ALL YOU EVER WANTED TO KNOW ABOUT CD RE-MASTERING
... BUT WERE AFRAID TO ASK
by Chris Burmajster
This article was originally written for Toccata, the magazine of the Stokowski Society and published in the autumn 2000 edition
In the Autumn 2000 edition of TOCCATA, Richard Luce, writing about the Cala Bach / All American Youth Orchestra CD wondered "how it was possible, given the shortcomings of these 60-year-old records, to manage such transformations". Well, as one of the engineers responsible for that particular disc - and several others in the Cala / Stokowski series - I can answer his question about how it's done. For the benefit of non-technical readers, I'll try to explain the technical side as best I can, as obviously it is a highly technical and musical process. Fundamentally, it's about knowing what the technical problems are and how to solve them but always being guided by your musical knowledge and good sense.
Obviously, one must start with the source recordings, which in this case were on shellac 78 rpm discs. The 78s were transferred onto DAT (a professional digital audio format) and were transferred 'straight', which means exactly as they came off the discs - warts and all.
I have been re-mastering old recordings for over thirteen years - ten of those for Decca - where I was responsible for a substantial portion of their re-mastered back catalogue. The way re-mastering is done has changed little in that time. It mainly concerns the growing role of the computer and the sophistication of new products aimed specifically at the re-mastering market which really took off with the advent of CD.
The first thing to be done is to transfer the analogue source to a digital format, usually tape based, although that is slowly changing. Because the work will be done in a very sophisticated computer and the audio will have to be transferred from the tape into the computer at some point, you might as well put the original directly into the computer in the first place wherever practicable. Until recently this idea of putting the audio onto the computer itself was expensive, but the fall in the price of huge hard drives (which store the audio) has made it much more of a practical proposition.
In the old days. at Decca, we would transfer the analogue source to digital tape (either Decca's own IVC based system or Sony U-Matic) and editing was done essentially by copying from one tape to another with very great accuracy in the cross-fades between the two. Naturally a computer was controlling this operation. This is known as 'destructive editing' because the audio on the tape is physically altered and, once you've done it, you can't undo it. In fact, to partially get round this, we would do what's called assemble editing. This means copying the first (good) part of the audio to a blank tape up the point where something needs doing and then adding the rest in stages by editing. This gave you a new, edited master tape at the end of the day. But the editing was still done by recording from one tape to another.
On today's modern computers editing is nondestructive because the original audio on the hard disk is not altered in any way. When you edit, you tell the computer to play this bit of audio on the hard drive. and then that bit, and then another bit, all of which might be on different parts of the hard disc. So playback is accomplished by the computer jumping from one part of the disk to another (in accordance with your instructions) with such speed that it sounds to us like continuous audio. However, the original audio is not altered in any way and you can undo what you've done at the click of a mouse. Toys for the boys!
In the case of the All-American Bach CD, Edward Johnson gave me the DAT on which had been recorded the original 78s. I loaded this into the computer, listening carefully to the material to see what it sounded like and deciding what might need to be done to it. Of course, these were 60 year old recordings and in many respects clearly inferior to modern recordings. The most obvious were in the areas of frequency response and dynamic range. Modern digital recordings have a flat response (meaning that all frequencies are replayed at the same volume) ranging from about 20Hz to about 20,000Hz. With older equipment, some frequencies are boosted or cut over others because of imperfections in the equipment which lead to a colored sound.
To give non-technical readers a point of reference the lowest note of a five-string double bass is 41Hz and the lowest note of a concert grand piano is 27Hz. The A to which an orchestra tunes is around 440Hz and, at the upper end, a piccolo is around 4kHz. These figures are for the fundamentals of notes - they don't include the all-important harmonics (or overtones as they are sometimes called) which go right to the upper limits of our hearing and beyond. It is mainly harmonics that distinguish between the same note played on a flute or a trumpet, so they are important for good sound. Because modern recordings can capture all these frequencies, the sound can be rich and lifelike. However, a good proportion of Stokowski's recordings were made when the equipment was incapable of both capturing and reproducing such a wide range of frequencies. In fact, recordings made onto shellac at 78rpm have little more frequency response than a telephone. This limits what the re-mastering engineer can do.
