Making transfers from 78rpm sources - a practical guide
By Roger Beardsley
With the increasing realisation of the musical value of historic recordings has come the need to improve the way in which these precious artefacts are reproduced and preserved. A record without the ability to hear it is useless; a record badly reproduced is not much better, for it gives a false impression of what is actually stored within the grooves. Imagine reproducing an historic document in such as way as to alter the text, or perhaps printing a copy of the music for Schubert’s C major symphony, but changing all the key signatures at random. Exaggeration? No, not at all, and you will come to realise this as we go through the guide.
Note, although I specifically refer to 78 rpm discs, much of the following will apply to early mono LP and rather less to reel-to-reel tape.
The first part of the learning process is not in fact in this document. Before even putting a record on the turntable, you need to learn how the sound got there in the first place: without knowing this, you will find it very much harder to satisfactorily recover the information stored in the record, and the results will be a hit and miss affair, rarely suited to what is required. Therefore, before going further you should at least read:
- A Brief History of the 78 rpm Record by Roger Beardsley with Daniel Leech-Wilkinson (2005)
- Defects in Gramophone Records by William D. Owen and H. Courtney Bryson (The Gramophone, 1931/2)
- High Quality Recording and Reproducing of Music and Speech by J. P. Maxfield and H. C. Harrison (Bell Telephone Laboratories 1926) (pdf file) [PDF]
- Speeds and Pitching of Gramophone Records by Roger Beardsley. (A specially edited version of an article first published in the Winter 1999 edition of International Classic Record Collector.)
- A Guide to Playing 78 rpm Records by Roger Beardsley. (A specially edited version that has appeared in International Classic Record Collector and elsewhere.)
The dedicated engineer will almost certainly look further into the art at every possibly opportunity.
What follows also assumes a certain level of technical understanding.
So, why not simply read the Guide to Playing 78 rpm Records, and go from there? The answer is in the word ‘play’. If you are a record collector, your records are played purely for your own pleasure. How you do that is up to you, and possibly your budget. You might prefer a rounded sound, or a bright one. You might possibly prefer the records of Signor X played at a speed contrary to accepted practice. If you have the appropriate equipment, you might wish to add reverberation to dryly recorded sound. In short, you choose how you wish to play them: the result should be what you enjoy.
TRANSFERRING RECORDS - THE DIFFERENCE
However, when you transfer a sound recording, the end result will be for someone else to listen to or work with: responsibility now enters the equation. It is now up to you to preserve the integrity of the performance, to the best of your ability within the parameters set by whoever commissions the transfer. Should those parameters be against your better judgement, then consider either attempting to persuade the principal of the error of their ways, or even declining the work. In the end, it will be your reputation that is at stake.
The practice and art of transferring—and it is an art, albeit one that depends upon science—is highly dependent on the various pieces of equipment required to carry out the task. So much so, that a detailed examination of that equipment will in itself act in part as a tutorial for the prospective practitioner. For the rest, you will need to need to gain experience by practice.
When asked what is the most important piece of equipment I always say ‘your ears’. A truism? Yes, but not to be overlooked. If you cannot hear the difference between the various shades of sonic quality, recognise the different types of distortion, or if you have no sense of relative pitch, then just stick to playing records for pleasure. Note, I did not say absolute pitch, and I will come to that later.
Now to the ‘kit’:
The starting point for any transfer of a disc record is the turntable. The requirements are:
- Speed stability
- Low wow and flutter
- Variable speed, and a way of measuring it
- A high quality, rugged pick-up arm with plug-in head-shells
- A suitable cartridge
- A variety of styli (needles!)
Speed stability is vital, otherwise your transfers will start in one key but perhaps end up in another. This is not the same as recordings that start at one speed and finish at another.
Wow and flutter or the various types of slow and rapid cyclic speed variation, produce unpleasant effects. For this reason direct drive turntables are the sine qua non for the job. They also tend to incorporate.….
Variable speed. The expression ‘78 rpm’ is a misnomer. Everything recorded at from around 60 to 100 rpm is termed a ‘78’. Unless you are just producing basic transfers for a professional engineer to work on, you will need to get the pitch, and therefore speed correct. To gauge the speed of the turntable, you either need a set of stroboscopic discs, or to have a unit with, or the facility for a digital speed read-out.
The pick-up arm should be capable of hard work. The average ‘high-end’ hi-fi arm is totally unsuitable. Not only are they too delicate for the playback conditions encountered with the 78, they are almost impossible to match, in terms of mass, with the typical cartridge/stylus assembly used for 78 rpm playback. A removable head-shell system is needed because you will require a wide range of different styli. Yes, you can just have one head-shell and change the stylus assembly in the cartridge instead. When you write out the cheque for £100-odd pounds for the first stylus-cantilever you damage (and you most surely will), the point will become obvious and painful.
Cartridge. As with the arm, hi-fi units are not suitable. In fact, DJ types are some of the best for the purpose, being rugged, reasonably priced and capable of very good, clean sound. They will also allow you to use extra playing weights such are frequently required for ‘difficult’ records. Shure are generally thought to be the best for our purposes.
