By Roger Beardsley

An updated and edited version of the article which appeared in Classic Record Collector magazine in the Winter 1999 edition. By kind permission of the Editor.

Of all the tasks to be carried out when transferring 78rpm discs, perhaps the most difficult (and potentially contentious) is that of pitching the recordings correctly. The term '78rpm' is somewhat misleading because, as many readers will have found to their cost, discs in the days before LP were delightfully if frustratingly non-standardized, in both recording characteristics and speed. Much of the problem stemmed from the fact that recording turntables, like all mechanical devices, were subject to manufacturing tolerances to a greater or lesser degree, which in turn affected performance. On top of that there was the question of maintenance. Even the wrong grade of lubricating oil or grease could have an effect. I well remember that my 1950s EMI studio tape recorder, the famed BTR2, required a certain grade of oil, otherwise it would not run correctly; so it is a perennial problem. Perhaps the engineers were not terribly worried about the speed question; they may not have been able to measure it accurately, in any case.

Many of the earliest discs from the beginning of the century were recorded at 68-70rpm. Some of Caruso's first recordings play at these speeds: to play them at the ‘standard' 78rpm produces a travesty of the original recording, with the great tenor sounding more like Demis Roussos. Luckily, within a couple of years, speeds tended to rise to figures much nearer the 'standard', and in some cases exceeded it. There are numerous examples of pre-electrical (pre-1925) discs that play at speeds in the mid 80s. Early HMV lists used to indicate the correct speed but as to their veracity, I leave you to guess!

Eventually the early discs remaining in the catalogues were marked as 'above' or 'below' 78rpm. Buyers were advised to adjust the speed controls on their gramophones accordingly. These were times, of course, when many households possessed a piano, so that if necessary those who were musically competent could adjust the gramophone's pitch to the instrument. Indeed, recordings of Dame Clara Butt carried details of the keys in which the songs were sung, so that, as the catalogues stated, ‘Dame Clara Butt may be accompanied on the piano'. Such details are very useful indicators for the engineer today, especially as most of the discs seem to sound right when played at the speed which gives the indicated key. What a pity the idea was not universally adopted. The Gramophone & Typewriter Company (which became HMV) once tried putting a tuning band on some instrumental records, including a number by Jan Kubelik. For a short time Italian Fonotipia did the same. Why could they not have done that on all discs?

Mention of the great Dame Clara brings us to the question of Columbia speeds. For many years their labels specified 'speed 80' even though the generally recognized speed was 78. Those of Dame Clara's records made during Columbia's 80rpm era seem to play at anything between 76 and 83rpm. It is most unlikely that only her records varied in this way - my experience suggests that others were equally variable. So much for Columbia's standard.

The introduction of electrical recording in 1925 should have simplified matters. The universal recording system was the Western Electric process (even DG/Polydor fell into line, after trying its own method with mediocre results). Use of the system was signified on Columbia's records and those of their associated companies by a small ‘W’ in a circle, on HMV by a delta symbol next to the matrix number, and on Victor by a small ‘VE' enclosed in an oval. All the recording and replay characteristics were carefully calculated and controlled but not the speed, even though much of the recording world recognized 78rpm as the standard. Many early electrical Victors and their HMV equivalents were recorded well below 78. The famous Schipa/Bori duet from La Bohėme (HMV DB900) plays correctly at around 74rpm. At 78 it is quite distressing to listen to, particularly as Bori's voice suffers more than most when her discs are played too fast. Many of Martinelli's early electrics require a speed of 75 or 76rpm. As for Columbia's 'speed 80', it was still no better. While British-made Columbias tended to be recorded at a speed between 78 and 81rpm, those made in America often played correctly at 76 or 77rpm.

When we move on to Pathé Frères' hill and-dale discs, a whole new area of difficulty opens up. The earliest were nominally recorded at 90rpm and were re-recorded from large master cylinders by a pantographic duplicating system; thus the correct speed was the result of two mechanical processes, each with a possible speed variant. Later Pathés were recorded nearer the 80rpm mark. However, it is noticeable that with this label, disc size can have a significant bearing on recorded speed. Why, is anybody's guess.

How different it was, of course, when the LP came along. Or was it? Although the LP was more reliable, it did depend upon both the speed of the original tape recording and the speed of replay, in addition to any speed variation in the cutting lathe. Early tape transports such as the EMI BTR1 and 2 and the Ampex 200 and 300 series were essentially mechanical devices without any of today's computer speed controls. Even in the 1960s mistakes were made. For example, the well known EMI transfer of the Elgar Violin Concerto with Albert Sammons plays correctly at almost 34rpm.

