Though magnetic audio recording has been around since before the Second World War, tape recording in the form we know it today is really a post-war phenomena. The first generation of paper-backed tapes made by BASF have long since been superseded by modern polyester designs with vastly improved magnetic coatings. The mechanics and electronics of the recording apparatus have also benefited from technological developments, to the stage where, to all intents and purposes, current magnetic tape recorders represent the very pinnacle of the medium. It is hard to see any means of making further significant improvements -- at least cost-effective ones -- to the medium, and digital formats are already coming to the fore, as we all know.
The inherent properties of magnetic tape recording mean that it is far from 'perfect'. Harmonic and dynamic distortion are part and parcel of the medium, but these qualities are so ingrained in our audio culture that technically more accurate digital recordings often seem 'flat' by comparison. Inevitably, the great open-reel tape machine manufacturers, such as Studer, Tascam, Otari and the rest, will gradually phase out their production of analogue recorders in favour of digital systems, so this is probably the best time to acquire an analogue machine if you don't already have one!
Before we look at how the tape recorder works, perhaps it's worthwhile to spend a few moments considering the actual recording medium -- the tape. Recording tape is coated with a metallic compound capable of storing a pattern of magnetisation. This magnetisation is varied in strength and direction to represent the original sound waveform. In other words, the magnetisation pattern is analogous to the air pressure variations of the original sound.
Modern tape formats use polyester as the tape base, with a suitable magnetic coating. Polyester has the ideal combination of strength and flexibility, such that it will not snap or stretch under normal usage. The thickness of the backing material is important: if it's too thin, the tape transport may have difficulty in handling it (the C120 cassette being the most obvious and notorious example). If it's too thick, the amount of tape which can be wound on to a given spool is limited, reducing recording and playing times.
Typical 'Standard Play' open-reel tape is about 50 microns thick, with a full 10-inch reel of tape (the standard professional size) lasting a little over 30 minutes at 15 inches per second, or ips (a common speed for a professional machine). However, other thicknesses are available, notably the 'Long Play' version, which is only 35 microns thick. This type of tape is most commonly used on machines with particularly twisty tape paths, or where the spool size is restricted, the most obvious example being the infamous Nagra tape recorders used for location sound recording. In its portable mode, the Nagra will only accept 5-inch spools, and its tape path involves a couple of very sharp 90 degree bends, so Long Play tape is the ideal choice.
Most modern tapes are 'back coated', which means that they have a slightly matt and rough-feeling surface, as opposed to a shiny, slippery one. The back coating is to improve how the tape winds (by offering improved friction between layers) and to exclude trapped air during fast winding. Neat and even winding is important in avoiding edge damage to the tape -- the most common cause of drop-outs and poor head-to-tape contact. On the subject of edge damage, a correctly wound tape should not touch either spool flange but should sit between them, and any tape spools which are bent, so that they rub against the tape during recording or playback, should be thrown away and replaced with new ones immediately!
Depending on the requirements of the tape, the detail of the magnetic material used to make up the magnetic coating may vary, but is usually either a metal oxide or metal alloy compound. The most common material is gamma-ferric oxide (gamma referring to the shape of the ferrous particles) but chromium-dioxide formulations are also used. In the early to mid-'80s, metal-particle tapes were developed, primarily for compact cassette recorders, but few machines were equipped to make best use of the formulation, as metal tapes need extremely high levels of magnetisation (they have a high 'coercivity', which is the ability of the tape to become, and remain, magnetised). The majority of domestic cassette machines simply could not achieve the required field strengths, although those that could benefited from better distortion and signal-to-noise ratios. In fact the metal-particle tape has become far more commonly used for digital audio and video recording, where its very high recording density is ideally suited to the high-frequency requirements of digital formats. It should be noted that tapes intended for digital recorders have radically different compositions to those for analogue recorders, because the nature of the recorded information is entirely different -- never use digital-formulation tapes on analogue machines!
Fundamentally, recording tape is little more than a highly sophisticated rusty ribbon, the nature, shape and depth of the 'rust' particles bestowing the combination of properties the manufacturer is seeking.