Styli. Unlike the LP (although there are exceptions with early ones), the groove-wall geometry and dimensions of the 78 rpm record were not standardised. For this reason, to enable proper tracing of the groove, you will need a wide variety of styli, each with different specifications. They are usually bi-radial or elliptical in shape, with lateral dimensions measured in thousandths of an inch. Ideally you should have:
- 0.0080” (for certain vertically recorded discs)
As to deciding which may be correct for a given record, all I can say is that there are no rules, only a guide.
This is probably the right point to state that, whatever you may read in print about the correct stylus for this make and era of record (or for that matter the right equalisation for that make of record), ignore it. There are far too many variables to produce sensible charts giving the correct data for the various makes and dates of records. There is only one correct stylus for a record, and that is the one that gives the least noise with the cleanest sound. Let me give you an example.
That scrape is typical of the problems encountered with Odeon, Columbia & Parlophone.
In reverse this time with a 1927 unpublished Chaliapin side:
The 0.0025” highlights the scrape, which is endemic to the master, but the 0.0032 virtually eliminates it.
The correct stylus can vary according to the pressing. I have encountered various pressings of a particular record by a singer. The 1926 pressing required a 0.0035” stylus. A slightly later pressing had some slight wear—that needed a 0.0030”. Both were HMV pressings. The Victor pressing, from exactly the same master, is best with a 0.0032”. I also have a recently made vinyl pressing which is perfect using a 0.0025”. All were pressed from stampers made from the same master.
Of course eventually, with experience, you will get to know a likely starting-point. Most HMV or Victors will track well with a 0.0032” or 0.0030”. Pre-revolutionary Russian HMV like a smaller tip—around 0.0021”. Parlophones and Odeons, together with pre-1931 Columbias, seem to like a 0.0028”, but as always there are wide variations. Experiment, and then listen and compare the results: that is the answer.
There will be times when through damage, wear or pressing problems, you will need to use the signal from only one groove-wall, or sometimes a non-equal mix of the two. The former is usually accommodated at the pre-amp stage and the latter with a mixer or when signal processing.
PATHÉS AND OTHER VERTICALLY RECORDED DISCS
Whilst almost all the records you are likely to encounter were recorded laterally, some, most notably Pathés made in France during the acoustic era, were made with the modulation in the vertical plane, in similar fashion to cylinders. To replay these successfully (and success is very relative with these discs), you must first activate the ‘lateral/vertical’ switch on your pre-amp. This changes the phase of one leg of the cartridge output. If the pre-amp does not have this facility, you need to do this manually by changing the pin connectors on your chosen cartridge.
Pathés do not have deep grooves that safely guide the stylus/arm assembly. It is essential therefore that you make sure the side-thrust compensation on the pickup arm is properly adjusted, otherwise the pickup will skate over the surface. The earliest discs were recorded at between 90 and 100 rpm, the later ones at a nominal 80 rpm. Many enthusiasts (and yes, there are some) of this type of disc, prefer to use a special ultra-wide stylus (around 0.0080” spherical): I have rarely found it necessary. A 0.0035” bi-radial usually does just a well, but you do find differences and a smaller one may be necessary. As always, experiment. Almost all Pathés are mechanical dubbings made from large master cylinders, which is why they have a characteristic tinny, nasal sound, with a worse LF response than lateral acoustic recordings. Trying to equalise for these discs is not easy, and no rules apply. Boosting the bottom end is difficult because of the rumble which is exacerbated by the lack of vertical component cancellation necessitated by the requirements of reproducing the format.
By way of comparison, a contemporary acoustic HMV:
The only other type of acoustic vertically-recorded disc you may come across is the Edison Diamond Disc. They are about a quarter of an inch thick and quite heavy, so, as E.L. Wisty said in another context, be very careful not to drop them on your foot. They have more of a groove than the Pathés but it is much finer, so a narrower stylus is required. Playing is around 80 rpm and despite the rather low recording level, they can be quite good: Edisons were not dubbings.
In the electrical era, the vertical format virtually disappeared except for some radio transcription discs, though as such it survived well into the 1960s. Usually recorded to a special NAB curve, they generally play very well.
Since we are discussing the tracking of the groove and recovery of the information held therein, it is perhaps appropriate to mention dirt.
Only play clean records. If a record is dirty, the stylus will be tracing the outline of the dirt and not the groove, and reproduction will suffer. Even if the record looks clean, dirt and the steel or fibre fragments from needles used in the past will lurk in the grooves. You would be amazed at the residues collected after records are cleaned by a machine. Such units wet-scrub the record and then vacuum dry. Top of the range is the Keith Monks used by studios. Much cheaper but effective are the Nitty Gritty, Loricraft and Moth machines, but before buying make sure you have a version meant for 78s. Some hold the record down by a variety of means, and this dishes the record slightly – not ideal for breakable 78s!
