Matching digitized vinyl iTunes imports

Thanks. I do get a little wordy at times though. You'll be doing us all a service by running the experiments you propose. You might well uncover some things that are common to the "won't match" subject, regardless of source material. Good luck with it. I look forward to hearing what you discover - Jim

P.S. Not surprised you'd prefer your vinyl. While every medium has its pluses, certainly there are issues with a lot of my CDs that make me prefer the vinyl copy.

Message was edited by: JiminMissouri

I'm thinking the mastering itself is a large part of the process, here.

I imagine so. Your tests using a CD that seems to be the same as the vinyl one would think off the bat would have a much greater chance of their both matching and apparently in this case they did. In that context, getting songs originall recorded in 1956, re-issued by Verve probably sometime in the 70's to match iTunes copies is pretty amazing, but since I've done it, we know it's possible.

These tests are of interest. The beginning of the song seems to have more weight than the end of the song, at least in this case.

That's along the lines of what I expected. My logic's probably twisted, but I'm thinking the process starts by identifying what may be a match, then checking the first 10 seconds or so. But if it's not easy to find those first second, forget it. Actually may be a lot less than a second would kill it. You killed part of the track, but did you try one where the open was sloppy? I'm still left wondering how important getting the tag close to the beginning of the track is.

On the other hand, for those of us in vinyl, when have situations like I've had, where see the last track on a side doesn't make it, well think about the difference. Longer fade outs were customarily put there. Why? Because it was enough of a difference to say "the side is ending, you need to get off the couch and change the record." My guess is that long fadeout isn't there on a lot of CDs. If that's true on the Glen Campbell vinyl, there's something to check out.

Added heavy clipping (+10db), enough to alter the sound audibly. STILL matched, which was a big surprise to me.

It surprises me a bit too, but think about it. just how awful does an MP3 have to be before Apple deems it ineligible? In my opinion, pretty darned awful. This may also explain though, why some things I've gotten to match really weren't in that great of shape. Other things that were "right" (pitch, song start point) might have been enough to get it through.

Pitch change, one note up (5%). It became an upload, so this is one I'd definitely watch out for.

5% is a huge difference though. I wonder if it wouldn't be unusual for a poorly performing TT to be off that much. Up a full note? I'm thinking of how management at the station would have the engineers bump up the speed. It was done for two reasons (1) to help support the "more music" pitch (2) well our stuff just sounded more upbeat (which it was). Along the lines with your varying the length cut off either end, I wonder what a 1% bump would do?

Final thought: All of this is very, very interesting and you may have figured out some important aspects to matching that will help. However, if you kept a copy of all the files that you used for the test, wouldn't it make sense to verify your results in a few days, just put them all to the match process again? I know easy for me to say, since I'm not the one putting all the time into it! I'd love to get into it, but we got lucky today with weather warm enough to do a little trim painting outside. This place is a work in progress and some things need to be done when they can be done. Or it's wait until Spring.

Thanks for all you're doing. I'm enjoying this little journey. Hope you are too.

Message was edited by: JiminMissouri

OK, if we accept that a 5% change in pitch is something not everyone would hear, then I can see that it could result in some people not getting matches, scratching their heads, wondering why. If they were thinking that because it sounded "clean" it should match, when based on your results, "clean" is less of a factor than pitch. It reinforces the importance of being sure you're at speed before you record.

So your Madonna results are adding credence to the idea that sloppy tags that put too much dead air before the track could well be a factor.

While I suspect in most cases, pitch problems are a bigger deal than noise (or distortion) I do think overall S/N ratio is looking to be a big factor, particularly at the beginning of a track. Between my experiences with Joplin's Pearl, the Zepplin track and others, noise may be less of an issue if the beginning of the track is, hmm, call it a "wall of sound," but if you've got something that's more simple, say voice and one guitar, nothing else, or if you have a very slow fade up, the S/N ratio may be bad enough to kill it. Do a good job in ClickRepair and that might do the trick.

roebeet wrote:

Just did another test - this time with an old LP, Nat King Cole's "Sings for Two in Love". I picked this one because I think it's the oldest LP I've needledropped - mono pressing with the Capitol rainbow label, probably an early 1960's pressing if not earlier.

I really expected it to fail to match at all. but I actually got 11 out of 12 matched!

The most important thing you've proven here is that you've got very broad interests in music (Nat King Cole vs Madonna) and for that I salute you (and suggest we ought to be talking outside of this discussion).

