Converting vinyl to digital, watch the tone arm


After going through the VHS to digital process, which I lamented earlier I started wondering about the state of digitizing old vinyl albums and tapes is.

There are a few turntable/cd-writer combinations out there, but like most people today, I'm interested in the convenience of compressed digital audio which means I don't want to burn to CDs at all, and nor would I want to burn to 70 minute CDs I have to change all the time just so I can compress later. But all this means I am probably not looking for audiophile quality, or I wouldn't be making MP3s at all. (I might be making FLACs or sampling at a high rate, I suppose.)

What I would want is convenience and low price. Because if I have to spend $500 I probably would be better off buying my favourite 500 tracks at online music stores, which is much more convenient. (And of course, there is the argument over whether I should have to re-buy music I already own, but that's another story. Some in the RIAA don't even think I should be able to digitize my vinyl.)

For around $100 you can also get a "USB turntable." I don't have one yet, but the low end ones are very simple -- a basic turntable with a USB sound chip in it. They just have you record into Audacity. Nothing very fancy. But I feel this is missing something.

Just as the VHS/DVD combo is able to make use of information like knowing the tape speed and length, detecting index marks and blank tape, so should our album recorder. It should have a simple sensor on the tone arm to see as it moves over the album (for example a disk on the axis of the arm with rings of very fine lines and an optical sensor.) It should be able to tell us when the album starts, when it ends, and also detect those 2-second long periods between tracks when the tone arm is suddenly moving inward much faster than it normally is. Because that's a far better way to break the album into tracks than silence detection. (Of course, you can also use CDDB/Freedb to get track lengths, but they are never perfect so the use of this, net data and silence detection should get you perfect track splits.) It would also detect skips and repeats this way. You could also do this with a small LED and optical sensor on the cartridge, if that can be non-standard. The inter-track area is of course very visible to our eyes, it's how we are supposed to find tracks by eye. I could also imagine such a sensor having an inside edge on detecting scratches. (Having out of band information is very useful in elimination of scratches and pops. That's because a general filter for these will distort the audio, so you want to apply it only over sections that you are pretty sure need it.)

The other thing I want is something I had as a kid -- a record changer. On the record changer you could stack 6 albums, and it would play them in sequence. Most of these came with very cheap turntables and cheap electronics, so they were totally disdained by audiophiles, but there is no reason you could not combine them with a decent cartridge, arm and turntable. That's what I want. To put 6 albums on, come back in one hour, flip the stack of six over and put it on the changer, and get the other six sides. (Except on Monty Python's Matching Tie and Handkerchief. Boy were we shocked the first time we discovered the gimmick on that.)

Come back in one hour for six sides? Yes, because we'll be playing all these albums at 78 RPM, or frankly as fast as you can safely play them without the cartridge skipping off (something it should detect by the way) and the digitizer should be good enough to sample at 44K samples per second of regular speed audio. (Oh, even 96K samples, which sound cards can do, at 78RPM gives you 41,025 samples/second which good resamplers can convert to the CD rate.)

Perhaps I'm asking too much, but it would be cool if the player noticed the arm did fly off or skip, and go back and retry at a slower speed. A basic retry of the whole album is easy. Being able to seek back in the album to retry requires a bit fancier mechanical control and would increase the cost. Indeed, since the changer gives me plenty of time (it may be simpler to view this as an overnight thing, so you have 8 hours to digitize 2 hours of music on 6 sides) it may make more sense to forget about the 78RPM mode and just do it super slow, which could produce better quality with less skipping. Both modes may make sense.

Now I say six sides because this was typical for the changers. In reality perhaps we can do much better, though after a while the weight gets high. But if the cheap-ass changers we had in the 60s could do it, I don't see a problem today. Of course if you could duplicate a jukebox mechanism and put large numbers of albums in, it would be great. In that case it might make more sense to make it expensive and rented rather than cheap and sold to everybody.

It should be possible just from the ratio of track lengths to find the closest matches to a set of track lengths in the CDDB database and provide an easy menu to identify albums. Or even a single pixel optical scanner in the tone arm could, if lifted, scan the label in the center of the album to provide an image and even do some pattern matching and OCR.

Other parts of the turntable can be very cheap. For example, we need not care a great deal about wow and flutter. If we have dots on the outside of the platter (similar to those used for strobe measurement) and an optical sensor, we can measure the exact speed of the platter very precisely at any given time. With that (and a high sample rate) we can resample dynamically (or adjust the sample clock in real time) to produce a result as though we had the most expensive, stable turntable in the world. We could get a result better than you could hear from the turntable in analog mode.

I know there's been research on optical cartridges that use a laser to read the physical bumps in the tracks, instead of a needle. There are even some very expensive laser turntables, though they are very sensitive to dirt in the grooves. That's great, and if it becomes cheap, I want it. But to make this box cheap, standard cartridges are the way. Let's face it, I'm not going to be playing any of these albums again, I would even tolerate it if I got a good recording and it made a little damage -- other than my desire for retries. But I would venture that most of the money would go into the cartridge and the tone arm. A2D at this sample rate are cheap, as are the optical sensors, and as were the record changer mechanicals.

(You can get record changers today for about $150, but they come with pretty low grade components. I would be curious as to how tolerable they are.)


Having fscked about with this for quite a while, and seen how much hassle it is for audiophiles who care a lot more than I do, I suggest giving up while you're ahead. My approach when I finally got down to a small number of LPs that couldn't be found either in shops or online was to pay one of said audiophiles to rip the last 10 or so for me. Much cheaper overall than the time and hassle involved in ripping them all myself (I suspect the record companies are counting on this... and hoping we don't just download the illegally instead).

