Signal Processing, Part 2
Introduction
A number of effects and signal enhancement techniques revolve around the use of time delay. These include the basic effects of doubling and echo, and also the complex processes of reverb, chorus, and flanging. The history of some of these effects is quite interesting as is the manner in which they were first implemented. With the advent of digital technology these processes are literally at the audio engineer’s fingertips. The following processes are normally created with the intention of being added back to the original signal. The original signal is referred to as the dry signal while the effected signal is referred to as the wet signal.
Time Delay and Echo
The simplest effect and the one that many more complicated processes are based on is a simple time delay. The idea is to somehow delay the signal for a certain amount of time and then add it back to the original signal, perhaps at reduced amplitude. If the delay is on the order of 30 to 50 milliseconds or more, the two sounds will be heard distinctly, creating the illusion of two instruments where there was just one. This is known as doubling. This process can be made more realistic by slightly varying the time delay over the course of several seconds, for example from 40 to 55 milliseconds[1]. If the delay is just a bit longer the process is sometimes known as slap echo because it mimics the effect of sound slapping off of a large flat boundary that is not too distant. This is used to “thicken” a sound, that is, to make it more prominent without having to increase volume.
Years ago, delay was widely implemented using magnetic tape recorders. As the record and reproduce heads would be staggered, the finite speed of the tape produced a certain amount of delay. If the effective distance was increased (through an arrangement of pulleys and so forth), the delay could be increased, certainly to the level of seconds. If the delayed sound was directed back to the record head, a series of discrete echoes could be created. Normally, the delayed signal would be reduced in amplitude so that the echoes faded away. The greater the attenuation, the quicker the fade. Sometimes musicians would create very long loops of tape to achieve a rhythmic repetitious effect as well. Another technique for creating delay was used in the Cooper Time Cube. This device used a small transducer pair in an insulated box with the delay coming from the air propagation delay. It was limited in what it could do as the delay time was set by the physical separation of the transducers, but it was simpler overall than setting up a tape recorder solely for echo. In the mid-late 1970’s, analog “bucket brigade” ICs appeared, making delay via analog sampling possible. This was supplanted by digital sampling. The basic idea is that the sampled sound is stored in RAM in a circular buffer. New values are recorded at one end while delayed values are pulled off at the other. It is possible to implement multiple taps on these systems so that complex echo patterns can be created.
Reverberation: The Quest for Artificial Ambience Reconstruction
Recording studios are referred to as having “nice” acoustics. But what if you need to create the sound of an entirely different acoustic space? This is where reverberation, or reverb for short, comes in. In the old days, reverb was created through the use of reverb chambers. These are very live rooms (think “concrete”) containing a microphone and loudspeaker. The signal to be affected is sent to the loudspeaker and then picked up by the microphone. It is then mixed in with the main (original) signal. The microphone will pick up the myriad reflections in the room, creating a very complex signal. It can be though of as innumerable echoes of differing amplitude and delay. The individual echoes are so closely spaced that you don’t hear them as separate entities, but rather as single “wash” of sound (think “cave”). Many pop recordings of the 1950’s and early 1960’s sound almost buried in reverb to today’s ears. While large companies could afford specialized chambers, many studios could not. To solve this problem, engineers came up with a variety of ways to simulate reverb. Two popular techniques were spring and plate reverbs. In a plate reverb, an output transducer was afixed to a large metal plate. The sound would propagate through the plate and reflect at its boundaries, eventually reaching a pickup. The pickup signal is analogous to the microphone output in the chamber. In a spring reverb, an output transducer would drive a spring and the signal would propagate along it, reflecting at multiple discontinuities, and eventually reaching a pickup at the far side. Spring reverbs ranged in sonic quality from extremely cheesy to very good. The cheesy units were often found in guitar amplifiers.
By the mid to late 1980’s digital technology and RAM became inexpensive enough to implement all-digital simulations of reverb. The initial digital reverbs normally included echo and delay functions as well because the digital reverb was basically an extension of the earlier technology.
The idea of the multi-tapped delay was extended to include frequency sensitive feedback parameters and decorrelation (essentially phase shifting) to create convincing reverb simulations[2]. These units also offered greater control over the reverb parameters than the early reverb units. By altering the reverb program, the user had control over reverb decay time, diffusion, pre-delay, and other factors making it possible to simulate a wide variety of acoustic spaces.
Eventually, by the late 1990’s, direct impulse response convolution became affordable. In this technique, the impulse response of an existing space was recorded and used to shape a sound. Instead of simply simulating the reverb process, this new technique literally computes the true reverb effect at a specific location in a certain acoustic space. It is quite possible, for example, to obtain the impulse response of say, your shower, and then apply this “sonic fingerprint” to a sound recorded in the studio. Of perhaps greater interest to the researcher is the ability to remove reverb with this technique. If the precise impulse response is known, then it is possible to compute a function that effectively “subtracts out” the reverb signal, leaving the original dry signal. This goes beyond acoustic spaces. Impulse responses of specific devices such as microphones and loudspeakers can be obtained to reproduce their “character”.
Chorusing and Flanging
Flanging and chorusing were discovered by accident. These are popular time-based effects heard on many pop recordings. As the story goes, an engineer once tried to run two tape recorders at the same time. He discovered on playback that the two machines would never be exactly in sync on playback and if the same sound was recorded to both machines, the subtle time differences between them created an interesting sound. As the delay was fairly short, an echo was not heard, but rather, a phase cancellation. The engineer had accidentally created a comb filter, which can be thought of as series of notch filters spaced through the spectrum. This imparted a curious characteristic to the sound. The effect became more pronounced if the engineer pushed on the flange of the tape reel as this created a variable time delay, causing the comb filter notches to sweep up and down the spectrum. The result is a sort of swooshing sound that can be heard on the runways of airports, and therefore sometimes referred to as a “jet plane sound”[3]. The delay sweep is often set for several seconds and the wet signal is sometimes inverted to increase the effect. The delays are usually in the range of 1 to 10 milliseconds.
If the flanger delays are increased to around 20 to 30 milliseconds, the effect is part way between flanging and doubling. It is known as chorusing with the name coming from achieving several voices from one. Chorusing achieves a thicker sound, as does doubling, but with a different character. Flangers and chorusing units, while once available as stand alone boxes, have largely been absorbed into the current all-in-one digital multi-effects units. These devices offer delay, echo, reverb, flanging, and chorusing along with other processing capabilities such as compression and purposeful distortion.
Example Problem
1. Q: A digital delay line samples at 50 kHz using 16 bit linear PCM. How much RAM is required to create a 2 second long echo? A: Each second requires 50 k samples, where each sample is 16 bits or 2 bytes. Thus each second of delay requires 100 k bytes of RAM, or 200 k bytes for a 2 second delay. (Dirt cheap by the mid 1990’s, but pretty expensive in 1979.)
3
ET163 Audio Technology Lecture Notes: Signal Processing, Part 2
[1] This is because it is impossible for musicians to play perfectly in sync with each other. There will always be slight variations in timing.
[2] Quality reverb cannot be created by using a very short echo time with a very long decay time. This will result in a very unnatural robotic sound.
[3] The same notch filtering takes place on the runway as the plane moves about. There is a variable delay at your ear between the direct sound from the engine and the reflected sound off of the tarmac.