Posted by PTravel on 01/12/73 11:52

"Richard Crowley" <richard.7.crowley@intel.com> wrote in message
news:e8v12a$jbg$1@news01.intel.com...
> "PTravel" wrote ...
>> Yep. Good examples of drop out that results in uncorrectable errors.
>> I've never, ever had that happen, and I've not heard reports from anyone
>> else saying its anything but highly unusual.
>
> I hope so. Unfortunately, I have seen three of them in the last
> two months. Keep your fingers crossed. :-(

Interestingly, half of the problems on that page were caused by clogged
heads (or so the website said). It goes back to what I said about
maintenance.

>
>> The question remains how often this happens. If it is so rare that it
>> happens once in 10,000 hours then, for all intents and purposes, it's
>> zero.
>
> You must have missed the reference provided by Mr. Heffels...
> http://www.creativevideo.co.uk/public/pdf/sony_dvcam_2005-6_inc_xdcam.pdf
> Page 4, middle of left column: "100 counts/minute" for DV.
> (50/minute for DVCAM). Having seen the RF envelope
> from the rotating heads on my oscilloscope, and the "eye-
> pattern" from the resulting bitstream, those numbers seem
> quite plausible. Likely 99.99% of these are completely and
> accurately correctable using ECC.

Actually, I did miss it but I've looked at it now. And, of course, it's not
an answer to my question, which is "how often does an uncorrectable error
occur"? You say, it's likely 99.99% of the time it's correctable. I say,
I haven't seen the statistic anywhere, so it might be 9% or it might be
99.99999999%.

>
> We have really done a good job of making digital look perfect
> from the POV of the end user. Most people really have no
> appreciation for the complexity of what is happening under
> the hood. (Or for their automobiles, either, for that matter.)

It does, particularly in comparison with, for example, Hi8 with which I
experienced multiple, visibile dropouts on each tape.

>
>> Then tell me what the uncorrectable error rate is for DV tape.
>
> There is no single number. It depends on a great many factors.
> For ANY media, analog or digital; audio, video, or data.
> Magnetic, optical, holographic, etc. For our DV recordings,
> many of these factors we can keep at bay by using new tapes,
> not switching brands, keeping everything clean, keeping the
> equipment properly aligned for interchange, avoiding the long-
> playing modes, etc.etc.etc.

Sorry, but that's a cop out. Again, the context of this discussion is
whether DV experiences generational loss. All your (and Martin's)
discussion has been theoretical and unquantified. This is no different than
Martin's mpeg2 rants, when he claims that mpeg is "capable" of better video
performance than DV-codec. Of course, since the discussion is always in the
context of whether a consumer DVD camcorder will produce better video than a
comparable miniDV camcorder, his answer, while technically correct, is
misleading and inapplicable to the question. Similarly, here, the question
was whether DV experiences generational loss in the context of making dubs
and as compared to analog video. Without quantifying the uncorrectable data
rate, claiming that "mitigated" data correction (for lack of a better term)
equates to generational loss is, while technically accurate, misleading and
inapplicable to the question.

If the uncorrectable data rate is sufficiently high, then describing the
effect as generational loss is accurate. If the rate is miniscule, then it
is not. Unless I know the rate, I can't tell which it is.

>
> If you held a gun to my head, I would guess that on average,
> under good conditions, there is an uncorrectable error every
> 10-30 minutes.

But what is that based on? If you held a gun to my head, I would guess
that, on average, they occur less than once every 200 hours of DV, since
that's what I've shot and I've never seen a perceivable drop out. Now, I
haven't examined all of it frame by frame, but I spend enough time editing
my projects so that I'd notice a signficant error -- I certainly have no
problem seeing them when I edit my old Hi8 tapes. Now, I understand that
mitigation routines are very effective, but you'd think I'd have seen
something.


> Fortunately most of them are so small (<8
> pixels) and so short (<2 frames) that we don't notice them.
> Its good enough for me. Sounds like it is good enough for
> you, also.

Okay, time for some math:

Number of pixels in 30 minutes of D-25:

720 x 480 x 29.97 x 30 = 310,728,960 pixels/30 minutes

Total error in 30 minutes (using your estimate):

8 pixels

Percentage error:

..00000025% (rounded up)

That means that 99.99999975% of the time, the data is accurate, i.e. a
bit-for-bit copy.


>
>

[Back to original message]