Vermont Softworks

The whole process, from A to Z

This section describes in great detail not only the exacting process we follow for cleaning, dubbing, restoring, and copying audio, but also the equipment, materials, and software we use.

1. Cleaning

Though we can perform near-miracles in the noise-removing department (see “Restoration,” below), the best noise removal is not to have noise in the first place. Here‘s how we start:

1. We thoroughly wash the LP disk using the same record-cleaning fluid and brushes as those used by the Historical Sound Recordings Collection at Yale University, and as recommended in the monograph Sound Directions[1], Disc Doctor’s Miracle Record Cleaner.
2. We rinse the disk with distilled water and absorb most of the moisture with lint-free towels, finally allowing the record to air-dry for at least an hour.
3. We brush away any remaining particles — and also reduce static charge — with an AudioQuest carbon-fiber brush.
4. We clean the turntable stylus with Disc Doctor’s Stylus Cleaner. (Actually, this is done at the end of the previous dubbing process, so that the stylus never remains uncleaned for any length of time.)

2. Dubbing

Dubbing is the process of getting the audio from the record and copying it into a computer file. This is similar what you may have done in the past when you “taped” your LPs to listen to in the car:

The dubbing has to be done in “real time,” so it takes an average of about 40 minutes just to play the recording and capture the audio. Naturally, the better the turntable, cartridge, stylus, phono preamplifier, preamplifier, and analog/digital converter — and the more carefully the vinyl disk has been stored, cleaned, and played over the years — the better the results.

1. So, we “play the record” and capture the audio at a rate of 96,000 24-bit samples per second (96 kHz/24-bit) using the following equipment and software:

1. We again remove any new dust and discharge any static buildup on the LP by brushing it again with the AudioQuest carbon-fiber brush.
2. We carefully remove the disk from the turntable and slip it into a new three-ply, anti-static “Original Master” polyethylene sleeve made by Mobile Fidelity Sound Lab, which then gets inserted into the original album jacket.
3. While we still have the file open, we trim it slightly, so that it contains only the audio signal from immediately after the “needle drop” right through to the finishing groove on each side.

3. Restoration

Unless given instructions to the contrary by our client, all of our noise reduction and audio restoration work is performed with a very “light touch.” Essentially, there is very little corrective work that can be done without having an effect on the very signal that we’re trying to restore. We strive to reduce hiss, rumble, clicks, pops, and crackle just enough to bring the life back to a recording without harming it in the process.

The dubbing process has left us with one large AIFF computer file that digitally represents, with ultra-high fidelity, the exact sound represented in analog on original the vinyl disk. We keep this file — untouched — throughout the rest of our work, so that we can return to it at any point if need be. (This file will also become part of the project’s archive.)

Note that, because of its high sampling rate (96 kHz) and word size (24-bit), this file contains more than three times as much audio data per minute than would be used for the same music on an audio CD. (For each stage of our sound restoration process, we create a new copy of the audio data — each copy is typically about 1.5 GB in size!) While debate rages amongst audiophiles and as to the point beyond which additional information makes no difference to the human ear, audio engineers do agree that this data is helpful — vital, even — to minimize digital “artifacting” (such as that caused by rounding errors) during the intensively mathematical noise-reduction process.

A word should also be said about the noise-reduction software applications we use. All of it takes into account the relatively new science of ”pscho-acoustics,” which, among other things, dictates that less noise reduction be done during loud passages in the source material (and correspondingly that more be done during quiet passages), since we’re most aware of unwanted noise during quieter moments. Many of the software settings we adjust simply tell the applications the most that they should do; they will often then do less.

1. We use the AudioLeak application to create a visual picture of the volume of the audio file. This is basically a “before” shot which we’ll be able to compare to the final “after” shot at the end of the audio restoration.
2. Using Amadeus Pro, we search through the file and manually repair its loudest pops and clicks. Ideally, we would repair every flaw manually, but it would take several days of intense work to repair an average record album; therefore, we limit ourselves to the worst twenty or so pops.
3. If appropriate, we remove all of the subsonic noise (below 20 Hz) and much of the low-frequency noise (below about 150 Hz) using DeNoiseLF. As with each of the other restoration steps, the settings we use will depend on the quality of the original recording, its current condition, and material that’s been recorded.
4. We remove most of the clicks and pops using ClickRepair. Again, many different settings are possible; we use a combination of experience and close listening (with either Etymōtic ER•4S MicroPro Reference earphones or Sony MDR-V6 Monitor Series headphones) to do it right.
5. Depending on the state of the recording, we might choose to use ClickRepair again — this time with settings tuned to reducing LP “crackle.”
6. Next, we usually choose then to reduce the surface noise endemic to all vinyl records, using DeNoise. Here’s where we use the “silence” that we left after the needle-drop at the beginning of our file; DeNoise will listen to the silence and adjust its own settings to minimize the surface noise. We take especially great care here, though, because it can be tempting to try to remove too much surface noise — and this can have a disasterous effect on the very audio that we’re attempting to save. We always follow the rule, “when in doubt, do less.”
7. Regardless of whether the original recording was in mono or stereo, we’ve been working all along with a left audio channel and a right audio channel. Why we bother doing this with mono recordings may not be immediately obvious: though the left and right channels theoretically contain identical original source material, they have worn differently over the years — and the clicks, pops, and other noise are very often much stronger in one channel than in the other. We’ve cleaned up a lot of the noise, but at any given moment in the mono recording, one channel probably has better audio than the other. ClickRepair lets us merge the two theoretically identical channels, slowly panning from one to the other to obtain the best audio. Though this is one of the most effective things we do for mono recordings, we naturally skip this step for stereo.
8. We always then check the file again manually with Amadeus Pro to double-check our work. (Is the result a distinct improvement over what was originally dubbed? Could it be improved by using different settings along the way? Are there any rough spots that we hadn’t noticed before?).
9. Finally, we use AudioLeak to “normalize” the volume of the audio. This is not a form of signal or data compression, but rather of audio delivery specification[2]. It’s comparatively advantageous to send “loud” signals down the audio chain through your sound system: the louder the source material, the lower the listener will set the volume; this, in turn, means that any hiss or hum introduced by the sound system will be less audible. (As an interesting exercise, try briefly turning your system’s volume control most of the way up sometime when you’re not playing anything. Hear the noise? That’s what we need to “cover up” as much as possible by providing “louder” material.) But if our signals are too loud, they’ll be distorted, or you’ll need to adjust the volume whenever listening to a restored disc, or restored tracks won’t play nicely with other tracks when you’re shuffling your music. Therefore, we adjust the album’s gain based on an RMS-average weighted Leq level of the whole album, to -23 dB (or less, if that setting would cause any audio clipping). We use the new ITU-R BS.1770 weighting algorithm for LKFS measurement, which is more appropriate for music than previous methods.[3]
10. We create an “after” picture of the volume of the audio file using AudioLeak, so that to compare with the “before” picture we took at the beginning of the restoration process.

