MiniDV tape: Not just like a computer file

[David Grove]

I used to think that a MIniDV tape was like a computer file: i.e., a sequence of bits that, like a computer file, transferred bit-for-bit from the camcorder to the destination file on a computer.

I found a web site that would seem to dispute this, and offers strong evidence. Unfortunately, the site seems to be gone, and the only "residual" is Google's cache. Unfortunately, the video clips aren't available in the Google cache, but the text of the page is.

Maybe I was just naieve, and everyone else knows this. I'm interested in reaction from people who really know DV. (That leaves out folks like me, who just play at it as a hobby).

Basically, the assertion, supported with evidence, is that you don't really necessarily get the exact bit-perfect bit stream, when transferring from camcorder to computer file. The web site author did experiments transferring the same miniDV tape multiple times, and compared the files. Not bit-for-bit identical. It seems ECC in the MiniDV format is not quite sufficient to provide an exact file copy operation. Of course, once you get the computer file created, further copies are perfect (because it then is a computer file).

Here is the Google Cache link:


MiniDV Errors


Comments, anyone?

Thank you.

DG

[Shaun Roemich]

Well, without commenting on whether there ARE differences in captures normally, the first thing that jumps out at me is just how unscientific the approach to the comparison is:
The banding obvious in the images at the bottom of the page indicates there is some "talk through" of previously recorded material on the tape. This IMHO makes the "test" completely null and void based on the fact that the recording, digital or analog, to tape is based on magnetically exciting bits of magnetic particles that reside in the medium. It is my OPINION that by recording over a pre-existing track, it is certainly possible that on separate passes of ingest, the deck COULD be reading the magnetic information slightly differently due to minute differences from one pass to the next in the gaussian (or magnetic) strength of the signals (plural) that had been committed to tape.

Am I asserting that DV captures are bit for bit accurate every time? No - I'm avoiding that discussion completely EXCEPT to say (as above) the tape recording are prone to failings and changes (especially over time and with subsequent passes) in the magnetic recording MEDIUM itself.

Very interesting find, though.

[David Grove]

Thank you for your thoughts.

I am not a video expert, and don't have any grounds to disagree with you.

But, I'm just thinking... Regardless of whether the tape has been reused, or not, shouldn't it be just a stream of bits? And, either the bits are read accurately, or not. If sometimes a bit is read as a "0" and other times the same bit is read as a "1", then the transfer is not accurate.

OK, nothing is absolutely, perfectly accurate over a sufficiently long time. I get this. But, it just appeared to me that there might be a significant difference between transfers ("capture") from MiniDV tape to a computer, as compared to a file copy (or FTP or UUCP or similar) on a computer. I mean I bet I could read a tape backup 10 times in a row (or more) and get, bit-for-bit the same file off it every time. But, apparently this is not the case for a MiniDV transfer.

I wonder if there is, or could be, something like EAC for DV tapes.

On the other hand, in the practical, real world, I guess it really doesn't matter. Just an anal-retentive thing, I guess :)

Thank you again, for your knowledgable comments.

DG

Edit: When I speak of "tape backup" above, I do not mean MiniDV cartridge. I mean any form of historically used computer file backup (archiving). Whether small cartridges, of various forms, or ("antique") 9-track. I think I would get the exact same sequence of bits, every time, regardless of whether the tape was previously used.

[Giroud Francois]

digital tape are supposed to to give you back exactly what you put on it since many error correction scheme are involved. The fact is that DV is supposed to be like this, while HDV is not and they use the same tape.
HDV is based on a lossy transport system (mpeg2.ts) designed to be used for broadcast (cable, aerial, network) where the main purpose it to make sure you receive the stream even if part of it are missing or corrupted. CD audio are also like this. The wav file is a binary stuff and suffer no corruption, but the support (CD) is made to support some physical degradation , while still delivering data, and another step is the player, that should be able to compensate for drops the best he can.
So it could happens that some corrupted data could not be used by some device (like computer that are pretty finicky about data) where other devices would play without a visible glitch.
Think about DVDs that you can often play on your standalone player but not copy on you PC because it is unable to read some sector.
Thera are many cases where people transferred HDV file that looks ok on disks or even playable with mediaplayer and just crash a NLE when you try to edit them, because they are slightly corrupted.

[Paul R Johnson]

Any helical scan recording mechanism won't give a bit for bit reproduction of what went in. I'm not certain that that ever was claimed for either digital audio or video. The error correction system was designed to produce a stable output even with significant errors. These errors are produced by either loss of head contact, or head contact with a non-magnetic surface. In practice, we're talking dropout - where the actual magnetisable surface is missing or damage. The gap in the data stream gets filled by the error correction system so data is continuous, but not authentic. With helical scan, there is also the switching issue as one head leaves the tape and the next takes over. With an incorrectly set up machine the switching point occurs in the wrong point - and with analogue we got the familiar band at the bottom of the screen. With digital, the error correction tries to 'repair' this too. As few recorders or players actually tell you how many errors are being corrected - you simply don't know how far away from the original you are. I've still got my Panasonic DAT audio recorders, and they were scary! You could prod a button combination and the display would tell you how many errors were being corrected each second. 44.1KHz sampling and the error correction let up to around 100 each second be repaired - and the tapes would play. I remember sitting in the studio, mesmerised by this display - seeing the errors beginning to creep up as the tape started to wear. It did give a bit of warning when the heads could do with a clean - and then the error count would drop, never to zero, but lower. Once head cleaning stopped improvement, then it was a worn out tape, or worse - worn out heads.