That the older stuff matches at all may seem counterintuitive, but there may be something at play here that we've only sort of touched on - the type/complexity of the music. Consider the difference between a single vocal with fairly simple instrumental backup and say, a heavy metal track. Look at the waveforms and on one you're not going to have any trouble spotting the beat, sometimes catching where the artist is singing, where he's not. In know we're only human, but still which might be easier to match?

But we also both know all too well that the good vinyl from the 50's & 60's, if in good shape is amazing to hear and a joy to own. Really top flight recording was being done back then. The master cutter didn't have to try and cram 20 minutes on a side and that sure changed in the 70's. 180 gram wasn't something particularly special, it was just done. Wonderful "advancements" like "microgroove"or RCA's "dynaflex" sure didn't help matters much. So the truth is on the old stuff, we might well have vinyl that's more true to the digital copy than something pressed, say, during the arab oil embargo, when the cooties in recycled vinly had a lot of us returning records that just wouldn't track.

I'm recording a fresh copy of Pearl, will put it through all the steps we've discussed. I think my Audacity file on Brubeck's Time Out hasn't been toyed with, so I'll just pick up with it and run it through ClickRepair. I'm using 20-25. I'll let you know how it goes.

Message was edited by: JiminMissouri

Like I said, we need to talk, sir! Send me a PM sometime. I think my Apple account allows it, but if not let me know and we'll work something out. I say that because we seem to look at music much the same way and to me that's significant. I get a kick right now out of hearing Billie Holiday's "You've Changed" under a Chanel spot. Wonder how many people even know who they're listening to. I'm re-recording Lady In Satin right now (not her best, but it's what I've got) and if any tracks on the A side get matched I'll be amazed. It was so trashed I didn't even bother keeping the original Audacity file, let alone break out tracks. I've gotten ClickRepair since I first recorded it, so if it can work magic, great.

Monk/Mingus/Coltrane/Davis, well I know what you mean about finding mint copies. It's not that there aren't other artists deserving of the same interest, but you've hit on several of the hot names, so if a clean copy shows up, it's high dollar. I've got a few, but honestly based on visual grading, just put them on the shelf. I did get an excellent copy of a Miles Davis, but it was a later work, not one people would kill to get. But this is exactly why I'm looking to get good matches. If the file from my source disk can be made good enough for a match, I'll have access to music I'd really like to explore more. After all, I've got a row over two feet long of old jazz upstairs.

When I get a minute, I'll look you up over in the Hoffman forums. For what it's worth, anything over $3 is hard for me to even consider. Most of what I pick up in the Kansas City areas (which for me includes the small towns within 100 miles), is under a buck. I pass up a lot of buck stuff simply because I know if I'm patient I'll pay half that, or depending on the place, a tenth of that. I really do love hunting for bargains. That may be why I've got so many shelves full of stuff I've not even found time to catalog, let alone clean and play. 25k songs doesn't seem like much. 25k albums, well that's entirely different.

The MIles Davis I got was Amandla. I like it.

Think of it this way -- Shazam, Soundhound and others are doing their darnedest to find you the song being presented. The goal is 100% match -- with much less significance given to the accuracy of the match.

Apple's goal is to match as much as possible -- while giving the RIAA 100% assurance that Apple is not giving away music by mismatching. i.e. the goal is to have 100% accuracy on songs matched.

The difference is huge. Statistically speaking, a 100% confidence level is impossible, so Apple has probably convinced the RIAA (or paid them) to accept something less -- but probably still at least 99%.

Total conjecture, here, but when I start thinking about iTunes Match in this context, I start thinking that the time-distance between the start of the track and the start of the song might be the nost critical, as there are probably 2 or 3 waveform snippets randomly selected and compared to the waveforms of what might be a selection of songs that meet a lower standard of "possible" matches. But, in order to actually score a "Matched", the waveforms have to be extremely close. Clipping, might be overlooked, as long as the unclipped portion of the waveform matches, and clicks/pops might be able to be filtered out during the match process, but a .2 second difference in the start of the song might mean a completely different waveform being compared, and thus rejected.

But that would make 40-60% success rate in matching LP rips seem exceedingly high....

I don't know. Maybe I should spend my time ripping some music and trying to get the best I can out of it. Though I think I've learned something of how to do that, by reading the threads associated with it.