4. Creation

At this point, we are still working with a single monolithic 96 kHz/24-bit audio file representing all of the audio from both sides of your LP. And even though the album is probably only about 40 minutes or so, this single file is so large that it would take four CDs to hold it! Obviously, we need to present you with something more useable.

1. The first step, then, is to break this huge file into separate 96 kHz/24-bit files for each track, using Amadeus Pro. Unlike inexpensive LP-ripping services, we do this manually — which yields much better results for most classical, jazz, and art-rock albums. If appropriate, we usually create a 0.5 second fade-in and 2.0 second fade-out for each track, replacing the inter-band LP “silence”; these figures represent a compromise between the (usually) shorter pauses between bands on CDs and the (usually) longer ones on LPs, and generally produce good results. Let us know if you have special needs or preferences, though, as there are no hard and fast rules here. Also, do let us know if you’d like any extra-long LP bands divided into multiple tracks.
2. We then import the individual track files into iTunes. (Yes, that iTunes!) It turns out that iTunes is actually a pretty good way to get metadata stored in audio files and also a good way to create CDs — so it’s a logical next step.
3. So, here we’ll add metadata, if we’re providing you with files: track name, album name, artist, track number, and total track count; also, composer for classical music.
4. Then, we finally convert the large 96 kHz/24-bit files into something more useable. If we’re providing you with a CD, we convert them to standard 44.1 kHz/16-bit stereo files (yes, even for mono material — the IEC 60908 “Red Book” specifications for audio CDs are quite explicit) using QuickTime conversion directly within iTunes. If we’re providing you with files, we’ll convert the 96 kHz/24-bit files into whatever you specify (Apple Lossless is our favorite, but AAC, MP3, HE-AAC, FLAC, Ogg Vorbis, etc., are all possible), either using QuickTime in iTunes or using other tools if QuickTime doesn’t directly support your requirements.
5. If you’ve ordered a CD, we then burn the 44.1 kHz/16-bit stereo files onto a MAM-A Gold CD-R per Orange Book specifications, guaranteeing that the disc will be fully compatible with all modern disc players. MAM-A Gold CD-Rs are manufactured to excruciatingly high standards, and are projected to last nearly 300 years when cared for properly (e.g., don’t leave them — or any CD — in direct sunlight!). We should note that we do not use the MAM-A Archive Gold CD-Rs, which are expected to last somewhat more than 300 years; these “archive” discs cost considerably more than the MAM-A Gold CD-Rs, and probably aren’t worth the extra cost we’d have to pass along to you, unless you’re pretty certain that you need discs that will still be viable beyond the year 2309. Isn’t 2299 good enough?
6. But our work isn’t limited to providing you with the audio of your LP. Though we know that nothing else will replicate the experience of poring over a 12" × 12" record album, we will nonetheless scan the album covers (up to four sides for no extra charge) at 600 dpi, 48-bit color using our Epson GT-20000. We repair most scratches and marks and correct any color-shifting and fading. If we’re providing you with a CD, we’ll then print the front and back covers and insert them as front and back panel inserts for the CD’s jewel case; any other covers and liner notes will also be printed and included inside the case. If we’re providing you with audio files, then we’ll also include PNG (or JPEG, PDF, or TIFF, if you prefer) files of the restored artwork. And, in either case, we give you a listing of tracks, as well.
7. We also produce a three-page itemized description of the specific steps we’ve taken to clean, dub, restore, and transfer your audio, so that you will know exactly what we’ve done.
8. As a final step, we create a DVD archive snapshot of several steps of the restoration process, so that if there’s a problem (your new CD is lost in the mail, say, or you decide after-the-fact that you’d like a large LP band subdivided into different tracks), we won’t have to begin the process over from scratch.