Manual cleaning is not recommended. It is time consuming, likely to leave liquids on the surface for too long, and can damage the record. A proper cleaner is a necessary part of the tool kit.
RECORD SPEED AND PITCHING
A (literally) key decision is that of record speed and pitching. There is a separate introduction to this topic, Speeds and Pitching of 78 rpm Gramophone Records .
If you have got this far, you now have the initial equipment for playing a 78, and are able to cope with the mechanical variables you are most likely to find, with one exception, the swinger. No, not a 1960s hangover, but a record where the centre-hole is not exactly centred—it manifests itself by causing the pick-up arm to wander to and fro whilst tracking the disc, and it produces audible ‘wow’. The only simple and non-damaging way to deal with the problem is to use DJ slip mats to build-up the platter height to a point where a record’s spindle hole is almost above the high point of the turntable spindle, thus giving some freedom of movement about the spindle. With the pick-up tracking the record, note the movement in direction towards you, and then simply tap into alignment. Try it, and you’ll soon become used to getting the disc centred.
By this point, you will have also spent quite a lot of money. But worry not, it gets worse!
Whilst this really refers to the pre-amplifier, which prepares and amplifies the signal from the cartridge, just as important in this section is the monitoring equipment—namely power amplifier and loudspeakers. A good quality amplifier is essential, as is a pair of neutral-sounding speakers. As a general rule, ordinary hi-fi speakers are not suitable, often being highly specific in character. If your speakers colour the sound, how can you tell whether you have got the right result? Of course, you cannot. Use proper monitors and if possible, near-field monitors—they are less susceptible to room acoustics. The listening environment is important. Make sure the room has plenty of furnishings and other items, especially on walls, to break up the inevitable sound reflections.
EQUALISATION AND THE PRE-AMPLIFIER
From the stylus/cartridge assembly the electrical signal must be fed to an equalising pre-amplifier. This is something required for all disc replay systems, and its performance & specification will have a great impact on the final result. Since it is also dealing with very low level signals, system noise becomes extremely relevant. Why is equalisation required in addition to amplification? The answer lies in the way in which records were (and still are) made, and how system limitations were overcome.
Disc equalisation involves increasing or decreasing the amplitude (loudness) of certain frequency bands—usually gradually—over a range of frequencies, in order to compensate for planned reductions or increases introduced during the cutting of the master. This of course only relates to electrical and not acoustic (mechanical) recording. For example low frequencies cause the cutting stylus to cut a groove with a greater lateral excursion than high frequencies, thus the stylus on playback has more difficulty tracking it. It also means fewer grooves (less music) per inch can be fitted onto a disc. It was the practice therefore to progressively reduce the volume level at lower frequencies when cutting, and for the playback system to restore the relative levels in the first stage of the reproducing amplifier.
At the other end of the spectrum high frequencies produce less signal strength compared to the amount of noise. In the last days of 78s, this property was use by some companies, notably Decca, to apply a pre-emphasis at high frequencies. On playback, the necessary reduction in the HF resulted in reduced surface-noise levels. This idea, along with modified low frequency equalisation, became the standard for LP and after several changes resulted in the RIAA standard curve for recording and replay of microgroove LP.
The pre-amplifier that you use, therefore, needs to be able to accommodate and equalise for a variety of different ‘published curves’, for they are the starting point. Names for these might include ‘Westrex’, ‘Western Electric’, US Columbia, English Columbia, HMV, RCA, Orthophonic, NARTB, Decca, FFRR and a host of others. Many are very similar to each other so don’t panic if your pre-amplifier does not list them all.
As to which pre-amplifier to use, let me say first, that original valved (tubed) equipment is not suitable (and indeed may now be dangerous) even if re-built and serviced. Fine for listening pleasure (if safe!), but for technical reasons connected with noise reduction systems they are to be avoided. For serious transfer work, the Danish Elberg is very good indeed with all the pre-set curves you are likely to need. At around a £1000, it is not inexpensive. There are other professional units with even more control and variability, but their much higher cost means that only a professional facility could justify them. The prime example is the famed Ted Kendall 'Front End'. With highly accurate replay curves (correctly calibrated and described in μseconds), it has the phase integrity and rise times of the standards required for CEDAR to operate to its maximum capability. That is why all the top disc restoration facilities use it. How much? Believe me, it is best not to ask just yet. An American company, KAB, makes a cheaper unit than the Elberg, and although very good, you do get what you pay for. A lot of work goes into designing low level audio electronics that will handle considerable amounts of equalisation with good phase characteristics. And anyway, who said audio restoration was a cheap business.
WHICH CURVE TO USE?
The answer is simply, the one that sounds right. Why cannot one just simply look up the make and year of the record and follow the published curve? Read on.
The equalisation to which we have been referring, is that which was built in to the recording system. For example the system might be designed to be constant level from 6,000 cycles per second down to 250 cycles per second, and then reducing at a set rate down to 60 cycles. This is the published curve. But there is a another type of equalisation in the recording process - that introduced arbitrarily by the engineers either as a one-off at the recording session, or by the record company as a general policy so as to produce the sound they wanted. It is similar to, if simpler than, the range of equalisation options available in today’s mixing consoles. One should also realise that in the 78 era, microphones were not as uniform in response as today, so a further response characteristic is added to the mix!