The magnetic properties of recording tape are far from linear, and, if used 'raw', would produce very quiet but heavily distorted recordings. To 'linearise' the medium, a high-frequency signal (typically about 150kHz) called bias is used. The amount of bias needed to produce optimum results depends largely on the precise construction of the magnetic layer, and will affect output level, noise, distortion and frequency response. This is a point I'll return to later.
Another characteristic of recording tape is that high-frequency signals tend to be retained by the top surface of the magnetic layer, whilst lower-frequency components tend to be recorded throughout its full depth. This has a bearing on the requirements of the recording heads and the longevity of recordings.
There are many analogue tape recording formats available today, all broadly standardised around the world. The largest professional format uses 2-inch tape and the smallest uses eighth-inch:
2-inch: originally recorded 16 tracks, but now the international standard is 24 tracks on 2-inch tape and is probably the most common multitrack format in use around the world.
1-inch: an 8-track format which was also available as 16-track from some manufacturers.
Half-inch: widely used for mastering (in stereo), or occasionally for 4-track work.
Quarter-inch: there are a number of common versions: stereo (also known as DIN format); 2-track (also known as NAB format); 4-track; semi-pro 8-track and occasional 16-track machines (with noise reduction systems built in).
All these tape formats record across the whole width of the tape, allowing blank areas (guard bands) between the different channels or tracks, and therefore operate in one direction only. The upper edge of the tape is always track 1 (or the left channel in the case of a stereo machine) and the lower edge corresponds to the highest numbered track (right channel in a stereo machine).
There are also a couple of domestic formats, and these all use the tape as a bi-directional medium -- you turn the tape over when it reaches the end of one side, and play it back the other way. The compact cassette works in this way, of course, as does the now very rare domestic 4-track (two stereo pairs) on quarter-inch tape. The latter uses tracks 1 and 3 in the first direction (for left and right respectively) and tracks 4 and 2 (ditto) in the reverse direction.
In general, background noise and susceptibility to drop-outs (areas of tape where the recording material is absent or inefficient) is reduced as the recorded area of tape is increased. For this reason, multitrack formats used to maintain the recorded area ratio established by the 2-track quarter-inch format: two tracks on quarter-inch, four tracks on half-inch, eight tracks on 1-inch and 16 tracks on 2-inch. However, noise reduction systems allowed a reduction in recorded area for similar noise performance, hence the advent of the 24-track format. For the ultimate signal-to-noise ratio, though, greater recorded area is the only way to go and consequently the half-inch stereo mastering format is very popular, as is increased tape speed to 30 or 15ips as opposed to 7.5 or 3.75ips.
Where the recorded tracks are entirely related (as in stereo recorders), the guard band between tracks can be reduced, allowing larger recorded areas (decreasing noise), but increasing inter-track crosstalk. In the case of stereo material, provided the crosstalk is reasonable, stereo imaging will not be affected, so this is a reasonable compromise. However, on machines where the recorded tracks may not be related, or where absolute separation is mandatory, wider guard bands must be adopted in order to keep magnetic crosstalk to a minimum. Hence the key difference between a stereo (DIN) quarter-inch machine and a 2-track (NAB) machine is in the width of the recorded tracks and the guard band separating them.
The tape transport is critical to the performance of the entire tape recorder. An analogue tape machine must be strong enough to be able to support the heavy spool and capstan motors, and the tape guides, control panel and interconnection panels must be robust enough to withstand typical operational abuse.
The tape path involves a collection of guides, tension-sensing devices and anti-flutter rollers, as well as the capstan and pinch roller, all of which are constructed from non-ferrous (thus non-magnetic) materials. Some of the guides will be fixed pillars, accurately machined to support and control the alignment of the tape, while others are free to rotate on high-quality bearings. A typical tape path might include a rotating guide on a movable arm (used to regulate back-tension from the feed-spool), then a large-diameter rotating guide linked to the tape counter or timing mechanism. As the tape enters the head block, there will be a precision tape guide before the heads, and then a motor-driven capstan pulls the tape through at a controlled speed. A rubber pinch-roller, mounted on a moving arm, holds the tape against the capstan when the machine is recording or replaying.