If you bring in a DV tape into your editor, and then dump it straight back to tape - the visible result is identical - assuming the errors the original contained were within the correction window.

You now have a backup that functions - BUT it is not identical, it contains (hopefully) intact data, probably repaired, and plays with no dropout. The actual contents aren't exactly the same. If you do this repeatedly, the artefacts start to creep in where the critical data is more corrected than original - but exactly what the errors look like is down to luck.

Has any manufacturer or web article suggested digital copies are identical? I've missed this.

[Shaun Roemich]

Originally Posted by David Grove

But, I'm just thinking... Regardless of whether the tape has been reused, or not, shouldn't it be just a stream of bits? And, either the bits are read accurately, or not. If sometimes a bit is read as a "0" and other times the same bit is read as a "1", then the transfer is not accurate.

This is exactly what I am saying - BECAUSE of the obvious banding that indicates Print Through, the signal is compromised - I suggest at the magnetic level - due to there having been something there before that is interfering. Could be a head alignment issue, could be one set of the four heads on the head drum assembly is laying down a less magnetic impression than the others due to residue (or "clogging")... just saying the second I saw the OBVIOUS banding and Print Through artifacting, I would have cleaned heads, used fresh tape and tried again.

And Paul did a wonderful job of explaining what I fudged my way through. BetacamSX studio decks actually have a front panel display that shows errors (normally 3 lights - green, yellow and red if memory serves) to indicate when tapes are starting to generate read errors. It's the magnetic surface that is suspect IMHO, not the integrity of what was written that is generating errors.

[Shaun Roemich]

Analogy time:
A BOOK printed by a high quality printing press SHOULD contain every single word that was inscribed on the plates that contact the ink that then contact the paper.

It is possible that after many impressions, the plates could warp or wear, causing sections of the plate to not contact the paper, leaving "dropouts". It is also possible that ink MIGHT not have fully filled every letter, leaving dropouts. Perhaps there was an imperfection in the paper itself.

As well, after owning said book for many years, leaving it open on my nightstand every night, I may have spilled water onto it and closed it up, causing pages to stick together and tear slightly when opened, removing several words.

Does this invalidate the statement that the original plate contained all the words of the book? No. The issue is with the medium.

Make sense? Or do I need more (or possibly LESS) coffee?

[Giroud Francois]

Any helical scan recording mechanism won't give a bit for bit reproduction of what went in. I'm not certain that that ever was claimed for either digital audio or video. The error correction system was designed to produce a stable output even with significant errors

Sorry that is wrong, many helical systems are used for computer backup and they give you bit for bit reproduction (or they would be useless).

Again is not the media or the way it is encoded that counts, it is the way it has been designed.

Some system are designed lossy , other not and they use the same technology or media.

[Duncan Craig]

An interesting thread so far, I just wanted to throw a few things into the mix, just from casual observation.

Firstly, digital system have error correction and parity bits and all sort of stuff to keep a weak signal from falling apart.

Secondly, assuming that tape stock can suffer from tiny dropouts due to it's magnetic nature, surely so can mechanical hard drives. But you wouldn't be so concerned about the integrity of copied from hard drive to hard drive would you? If you were, you would run a CRC check, but you would rarely if ever find an error, I'd hope the same can said of DV captures.

Lastly, DV is a compressed system, perhaps the tiny data discrepancies seen in different captures are from the decompression used at playback in the NLE?

Those are my initial thoughts, without spending too much time looking into it all.

Cheers folks.

[Duncan Craig]

Strange I posted before Giroud, but it's shows afterwards. Anyway, he's said it better than me.

[Andrew Smith]

Actually, it is possible to record / backup your files to DV tape. Just didn't seem to take off, though.

From waaay back in 2005 ... Backup Files to DV Camera

Andrew

[Kent Frost]

Any digital system which uses components that have to physically touch each other and move (as opposed to simply using electrical pulses through conductive materials) will inevitably give less than 100% copy results. It's not that it would read a 0 as a 1, but rather, it wouldn't register at all, and thus causing blocky tracking issues that those of us who do/have use(d) MiniDV are familiar with.

The reason backup systems work so well is because you're not devoting multiple sections of the tape for different tracks. You get a much wider space on the tape to record those bits, making them more efficient and error free until the tapes wear out. Also, I doubt there are many people taking their tape drives out in the field as one would with their camcorders, resulting in less debris such as dirt, dust, sand, etc.

[John Craddock]

The reason DVCAM tape exists is that it has a 4x to 5x improvement in errors and dropouts over traditional DV (and that still doesn't make it perfect). These errors occur on playback as well as capture, so when there are errors and dropouts on playback that the error correction in the deck/camera/etc. are compensating for, logic dictates that s there would not be a bit-for-bit match in the resulting files after multiple captures. This does not mean that the data is not being copied correctly or even that the image would be visibly corrupted, it means that there is data that couldn't be read, but that it was within the error tolerance of the device, so it made something up instead.