The Grieg 1903 Recordings

On May 2nd 1903, Edvard Grieg visited the Paris studios of the Gramophone and Typewriter Company. Virtually nothing is known about this historic day except that by the greatest of good fortune Grieg left the building having recorded nine 10-inch wax master discs, each containing one piece, making in all about a quarter of an hour’s music. It seems strange, given Grieg’s extraordinary popularity, but very few copies of these masters were pressed at the time.

Only one of the nine was destined to be re-issued again during the 78 era and for only one title of the group does the metal master still survive. The remaining commercially-issued shellac discs are of the most extraordinary rarity and no one collector in the world has all nine; indeed in some cases only a single copy is known to exist. In point of fact it has only become common knowledge in recent years that there were nine pieces recorded that day. With the advent of the LP three of the titles appeared on collector’s anthologies but they were never readily available and it was not until they were all brought together and issued as part of a three CD set by the Norwegian company Simax in 1993 that anybody outside a rather small circle had heard the whole set. The Simax release did pioneering work in enabling these incredibly important recordings to be available to the general public for the first time. Available, but still not heard for the most part as the primitive nature of the original recording process obscured the treasures beneath, and treasures they indeed are.

Reviews acknowledged the historical importance of the release (how could they do otherwise) but very little more – one memorable review summed them up as ‘often grim listening’. However, Will Crutchfield in a highly perceptive review put his finger right on the button when he said: ‘These are dashing, spirit filled, tremendously rhythmic performances… In eight of the nine pieces, he plays faster – often far faster – than any subsequent pianist whose recordings I know, and in the ninth he is still faster than most… The tempos tell us a lot, and what he does within them tells even more. The hesitations and rushings are High Romantic pianism at its most confident and appealing… if the records had not survived (and they barely did!), our understanding of how music sounded in the 19th century would be significantly diminished.”

Only six pianists born before 1850 made surviving records, and apart from Brahms (just 59 seconds of whose playing can be heard), the Leipzig-trained Grieg was the only one of them who had any connection to German traditions of performance.’

Given this explosion of enthusiasm, these performances should have generated quite a flurry of interest, but not a bit of it; few musicians and musicologists seem to have taken the trouble to look at the recordings in any real level of detail, even in the composer’s own country.  In their defence, it is true to say that these nine 10-inch discs are not easy listening for people who do not have an interest in the early years of recording history. They need approaching with some knowledge as to how they were made in the first place in order to sort out in one’s own mind what is Grieg and what are the distortions introduced by the recording process, then only a few years old and hardly perfected.

All these discs exhibit an astonishing array of technical recording problems. For example, a characteristic of early Parisian recordings emanating from this studio (not only Grieg’s) is instability of pitch. The cutting lathe, which was directly connected to the recording horn, had a turntable on which a heated blank wax master was placed before recording began.  This was incised by the cutting needle attached to the horn during the recording process, with the sound from the piano travelling down the horn to the tip of the cutting needle and on to the wax master in the form of lateral grooves. For obvious reasons this turntable needed to turn at a constant rate for the whole of the side being recorded. This was simply not the case with the lathe being used in Paris, which was constantly fluctuating in speed. The result is both a ‘watery’ piano sound and, much more destructively, larger scale changes of speed (and thus pitch) throughout the side on replay. This pitch/speed variation occurs without rhyme or reason and is of such complexity as to be impossible to correct without the use of some kind of computer program that could be made to lock to some known pitch constant. We are lucky that this is now possible and Marston Records issued all of the Grieg recordings for the first time in a new pitch stabilised version with restorations done by Ward Marston in 2008.

While on the subject of pitch it should also be said that the general pitch here needs a lot of thinking about. The Simax transfers are not always constant in their general pitch centre but they surely should be, given that all the single sided discs were recorded on the same day. However they are almost certainly correct in opting for a low pitch. Pitch did vary a good deal at this time throughout Europe with the ‘official’ level in France still being A=435. No documentation will ever tell us exactly what pitch the piano was tuned to (and so what speed the turntable was set to run and how fast it all sounded) so this remains an artistic judgement and a crucial one. Running these discs at A=440 is plainly silly, Marston opted for A=435 whereas we think a slightly lower pitch/speed of  A=434 feels right. The original Simax is mostly lower than either of us and probably on the lowest limit of what is at all possible.