Most professional machines employ three heads -- erase, record and playback. However, not all machines have three heads, as some combine the functions of the record and replay heads into a single head. This compromises the quality of both the record and replay functions but reduces cost significantly, and can simplify some other aspects of the machine. The advantage of a three-head machine is that recording quality can be checked by listening to the signal replayed from the third head. This is often called confidence monitoring: not only does it confirm the technical quality of what is going to tape, but the small replay delay due to the spacing between record and replay heads also verifies that the machine really is recording something!
The tape must be kept in constant contact with the record and replay heads to ensure good high-frequency response and uniform levels. This task was originally performed by felt pressure pads, but fortunately this crude and unreliable technique has largely been discontinued in favour of accurately controlled tape tension determined by the feed-spool motor. However, cassette formats and some older domestic open-reel machines still rely on felt pads. Any unsupported length of tape under tension is likely to resonate or vibrate as it is pulled across the heads and this may become audible as a modulating tone, or as speed variations. Many professional machines use a roller guide between the erase and record heads to help control this 'scrape-flutter', as it's called. As the tape leaves the head block, it might pass over a second tensioning arm and, finally, on to the take-up spool.
Most modern machines employ three motors, one for each tape spool and a third for the capstan, all with sophisticated servo-electronics to regulate their speed, torque and direction. The spool motors normally govern the tape tension between feed spool and capstan, and between capstan and take-up spool. The former is critical for good head-to-tape contact, and the latter for neat tape packing, although neither is made easier by the continually varying diameter of the tape spools! Some machines employ two capstan and pinch-roller assemblies, one positioned either side of the head block, with the aim of controlling the tape tension across the heads very precisely indeed. While this approach undoubtedly works extremely well, its accurate alignment is critical, and few manufacturers have felt the need to go to such lengths!
The speed of the capstan motor is possibly one of the most important aspects of the entire transport. If the overall speed is not correct, replaying a tape on another machine will result in pitch and timing errors. If the speed is not accurately maintained, short-term pitch variations will be heard as wow (low speed variations) or flutter (high speed variations), neither being desirable on a modern machine, of course. However, there are many causes of wow and flutter, not all of which are directly attributable to the capstan motor. Eccentric pinch-rollers and badly worn rotating tape guides are actually the most common culprits.
The vast majority of open-reel tape recorders are arranged so that the tape heads face the front of the machine and the backing side of the tape faces the operator. This is very convenient for mechanical tape editing, but means that the recording oxide comes into contact with all the guides and rollers, sometimes leading to scratches and drop-outs. In contrast, many European tape recorders are arranged so that the tape is used 'oxide-out', the head block facing away from the operator and the tape backing coming into contact with most of the guides.
The three magnetic heads of a professional tape recorder are each optimised to perform their own particular job, but fundamentally consist of what can be thought of as an iron C-shaped core, wrapped with a coil of wire. When an electric current is passed through the winding, a magnetic field is produced across the gap in the iron core, the latter being arranged so that it is in contact with the recording tape (see Figure 1). If the current through the head varies in direct proportion to the sound signal, a varying magnetic field is produced, and as the tape is pulled past the head, its oxide layer becomes influenced by this magnetic field, effectively retaining its magnetic state as it departs the head. The replay head is constructed in a similar way but operates in reverse, so that the magnetic field embodied in the recorded tape sets up alternating electric currents in the head winding. These currents are then amplified and processed to recreate the original audio signal.
The erase head may cover the entire tape width (normal on a true stereo machine), or it may be split to allow independent erasure of individual tracks. During the erase process it is fed with the very high-frequency signal (typically in the region of 150kHz or so) called bias, which I mentioned earlier, and the magnetic field this creates becomes weaker with increasing distance from the centre of the head gap (in both directions). As the tape is pulled towards and past this head, the magnetic particles in the tape experience gradually increasing levels of high-frequency signal, reaching a maximum value and then gradually decreasing to zero. The tape is not capable of storing this high-frequency bias signal, and so leaves the head with its magnetic particles effectively randomly magnetised. Thus any previous recordings will have been erased.