It is thus important to remember that system equalisation is quite different to the response modifications introduced by engineers or microphones. In reality of course, it is the same as for a modern LP. Two recordings made in the same studio on the same equipment with the same performers can sound totally different, and were meant to, even though each was cut to the RIAA curve.
Not understanding this principle is the reason why there is such a lot of rubbish talked about replay settings. Anyone who tries to state that, say, HMV for 1929 was ‘this’, and for 1932 ‘that’, just does not understand the principles and the reality. A knowledge of recording characteristics is useful, indeed essential, but the ear is the only arbiter. Any guide that purports to state what curve should be used for each make and era of 78 should be avoided. Your ear and sonic experience should be the guide. Start with one curve and then try others. Choose what sounds best, then if you feel the need to tweak it, that can be done manually at a later stage – see ‘Signal processing’.
ACOUSTIC (MECHANICAL) RECORDING
For acoustic recordings, replay equalisation is a matter of attempting to correct for unplanned and possibly random mechanical deficiencies rather than calculated electronic characteristics, since the original engineers had no means of altering the balance of frequencies during mastering. But the nature of the equipment, especially the horn, introduces changes to the balance of frequencies, and modern equalisation can be used to compensate for that. This can only be done by ear, but in general, you should avoid too much low-mid frequency boost in an attempt to compensate for the poor LF response of acoustic recordings. Better to choose a curve similar to the Western Electric, and then manually make adjustments later on.
One thing to remember when using any pre-amplifier in a non-stereo mode, is to use the ‘mono’ facility to electronically parallel the left and right outputs from the cartridge. This will help to reduce distortion and low frequency rumble which may be part of the signal. Never wire across the cartridge pins. Some Shure 78 rpm cartridges have this automatically done with wire twists – remove them. Someone at Shure forgot a basic principal relating to impedance. Having said that, there are times when you need to extract the signal from just one groove wall due to damage or pressing problems. You can of course use a mixer to do the summing of the two channels.
I have naturally assumed that you have also connected your pre-amplifier to the monitoring system, so that you can hear what is going on.
Like it not, we are in the digital era, so the next stage is conversion from analogue to digital. The sooner the signal is in the digital domain the better: less chance of added noise and other analogue problems. There are two ways to achieve this, a standalone Analogue to Digital converter or a computer soundcard. Unless the soundcard is a full specification unit, probably costing £500 or so, it will not be suitable for the purpose. Indeed, many practitioners, rightly in my view, prefer to use a separate and purpose-built A/D converter. At the bottom end of the professional ladder, the Midiman appallingly-named ‘Flying Cow’ (don’t ask) is a very good 24-bit converter capable of external clocking (needed to ensure uniform sample rate throughout a system). It has a D/A converter as well. With a larger budget, the Apogee range is first-rate. There are others that cost as much as a half-decent used BMW!
We have now come to the first point in the chain where volume level becomes critical. Although digital systems have an enormous dynamic range compared to analogue, they do have an absolute ceiling. Effectively this occurs when all the ‘zeros’ are ‘ones’, and that peak level is expressed as ‘0 dbfs’. Try going much over that, and you will have no doubt that you have reached the brick-wall. Small ‘overs’ may be corrected by a D/A converter’s error correction system, but don’t rely on it. Remember, all digital audio has errors, and these are dealt with by error correction systems. You might see in the specification for a digital replay system: ‘double-encoded Reed Solomon error correction’.
In view of the importance of keeping within the level limits, high accuracy metering is essential. Whether you are using a stand-alone A/D converter which will have usually have a small LED bargraph, or a computer based system with an on-screen bargraph, the included metering is rarely very accurate. The best solution is a purpose-built meter. There are three recognised makers: RTW in Cologne, NTP in Denmark and DK Audio, the latter being more used in television for their on-screen metering. RTW meters are as good as any and have been the mainstay in studios for years. Choose a digital model, preferably with peak hold and display. That enables you to see at a glance the actual level reached: for many tasks that figure will be important, so make a note if you have such a meter. The 120 series desktop meters are ideal, especially if you use computer-based systems.
In general it is best to drive the converter to within 2 db of ‘0 dbfs’ or in other words, -2db. That 2 db is a useful safety margin. If you peak a lot lower, you will effectively increase the noise-floor when the signal is increased to bring it into line with normal practice. Low level distortion will also increase.
Now that you have a good, equalised analogue signal converted into digits, what do you do with it? If your mission is to provide flat and untouched transfers for a professional engineer or studio to work on, then a straight digital recording on to CDR or DAT will be then the end of the process. Many archives will want the transfer in this state, although they may also want a fully processed transfer for listening purposes.
