Wednesday, June 16, 2010

This strange hobby

Anyone want to know all the stuff I do to rip LPs? Didn't think so.

In the olden days I used my wife's inherited Zenith stereo (which, btw, can play 78s!*) plugged into the computer via a device called an iMic. Although the iMic eliminated a lot of noise, since it inputs into the computer via a USB port, the stereo itself was really noisy. It's also quite old, and the left channel had a tendency to drop out quite frequently. So a while back, as you may recall if you've been reading this blog for a long time, I bought a USB turntable. It's not audiophile quality, but it's good enough for me.

The software I use is Goldwave. I tried a few other audio programs, but Goldwave is the one I really liked and eventually bought. So after cleaning the record with some stuff called GruvGlide, I play it and record a whole side at a time into a big wav file. I break this file into individual tracks and more or less give them all a fairly uniform beginning and ending silence, and save each of the individual tracks as separate wav files. This is not always the case, but it is for all "normal" albums.

Following that, I run a batch with Goldwave that runs each wav through a hum filter which eliminates pretty much all the hiss & rumble you get playing records, then it normalizes (or maximizes, in Goldwave's terminology) each file and saves them all as new wav files in a different directory.

After that, I turn the volume up some and listen to each track all the way through, carefully listening for pops. In most cases, when I hear a pop, I stop the playing and select just a brief section before and after the pop (including the pop) and run Goldwave's pop filter on that selection. Most of the time, this eliminates it, although sometimes it isn't quite as simple as I've made it sound--sometimes the filter itself creates an artifact, or a new pop. Sometimes it doesn't work well enough--or at all, so I "zoom in" on the pop until I'm viewing just the split second that it occurs and manually redraw the sound wave to eliminate the pop. This also (usually) works well. Sometimes there's a noise buried so deep in the music that I can't pinpoint it; in this case I will try to delete the fraction of a second in which the noise occurs, and if this doesn't work I just give up and go on. I've found that I can delete up to about 0.05 second without making it sound weird (sometimes more, but if you get close to a tenth of a second you will probably hear it). Most pops occupy much less time even than that--the main problem is pin-pointing them. Most of the time, I'm listening to these wavs at a higher volume level than I do for "enjoyment listening," so when I turn the volume back down to my normal level, I can't hear the noise anyway.

You may wonder why I don't just pop-filter the whole track at once. Well, I tried this when I was first starting out, and learned that it doesn't work as well that way. For one thing, it reduces the effectiveness of the filter when it's run on too big of a selection. For another thing, it tends to create distortion. The distortion is especially bad when filtering music that's made mostly with acoustic instruments such as jazz or classical (and by "acoustic" I am including the human voice). It works much better on pop/rock and so forth that use mostly electric or electronic instruments. I don't know why--it's just something I've learned.

When I think I've done all I can I save the wav file again and go on to the next one. When I'm done, I batch convert this directory of cleaned up wavs into mp3s using Goldwave. Then I use a little freeware program called MP3 Tag Tools to tag all the files properly. This includes tagging each file with an embedded graphic of the cover art. By "embedded" I mean it becomes a part of the mp3 file itself. I started out just snagging thumbnails from Amazon or Wikipedia, but I wasn't always happy with them and many times I couldn't find the exact art I was looking for, so lately I've been scanning the covers and resizing them down to a 400x400 pixel graphic. Of course my scanner isn't big enough to scan a whole record cover at once, so it takes me 4 passes to stitch all together to create the full cover.

I started out saving all the mp3s at 128 kbps, which is considered "FM radio quality" and which I had decided was good enough. Later on I became a little pickier (or snootier**) and now I save them all at 320 kbps VBR (variable bit rate), which is considered "CD quality." The files are larger, but I have room for them (so far).

A strange hobby, I suppose, but otherwise I'd just be out back building scale model fortifications and blogging about the difference between hornworks and bastions.

*It would be quite simple to play a 78 on my turntable at either 45 or 33 and then use Goldwave to speed it up to the proper speed. However, I'm not entirely certain playing 78s would be good for the stylus. I think if the situation ever arises, I'll just go back to the old Zenith.

**"snootier" is in Firefox's spell checker!? But "hornworks" isn't?!

3 comments:

  1. "... but otherwise I'd just be out back building scale model fortifications and blogging about the difference between hornworks and bastions."

    And reading about Tristram Shandy, perhaps?

    ReplyDelete
  2. This is a huge process. Where do you get the time?

    I purchased one of those USB Turntables with dreams of converting my collection of 45's (several thousand from the late 70's to mid 90's) into MP3's but did about 10. That turntable is sitting up in my attack and my 45's remain in their protective cases. I guess I'm a slacker.

    I think one of the issues I had was attempting to adjust the sound quality. I don;t even mind the occasional pop, I just would like to play three or four songs in a row without having to turn up or down the volume between each song.

    I do like your idea of embedding the covers into the track.

    ReplyDelete
  3. Albatross: You have passed the test.

    Dave: I do this instead of watching TV. There are two ways to make everything the same volume: normalization, which is easier but rougher, and volume averaging, which is a little trickier but turns out more reliably steady volume levels. I just use normalization, which works well enough for me.

    ReplyDelete