Posted by Toby on 07/13/06 02:15

"PTravel" <ptravel@travelersvideo.com> wrote in message
news:aSusg.63521$fb2.15301@newssvr27.news.prodigy.net...
>
> <mmaker@my-deja.com> wrote in message
> news:1152543082.503878.45400@h48g2000cwc.googlegroups.com...
>> ptravel@travelersvideo.com wrote:
>>> Simply saying so doesn't make it so.
>>
>> Look, you can claim all you like the data dropouts don't happen on DV,
>> but some of us have been working with DV for the best part of a decade,
>> and have seen plenty of them. Which are we supposed to believe: your
>> claims or our experience?
>
> I didn't say they don't happen. I said that they're such a rare event
> that considering them a source of generational loss is incorrect.

And your evidence for this is?

>
>>
>>> > > In the
>>> > > 200-300 hours of digital video that I've shot, I've never had a
>>> > > single
>>> > > instance of drop-out -- not one.
>>> > Really? Your deck actually tells you when it gets an uncorrectable
>>> > error?
>>> No -- drop out is obvious when it occurs.
>>
>> How? The only obvious dropouts are when a) the dropout is in the audio
>> track, where the DV deck can't mask it and drops in a frame of silence
>> instead or b) when there's too much motion in the frame for it to hide
>> the dropout by replacing a video block with the same block from the
>> previous frame. Otherwise you won't notice dropouts when they occur,
>> but you've still lost data from the original file.
>
> Decks may do this kind of correction, but I don't use one. I use an old
> camcorder for transfer -- if there was drop out, there would be a
> noticeable block of corruption on a single frame. Though I've not
> experienced any drop out, I have had instances where there was a break in
> timecode because, for example, I had removed the tape, then placed it back
> in and not bothered to find the final frame. The break was obvious when I
> captured or tried to view the tape.

Cameras also do this kind of correction. That's the beauty of ICs, design
one that does what you want and mass produce them for pennies. They are so
small you can fit them anywhere, even in one of those little camcorders.

Take the mentioned "block replacement" for example. Let's digress for a
moment and think about Mpeg2 interframe compression, because the techniques
used here are similar to various schemes of error correction. It ain't all
black and white, as you lawyers would like to believe.

In interframe compression one of the main ways of saving space is to reduce
the data actually recorded and use various algorithms to reconstruct what is
missing in some "acceptable" way. Basically what happens is that an I frame
is recorded in its entirety, and following that a whole series of P frames
are recorded which contain information only on what has changed temporally
since the I frame was recorded. This is what is being done in all those new
little palm sized HDV cams, so you have some real fancy processing going on
in that little package, something that its quite difficult to do on your PC
unless you're running dual Xeon 3.2s or better, but then it's pretty easy if
it's all hardware-encoded...

But I digress. The point is that even in intraframe temporal compression if
you had to analyze every single pixel it would be a huge burden on the best
hardware. So there are various schemes for reducing the number of pixels
that have to be analyzed and reconstructed. Blocks of varying sizes are
constructed based on change between frames and lots of other very ingenious
stuff. For a full treatment see here:

http://www.bbc.co.uk/rd/pubs/papers/paper_14/paper_14.shtml

Anyway, the point is that a huge number of tradeoffs have been made, MAJOR
compromises that can be gotten away with only because of the way the brain
processes visual data. And because the engineers have done their job well,
basically you never notice the degradations when the family is oohing and
aahing over your vacation video. But the losses are very real and they are
there, although they usually only become apparent in scenes containing a lot
of motion, and only if you are paying attention and know what to look for.

Think about jpeg compression. You would be hard pressed to tell the
difference between a lossless image like a bitmap and a high quality jpg -
even at 100 percent with close scrutiny - although the jpg might be an order
of magnitude smaller. However open and resave the jpg a number of times and
you will soon start seeing the image quality degrade, because it is a lossy
compression scheme, and therefore something is lost each time, which adds
up.

If the error correction on video can be considered a lossy scheme, and it
certainly is if the correction is not bit perfect (and I believe, couselor,
that you have already conceded this possibility), then there is no question
but that at some point, after a finite number of copies, these losses will
exceed the visual threshold and become noticeable, just as one notices a
general lack of quality in a jpg after repeated copying.

Please don't misunderstand. I'm not saying that direct copying of the dv
bitstream is lossy, only that non-fully-corrected errors will act lossy
after a given number of copies, and small, undetectable losses will build
upon each other with each generation until they become detectable after a
given number of copies.

Toby

[Back to original message]