THE FULL MONTY
So, you want to go further, the whole shebang. Are you sitting down? Good. Now prepare to re-mortgage your house, if you haven’t done it already to get this far.
Once the signal is in the digital domain, the next consideration is noise reduction: that is to remove as much of the surface noise as possible without significantly or audibly degrading the musical signal. There can be a very great deal of noise: the result of the recording and stamping processes, the pressing material, and through wear and damage to the disc. The background noise from a 78 is a mix of hiss, (although very different to tape hiss), crackle, and clicks. and it can be relatively loud. Indeed it occasionally overwhelms the musical signal so that, not entirely in jest, we sometimes refer to noise-to-signal ratio instead of the other way around. To most listeners it would be totally unacceptable to leave everything untouched with all noise intact: it is the musical performance that is wanted, not the distraction of what at best is like gently sizzling bacon (frying cumin seeds for non-carnivores), or at worst a hail storm. Noise-reduction is a whole subject in itself, and I have simplified things in this description.
Noise can be simply, if only partially dulled by using an analogue filter to remove all the frequencies that could not have been recorded on the disc; and for practical purposes, for most 78s until quite late on, any signal above about 6 – 7 kcs is almost certainly mainly noise and not content. Cutting it out should produce an immediate improvement in the sound. Unfortunately, it doesn’t, mainly through the effect of electronic filters not behaving as perfectly as we might wish.
This is perhaps the appropriate point to think about, and delineate the nature of the noise that is part and parcel of the 78 rpm disc. There are many components, but for our purposes, the most significant are:
- Clicks - The click is a short-duration burst of non-musical, initially analogue information. It is usually caused by damage to the groove surface by mishandling – for example a deep scratch. Providing the replay stylus can read it correctly, and the following electronics handle the signal there-from with sufficient speed or waveform rise-time, then digital noise reduction systems such as CEDAR can generally eliminate it as being non-musical information. Natural music wave-forms do not have the same vertical beginning as that produced by a click. Very large and long duration clicks however, caused by serious damage to the grooves, need to be removed by a manual editing process. See section headed ‘Editing’.
- Crackle - In essence, crackle consists of a dense series of very short clicks. Its cause is surface imperfections in the record. This may result from defects in the original wax, cutting stylus, electroplating process, stamper wear, and pressing-cycle faults, but the largest proportion is related to the composition of the pressing material. Like clicks, when properly reproduced the initial rise-time of the waveform is to all intents and purposes vertical, rather like the beginning of a square wave. As with clicks, true noise reduction systems can recognise that energy surge as not being music. At this juncture, it is necessary to re-iterate that both stylus and electronics must be capable of reproducing the clicks and crackle with the greatest accuracy. This is where high performance ICs become a necessity. Valved (tubed) electronics may give a warmer and more rounded sound, but that is not what is required for digital recognition of the instantaneous energy changes that characterise clicks and crackle.
- Hiss - After clicks and crackle have been removed, we are left with the residual noise caused by the stylus tracing imperfections in the record, which have a wide variety of causes. Depending upon the record material used, this may be at best a very unobtrusive background ambience, or at worst, a very dense ‘shush’ noise. Although we tend to call it hiss, it is not really that. True hiss is random electron noise of the sort generated by electronic systems and mimicked by magnetic tape. Residual shellac noise is very different and is more akin to an ultra dense crackle but without its velocity or amplitude. Thus when de-hiss systems designed to deal with tape hiss are used to try to remove residual shellac noise, the results are less than satisfactory. This is usually where unpleasant artefacts begin to show through. ‘Footprint’ systems that sample the noise and then digitally remove anything corresponding to the sample, are no better since although the noise may appear aurally to be constant in character, it is not.
- Rumble - Usually associated with poorly designed replay turntables, rumble or low frequency interference, in the recording chain, stems in the main from less than ideal bearings in recording lathes. Whether caused by poor or incorrect maintenance (and the wrong grade of lubricant can have a devastating effect), the result is an audible rumbling that has many harmonics. Certain makes are known to be worse than others. In particular, the early Fonotipia recordings are often badly deficient in this respect. Of course, if the rumble is low enough in pitch, contemporary acoustic replay would not have shown up the fault since its LF response was incapable of replaying it. The cure is exceedingly difficult! The random and relatively wide-band nature of rumble makes it hard to remove entirely. Various harmonics can be notched-out with filters, but some will always remain, unless of course one wishes to go to forensic lengths, in which case the music will suffer badly. Essentially, the worst is ameliorated and hopefully what is left can be overlooked!
CEDAR AND REMOVAL OF NOISE
There is no question that CEDAR noise reduction systems are the industry standard. If you cannot afford CEDAR, then you cannot compete at the highest level. With other systems, almost all computer based, you run the risk of signal degradation when attempting to remove more than often minimal amounts of noise.
You can go down the CEDAR route two ways, either with standalone units such as the Duo series or as part of a computer-based audio restoration, noise suppression and signal processing system. Either way, use the system to remove clicks and crackle. Do not try to use any de-hiss feature unless removing tape-hiss. At best there will be subtle and easily overlooked changes to the music, and at worst you will introduce terrible artefacts. Most listeners easily accommodate low-level constant-character hiss behind the signal. Indeed, it can be beneficial in giving an aural impression of a wider frequency range.