Dynamic range is the difference between the loudest and the softest sounds. In real life, a large orchestra can manage over 100dB and only recently could any recording medium capture that without some sort of fiddling by the recording engineer. Ironically, capturing the full range of dynamics is not really desirable for most home listeners, because in the small space of the home, the quiet bits will be inaudible and the loud bits will blow you and your neighbours to kingdom come! However, this is not a problem with any 78 disc, as the dynamic range will have been severely compressed by the original recording engineer to fit the constraints of the medium. You might think that this would make life easy for the re-mastering engineer, but in fact they limit what you can do.
So, once you've loaded the audio into the computer, the first thing is to edit the sides together to make the program run as desired. This presents the first problem, as 78s do not always run at exactly the same speed, thus giving problems of pitch and speed differences between the sides. There are sophisticated computer software programs (known as plug-ins) that can deal with this. Another problem, as with the Bach CD, was that some of the sides overlapped ever so slightly and weren't quite the same. This meant editing very carefully indeed for the edit to make musical sense.
After the editing is done, the clicks and ticks need to be removed. There are two ways of doing this, depending on what kind of clicks they are, how frequently they come, and how loud they are. In a lot of cases, another of those clever computer plug-ins is used to de-click the audio. However, in the case of the Bach CD, there weren't that many clicks, so I removed them by hand, so to speak. This is done by placing an edit point on one side of the click (clearly visible on the computer's monitor screen) to break it up into two streams, then zooming in on it to a huge level of magnification, and altering the audio streams so that the cross-fade comes just before it. Then it shouldn't sound.
One very important point about all these clever processes is that their success is to a large degree dependent on the way that they are used. For example, with a de-hissing plug-in, you can remove the hiss almost entirely, but the sound will be awful because the processing affects the basic sound as well as the hiss. This is much easier to demonstrate than write about, but I'm sure you'll take my word for it! Therefore, the engineer must find a compromise between a degree of hiss removal and not altering the basic sound. As always, your ears and musical sense must be your guide. Part of the criticism of these plug-ins - when they were new - was due to over-enthusiastic engineers throwing out the baby with the bathwater. This also explains why the same recording can sound different on different labels - it's largely down to the taste and amount of fiddling by the engineer.
Once the basic audio is clean. then comes the icing. There are two things that can be done here, equalization and adding artificial reverberation. Equalization (EQ for short) is just a technical name for treble, middle and bass adjustment, although it's a lot more sophisticated than on home replay equipment and - as it's done in the digital domain inside the computer - it's more accurate and has fewer side effects than analogue EQ. As I mentioned earlier, because the frequency response of source material from 78s is very limited by modern standards, you really don't have much leeway and you often have to apply levels of boost and cut that would give you a heart attack on more modern material. Somebody once pointed out that most award winning modern recordings never had more than 3dB's worth of boost or cut on them. On the Bach CD, I ended up applying almost 18dB at some frequencies! My goal was to make the sound as rich as possible, because I believe that this is what Stokowski would have wanted. So I cut (?) the treble by a few dB to reduce the harshness and substantially boosted what little bass there was on the recording to add richness to the sound. Had it been a Toscanini recording, my approach would have been different.
Artificial reverberation can also be added. Again, good musical taste must dictate how much and of what type sounds right for the particular program you are working on. In the case of the Bach CD the originals were very dry and boxy, so 1 added a Royal Albert Hall type of reverb which gave more richness to the sound, together with the EQ which both Edward and myself thought Stokowski would have liked.
As always, you are guided by your ears and your musical sense of what sounds right and what doesn't. Whilst you can teach a monkey to operate the computer, you cannot teach what feels right musically speaking, or at least it's very difficult. Ironically, I personally have a horror of EQ and artificial reverb. certainly as far as more modern re-mastering goes. But this program just didn't sound right without all the fiddling and, judging by the reaction of critics and Society members, I think I got it right.
© Innocent Ear 2002.
