Maria Callas' 1955 La Scala Norma

What can be done with a poor recording of a superlative performance?      

It's taken me considerably longer than just about any other recording on Pristine to remaster - is this as good as this recording can ever get when no amount of time and attention to detail can put back into a recording what was never there?

Maria Callas - Norma, La Scala, 7 December 1955

This week's new release

The response to our reissue in July of Maria Callas's classic 1953 Tosca was one of the most enthusiastic I can remember, and amongst the correspondence received in the days that followed was one e-mail requesting my services with regard to her classic Norma.

The only problem was that the e-mail did not mention which classic Norma might be the classic. Was it one of her two studio recordings? Perhaps the 1954 recording when the voice was in top condition - or perhaps the 1960 recording where, even if the voice was starting to wane, the musical experience of the artist brought something extra?

Or could it have been one of the live performances that fell between these two studio classics? Research online led me to suspect one of two 1955 concert performances - and of these the 7th December La Scala concert seemed to garner the most enthusiastic comments from the cognoscenti.

Mario Del Monaco

This may not be the recording my correspondent had in mind, but it certainly fitted the bill for an attempt at sonic overhaul. Poor sound quality was the criticism which came up over and over again in the comments sections - a brilliant performance if only you could get back the sonic deficiencies seemed to be the general consensus. Time to investigate...

The recording certainly isn't the best. It suffers a limited frequency range. It has wow and flutter a-plenty. It is noisy. It is hissy. It's harsh, thin of bass, and has bits missing (the first quarter of an hour, for example). Yet for some it's the finest Norma on record and among Callas's very finest two-and-a-half hours. What might be done?

Giulietta Simionato

Having assembled a variety of sources I got to work. One of the most time-consuming aspects of this restoration was sorting out the pitch problems. The basic pitch of the performance appeared to be just under a slightly low A=435Hz. This surprised me, until I carried out research which indicated that it was far from unusual for opera - Verdi's preference (the so-called Verdi Pitch) is even lower, at 432Hz, and a slightly lower that standard pitch of 440Hz allows both for a richer vocal tone and for easier access to the very highest notes in the score.

There was certainly plenty of residual mains hum to point pretty conclusively at this pitching (which is thus lower and slightly slower than some restorations would have it). But there was also considerable wow and flutter. The latter was very noticeable in the numerous sections of the score where the orchestra played on its own - the vibrato of the singers happily masks flutter very well. Unhappily for me, the score is riddled with single orchestral notes, as a kind of punctuation between often-unaccompanied vocal lines. Trying to correct all of the orchestral sections (most of which required specially-tailored settings in a note-by-note or section-by-section approach) whilst leaving the vocal vibratos intact was incredibly time-consuming - yet at the end of it the overall sound was the same, just less wobbly where it needed it.

So a lot of work for apparently little gain - with so much of the opera involving singing (as they do!) - but the instrumental parts now sound firmer and more natural, which I regard as a good thing. Next up was XR re-equalising. Here I wanted to try and reinvigorate the thin bottom end and lower mid-range. I'd have liked to find something more in the top treble, but there really wasn't much more there. Meanwhile the harsh upper-mid needed taming - after days of re-pitching work a distinct ringing in my ears suggested this ought to be a definite priority!

I've written in the past about "extending" the top end or bottom end of a recording, and was recently taken to task for this in a magazine, where the reviewer rightly pointed out that "you cannot add fundamental frequencies to an already pre-recorded sound source". This is correct - unfortunately the reviewer in question then goes on to state that you can boost the treble and bass with an equaliser to give the same impression. Well no, you can't, for the reasons he'd already stated. Perhaps therefore I need to be a little more clear when I make these kinds of statements.

First of all let's get the top end out of the way - in this particular recording there's not a huge amount more to be brought out of the background noise, thus "extending" what's audible. But the opposite is true of the bottom end, and although nothing has been added here, a lot (of rumbly rubbish) has been taken away allowing a considerable (of musical information) to be boosted back up to a more realistic level.

