious typos on
some of the diagrams, and I suspect it has been withdrawn to correct them.
Perry Mitchell
Video Facilities
http://www.perrybits.co.uk/
'That' Panasonic booklet - "Perry" nospam-perry.mitchell@btinterne
Jan:
This is getting a little bizarre and I can only suggest you do a little
homework over the weekend with the famous booklet!
To avoid everybody else getting very confused I'll just say that:
1) Raster Transformation is the process of irretrievably throwing away some
of the colour information, to give either 4:2:2 or 4:1:1 or 4:2:0 color
mapping. You CANNOT get from 4:2:2 to 4:1:1 to 4:2:2 and have the same
result as staying in 4:2:2. (assuming of course that the color information
is there to start with!*)
2) AFTER AtoD, the signal goes through a transform called DCT (Discreet
Cosine Transform). This process in itself does not compress and is
theoretically reversibly transparent (but subject to rounding errors).
3) The compression is achieved by then manipulating the data after DCT.
4) The compression per frame for DV is more than it is for the I frame of
MPEG2 422@ML.
*note* In the earliest days of digital, one of the party tricks was to
remove 4 of the 8 bits when viewing color bars. The effect on the picture
was almost invisible.
Perry Mitchell
Video Facilities
http://www.perrybits.co.uk/
EIDE tech questions - Adam Wilt nospam-adamwilt@flash.net
> Excuse me, but a SCSI cable only has ONE set of data lines, therefore only
> ONE SCSI device on the cable can be putting data on the cable.
Excuse me, but SCSI supports device disconnect and overlapping I/O. A
controller can ask one disk to fetch some data, disconnect, and ask another
disk to fetch some data. At some later time, one disk comes back and returns
its data, then the other one does.
With IDE, there is no overlapping I/O. A controller can ask a disk for data,
but it must then wait for the data to be returned before it can initiate a
request to another disk.
So while there's no simultaneity, SCSI allows useful work to happen in the
latency period between request and response. More efficient bus usage results.
Medea VideoRAIDs use multiple EIDE disks, but each is on its own controller,
thus allowing for simultaneous access to as many disks as are in the RAID. A
clever trick!
Cheers,
Adam Wilt
a DV FAQ: http://www.adamwilt.com/DV.html
Filming sunset. How the pros d - "Perry"
>How do the pros film sunset? Is a more professional camera necessary
for satisfactory results? Filters? Did I do anything wrong?
The best technique is to use a film camera! Using a video camera it is very
difficult to get a sunset in the tropics (I used to live in Seychelles)
because the light changes so fast. For the uninitiated, a sunset near the
equator goes from full daylight to full darkness in a couple of minutes,
there is almost no twilight. The spectacular red sky can be there for
literally just a few seconds.
I use graduated ND filters and manual exposure, but most importantly divide
the sunset up into short sequences, compose and expose each to suit, and
then edit them together later.
Mostly it is luck anyway, I went back last summer with a DVCPRO50 and never
got a decent sunset in 10 days! The day after I got back to UK we had the
most spectacular sunset that I ever remembered seeing.
I also used to live in Greece, and there is a very famous sunset destination
at a place called Cape Sounion. You've probably all seen it on travel
brochures and posters, with a temple and the sunset behind. I once took
some visitors there and literally about 5 minutes before the sunset a BBC
film crew turned up and managed to clear the site of about 100 tourists and
shot a documentary insert involving a violin player with the sunset behind.
It's still the most amazing example of people management that I have ever
seen, but they were shooting 16mm and I suspect if it had been on video they
would have stood no chance of lighting it in time.
Perry Mitchell
Video Facilities
http://www.perrybits.co.uk/
Filming sunset. How the pros d - "Perry" nospam-perry.mitchell@btinterne
Ton Guiking posted:
>Two questions:
- Maybe I've got an off-day, but why would it be that much easier in film?
(apart from the latitude in exposure)
- This people mangagement makes me curious: did they do it the soft-polite
way or the crude-hard way?
Not only latitude, but dynamic range which to some extent can be adjusted in
post. If you get the exposure roughly correct then you can get what you
need later. If a video camera is not correctly exposed, with the dynamic
range reduced by suitable filters, then the image will be clipped and lost
forever.
The BBC chap treated everybody like an army officer would talk to his
troops, and they all obeyed him with hardly a complaint. They were tourists
from all over the world, I suppose maybe it was more like a shepherd dog and
a flock of sheep! It was simply amazing, I talked about it for weeks!
Perry Mitchell
Video Facilities
http://www.perrybits.co.uk/
PAL resolution - "Perry"
dani-el landau posted:
>does anyone know why PAL DV ratio is 5:4 (720*576)
where as most PAL TV's and projectors are 4:3?
Don't we lose this way part of the image?
Don't confuse the sample rate with the image size. In PAL the image size is
768x576 which is 4:3. When most of the current digital video formats were
specified, it was decided to use the same horizontal sample size for both
525 and 625 line video, and that was 720 samples per line. Actually this is
a little wider than the broadcast picture, so of the 720 samples only 702
are actually used for the finished image and this should be related to the
4:3 picture.
If therefore you wanted to make a graphic still in say Photoshop, you should
compose it at 768x576 and when you have finished 'distort' it to 702x576 and
then add some black to the sides (change 'canvas' size) to make it up to 720
width. If you go directly from 768x576 to 720x576 then the error is only
about 2% and very few viewers will notice the difference!
Perry Mitchell
Video Facilities
http://www.perrybits.co.uk/
videoke chargen - Adam Wilt nospam-adamwilt@flash.net
> Can anyone fill me in with regards to doing videoke chargen in premiere?
If videoke is like karaoke, do two versions of the text lines, one in the
"before" color and one in the "after" color, and do a slow horizontal wipe
between them.
Build one line of text. Save it, change the name, and then *carefully* select
the text (so you don't move it) and change the color. Now you have both the
"before" and "after" lines, which should superimpose perfectly.
You can put the "after" chargen on a higher layer than the "before" chars, and
do a reveal of it (perhaps with an alpha wipe) so that it overlays the lower
layer exactly.
Or, put both text lines on video 1a and 1b in a virtual clip (which you then
super on the underlying background video) and simply do a horizontal wipe from
"before" to "after".
Cheers,
Adam Wilt
(diese posts stammen von der DV-L Mailingliste - THX to Adam Wilt and Perry Mitchell :-)
[up] |
|
|
