Modern digital effects units always include emulations of analogue effects such as tape delay and flanging — but none of them ever seem quite like the real thing. Paul White explains how these vintage effects worked, and offers insight into how our modern attempts could be made more accurate.
We live in a digital age, but you can't go into a studio without hearing stories about how good things use to sound in the old days. But were the old electromechanical effects really that good, or have we fallen foul of the 'nostalgia isn't what it used to be' syndrome? Certainly some old effects were noisy and unreliable, but in most cases, there was an element to the sound that the human ear found especially pleasing in a musical context. Sometimes the reasons are fairly obvious, but other times the magic element remains elusive, nowhere more so than in the case of tape flanging or phasing.
Tape phasing, commonly known as tape flanging, is a unique effect, and though some digital flangers have managed to approximate it, I've yet to hear a truly convincing emulation. If you've never experimented with tape flanging, the effect is created by running two identical copies of the same recording on open‑reel analogue recorders (usually in mono) and then summing the two outputs together via a mixer at exactly the same levels. The two recordings are started together — a hit‑and‑miss business at the best of times — then the speed of one of the machines is slowed slightly by using hand pressure on the tape reel. The idea is not to get so far behind that you can hear a tangible ADT‑style delay, but simply to produce a comb filtering effect. Comb filtering occurs by virtue of the addition and subtraction of frequencies that end up being in‑phase or out‑of‑phase as determined by the delay time. Whichever machine is leading is then slowed down so that the delay decreases until the point where the other machine takes the lead. As the relative delay between the two tapes changes and finally passes through zero, the familiar whooshing effect is created as the comb filter sweeps through different frequencies in the source material. And so it continues with the leading machine being slowed manually so that the two recordings drift in and out of phase with each other. Because tape flanging is literally hands on, the effect is different every time. Some of the more sophisticated studios used electronic speed control instead of hand braking to create the effect.
Modern flangers seek to emulate this effect by digitally delaying one signal relative to another (by just a few milliseconds), then modulating the delay time using a low‑frequency oscillator, or LFO. To make the effect stronger, some of the output is fed back to the input, which adds resonance to the comb‑filtering effect. Note that with tape flanging, no feedback is used. Artificial flanging of this kind sounds different to tape flanging for a number of reasons — the LFO‑controlled modulation is regular, feedback is used to add depth to the effect and the delay between the two signals never passes through zero, as it does when two tape machines are used.
A refinement of this method is to delay one signal by a very small amount (say 5mS), then modulate the delay of the other signal path so that it slowly changes from less than 5mS to more than 5mS. This provides the 'through zero' element of the effect but does nothing to break the regularity of the modulation unless the delay time is adjusted by hand. Furthermore, if feedback is applied, it doesn't create the desired effect, as the feedback‑induced resonance will be a function of the whole DDL delay time, whereas the comb‑filtering effect itself is related to the difference between the two delay times. A simple setup for through‑zero flanging is shown in Figure 1 alongside the original tape‑based arrangement.
In theory, it should be possible to emulate tape flanging much more closely by using techniques such as physical modelling. For example, one reason the effect sounds the way it does with analogue tape machines is that analogue machines don't have the precise phase response of a digital system. For example, put a 1kHz square wave into a digital recorder or effects processor and what comes out will be recognisable as a square wave. Not so with analogue tape — the necessary frequencies are all there, but because of phase shifts in the electronic and magnetic components of the system, their time relationship is disturbed, which is why the waveform looks very different to the original. The simulations might be significantly closer if we were able to emulate this smearing before delaying the signals, as well as introducing more randomisation into the modulation.
During the '60s and '70s, most artificial reverb used on recordings was generated using a reverb plate, sometimes called (inaccurately) an echo plate.The reverb plate is an ingeniously simple device, but it takes a lot of tweaking at the design stage to get it sounding right. Plates work by suspending a thin sheet of metal under tension within a rigid frame via springs or clamps attached to the corners. A transducer similar to the voice‑coil of a cone loudspeaker is used to inject audio energy into the plate and two or more contact mics fixed to the surface of the plate then pick up the vibrations inside it and feed them to preamps connected to the console effect returns. By feeding the different contact mics to left and right channels, a pseudo stereo reverb output is created.