When first using CEDAR, take some time to experiment with settings. In particular the de-click function can ‘growl’ on certain voices, especially with speech. Orson Welles for example, gives a lot of trouble – nothing new there I hear you cry! On a wide-range recording, the attack on brass instruments can catch. If you get these effects, simply back-off during the parts that cause the problem. Just as with a motor car, you have to learn to drive the system, then practice. One day, just as with some of the processes to follow, you will think you have mastered it. Shortly thereafter, a record will present itself for transfer with problems to make you realise that you have not. You never will. Some of us have been doing it for many years and still find fresh problems appearing.
Having de-noised the signal, you will probably want to process it further, perhaps some gentle response shaping or possibly horn-resonance removal. The record may have hum that requires treatment. There may be a host of things to do, or very little.
Whilst all these and many other tasks can be achieved with computer-based recording systems, there is no doubt that a suitable digital mixer is the simplest solution. Mixers are far easier to use than computer programmes for signal control, distribution and processing. That is why there have been substantial moves to add a conventional mixer-type control surface to computer systems. I rest my case.
Whichever system you use, always remember that if you need to make drastic alterations to the overall response at this stage, then the chances are that something is wrong at the pre-amp stage – possibly the wrong curve has been selected.
Dealing with hum can be tricky. 50 cycle (60 in USA) hum is not too much of problem to notch out with a tight ‘Q’ filter. ‘Q’ relates to the width of the notch. The higher the ‘Q’, the tighter the notch, and thus it will have less effect of frequencies either side. However, hum is usually most obvious from its harmonics say at 100 or 150 cycles. These are bang in the middle of the music’s bass. Just be careful not to throw the baby out with the bathwater by overdoing it. Cedar do have a module for doing this, but it is expensive and not always ideal our purposes.
This is also the time to deal with rumble:
This is also the stage where you can use a low pass filter to reduce some of the residual HF noise. Be careful not to damage the top music frequencies by overdoing it. Also at the HF end is ‘cold wax chatter’. Caused by a too-cool wax being cut, it usually occurs at the end a side and is usually around 10 kcs. A notch filter will usually get rid of most. Best to note where it begins to be audible, and then gradually apply the notch.
Whenever making changes to the sound, always use the A/B comparison button, and use it often. Make sure of the integrity of the original. The ear quickly makes allowances and can fool you into thinking all is well. You are now in an area where such concepts as taste, musicality, even musicianship become pertinent. When you come to the editing process, these become even more important.
This covers a multitude of sins. At its most basic, it is simply topping and tailing each track. Even here, there are pitfalls. Always try to ensure that the opening surface noise is faded up fairly slowly: 2 seconds is about right. That gives the ear time to get used to the background: a sudden ‘in’ on the initial attack of the first note does not. Similarly with the fade out. Here I would recommend using a manual fade at the (mixing) desk if you have one. You can adjust it in a way that automated fades can never mimic. One problem you will inevitably face is the recording that goes into the (noisy, and often added later) run-out groove within a fraction of a second of the music finishing, and occasionally before! All you can do is to add a fade down from some similar blank surface noise. You will need to build up your own library of these.
Large clicks, clunks and other noises not removed by De-click, must be treated by manually. This involves wave-form editing. There is no substitute for practice here and there are few rules to help you. What works on one click will not work on another. The only thing one can say with any certainty is that you should always try to edit on ‘null’ points, that is the where the wave is on the centre, ‘zero’ line; and always try to join by finding a similarly shaped wave-form to join up to. It frequently depends upon what is going on around the click. If relative silence, it is easy. When in the middle of very ‘busy’ music, it can be a different matter. With singing, speech or rapid note changes, it is always possible that the click straddles the attack or even the note. A value judgement needs to be made.
- 20. An example of damage – in this case a bad scratch. Germaine Lubin, unpublished Liebestod (Tristan 1938) (mp3 file)
- 21. As 14, but using CEDAR de-click. Note two clicks remain. (mp3 file)
- 22. As 15, but the two clicks manually removed. Note, the second still leaves a faint trace, but further editing will cause a timing problem. See next example. (mp3 file)
- 23. As 16, but with the second click fully removed. The result is not acceptable. (mp3 file)
Many classical pieces are longer that one side of a 78. Occasionally movements may simply fill a side each, but that is rare. Here, simple cross-fading between sides should be sufficient, although be sure to keep a reasonable gap between. 4 or 5 seconds is generally enough. If there is not enough blank background then you will need to duplicate some and add it in. Cross-fading seamlessly is not easy, and sometimes it seems impossible, especially if a work was recorded using a variety of takes from differing balances. But persevere: practice and experience is the only way, and eventually it will get easier. One thing never to do is to fade down to digital silence between movements of a work. Nothing is more annoying – keep the background constant.