The record head has to impart an audio frequency signal to the tape, but in its raw form this signal is found to be grossly distorted on replay, a phenomenon caused by the non-linear characteristics of the oxide particles within the tape. The solution is to add a high-frequency bias signal to the audio signal, which effectively forces the magnetisation process to become far more linear (note that the bias signal is not retained by the tape -- it merely enables the recording process). The size of the gap in the recording head is not particularly critical, provided a sufficient field strength can be created, so relatively large gaps tend to be used to ensure that the magnetic field is large enough to reach the full depth of the magnetic layer in the tape. In fact, the effective recording zone is not in front of the head at all, but is actually at the point where the tape leaves the influence of its magnetic field.
The replay process relies on the generation of electric currents in a coil of wire wrapped around an iron core which gathers the varying magnetic field embodied in the tape. The inherent sensitivity of the replay head to magnetic flux means that it must be properly shielded from fields generated by the transport motors and mains transformer, so the head is usually enclosed in a protective case, often with a fold-down front piece which can be raised after the tape has been laced.
The physics of the situation are such that the voltage induced across the head windings increases in direct proportion to the rate of change of the magnetic field, thus the head produces a much greater output for a high-frequency audio signal recorded on tape than it does for a low-frequency one. This increase in output voltage with increasing frequency can be compensated for with a 6dB/octave equaliser, but at very low frequencies various magnetic anomalies result in an uneven frequency response, often referred to as 'head bumps' or 'woodles'.
The size of the gap in the replay head has a critical effect at high audio frequencies. When the recorded signal on the tape has a wavelength equal to the width of the head gap, there will be no net magnetic flux, so the electrical output from the head will be zero. This is called the extinction frequency, and the frequency response of the head falls rapidly as the extinction frequency is approached (see Figure 2). Clearly it's desirable to design the head such that the extinction frequency is well above the highest required audio frequency, so a very narrow head gap is used. While this was difficult to achieve in early tape machines, modern engineering has allowed the sub-micron sized gap.
Note that if a lower tape speed is used, the recorded wavelengths will be correspondingly shorter, so the extinction frequency will fall, potentially reducing the frequency range on replay -- a common problem with cheaper cassette machines, for example. The dichotomy between the head gap requirements for recording and replay functions also explains the preference for three separate heads in professional machines!
An interesting point related to the reduction in replay head gap sizes over recent years is that modern machines can extract a far better frequency response from old archive tapes than was ever possible at the time of their recording. As the recording process is largely independent of the record head's construction, recordings made in the late '50s and early '60s are frequently found to be of extremely good technical quality when replayed on modern machines. The quality limitations of early tape recorders were generally related to their large-gap replay heads and antiquated amplifier circuitry; the recordings themselves were often to a far higher standard than could be replayed at the time!
One of the most important elements of the complete head assembly is the accurate alignment of each head to the tape, the other head(s), and the international standards. Poorly aligned heads will result in frequency-response errors, level mismatches, phasing problems, and increased crosstalk or noise.
There are four critical head adjustments (see Figure 3):
Azimuth is the angle of the head gap relative to the tape, and should be exactly perpendicular.
Height governs the vertical alignment of the head gaps within the recorded tracks.
Wrap governs the position of the head gap within the contact area between the tape and head, which should be central.
Zenith is a related adjustment to make sure that the head does not lean forwards or backwards, but matches the vertical orientation of the tape, such that the Wrap is consistent for all tracks.
Precise and accurate head alignment is crucial to the overall performance of the machine and normally requires special line-up tapes and tools.
At the simplest level, the audio electronics within a tape recorder can be broken down into just seven functional blocks: the bias oscillator, record amplifier, record equaliser, bias filter, replay amplifier, replay equaliser, and output amplifier/metering.
Record signal path: This involves a record amplifier to determine the appropriate drive level for the input signal, followed by an equalisation circuit, which is used to ensure that the recorded signal adheres to a standard frequency response when replayed by a 'perfect' head. This equaliser counteracts the effects of magnetic losses within the record head, as well as the effects of self-erasure at high audio frequencies, and also applies a little bass boost to offset the inherent low-frequency problems during replay. As always in the professional audio industry, we like standards so much that we have lots of them, so there are four commonly used tape equalisation standards: AES (used for tape speeds of 30ips); NAB (defined for 15 and 7.5ips); and two IEC curves (one for 15ips and another for 7.5ips).
