|
|
Posted by PTravel on 09/29/47 11:52
"Richard Crowley" <richard.7.crowley@intel.com> wrote in message
news:e8u41k$439$1@news01.intel.com...
> "PTravel" wrote ...
>> But that's the whole point -- if it's corrected, it's not a drop out, and
>> there's no data loss.
>
> That is a philosophical difference and you appear to be using
> the existentialist definition. You can be sure that the engineers
> who design the guts of these things take a far more "rationalist"
> view where they are indeed dropouts. :-)
I'm just looking at the resulting copy. I take a black box approach to
these things, i.e. if the output is the same as the input, it doesn't matter
to me what happens inside the black box to make it so. In this discussion,
if I wind up with a bit-for-bit accurate copy, it doesn't matter to me
whether the ECC guessed, used psychic prediction, or had employed a time
machine -- a bit-for-bit accurate copy means no generational loss.
>
>> There's no generational loss, because, by definition, the data was
>> corrected.
>
> If you are talking about computer files, you would be exactly
> correct. But I believe the context here is video (and audio)
> data on DV tape, where the rules are different. Some data
> failures can be corrected exactly (as they are for computer
> files). But then if the data cannot be corrected, the failure is
> mitigated with other kinds of recovery.
We're not disagreeing about that. My question is this: how often, in duping
from a DV tape, are there data failures that cannot be corrected exactly as
they are for computer files? I've never heard anything other than, "so
rarely that it's not a concern."
>
> For example, even audio CDs, which contain uncompressed
> 16-bit samples of the audio very very rarely play back perfectly.
> Data errors are routinely mitigated by extrapolating the missing
> values and you very rarely hear the difference. But the result is
> NOT bit-for bit. Same with digital video.
All I know about audio CDs is that my computer employs error correction when
it reads them. However, that's still audio CDs -- I'm asking about digital
video tape.
>
>> We're talking apples and oranges here. On the one hand, there's
>> everyone's DV deck that can invent data if it's missing, and on the other
>> hand there's ECC which can _repair_ data, i.e. restore it to its original
>> state, if there's an error. The former will result in generational loss
>> (to the extent that it happens). The latter will not.
>
> DV tape equipment depends heavily on BOTH levels of error
> detection, correction, and mitigation. DV wouldn't work otherwise
> (or it would hold only 30 minutes of 100% reilable video where
> we now get 60 minutes of "very good" reproduction).
Are these hard numbers from somewhere, or offered only as an example?
>
>> The big difference, though, whereas generational loss is inevitable with
>> analog, loss due to drop outs or other uncorrectable errors is not
>> inevitable with digital.
>
> It actually is inevitable with digital also.
Well, we're just going in circles here. I'd like to see hard data that
quantifies the frequency of uncorrectable errors in digital video.
> Digital relies on the same
> analog technology for recording as analog video did. Extensive
> ECC in hard drives keeps this inevitable loss to a vanishingly low
> rate. And digital video tape uses mitigation to mask the higher
> rate of failure in DV recordings. But the mitigation is usually so
> good that we think we are geting 100% reliable data storage and
> recovery.
>
>> Okay, this gets into statistics. If you agree that, for a given digital
>> tape, drop out is not inevitable,
>
> We will have to agree to disagree at that point. No playback
> of digital tape is 100% accurate. Dropouts ARE inevitable, but
> the built-in ECC of DV makes us think we are geting 100%
> back.
>
>> then it is solely a question of the uncorrectable error rate for a
>> particular digital tape medium which, I would think, is incredibly low.
>> If you can copy a 60 minute tape to a computer and then back to another
>> tape without an error,
>
> You can't do that "without an error". You can do it well enough
> that most of us won't see a visible error, but guaranteed it is not
> 100%. But if it is close enough to 100%, you can go many
> generations before you start seeing the effects of all those
> corrected errors.
>
>> It also has to be more robust because (1) hard drives are constantly
>> re-written, (2) data is packed more closely on hard drives, and (3) data
>> is read more quickly on hard drives. The ECC on miniDV is sufficiently
>> robust that the likelihood of losing any data on a transfer is very, very
>> low.
>
> But not low enough to be "zero" (or even invisible) after 20
> generations of digital dubs.
I never said it was zero, as people in this ng have reported the occassional
data-losing dropout. Heck, cosmic rays could probably cause it, but how
often do cosmic rays effect video playback (answer: not often).
>
>>> The expectation
>>> for data is that you will get one uncorrectable error in several
>>> trillion bits whereas most of us likely have seen and/or heard
>>> uncorrectable errors in digital audio/video media in recent memory.
>>
>> As I said, they happen. Just not often enough so that they should be
>> considered a source of generational loss.
>
> Then we have come full circle back around to semantics.
It will stop being about semantics if the amount of uncorrectable errors,
i.e. those that result in data being lost, can be quantified.
>
>
Navigation:
[Reply to this message]
|