Of course, not all movements end exactly at the side ends. Here you face a multitude of problems, some of unsolvable. Why?
It usually comes down to the way in which sides were ended. Occasionally, though not often enough, there is an actual overlap where a phrase is repeated at the end of one, or the beginning of the next side: cross-fading is fairly straightforward here, but you will find that most sides end slightly unnaturally, with ritardandi, and even occasionally with a cadence not written by the composer. These practices were an attempt to ‘finish’ each side less starkly. Whilst you can simply cross-fade, the performance at this point does not represent that which the musicians would have given if playing continuously. One solution, although it does not always work by any means, is to digitally shorten the ritard to the correct length. It can be a complicated process involving much trial and error, and sometimes mimicking the studio or hall reverberation to ‘naturalise’ the shortened note, depending on the way you have to shorten it. As to the ethics of this practice, it will be up to you to decide, but there are times when doing nothing about it ruins the performance. What would the original performers have wanted you to do? It is a dilemma, and whatever any one tells you, there is no right or wrong answer. It is right for the listener who wants it that way, and wrong for the one who does not. One can argue that it is being more faithful to the composer’s intentions.
For some instances of editing situations, the following examples have been prepared.
- ‘To Keep my Love Alive’ (late 1930s Broadway)
A frequent problem is having to join sides that don’t properly follow on. In the example you’ll hear, the first side is given a false finish (Editing Example 1 (mp3 file)) and the following side starts oddly rather than following on (Editing Example 2 (mp3 file)). Be warned, this is not your normal fare!
The final version (Editing Example 3 (mp3 file)) involved removing the false end and start of the next side, adding a phrase repeated elsewhere, and then cross-fading them together. This is how it turned out, and is reasonably faithful to the original score.
- ‘Look Around’ (Acoustic musical comedy ca. 1914)
A more unusual problem, is having to put right a previous engineer’s mess on a transfer. In the next example from an LP by a well known company, a recording of a musical comedy piece, the pick-up has jumped a groove and it’s not been noticed by the engineer. How, I cannot imagine, but perhaps the pub was about to open. (Editing Example 4 (mp3 file))
Putting this right involved finding the same phrase elsewhere in the record and then inserting it instead of the damaged section. I recall that when I did this one, pitch correction had to made, as the key had changed. Thankfully it turned out alright, but it was over 2 hours’ work. (Editing Example 5 (mp3 file))
- Antonina Neshdanova: (Les Huguenots) 1914
Back to the good stuff! And an Historic Masters issue, HM 122. On one side a fabulous version of a Huguenots (Meyerbeer) aria. Here’s a short extract (Editing Example 6 (mp3 file)).
Faultless? Listen to it again: this is how it really went. (Editing Example 7 (mp3 file))
If I were going to use that on a CD, I would have to correct that fault, but explain in the notes what had been done. Why it was issued like that I cannot understand. I’m sure the artist wouldn’t have agreed, so I’m happy with assisting; and it is a fabulous performance. It simply required the slip excising.
- Ernest Lough (of ‘Hear my prayer’ fame) 1927
The next one’s a rather more difficult problem ethically. Here Ernest Lough unusually doesn’t quite get right the pitch on the attack of the solo entry. (Editing Example 8 (mp3 file))
It’s only the merest fraction, but it annoys me. It can be put right. (Editing Example 9 (mp3 file)). That was done by manually adjusting the pitch in real time by just fraction. Whether one should do that is another matter. Kirsten Flagstad required help in this way with a poor entry at the start of a two sided disc. The join stood out like a sore thumb.