Actually, the best bit was tracking down and emailing a couple of the artists and in one case getting a CD directly with everything he had from the band he used to be in. Years later the album was released with 15 tracks instead of the 24 that I have :) The other group decided to put out a "when we were young" CD after all, since it's pretty cheap these days even with remastering costs. Both times I was happy to pay the artists for the CD.

Of great benefit there has been the recent Flying Nun retrospective collections and a couple of other compilation projects, because now I even have music videos from the olden days (New Zealand music from the 80s). Hopefully the more luddite record labels will eventually realise that this is a cheap source of profit too (remaster to CD, sell double CDs for $20... it's like printing money).

Which does limit how much money can be put into products like this. I just like to figure out how you could make it good if you were building it.

Though indeed, a lot of it is not on CD. I taped about 30 of my albums onto VHS-HI-FI in the 80s. VHS HI-FI is actually pretty good quality (much better than cassette but not reel to reel) but the problem is getting a different VHS unit to track properly and not gum up the audio.

Also interesting is the ability to, with the combo of wow and flutter elimination, click and pop removal, a high sample rate and FLAC compression to make a recording that sounds better than what you can get from a turntable, if you had an audiophile cartridge and A2D. There are those who disdain CD with 44K 16 bit samples and certainly disdain lossy compressed music, but they should not disdain 96K sampled 24bit or floating point sampling with lossless or much less lossy compression. With disk space costing 25 cents/gigabyte today, the economics that governed the choices of CD are no longer present.

The technique I describe for removing wow and flutter could even be done without resampling, if you could have the clock on the A2D adjust in real time based on signals coming from the optical stripes on the platter. It might be able to make an audiophile digital copy with inexpensive equipment. Though of course, you never make cheap audiophile equipment!

With my VHS to DVD transfers, there was a lot to be said for the "one touch" mode of the box (which really was far more than one touch.) Though I did decide, in the case of various old movies and TV shows I had on VHS, that I would just buy such things using amazon or peerflix if I really wanted to see them again, because the quality would be so much better as well. So I actually restricted my transfers to personal stuff, stuff you can't get on DVD, and stuff I only cared modestly about. There was a temptation which I should have given into to use the overnight one-button xfer to DVD+RW as a means to simply find out what was on a VHS more conveniently. I had a ton of old VHS tapes, 6 or 8 hours long, full of random shows, unmarked. For some of these I may do this technique. Since doing 6-8 hours means going to 352x240 resolution, it would mean if I actually really wanted something I would have to seek to it and re-do it. I suppose that might be a clever mode for the recorder -- you seek to a spot on the DVD and say, "I want this piece of the VHS for 30 minutes" even without index marks, and it could do it.

No argument about the desirability and technical possibilities of ripping vinyl that way, and if I had the time I'd possibly be interested just as a geek project. But I suspect it wouldn't be quite as simple as we'd hope (having perfect pitch means disliking cheap CD players just on general principles).

Speaking of cool ideas, when is someone going to release a chainable NAS box? This hassle of spanning my stuff across multiple drives is annoying, it's like DOS all over again. Every time I fill up the array I have to split stuff to free space on the full unit and move it to the new unit. Why can't I just plug them together and say "ok, you are one drive now, all 8 or 12 or 20... hard disks worth of you". Well, for less than $50k, anyway. It can't be that hard, especially since half of them are running linux and software RAID anyway... ideally getting a RAID6 6+2 array out of it, but 2 RAID5 with JBOD NAS boxes would be acceptable. It might even be possible to homebrew using eSATA NAS boxes and motherboard JBOD... fragile though.

There's a nasty tradeoff between having it all look like a single disk and letting the user see the disk structure. Disks fail, or go into pre-fail state, and you want to be able to pull them and replace them. You also want to be able to move them sometimes. In addition, you don't want runaway applications that fill up the disk to stop other critical apps from working, like getting E-mail.

For a giant archive, one big volume may be fine.

There is, of course, another way to do the vinyl job, the way did it before they were sued out of existence for doing it. Namely you prove you have the vinyl and you get MP3s or Flacs back. It's not legal, but the studios could do it themselves, and there was even a program for a while where you could bring in your old disks to record stores and get replacements for a discount.

The latter method solves the problem the RIAA worries about of people borrowing a friend's disk and thus getting free music, in that the store takes the vinyl or marks on the label with a pen. But it only would work for stuff that has been digitized already, and only for studios that would participate. And right now they would rather sell you your music again at full price (to fewer people) than let you have it at a discount.

I would think an image of the disc at a very high resolution and an algorithm to translate the dics image is a .wav file. Like the laser/needle, but just a complete image and process it into a sound file in one readthrough.

This would be a new level of very high resolution, I think. Distilled to CD-quality, a disk has 220MB of data, but as I said, that's distilled from the middle of the groves. An image has lots of extra information.

There is a fellow who tried this:

And this:

(You can google for more hits on this, people have done a bunch of work.)

Near the end of the record, we see about 7 inches per second. I figure your image would need immense resolution because the height of the groove isn't something that will translate that well to light intensity, but perhaps something could be done. If you needed just 200,000 samples/second, that's 28,000 dpi, beyond the range of any scanner we have around. You could get 44,000 samples/second from a 6400dpi scan going perpendicular to the grooves. For this scan you would scan inward very slowly, as you would want 64,000 samples for each edge of the groove -- that's incredibly dense. (How "high" is the waveform on the side of the groove?)

If you could make it so a single sensor could measure the width of the groove based on light intensity, you could do amazing things, but so far the experiments in that area with lasers work to some degree but pick up dirt too much.

Add new comment