After the pre-equalisation stage (and some machines also include a pre-distortion stage in an attempt to further linearise the recording process), the signal passes through a filter circuit, before being combined with the high-frequency bias signal and finally passing on to the record head itself. The filter is called a bias trap and is purely to stop the bias signal from interfering with the preceding circuitry.
Replay signal path: This starts with a high-gain amplifier to boost the small signal extracted from the replay head and then passes through the replay equalisation before an output amplifier conditions the signal for the output connectors and drives the metering circuits. The replay equalisation compensates for the inherent rising 6dB/octave frequency response of the head and the fall in HF response as the extinction frequency is approached (particularly important for low-speed operation).
Bias Oscillator: This generates a pure sine-wave signal somewhere between 100 and 200kHz, which is used to feed both the erase and record heads during the recording process. The level of bias is critical for a given tape type and affects output level, distortion, frequency response and noise. As the bias is increased from a very low value, the output level at 1kHz also rises, eventually levelling and then falling (see Figure 4). At the same time, the distortion falls in a mirror image to the rising output level, rising again slightly as the output level falls at high bias levels. The background noise level of the tape rises with increasing bias but, fortunately, the build-up of noise is much slower than the gain in output signal level. The noise reaches a peak value at the same point as the maximum signal output level is reached, and falls very rapidly thereafter. The effects of biasing suggest that the best distortion figures would be achieved at the point of maximum signal output level. However, a slightly higher level of bias (over-bias) is considered optimum because this improves the signal-to-noise ratio by a useful amount with only a small increase in distortion.
At this biasing point, the tape sensitivity to high frequencies (say 10kHz) will be slightly reduced compared to the level at 1kHz, mainly due to partial self-erasure effects. This is often used as a measure of correct biasing. The optimal bias needed will vary with different tape formulations and constructions, and also with different batches of the same tape (although rarely by much in the case of the latter). It's therefore advisable to select one type of tape, align your machine for that brand and stick with it! If the tape type is changed, the machine should be realigned if the new tape is to be used to its best advantage.
An increasingly common system used on many top-of-the-range cassette machines and some professional open-reel recorders to address the problem of self-erasure at high frequencies is the Dolby HX process. This is not related to noise reduction in any way, but actually varies the bias level so that high-level, high-frequency signals are not over-biased.
Obviously, one of the requirements for a multitrack machine is that new tracks must be recorded in sync with previously recorded ones. Replaying earlier tracks from the replay head will result in non-synchronous recordings because of the physical spacing between record and replay heads within the head block. To overcome this problem, multitrack machines use the appropriate tracks of the record head to replay tracks at the same time as new ones are recorded on other parts of the tape, a process normally activated through head switching (labelled Sync, or Sel-Rep). On budget multitrack machines, electrical or magnetic crosstalk within the head assembly may result in howl-rounds if adjacent tracks are both recorded and replayed, but there are no such limitations on professional recorders.
The frequency response and noise performance obtained from the record head are rarely as good as those from a separate replay head, but are perfectly adequate for guide tracks during overdubbing. This aspect of tape machine design may impinge on the process of mixing bounce-downs, since making a sync bounce-down from the record head often results in a quality loss, whereas bouncing down from the replay head causes the mix to be delayed relative to the other tracks. With careful planning of the tracking and mixing order you can usually come up with acceptable results.
As I've explained, the workings of a modern tape recorder are complex in detail, but relatively simple in concept. The magnetic recording medium has its limitations, particularly in terms of signal-to-noise ratios and distortion, but also has an integral quality which many consider to be an important part of the whole recording process. Tape recording technology has long since reached the point of diminishing returns, where minute gains in quality are disproportionate to their expense, so it's predictable that low-cost digital formats will largely replace the analogue machine in the coming years. However, if you already have an analogue machine, don't throw it away -- it can produce sounds that no DAT recorder can even approach and can be put to many other uses too. If you don't already own one, perhaps now's the time to be on the hunt for bargains, as the foolish discard their valuable workhorse machines in favour of under-developed digital toys (controversial or what?)...