Because metal plates have a tendency to 'ring', getting the plate thickness, size, material and tension right is quite an art, and some pre‑ and/or post‑reverb EQ is invariably needed to fine‑tune the sound. Furthermore, because the plate is very sensitive to external sounds and vibrations, it has to be mounted in a soundproof box, ideally on shockmounts.
Unlike digital reverbs, which have innumerable adjustable parameters, the plate reverb relies purely on EQ for tonality and physical damping for decay‑time control (usually via a motorised felt pad). Pre‑delay was often added by using an open‑reel tape machine on the input, and replaying the input signal via the replay head to exploit the time gap between the record and replay head. You just had to hope the tape reel didn't run out during a mix...
Because a typical plate may only be between one and two square meters, and because sound travels much faster in metal than in air, the reflection density within the plate builds up very quickly following an impulse. The sound from the input transducer spreads rapidly across the surface of the plate in all directions until it encounters the edges of the plate, whereupon it is reflected and re‑reflected back into the plate. To get the most random reflection build‑up, it's best to have the transducer mounted a little way off‑centre, and to get a wide stereo image, the two pickups are sited at slightly different distances from the plate edge, as shown in Figure 3 (right).
The characteristic plate sound is bright and extremely dense, with little or no impression of individual early reflections. The reverb builds very quickly and decays smoothly with a maximum undamped decay time of several seconds. Digital reverbs can provide a reasonable emulation of the coloration and envelope of a plate reverb unit, but only the more processor‑intensive models produce the speed of reflection density build‑up required to be truly convincing. Plate reverb was very popular for vocal and drum treatments, and though digital reverb has largely replaced it, many purists still prefer the 'real' plate sound for certain applications.
Those who couldn't afford plate reverbs used spring reverb — a system still used today in guitar combos. The physical principles are similar to those of the plate reverb, except that sound is injected into one end of a loosely coiled metal spring rather than a metal plate, usually via a small magnetic transducer. A pickup transducer at the other end picks up the sound as it reflects back and forth along the spring.
Springs invariably impart a metallic coloration to the sound and they also tend to have a cyclic characteristic as percussive sounds cause vibrations to bounce back and forth along the spring in a fairly regular manner. Excessive input levels cause the springs to 'twang', so some systems incorporated an input limiter. To help even out the coloration and cyclic modulation, it's possible to use two or more springs operating side by side, each with slightly different mechanical characteristics and serviced by individual transducers. Using two similar sets of springs to treat left and right channels produces a convincing stereo effect due to the non‑correlated nature of the two spring outputs.
Like plate reverbs, their main advantage is that they don't produce the gritty 'shattering' early‑reflection effects of digital reverberators. Figure 4 shows a stereo spring reverb.
Before digital electronics and charge‑coupled delay lines (analogue echo), delay effects were invariably created using tape‑loop echo units such as the Echoplex, Roland's famous RE201 Space Echo, or, at a somewhat lower cost, the Watkins Copicat. They all worked on the same principle — a loop of tape passes around a series of heads starting with an erase head, followed by a record head fed from the signal to be treated. Playback heads are positioned after the record head to provide the echoes, and some of the delayed signal is fed back into the record circuitry to create decaying echoes. Delay time is varied by switching heads or varying the tape speed, and models with multiple heads usually have switching systems for setting up different delay patterns (see Figure 5).
Basic digital delays only approximate tape‑loop devices, even if multiple delay taps can be set to different delay times. There are various reasons why tape‑based systems sound so distinctive; one of the main ones is the restricted frequency response of a loop of tape that's been dragged over a set of tape heads thousands of times. The tube circuitry of the original models also had a limited bandwidth and introduced a significant amount of harmonic distortion. This, combined with tape's tendency to saturate meant that when feedback was used, successive echoes became less bright and more distorted, creating a sense of the sound receding into the distance. On top of that, there was instability in the tape path caused by worn rubber pinch rollers that translated into low‑level pitch modulation. As with distortion, this type of modulation becomes cumulative when feedback is used. Another feature that most of us would rather forget was tape noise, but there's no denying that a good tape‑echo unit has a much more 'organic' sound than even the best digital emulations.