The other problem with these discs is the piano’s lack of proximity to the recording horn – the piano is just not close enough. Lots of experiments were done with recording at different distances from the acoustic recording horn (and even using more than one horn in the case of piano, to good effect). Here is not the place to go into a lot of detail about this, suffice to say that the real skill was to place the voice or instrument close enough to record as much as possible of the frequency spectrum and to achieve the best signal-to-noise ratio by getting as much volume as was possible in the grooves of the wax master. Against this was the ever-present problem that if it was too loud many very nasty things could result that would mean the whole take was completely ruined. It was a fine line that the best engineers of the time trod with great – and I would say rather unsung – skill. Piano recordings made in Germany at the same time as the Grieg’s by the great pianist and friend of Brahms, Alfred Grunfeld, are a wonderful example of how to get the most out of this early acoustic process and many of them still sound remarkably fresh today. As a result of all this the Grieg discs have a very quiet piano sound and a lot of surface noise. If we add to that the acoustic horn’s normal characteristics of a severely limited frequency range and the over-emphasis of certain frequencies at the expense of others then it is clear we are moving ever further away from the sound that Grieg himself made and heard in May 1903. The Holy Grail here must surely be the sound itself before the intervention of the process of recording.

Digital Restoration and a little history

Simply put, there are three areas that need addressing if we are to move closer to that Holy Grail:

 1.  Stabilising and fixing the pitch

2. Flattening out the distorted frequency envelope that the acoustic recording process via a horn introduced

This can be divided into two areas: we might call these, for now, the possible and the impossible. The possible resides in the area that the acoustic recording process actually could record – perhaps 100Hz to 3kHz, very roughly. The extremes are hardly recorded at all and often well embedded in surface noise, but the middle area (say 400 Hz to 1 kHz) recorded very strongly in relation to them.

This is an awful lot of numbers that probably means little to anyone not in the recording industry. To simplify things, 20Hz to 20kHz is roughly the audio spectrum that the human ear responds to (A modern CD recording records all of this).

A little simple maths tells us that the acoustic process could record something like 1/7th of that total spectrum although it is a very important seventh. What was lost was simply not recorded and cannot be recovered. It is just not there.

3. Removing the surface noise

This is the noise that is created by the materials that were used as the carrier and recorder of the original analogue signal itself, the wax compound that made up the wax Master and the subsequent material, commonly called shellac, that the commercial disc consisted of.

The first point, stabilising and fixing pitch, is a problem on the Grieg recordings that has been largely solved as we said above.

The second point has a partial solution. Digital equalisation does enable us to do more than we could before and it is here where the different versions of the recordings on CD will begin to vary. Our new versions issued do, we think, sound significantly different to both the old Simax and the Marstons.

Although much information was not recorded on the original discs in 1903 due, as we have said above to the limitations of the process, much important information was captured, and in surprising detail. It is our view that the important thing is to use equalisation to straighten out as much as possible the all important mid range area (which the acoustic process does record but in a very unstable fashion). The ear seems very sensitive to problems here. It is an odd fact, which some of us have noticed, that when the mid range area (say 200hz, around a violin’s open G up to 3 khz) is not right, creating a frequency distortion,  we tend to hear the recording as ‘dull’ and lacking ‘air’ and in extreme cases (like the Grieg’s) unnatural. It is normal to dive for the upper frequencies and lift them up when that particular alarm bell rings. On these old recordings all that does is make a wall of surface noise, two walls of surface noise. The other tendency when dealing with old recordings (and this is a massive simplification) is to lift up the middle area to try and get more music and less noise. We have all done it and we, along with a some other people working in the field, now think this is wrong.

Our approach after much experimentation has been to use advanced frequency analysis and equalisation software to analyse the precise frequency spectrum of each piece as played in 1903 by Grieg and Sigurd on Grieg’s piano. From this we get a complex equalisation ‘curve’ – a kind of fingerprint of each recording. The same music, played on two pianos of the same period in exactly the same way but recorded in radically different ways. The first step was to apply the ‘fingerprint’ of the new recordings to the old. Then we put the analysis away, went back to our ears and refined and developed to get a sound as close as possible to a natural one, given the massive limitations, to the sound in the room in 1903.

This leaves one very large elephant in our room, the third category, noise removal. This is where there has been far and away the most development in recent years. Simply put, up until the computer arrived in audio post-production (lets say from the late 1980’s/early1990’s) the established method for reducing surface noise was to filter out specific frequencies. These were the frequencies where the surface noise was at its strongest. This reduced (though did not entirely eliminate) the surface noise, but of course completely removed all the musical content that was also present at those frequencies, affecting the sound to a lesser or greater degree. Clicks were cut out of the tape with a razor blade by removing a tiny amount of time from the tape. This too affected the musical result and certainly makes those earlier transfers inadequate source material for any kind of musical analysis. Digital processing promised much more. The promise remains: to remove what you don’t want and leave untouched what you do. But of course life is never like this and there is a serious trade-off. Whilst it is true that certain kinds of clicks can be removed often very effectively and without apparent trace, it certainly has its limitations. Not all clicks can be removed this way and the more clicks you remove – sometimes hundreds of small ones every second are removed in automatic de-clicking programs – the more the sound is compromised. It is our view that even the best de-clicking programs available, remove critical performance information that our brains are very good at locking on to. So we have used no processing of this type in these recordings.