Stereo or 2-track quarter-inch tapes running at high speed are capable of signal-to-noise ratios of up to about 60dB with modern tape formulations, half-inch machines achieving even better performance. However, tape formats with smaller recorded areas will be correspondingly noisier, so noise reduction (NR) systems are frequently employed to help improve the performance of such systems. Multitrack formats and cassettes are the most obvious examples which generally rely on noise reduction, and the two most familiar NR systems are Dolby and dbx.
The dbx approach is fairly simple, using a fixed-ratio compressor to reduce the dynamic range of the signal to be recorded (primarily by increasing the level of quiet signals). On replay, a matching expander pushes low-level signals even lower, with the aim of recreating the original signal dynamics and reducing the tape noise correspondingly. This simplistic approach suffers from a few problems which are outside the scope of this article.
The various Dolby systems attempt to address the shortfalls of simple companding by dividing the audio spectrum into separate bands and processing each independently. At a fundamental level, Dolby B uses a single band restricted to the upper portion of the frequency range, whereas Dolby C uses a pair of overlapping bands. Both systems are designed for the domestic cassette medium (although Dolby C has been used on a variety of other formats, including some professional audio-video equipment). Dolby A is a good but elderly professional system using four signal bands, and Dolby SR is its successor, with a very complex arrangement of 10 bands, some fixed and some which slide over pre-determined frequency ranges. A simplified version of SR, Dolby S, is used on some recent cassette machines and some semi-professional multitrack formats.
With the exception of dbx and Dolby B, accurate calibration of the record and replay characteristics of the tape machine with the noise reduction system is absolutely essential. The better systems, such as Dolby A and SR, incorporate an alignment signal from the encoding processor (a warbling tone or pink noise signal respectively) which is recorded at the start of the tape and used to calibrate the decoding process on replay. Small level and frequency-response errors in the tape machine can cause severe decoder errors if the alignment procedure is not performed accurately.
The maximum level that can be recorded onto a tape will depend on its construction and the correct setting of bias and record drive. The maximum operating level, or MOL, of a tape is defined as the level of magnetisation (measured in nanoWebers per metre) which produces 3% third harmonic distortion of a 1kHz sine wave. This typically lies somewhere in the range 700 to 1400nWb/m and unless peak distortion is particularly desirable, it's normal to align the MOL for a given tape slightly higher than the normal peak operating level for the studio, so that the best possible signal-to-noise ratio can be achieved with little peak distortion.
The electrical signal reference level is 0dBu (PPM4 or -4VU) and during tape machine line-up, this is equated to a magnetic reference level on a given tape (the figure for which should be quoted on the tape label). For example, a common reference level for modern tapes is 320nWb/m (the so-called IEC level), and if this was aligned to 0dBu input and output levels, a signal peak level of +8dB would produce a recording level on the tape of about 800nWb/m, resulting in perhaps 2% third harmonic distortion on a good, high MOL tape. Other common reference levels are 200nWb/m (the NAB level) and 250nWb/m (DIN level). A tape set up at the 200nWb level would be running about 4dB quieter than one set up at 320nWb/m.
Print-through is the process of a strongly magnetised portion of tape partially magnetising regions in adjacent tape layers as the tape sits on its spool. Depending on the way in which the tape is wound (head out or tail out), this will result in pre-echos, where the start of a track may be heard faintly shortly before it actually begins, or as post-echos, where an abrupt loud ending is repeated faintly. The strength of print-through will depend in the construction of the tape (thicker backings tend to reduce the effect), the bias level (stronger bias causes deeper magnetisation of the recording layer), the record head gap (larger gaps cause the magnetic field to penetrate the tape deeper), the MOL, and the manner in which the tape is stored.
It's normal to store tapes tail out, which tends to result in more acceptable post-echos rather than pre-echos. This practice also leaves a played tape with a much neater wind on its spool than is likely to be achieved from fast rewinding, and this helps to maintain the stored tape in good condition.