I am not a technical as you guys, but some old vinyls have a completely different post production mix than the current distributions of the same albums. For example, the very first Crosby Still and Nash album release in early 70's has a much better mix than the later CD's. If you compare original vinyl mix to the later CD's you would never choose to listen to the later digital releases . . they are terrible by comparison. I was wondering if iTunes-Match would identify my original Vinyl version of this 40 year old album.

Whether iTunes would find your CS&N album similar enough to match, you'd really have to try it to find out. Given it has for Roebeet, matched stereo Beatles albums to his monos, I'd say there is no guarantee that a different mix will in and of itself be enough to preclude a match.

Roebee'ts described in some detail how he's managed to get some tracks to force upload. Once a song has been matched, getting it out of the cloud and then getting an altered copy to uploaded instead can be done, but requires additional steps to remove and replace the track with altered copies. That's been described here as well.

If the CSN vinyl is actually a different mix than whatever is in iTunes (and not just a different master) then there's a higher probability that the vinyl rip will be an upload and not a match

I agree. I just wouldn't assume it'll be a upload. That said, if somebody wanted odds on it, given all the experience we've had, combine it being a different mix with it coming from vinyl and I'd say more likely to upload than to match.

A Match Question: I have folder full of ripped Burning Spear but they are only labeled as Track 1, Track 2 etc in iTunes.

All of the Tracks can be identified by Shazam on my phone. When I upload them to iTunes Match, delete them on my PC, then Download them back to my PC, they are still the original files, not the Matched AAC files. I manually labeled Song and Album but that did not help;. How does iTunes match a music file ??? If you guys can rip Vinyl records and get matched . . what is the difference ???

+2 on what Robeet says. Getting something to match in Shazam really isn't relevant, for reasons that have been discussed in other posts. For what it's worth, I had several "track 01" files that came from CD, but somehow all the metadata got screwed up. iTunes matched them anyway. After they matched, I used Tuneup on them and was able to restore the tags.

Correction: The description I gave for DC offset in my original post is totally wrong. In essence, if you look at a waveform and it's shifted off the null line, that's what DC offset does. So if you are looking at what is supposed to be silence and it's not on the null line - there you go.

Turns out I did not have this problem at all. The Macbook Pro's sound card is fine in this respect. What I thought was a DC offset problem turned out to be turntable rumble below 20Hz, most beloe 10Hz. For some reason the DC offset remove filter in Audacity was taking a lot of this out, which just added to my confusion. I'm currently using the Apple LowShelfFilter in Audacity to remove the low-end noise.

Also, my workflow has changed in the past month. I now record 96/24 in Audacity (though I understand I might not really be getting it). I run the LowShelffilter, then export a 24 bit AIFF, which I run through Clickrepair. The cleaned up file is moved back into Audacity, where I break it down into individual tracks. Those tracks get exported as as 24 bit AIFF, which are then added to the iTunes library, with iTunes Match turned off. I then create 256 AACs of the AIFF files, delete the AIFFs and turn on iTunes Match.

Just wanted to jump in here and say thank you to you guys for a very informative discussion. I recently started playing with iTunes match in the hope I could short circuit some of the time-consuming activities in Audacity when it comes to digitizing my vinyl.

So far I have digitized 16 albums and only 6 tracks matched (sorry didn't count the total tracks). I was hoping to be able to match because the sound quality of my rips (using a cheap USB turntable) are awful. The frequecy response is way off and the wow and flutter is very evident. So, picked up ots of toips from you guys there.

I now have my B&O turntable repaired so I'm hoping for betting quality rips, and matches, going forward. Interesting comments on the space at the beginning of the track, but if I have nothing to compare it to how do I know the optimum value? (i.e. in the space between tracks when does one track end and the next begin?).

Very interesting comments on the Audacity click filter, looks like I need to be trying ClickRepair - and TuneUp as well.

Thanks again!

Hey folks, I realize that this thread is kind of old, but I just encountered something I hoped y'all could help me with. The White Stripes's album Icky Thump is on the iTunes Store, and I have it, but Jack White's record label has a record club where you get limited vinyl every quarter, and their first issue included a MONO mix of the album Icky Thump. It only exists on vinyl and is fairly rare, so I was hoping to digitize it and have it upload to iTunes Match. The first two tracks matched, and the rest uploaded. I was hoping it would upload it all, because the things it matched to are the STEREO tracks from the iTunes Store, which i already have in my iTunes library.

How do I force iTunes Match to recognize these as tracks that are new, rather than tracks that the store already has?