- Tamagno, Otello death scene. 1903
Sometimes there are inherent faults in the record, if serious, often the reason for non-publication. This next one is such a record, a vivid unpublished recording of ‘Nium mi tema’ by Tamagno made in 1903. He was the creator of the role and was specifically chosen by Verdi. There are several faults as a result of the cutter not operating properly. This is one of them. (Editing Example 10 (mp3 file))
Correcting it involved copying the syllables spoilt by the damage from the undamaged section before and after, and then substituting. I see no problem in doing that providing an explanation is given as to why. (Editing Example 11 (mp3 file))
- Lotte Schöne and Willi Domgraf-Fassbaender (Don Pasquale) 1931
Here, the engineer has not left enough blank grooves at the end of the side, making a good fade impossible. (Editing Example 12 (mp3 file))
Whilst I could have copied some identical surface noise and added that, what I did was to use orchestral ending from the Italian version of that side recorded the same day. (Editing Example 13 (mp3 file))
- Pastorale Sonata, Artur Schnabel
This precious recording from February 1932 came from a sadly damaged set of tests once owned by the artist. Originally to have been part of the complete Beethoven Sonata Society series, unaccountably it was remade a year later just prior to its scheduled release. EMI files can tell us nothing as to the reason. The damage consisted of small half-moon chunks out of the beginning of each disc. The first track is in totally un-processed form. (Editing Example 14 (mp3 file))
The clunks and missing parts are far too large to manually remove without trace, so those pieces were taken from the 1933 version. (Editing Example 15 (mp3 file))
- Missa Solemnis
In the late 1930s, Victor decided to record two live performances of the Missa Solemnis conducted by Serge Koussevitzky. For various reasons the resulting sides were not all suitable for publication as they stood. Victor’s Musical Director, Charles O’Connell decided to make transfers in an effort to correct the problems. This was done, but in view of the results, it is rather assumed that the task was an ‘after-lunch’ job, and since Charlie like to uncork his meals…. The main problem was that sides were either faded down too soon or faded up too late. Missing bars of music were the inevitable outcome. Two of the offending sections show the problem (Editing Example 16 (mp3 file) & Editing Example 18 (mp3 file))
It had been originally intended to issue a CD version exactly as it was, but that really would have been unacceptable to the average listener. However, with a bit work, the gaps were filled by finding similar bars elsewhere. (Editing Example 17 (mp3 file) & Editing Example 19 (mp3 file))
- Kurt Weill. ca.1929
This was an existing CD release from which a compilation was to be made. The original transfer had a very bad tape edit and the music jumped. (Editing Example 20 (mp3 file)) By a process of lengthening the affected note through duplication, the jump is almost imperceptible. (Editing Example 21 (mp3 file))
- Seinemeyer & Jung (Flower Duet - Madama Butterfly) 1929
This is typical of many of the side-joins you may face. The final note of the first side is lengthened. (Editing Example 22 (mp3 file)) The same note is repeated at the beginning of side 2 (Editing Example 23 (mp3 file)).
By cross-fading over the last note of side one, a seamless join is achieved. (Editing Example 24 (mp3 file)) Cross fade specification: 1 second, 50%, 1 second fade-down, I second fade up – both normal profile.
- Flying Dutchman overture, Beecham 1937
Much trickier, as you can immediately hear. Made on different lathes (takes 2 and 2A respectively), each with different settings. The end of side 2 (Editing Example 25 (mp3 file)) finishes with a truncated chord. Beecham also wobbles slightly just prior which upsets the drive. The beginning of side 3 (Editing Example 26 (mp3 file)) has the full chord, but begins scrappily in terms of attack. It also increases in level as the chord progresses. I corrected that on transfer (approximately 2db reduction midway before a gentle return to correct level).
The side-join will never be perfect, but by setting the cross-fade early enough and shortening it, a fairly good join can be achieved (Editing Example 27 (mp3 file)). Cross-fade specification: 550ms 50% 1 second fades.
I touched on levels earlier, and made the case for accurate metering. If you are producing a simple set of unconnected transfers, then so long as the levels are within say the -5db to -2 db mark, you have nothing to worry about. However, if your aim is a compilation designed to be heard at one go, then you need to make sure that apparent levels, or loudness, are roughly comparable in line with the music. With popular music this generally entails what is called normalising, or bringing all the peak levels in each track up to the maximum. For classical work, that is obviously wrong. It is not a good idea to have say a quiet Chopin nocturne sounding as loud as Finlandia played by Stokowski and his Philadelphia boys. That said, for normal listening, you cannot reproduce the true level differences one would hear in the concert hall. The simplest method is to get the loudest piece to the maximum and then set the others at the levels which sound comfortable and make musical sense.
Generally this is for the professionals, but for ordinary use, a CD writer is fine for the end result and is quite straightforward. Make sure track points are 1 second before the onset of signal, with 2 for the first track. Gaps between tracks for classical music should be at least four seconds, preferably slightly more. Whether your end product is suitable for replication will depend upon how good your CD writer is at writing with low error rates. The average CD writer is frequently built to a price. Replacing with a better quality one will be a good investment. You may also need one that supports various texts and recording codes; many do not. The Freecom Portable is one that does.
To transfer 78s (or indeed any analogue source) requires, as we have seen, specialist equipment, and this comes at a cost. If you wish to enter this rather strange business, and to do it properly, you will need to invest heavily in both time and money. In terms of the latter, a short summary of the likely cost (as at 2006) is given below. Not a cheap business by any means. It can be done on the cheap, but the results are a travesty of what can be achieved, and in many cases a downright disgrace. Always remember that you are the guardian of the artists’ reputations.
|Record cleaning machine||1600.00|
|Variable speed turntable||1000.00|
|CEDAR units /system||8000.00|
|Computer editing and recording system||2000.00|
|CD writing system||300.00|
If you add in racks, plugs, cables and hardware, it is not far short of £25,000. Many figures quoted are a minimum and can be easily doubled or more for certain items of kit. The famed Ted Kendall 'Front End' pre-amplifier will set you back £4500. Believe me, for the professional it is worth every penny. Of course if you wish to be able to work from reel-to-reel tapes, cassette tapes, DAT and other formats, then the cost will soon reach £40k.
Still keen? Wonderful, come and join us: there are too few who care enough. Some terrible things have been done in the name of transferring:
No further comment necessary?
© Roger Beardsley 2005