Editorial for November 2010

Sometimes a degree of intervention might have been necessary – a little roll-off of the treble frequencies to reduce hiss, perhaps; the manual removal of bad clicks through tape editing, where a master at 15 inches per second allowed sufficient editing precision to allow this painstaking procedure to work to the extent that these edits were pretty well undetectable. This was more or less standard procedure as recently as 20 years ago when I started working at the BBC, though we did also have access to an Packburn analogue de-clicking device in one specialist transfer suite, which produced results of varying and often rather dubious quality.

This was around the time of the first CEDAR digital audio restoration tools – they started making their commercial equipment in 1988 – which took the first steps in changing the game. Meanwhile, however,  other “tricks” were in use, techniques to “improve” sound by attempting to mask tricky sonic problems or to try and make a scratchy old disc sound more appealing, techniques which quickly got themselves something of a bad name, particularly in the classical music world.

Here I’m referring to the world of electronically generated stereo, of heavy artificial echo or reverb, of ham-fisted re-equalisation, and other dubious means of manipulating old recordings to the tastes of the engineer, producer or record company, often with hit-and-miss results. Some of these practices have carried through well into the digital era, and one only has to trawl through the sound samples of vintage recordings at the likes of eMusic.com to hear all sorts of sonic horrors being committed today in the name of audio restoration. Indeed it was one such CD which convinced me to move into the field myself…

But since those early beginnings in the late 1980s things in the world of digital audio restoration and remastering have advanced at an almost unbelievable pace. Some sonic problems which simply couldn’t be tackled five years ago – and hence might lead some engineers to consider the kind of “masking” solutions I’ve just alluded to – are now a breeze to fix. The output quality of today’s de-clickers and noise reduction tools is not only light years away from what was available at the beginning of the decade, it’s advancing in leaps and bounds, at a pace which mirrors the phenomenal increases in computing power available on your desktop and mine. Processes which take minutes today would have been infeasible in 2000 through lack of memory, drive space and processing power – instead of minutes they might have taken days or weeks to complete.

But hand-in-hand with this improvement are parallel developments in the very tools which have previously caused so much sonic damage, and as a result I wonder if it is necessary to begin to rethink entirely the final aims of a good transfer and restoration, and what should be “permitted” in order to achieve one. In fact, this week’s Toscanini Brahms German Requiem is an ideal case in point, and an ideal way to examine what is now possible and how it sounds in application.

In an ideal world one starts, of course, with the best possible source material, and in the case of the Toscanini  I’ve been exceptionally lucky with a source of remarkable quality. I’ve already made clear that today I can de-click that source much more quickly and effectively than I could have done a few short years ago, and indeed that’s what I have done. After this is the re-equalisation – as with many XR remasters, this is a significant intervention, but unlike those of the past this is carried out not by subjective knob-twiddling but by objective analysis and precise adjustment by computer, albeit at the hands of a human being. The aim of this is to correct both small and major frequency anomalies in the microphones and other recording equipment used to make the original.

This is then followed by digital noise reduction, and I’ve been using technology so advanced it’s still in the pre-release testing phase and won’t be available commercially for several weeks yet. It is astonishingly effective.

I then carried out basic editing to join disc sides together, something I could just as easily have done with tape and razor blade, but then began some seriously advanced restoration work to deal with an assortment of the type of flaws often found on discs – the removal of swish, clunks, blips and so on – whatever is required. Again, this aspect of the technology is very recent, and in its latest incarnation is more powerful than ever before – I’m glad to say I’ve been in a position to shape the development and application of some of the tools I’ve been using to fit precisely the kind of work I do. Thus defects that others might previously have been tempted to make “disappear” under reverb can be dealt with directly and permanently.

Next in my remastering process is the application of Ambient Stereo. Ever since its introduction it has proved immensely popular, with a clear majority of our sales now in this format. It’s a subtle intervention and produces very natural results which don’t mess with the original sound as previous technologies have, and makes listening to mono source material far more comfortable, especially on headphones.

By this stage it is clear quite how a recording’s inherent acoustic sounds. It was a great surprise to me, in the earliest days of XR development, to find that the re-equalisation process I had developed often revealed room or hall ambience and reverberation which had previously been buried almost inaudibly in the recording. I distinctly remember one recording, Weingartner’s 1936 Vienna Eroica, having a huge buried reverberation – I think this was picked up by one reviewer – which nobody had ever suspected was there, to the extent that I had to go out of my way to imply that I had not added it myself.

But here in the Toscanini German Requiem we have a problem, to my ears, anyway. That problem is well-known to any aficionado of NBC broadcasts of the era, and it is the sound of Studio 8H. It’s as dry as dust. It can make one of the finest orchestras in the world sound like a high school ensemble. It’s almost painful to listen to for many people. If it has any redeeming feature it is that it can be incredibly revealing of inner detail, but at what cost to musical enjoyment?

And so we return to the thorny subject of reverberation – which I dismissed some paragraphs ago as one of the horrors used to mask problem transfers. But when we look to the tools available to the modern remastering engineer and find that, once again, the game has changed entirely. I’ve previously experimented with digital reverberation on 8H recordings, going to the extent of releasing one transfer both with and without it. The reception to the “reverbed” recording was at best mixed. It still didn’t sound quite right. This is largely because traditional digital reverb is basically a randomisation of multiple delayed signals shaped by equalisation and lasting for a settable duration to create an impression of generic spaces. The software or unit to generate this offers presets called ‘large hall’ or ‘small room’ and so forth, and they kind of sound like somewhere and nowhere at the same time. In short, they are most definitely artificial.

Convolution reverberation, as applied to this Requiem, is a different beast altogether. We now have the computing power and technology to record and encode the sound of a specific acoustic space, and then apply this to another recording – and as far as the source recording is concerned the drier and deader the better, making an 8H recording ideal. The results of this are so natural and unobtrusive as to be undetectable – unless of course you were expecting a bone dry acoustic because you read the notes that said “Studio 8H”. Nobody listening to the German Requiem will think “a-ha, he’s added reverb to this then” unless they already know the recording from a previous issue or had expected something very acoustically dead. It sounds simply like a well-made, well-miked recording in a natural setting. Moreover, in combination with the other techniques I’ve described here, this particular recording now sounds like one that could almost have been made at any time between 1943 and today (though of course there is no stereo positioning information as it is in origin strictly mono).

So have I committed the same sins as some of my predecessors? Well, I’ve heavily re-equalised, I’ve created a kind of stereo ambience, and I’ve used artificial reverberation, all of which I’d already listed as some of the worst horrors of the past. But is this a sin today, given the very different approaches and reasons for doing so? I would counter that it is not – and suggest that perhaps it is time to re-examine some of the sacred cows of audio restoration and remastering in the light of recordings such as this. For while it is of course possible to use these ideas and techniques to do great damage to a recording, it is also now possible to use them in such a way that they become an intrinsic part of the restoration process, dealing with sonic issues beyond the clicks, scratches and hiss, and with an aim to complement the recording rather than mask problems within it.

The game therefore is changing, and I note that others are going along with that change – the latest Testament releases feature their take on Ambient Stereo (it’s non-optional, by the way), and elsewhere the basic principles of XR remastering are being used by other restoration engineers to deal with the same issues for which I developed it. It’s all powerful stuff, to be taken in moderate doses perhaps, but it seems to point to a future of continuing improvements in the results of restoration and remastering. As such I do think it’s a very good time indeed to think again about what is and isn’t “permissible” when preparing an historic recording for reissue, and hopefully a new audience, in the 21st century.