Sunday, September 27, 2015

iTunes Error Correction

I have always kept the iTunes "Use error correction" box set when importing CD's.  I have been unsure of how well this works, though I also use a TOTL external Plextor Plexiwriter Premium which has the most advanced features including error correction--in this case meaning that a track is read over and over until the internal validity checks show no hard errors.  I believe this combination of iTunes on Mac with "Use error correction" and a drive capable of re-read such as the Plextor gives you accurate reads or at least it tries hard to give you accurate reads.

Most discs read without a problem, especially recently.  I also manually read every disc twice and diff the files.  I have scripts which do this quickly for an entire album read twice.  People have asked: are you sure you aren't just getting the same garbage twice?  But I believe in most cases the answer is no.  This "twice reading" is exactly what the computer drives themselves do, and special high accuracy reading programs too (though nowadays the most special programs also check against online databases).

Today I had some kind of near proof.  While reading in an old disc, the Plextor was buzzing more than usual.  And it seems reads were not at the highest possible speed, only around 15x.  Finally, on the last track, the high speed scanning stopped near the end of the track.  From that point onward, it read very slowly, at something like single speed (during that time speed was not reported).  It did finish the track, however, and I ultimately read in the CD twice completely.

Sure enough, I got differences between the last two readings of the last track.

This shows that iTunes error correction does not guarantee a correct read, though it seems like with the Plextor it did seem to try (if that is what it was doing).  But in spite of best efforts, if a correct high speed read cannot be done, it goes with the best slower speed re-read it can do.  And I have one more datapoint that when this was obviously happening, my difference method caught it too.  I have caught other differences like this too, and it seems whenever the drive is going through fits like this I get differences, which isn't proof that my method works but it does seem like it is.

I read this disc many times with and without error correction.  Not once did the diffs for the last track match according to script.  When at I disabled "Use Error Correction" it read the last track straight through without slowing down, and all three times.  When I switched Error Correction back on, it continued reading the last track straight through for the next few reads, then went back to slowing down on the last track.

Doing this more than a dozen times, the last track checksum value did match twice, but only in the "second read" I discarded before getting the next one because my scripts work that way.  Really what the script should do is record checksum scores and keep all reads until you get two having the same checksum.  But once I saw the matching checksum, I couldn't get that same matching checksum in the next 4 reads and gave up.

All this tends to confirm my reading strategy of diff'ing files helps catch errors.  At least in some cases, the "fill-in" data after reading errors usually doesn't match in cases like this where they don't match once.  It's still possible in cases with different kind of errors they would always match, though I have no evidence for that.

MQA Reconsidered

Having read the introductory articles in The Absolute Sound, I like the look of the thinking behind MQA, Meridian Quality Assured.

In it's full encoding and decoding, when possible from an analog source, it preserves fantastic temporal response, so that tiny pulses are tiny pulses without the pre and post echo.

It does this by using more than one sampling rate, with the higher sampling rates having fewer bits because the greater resolution is not needed.

I was thinking about that sort of thing in my previous 'information" arguments against DSD and Delta Sigma in general.  Only a few times, if ever, I have noted my reservation that most of this "information" that DSD (and to a lesser degree, multibit delta sigma modulators) is at barely audible if at all frequencies.  So maybe I've been making a mountain out of a molehill.  I haven't known how to account for this before, but Meridian discusses this very issue (in their terms) showing that the information needed is a diagonal curve.

DSD's errors fall roughly in this range, but as I have noted before, the use of noise-shifting means the upper frequencies aren't really linear and accurate.

Meridian can achieve better quality, fully up to the required diagonal, in fewer bits than 44.k uncompressed.  Better than HD quality such as 192/24 they say…since they have better pulse accuracy still.

It sounds possible.  It certainly sounds better to me than DSD, and likely my older secret decoder ring, HDCD.

I'm a big fan of HDCD and I've always sought out HDCD versions and HDCD compatible players.  I have a way of transcoding HDCD into 24/96 by analog resampling, which sounds better to me than digital conversion, and I've heard experts say that.

While I was willing to pay the nickel for better sound, I wondered if others would balk (and, well, they did in the sense that HDCD became a very small player in the market) to pay the nickel.  Then, if you didn't have HDCD capability, you'd get a version that not only had less resolution than the encoded version, it might have an altered dynamic envelope.  So, ,in effect, a low-fi version in the unquestionably audible way.  That in effect makes it a secret decoder ring you must have to get the real music, which can be seen as  a kind of extortion.

Of course Dolby, Dbx, and other encoding schemes have been based upon those same principles.

And regards compatible SACD's…there has been much speculation that the CD layer is a different mastering, or dumbed down from true 16 bit quality.

And this even gets to audiophile releases…in many cases the main reason why they may sound different is different mastering.  And so too with LP's.

So ALL these are using the same extortive system as Meridian would be.

It's hard to pull these things off.  Dolby dominated cassette, though I was not happy with most dolby playback (except Nakamichi), done properly, it gave a nicer recording, though sometimes I preferred making my own recordings without dolby, especially if meant to be played on non-Nakamichi cassette decks.  I felt it was a useful but highly flawed system.  Actually back in the heyday, I hated Dolby B.  Now I'm inclined to use it consistently on my Nakamichi RX-505.

Elsewhere, I don't think Dolby was as successful, but I heard the claim they put a lot of pressure on producers to use Dolby products when the producers didn't feel it improved the products, or preferred others, such as Dbx.

I was glad that Meridian won the rights for DVD-Audio with MLP knowing that Meridian is, well I've generally had more respect for their thinking than Dolby.  I heard that DVD-Audio was held up by a battle between Dolby and Meridian.  I'm glad Meridian won.

And I glad and very surprised to hear that the "Dolby True HD" audio on Blu Ray is basically MLP.

So know the geniuses at MLP have given us a really really true HD format.  If it's anywhere as good as it sounded in the pages of The Abolute Sound, I wish them luck, and I am interested in getting the magic.

Sadly everything in my main systems must be converted to 44.1 to 96 kHz.  So I'm not sure I could even approach using MQA as intended, but I can get a version downsized to 24/96.  Or I can resample from the analog outputs, as I do with SACD, HDCD, and mostly DVD-Audio.