In conclusion, it is odd that vintage equipment is now valued because of its sonic 'imperfections', yet none of these were designed in deliberately. Modelling is still in its infancy, however, and I don't expect it to be long before vintage effects can be replicated much more accurately. Perhaps then we will be able to move on to inventing the future rather than striving to recreate the past.
When the Hammond tonewheel organ was invented, one of its limitations was that the mechanical oscillators were all phase locked with each other due to the fact that the tonewheels were mounted on a common shaft. This produced a somewhat sterile sound in comparison with pipe organs, where minor tuning errors across the numerous pipes produced a very pleasing natural chorus effect. Vibrato was added to try to stir the sound up a little, but it was the Leslie cabinet that really shaped the Hammond sound.
The Leslie is effectively a mechanical chorus device with some very interesting properties. Essentially, the Leslie comprises two speaker systems fed from a crossover with a small horn handling the high end and a conventional cone driver handling the low and mid frequencies. The low‑frequency modulation effect is created by a rotating baffle system rather like a hat box with an opening in one side. The speaker faces upwards into this baffle, which is turned via a motor, the speed of which can be switched from fast to slow by the organ player.
At the high‑frequency end, the modulation is created by rotating the horn itself, and to keep the system in balance, an opposing dummy horn is used. A simplified schematic of a Leslie cabinet is shown in Figure 2. The horn is spun by a separate motor so that the low and high frequency modulations aren't locked together, and because of mechanical inertia, changes in speed take place over a period of a couple of seconds rather than being immediate.
Though this arrangement is mechanically simple, creating a digital emulation is more complex than it first appears. Imagine what's happening with the low‑frequency rotor first. When the aperture is facing the listener or microphone, the sound will be at its loudest and brightest, although the tonality will be affected by the acoustic properties of the baffle, the speaker enclosure and any distortion added by the amplifier, which was originally tube driven. As the baffle turns away from the listener, the level will drop and the frequency response will change according to the off‑axis characteristics of the baffle system. To complicate matters further, an element of doppler shift will also be introduced as the baffle surfaces reflecting the sound are moving relative to the listener, although this is less pronounced for the low‑frequency driver than for the horn.
If the room interaction is taken into account, the reflections from the moving aperture will also be continually changing, and these will combine with the direct sound.
A similar thing happens with the horn, except that, as the sound source is smaller than for the low‑frequency end (the mouth of the horn), and because it is describing an arc, the doppler effect is more pronounced, producing a noticeable degree of pitch modulation. The horn is also more directional than the low‑frequency baffle system, so the amplitude modulation is greater. If anything, the room reflections contribute more significantly to the sound of the high‑frequency rotor because high‑frequency reflections tend to be more directional, the overall result being a very complex mix of amplitude and frequency modulation combined with dynamic tonal change and moving room reflections.
When you come to mike up a Leslie, the effect changes again because if the mic is close, the amplitude modulation components are emphasised while the room reflections make less of a contribution. There's also a certain amount of wind noise from the rotating baffle and horn that can be heard on some recordings as a kind of 'chuffing' or fluttering. Using a stereo pair with one mic each side of the cabinet creates a very deep stereo effect that includes an element of fast panning, while miking from further away reduces the degree of amplitude modulation and increases the contributions of the room reflections.
Electronic rotary speaker simulators range from little more than chorus/flangers with appropriate modulation rates and depths to more advanced modelling where the low‑frequency rotor and high‑frequency horn sounds are generated separately after first splitting the signal into two bands using a crossover. The processing usually involves both amplitude and frequency modulation, and because of the characteristic overdrive sound of a valve Leslie, some form of simulated tube overdrive is often included as well. Modelling the dynamically changing room reflections is a rather more complex business and is approximated only in more serious simulators.
One of the best Leslie simulations I've heard recently is part of Native Instruments' B4 plug‑in instrument where the user can edit a number of separate parameters associated with the high‑ and low‑frequency rates, tonality and acceleration as well as overdrive and virtual mic positions. As DSP power continues to become more affordable, even more ambitious emulations are likely in the future.