By comparison to the source material this does indeed sound quite a lot like an extension of the bottom end, and that's what I mean when I say "extended" in some of my technical comments. Yes the frequencies were of course always there, but they were often so buried in rumble as to be effectively lost, and a simple boost of the bass, using even the most advanced graphic equaliser, will only bring up both music and noise. All you'd hear is more rumble and an unpleasant mushiness. But by teasing apart the wanted from the unwanted I can then allow those frequencies to be restored to their rightful place in the overall sonic picture, thus effectively  extending the audible bottom end.

If that sounds a bit difficult to get your head around, let's draw a very rough parallel: Take an old painting that's become covered in a kind of brown grime. All the various shades of browns, oranges, yellows and reds in the original seem to have merged with the brown of the grime over the years and much of the subtlety and detail of the original colours has been lost.

An art restorer probably starts by gently cleaning away the years of brown grime, hoping to reveal the full palette of colours beneath. But if it's the Mona Lisa we want to see, nobody's going to let you anywhere near it with a jar of turpentine spirits! If you want to see what that particular painting really looked like when Leonardo started cleaning his paintbrush you'll have to try restoring a photograph of it - and how close you'll get to it will depend on the quality of the photo and the degree of grime, as well as your expertise with PhotoShop!

The photo is perhaps in this instance analogous to the recording. We can't be at the concert again (alas) and all we've got is a battered old snapshot of it, from which a lot is faded, a few corners are missing, and it's all a bit fuzzy. Traditionally the method used for "helping" Norma might have been something more like taking that photo and reducing the browns in it across the board - reducing both the muck and the content. AKA filtering out the bass.

Now that's not of course the best analogy ever. A picture restorer wishes the return the picture to its original glory. When I work on a recording such as this I know that the original "glory" was heard in the concert hall, not on the recording tape. Somehow I need to try and push the recording in the direction of the actual concert hall sound whilst cleaning away the muck, yet if I'm not careful, each push in that direction also enhances the muck, making for more cleaning work.

It's slow, painstaking, time-consuming work (as if the aforementioned re-pitched wasn't), and it's work in a tonal area which gives less instant gratification to the listener than finding you've got an extra octave to play with at the treble of the spectrum. Not everyone will appreciate it immediately or perhaps to the same extent, though I think anyone with an experienced ear for orchestral and choral tone will. It also mainly affects the frequencies below those that Callas's voice utilises - so I suppose this is more about improving her "setting" than her own sound, if you like.

Of course this is only a part of the story. There was a lot of background noise and hiss across the whole frequency range, and I've managed to reduce this by some considerable amount. An unnatural boost in the region around 3kHz which seemed to give a more "immediate" sound when first heard but quickly becomes a tiring earache-inducer has been tamed, letting the voices sit more comfortably in the mix.

Meanwhile thousands of bumps, coughs, thumps, sneezes, bits of electronic noise and cross-talk and radio interference have been excised or reduced.

And then finally yesterday I made a last minute, perhaps barely-noticeable last tweak to try and tease out just a little more sibilance and air around 5kHz - I actually made the adjustment whilst writing the first draft of this editorial, meant a further three hours of processing was needed prior to remaking all of the master files and uploading them once again to our servers. Listening back to it again today I'm glad I did - it's sometimes those tiny, subtle adjustments which can really make a difference.

In some ways this final point just goes to highlight the fact that with recordings like this one could spend half a lifetime on the restoration, with each tweak making less and less difference to the overall result, yet perhaps always wishing for something extra to help the sound match up qualitatively to the musical treasure within. The laws of diminishing returns soon come in to play, and one has to know when to stop, but unless a new and better primary source is discovered (something which at this distance seems highly unlikely) then this is probably as good as this recording will get.

And with a performance as compelling from (its delayed) start to finish as this Norma is, that's got to be worth the time, effort and technology that's gone into it. Take a listen to the lengthy sample from the finale and see what you think!

 

Andrew Rose
14 September 2012