Slashdot Mirror
Ripping Vinyl Via Your Scanner?
An anonymous reader writes "This site describes a method of extracting audio off of scanned images of vinyl records. Kazaa vinyl swapping is on it's way!" While this method creates exceptionally noisy samples, you can definitely hear the underlying music.
Do /. editors actually edit? Probably not the first to notice, but it's spelt Vinyl. V-I-N-Y-L.
Not that hard, folks. Especially when you get it right in the headline.
Matthew G P Coe
http://mgpcoe.blogspot.com/
Comment removed based on user account deletion
this is really cool and all, but whats the point, if you really dont care at all about quality then this may be a quick option for converitng your collection, but if you still have vinyl you probabably care enough to plug your turntable into the audio in jack on you computer
"Sic Semper Tyrannosaurus Rex."
Comment removed based on user account deletion
No, of course not.
MP3s are not like-for-like copies of CDs, they're extremely lossy, and you only get a tenth of what's on the CD.. but.. you can still get busted for swapping them! I believe the copyright laws specify that any 'likeness' to which a third-party could associate with the original, is covered as such.
Ditto for music encoded within images, though this is a hoax.
mogorific carpentry experiments
A vinyl LP is 12 inches in diameter, and has a label area in the middle that's about 4 inches in diameter. So the area containing the spiral groove is about 4 inches wide. That's about 10 centimeters. An LP side typically has a little more than 20 minutes of music on it. It rotates at 33 1/3 RPM, so the groove spirals around roughly 667 times. So the width of the groove is roughly .01/667 meters, which is 150 microns. The signal (on a monaural record, stereo is more complicated!) is recorded by wiggling the groove from side to side in that 150 micron space. To reproduce a signal whose dynamic range is 90 dB, the smallest excursions have to be roughly 1/30000 of the maximum amplitude. 150/30000 microns is 5 nanometers.
Think your scanner has that much resolution? Guess again -- 1200 dpi is roughly 21 microns, off by a factor of 100.
Note that 5 nanometers is way smaller than the wavelength of visible light (roughly 750 to 350 nm), so those laser turntables everyone is talking about don't work very well either, unless they've got x-ray lasers in them.
-Tom Duff
After visiting a pdf on LP technology and finding that LPs have around 290 lines per inch, and 45s 160 or so, it would seem that a reasonable scanner (say 1200 dpi) would pickup 7 pixels per track on a 45. This would make horizontal (along the plane of the record) resolution quite poor. However, if he's only tracking the vertical component (perpendicular to the plane of the record), and the varying heights translate well into light to dark gradients, perhaps there would be enough information to produce some bad sound. I too would like to see the code and perhaps some of the source images.
Standard rotational speed = 33 1/3 RPM
12" record
Circumference = pi * D
33.3RPM /60 ~ 0.5 R/second
12" * pi ~ 37" circumference.
0.5 * 37" = 18.5"
18.5 * 600dpi = 11,100 samples per inch, which gives a Nyquist limit of 5550Hz... a 2400 dpi or better might actually give full audio bandwidth, though in this case, the higher the better, since the area available for sampling decreases towards the center of the record, and for really high fidelity sound, more than 2 samples at 20K are necessary.
His model for how the record was encoded is *wrong*. The RIAA method of stereo modulation (back when they were mostly a standards organization) places the amplitude information on each wall of the V-shaped groove. It is intended to be picked up with a stylus connected to a something in the form of an Y , with channel information picked up by coil or magnet or other means attached to each upper leg of the Y.
Fixing his model should result in drastically improved performance if he's extracting stereo information. Cleaning the record would also help a lot.
His project actually *is* worth doing. An optimized algorithm should allow anyone or a museum with a good scanner to turn his vinyl (SPELLED CORRECTLY) collection into decent quality Red Book or MP3 tracks without any further damage to the records. The basic problem is to linearize the relationship between 16-24 bit gray scale information of reflected light and the depth modulation in each groove.
The suggestion of using software to extract 3D information from the grooves posted elsewhere is a good idea, but this is a good start.
Cool hack.
Tech Public Policy stuff
Much of microscopy work, which this is, involves fooling with the illumination direction vs. the viewing direction. Getting that right is a big part of doing it at all. This guy had to scan the record in four quadrants to get some halfway reasonable result. Obviously, you'd like a rotational scan, like a turntable with a stationary scan arm. The amusing thing is that you could read an entire vinyl record in one rev. Now, at last, the 1000x LP player!
Incidentally, the recording system for stereo LPs is called "45-45 Westrex", because there are two perpendicular tracks recorded 90 degrees apart (at +45 and -45 from vertical). Mono records, which have no vertical component, are thus backwards compatible. If all you can read is the horizontal component, you get a valid mono signal.
Tool's Opiate EP has a similar thing. There's a bonus track, but you have to put the needle down in the right groove. Pretty cool.
"If he thinks he can hide and run from the United States and our allies, he's sorely mistaken." Bush on bin Laden
Enjoy
http://arts.ucsc.edu/ems/music/tech_background/TE- 19/teces_19.html contains basic information on how the LP record works. I think the most important thing for the experimenter is called RIAA equalization, in order to limit the physical motion of the recording stylus that cut the record, bass was reduced and treble increased in a very precise way, in order to reproduce the original sound, the opposite must be done.
The RIAA equalization curve is a plot of amplitude boost/cut vs. frequency. Apply its inverse to the raw analog signal(s) that come out of your signal processing.
You can find it at http://www.tanker.se/lidstrom/riaa.htm.
Oh, and CLEAN THE RECORD BEFORE DOING THIS. The info in Part 14 of the rec.audio.* FAQ is as good a place to start to find out how as any.
Have fun and feel free to let me know if you get anywhere.
You might also want a look at my other post to this thread.
Tech Public Policy stuff
I don't really see the point, needles sound great, they don't scratch the record if you take care, and I've never heard of records getting deformed through contact with the needle. Even if they do, it seems to me lasers would be worse.
Slashdot: providing anti-social weirdos a soapbox, since 1997.
Some clarifications:
I am sorry so many of you thought this page was a hoax only
because no source code was supplied (I'm sure you'll all agree, now that
you can see the code, that it is both straightforward and crappy).
I guess I didn't do enough on the actual explanation side either.
The whole thing was done in a couple of late nights so I didn't really
have much time to gather all the technical details concerning phonograph
modulations. Moreover the "archeological" reverse-engineering aspect was part
of the fun.
I now know (thanks to some great replies) that the horizontal modulation (the only
one I did decode) is not a whole channel in itself but merely a delta between
the h-modulation and the depth-modulation which I did _not_ decode.
Some repliers seemed to be a tad confused as to what recordings were
the actual decodings. I'd like to stress that gramophone3.mp3 is a recording
while the rest (dneedle*) were decoded from the image.
Have fun,
Ofer Springer
The guy goes on to say;
Alison
"It is a miracle that curiosity survives formal education." - Albert Einstein