audio: record scanning not a fantasy

It turns out my armchair fantasy of digitizing my records by visual scanning rather than dragging a stylus through the groove has been done by researchers, including project IRENE, the visualaudio project, and other attempts. Carl Haber at IRENE in a long interesting presentation (80MB! PDF) says

We study the use of new, optical measuring and image processing methods to create a digital representation of the complete record surface, on the computer, and then “play” it with a virtual needle.

He points out in another paper that a 10" side becomes a 300 MB processed image scanned in 2D (so an LP might be a few GB), and, just as I perceived, that “[the process] can archive images for future re-analysis with new algorithms.” Awesome, ship it!

Alas, the sound-to-noise ratio is poor. It's a vital technique for non-destructive archiving of sounds from rotting wax cylinders, cracked shellac transcriptions, and broken 78s, but doesn't come close to top-notch vinyl playback. The Swiss visualaudio researchers comment

Reaching the same quality as with a good 33rpm record played on a modern turntable probably is probably utopic.

Oh noooes, say it ain't so.

audio: thoughts on digitizing vinyl

I have quite a lot of vinyl records. I play them on a Rega Planar 3, a fairly high-end turntable.

I'm not dogmatic about analog vs. digital. My well-recorded CDs sound great, though the loudness war means many are tiring to listen to. And many of my vinyl records sound fantastic, though some are warped or have irritating pops.

The thing about vinyl playback is the headroom for improvement is vast. Decent CD players sound very similar, but a better turntable solves real engineering problems and sounds much better than lesser turntables.

A big motivation to get a better turntable is to make better digital copies of my albums and singles. The digital library I cart around on music phones and PCs currently only has songs from my CDs

On that subject, I was struck by this review of a super-exotic Rockport Sirius III turntable:

I made two demo CD-Rs of various tracks using the Rockport, to A/B with the LPs in real time. Of course, the "live" LPs creamed the CD-R, which sounded slightly brighter and edgier but less immediate. Nonetheless, the CD-Rs did capture the Rockport's essence.

When I A/B'd the CD-R with the "live" LPs on the Yorke [a merely good turntable], the CD-R topped the LPs in overall presentation, dynamics, and especially solidity.

In other words, a digital recording from a super-exotic turntable sounds better than vinyl playback from a merely good turntable.

So, do I really want a $5,000 turntable? Not when there's someone with the same vinyl who plays it through a cartridge hand-made by the 90-year old sensei, mounted to a gazillion-dollar 1000 kg turntable hand-calibrated by its engineer, in an isolated underground vault, connected to the discontinued $25,000 Boulder phono preamp — I'll take his digital files instead of playing my own records!

The next issue is what kind of digital file to make. MP3 isn't enough, even FLAC lossless is only CD quality. I want future-proof. The consensus is zeroing in on DSD. The latest Absolute Sound has a review of the Korg DSD recorder capturing vinyl playback and makes the points of how close the DSD sounds to the original and how it's the best format for archival. Again, I want someone else's superb digitization of their vinyl. Her archival file will be better than mine.

(This still locks in the playback of the vinyl record on a particular turntable, cartridge, arm, and digitizer. What about future improvements to vinyl reproduction? So my most radical thought is to capture the record and read it in software. It should be possible to make a detailed scan of the record surface itself. Then software can reconstruct the audio waveform without the primitive mechanical operation of dragging a diamond along an undulating spiral valley. As long as the initial scan is high enough resolution, software can do a better job knowing what kind of cutting head made the vinyl, how vinyl deforms when stamped, where best to read the undulations in the walls. I have no idea if a 100 square inch microscopic 2½-D scan is remotely practical.)

Returning to reality, if the record companies were smart (OK, that's still unreality), then they would meet the demand for very high quality digital copies. From a rant

My concern is that we get every possible historical recording archived to DSD as soon as possible, AND that the artists (both performance and recording) can and should make DSD the authoring medium of choice, REGARDLESS of what happens to it subsequently.

Definitely! The DSD file becomes the fabled "master tape" that is the source for every released version of the music. But unlike a reel-to-reel tape, the music's owner can sell me the master DSD file, charging me more than a $0.99 MP3.

Here's Pete Townsend of the Who on DSD, from a great interview about recording music:

Genex DSD [was] what my Mastering Supervisor Jon Astley preferred. I preferred the sound of analog tape (1/2", Dolby SR at 15 ips) but they sounded so close it was almost impossible to tell the difference. ... It’s hard to tell whether going to tape would have produced better sound on CD. A CD is pretty difficult thing to get to sound “warm” (whatever that means, such a hard word to define in audio terms).

So the ultimate sound would be to play Pete Townsend's analog master tape. But even an analog fan like Pete Townsend says the original DSD file sounds fantastic, then gets mucked up when you turn it into a CD. So sell us that master file! 2L is one label offering this, as a test: a DXD file, as well as several lesser formats.