One problem encountered by collectors wishing to play their records at the correct pitch is the matter of knowing what the speed of their turntable is at any one time. The most reliable means for the amateur is a stroboscope, since its performance is directly controlled by the frequency of the mains electricity, which is fairly stable to within 1%. Stroboscopes carrying the appropriate number of segments for a range of speeds are available through specialist suppliers. Caution needs to be exercised, however, when using any in-built speed indication unless on professional equipment. For example, some users of the popular STD variable-speed turntable rely on its digital speed read out. But the STD, although useful, is very crude electronically and the drive for the read out is most unstable. Variations of 4 or 5 per cent are known.

So how should we determine pitch? 'Easy,' I hear you cry. 'Correct pitch is correct pitch, isn’t it? A=440 and all that.' If only it were so. During the period in which recording has been a serious medium, 'A’ has varied between around 425 and 445, a variation of some 5 per cent. On top of that, different pitches were standard in differing countries: at times even within a country. Hearing an LP transfer of a recording by his father-in-law and duo partner Adolf Busch, the pianist Rudolf Serkin said: 'That is too high. Busch used a lower A than that.' Yet surely, if we know the date and place of a particular recording, the problem is solved? Sometimes, but purely local conditions can have an effect on pitch. Temperature and humidity are two of them. The local piano tuner's tuning fork is another, especially in the case of instrumental and piano-accompanied vocal recordings.

Which brings me to the matter of transposition - and I am not referring only to aged singers who needed to transpose their music down when they made their records. Transposition was still common in the first half of the twentieth century - Tito Schipa, who had a notoriously 'short' voice, used to travel with a trunkful of transposed orchestral parts for his arias - and without a knowledge of the conventions, and of the singer involved, it is very easy to make serious pitching errors when making transfers. The pitch may be technically correct but the singer may not sound as he or she should.

Therefore it is all a little trickier than one may first have imagined. In the end it comes down to judgement. For instrumental music, a decision based on a knowledge of the place and date of the recording, as well as the musicians involved, can be helpful, although the basis for pitching decisions should always start at the score pitch, since transposition will be unlikely unless an arrangement for another instrument is involved. Then it is a question of how far, if at all, to deviate from today's absolute pitch. In the case of vocal music, however, we advance into a minefield. For the transfer engineer it is very much a case of putting your best foot forward and hoping there will not be an explosion.

First we need to get the record to something like a recognized pitch, based on what is known of the music and the record itself. Were other records made at the same session? If so, they may provide clues or points of reference. Then we must concentrate on the singer. As I mentioned earlier, singers in the 78rpm era were used to transposing, so as to suit their vocal range or individual circumstances: they thought nothing of it, whereas today such behaviour would be reckoned cavalier and unmusical. A number of tenors recorded 'Di quella pira' from Il Trovatore without the top C added by their predecessor Enrico Tamberlik. Quite rightly, they decided that they were much more comfortable with the semitone drop to B, and that they sounded just as thrilling - listen to Antonio Cortis, for one (HMV DA 1155). But there are those who still insist on playing these records at a speed which reproduces the note as a top C, with the voice consequently taking on an unnatural quality.

The crucial thing for an engineer or a collector listening to a new acquisition is whether the singer's voice sounds right. I know, for instance, that the Editor is not convinced by the low transpositions which some people inflict on the tenor Fernando De Lucia. Quite apart from the state of the voice, he feels that at those slow speeds the orchestral accompaniments sound false - a factor overlooked by some engineers. As you play the record, you have to ask yourself. ‘does the voice sound like the voice you know?’ - or, in the case of one you don't know, ‘is it consistent with the vocal range and the music?’ If it is, you can stick with it and be thankful; but if not, you can try moving a semitone on the appropriate side of the key. In the end, you must trust your instinct, in the knowledge that a certain amount of subjectivity is involved. Only one thing is certain: whatever your decision may be, someone will always disagree with you!

And you are still not out of the minefield. Not all records continue and finish at the speed at which they started. The recording machine could vary its speed considerably during a side, sometimes by 5rpm or more. Thankfully this variation is not too widespread a problem, and if you are lucky it will be a continuous, steady change, or one sudden variation. I did once have to transfer a Leo Slezak recording which had six different speed fluctuations of up to 4rpm, all within a 10 inch side lasting less than three minutes. When you are dubbing a set of records there is frequently a variation between the sides even if they were recorded on the same day. Frida Leider's wonderful Immolation Scene from Gotterdämmerung (HMV D 2025/6) and Lauritz Melchior's Forging Scene from Siegfried (HMV D 1690/91) are cases where not all the sides were recorded at the same speed - and yet on occasions they have been transferred as though they were. The results do not bear thinking about, let alone hearing.

What it all comes down to is that there is no magic formula for the engineer to follow, no easy computerized system, as with noise reduction. In the last resort it is down to an engineer's ear, judgement and experience - and sometimes a little luck. One thing I do believe, however, is that our forebears were more tolerant of the sort of pitch or intonation problems which today we find difficult to live with.