The Final Frontier?
Is what digital reverb is all about recreating. But how do the mechanics of reverberation work, and what relation do today's digital techniques bear to the real thing? Paul White reports.
As interest in digital reverberation continues to grow, how can we compare the machines available, and how do digital approximations relate to the real thing?
TO MANY MUSICIANS - especially those without huge salaries for whom the likes of AMS, Yamaha REV1 and Quantec Room Simulator systems have been nothing but a far-off dream these last few years - digital reverb is something of a new consideration. But now that budget digital reverb systems are here (and in growing numbers), the whole subject of what makes them tick - and what separates one model from another - is receiving a lot of attention for the first time.
Before embarking on a detailed discussion of the intricacies of digital reverb, though, you have to know something about its natural equivalent.
To begin with, imagine that somebody suspended above the ground - and well away from any walls or other solid boundaries - claps their hands. When this happens, the sound waves (usually screams, in this case) travel outwards in a spherical fashion and never return. But seeing as being suspended in mid-air is not normally a tenable position for the human form (due to the intrusion of gravity, among other things), we tend to spend most of our time close to the ground and, when in buildings, close to wails, floors and ceilings.
These walls, floors and ceilings interact with sound waves, absorbing some of the sound energy and reflecting some, with the reflections being subject to re-reflection when they encounter new boundaries. In a typical room, our handclap would bounce from the walls, causing multiple, closely spaced echoes. And within a very short space of time, the number of echoes would be so great that individual echoes became indistinguishable. Because of the absorptive qualities of walls and other boundaries, these echoes tend to die away fairly rapidly as the sound energy is converted into heat. And seeing as high frequencies tend to be absorbed more readily than low ones, the high-frequency part of the sound decays more rapidly than the low frequency part. Then there's the fact that because the air itself absorbs high frequencies due to the viscosity of the molecules, it's clear that the further the sound travels, the lower its high frequency content will be.
In a large hall - where the reverberant sound reflections travel hundreds or even thousands of feet before being completely absorbed - the tail end of the decay may contain little or nothing above a couple of kHz. In a smaller room, the reverberation time may well be so short as to be unnoticeable. But a trip to an acoustically dead, soundproofed room will soon convince you that a substantial part of the sound you hear in everyday life is in fact reflected, not direct.
Our brains make use of this reflected information to locate sound sources, and also to make judgements about the size and nature of the environment we're listening in - especially in the dark, where information from our eyes gives no clues as to the nature of the room.
But because we do all this quite subconsciously, every day of our lives, any artificially generated reverberation must be very convincing if our brains aren't going to reject it as an imperfect fraud.
From this short analysis, we can break the characteristics of reverberation down into more manageable sections which give us a clue what we need to do to synthesise it.
Going back to the handclap-in-a-room situation, the first thing that happens is that after a short delay (caused by the time it takes for the sound to travel to the nearest boundary and back), the sound starts to bounce from wall to wall. And because sound travels at around 1100ft per second, these early reflections can be discerned as separate echoes in a large room or hall.
The spacing and magnitude of early reflections depend on the physical characteristics of the room, but they decay in amplitude and lose high frequencies as time goes on. These reflections then build up into a dense clutter, which is what distinguishes reverb from simple echo. The high-frequency content decays further and the overall level decays, ideally in an exponential manner. The rate at which the clutter builds up and decays depends on several things: the size of the room, the materials from which it's built, its geometry, and objects within the room such as soft furnishings and people, which cause the decay to be more rapid than when the room is empty.
Research has shown that the brain needs a minimum density of between 1000 and 3000 individual echoes per second before it will accept a sound as true, natural reverb. So an unaided multi-tapped delay is hardly a practical method of generating the effect artificially, especially as you need a different tap arrangement for each different reverb type.
BACK IN THE 60s, an incredibly clever man by the name of Schroeder delivered an AES paper on artificial reverberation. He suggested that certain building blocks could be combined to simulate the effect, and that these building blocks could be generated by computer. This work was remarkable, not least because the computers then available were a good deal too slow to generate reverb in real time.
Later on, somebody equally clever called Moorer built on Schroeder's work, and the combined outcome was the definition of a series of different kinds of filter, which could all be used as building blocks to help simulate natural reverberation.
The first of these blocks is called an all-pass filter, used singly or in cascade to generate repeats of a signal without modifying the frequency response, an important factor in creating reverb that is free from unwanted colouration. This is simply a delay block with feedback, plus a feedforward path which makes the block different from a simple comb filter. Comb filters are used in parallel banks to create the clutter of reflections, while all-pass filters diffuse them further.
Other building blocks include the digital filter, to simulate the more rapid decay of high frequencies; and a multi-tapped delay or finite impulse response (FIR) filter, to simulate the early reflection part of the sound.
Exactly how these blocks are interconnected is a closely guarded secret of all digital reverb manufacturers, but most use Schroeder's and Moorer's research as a starting point. Many designers vary the arrangement and parameters of their building blocks in the left and right channels, so that a mono input can still give a convincing stereo output.
In the case of digital reverb, each of these building blocks exists only in software, so a new program will give a new reverb effect using exactly the same hardware.
The main problem is that to create the density of reflections needed while retaining a respectable audio bandwidth, the digital reverb's computing system needs to be capable of performing very fast calculations.
Top-end digital reverb systems generally offer good bandwidth, well-designed reverb algorithms, user variability of parameters, and user programmability. MIDI patch control is also fast becoming a standard, as are special effects such as gated and reverse reverb.
Let's take a closer look at the kind of variable parameters you can expect these machines to offer. First, the early reflection pattern and spacing is variable because this creates the basic character of the synthetic room or space. A variable pre-delay is also common, as this separates the reverb from the initial sound, to create a sense of space.
Next we come to the clutter section comprising dense reverb, and here we need to be able to vary the decay time and the high frequency decay characteristics. If we can vary these basic parameters, we can simulate anything from a small room to a vast hall. By increasing the high-frequency decay time we can make a room sound brighter to simulate, say, a tiled washroom. Conversely, we can damp the high frequencies heavily to simulate rooms filled with soft furnishings.
It shouldn't come as a surprise to learn that as soon as you get below state-of-the-art price levels, you come up against machines that offer slightly less than state-of-the-art performance - though as with everything else, each different design team has its own set of compromises which it chooses to adopt.
Now, all these compromises incur penalties of one sort or another, and don't let any salesman talk you into believing otherwise. It's really just a question of which compromise makes the smallest impact on your personal requirements.
FIRST, YOU CAN opt to use less than 16-bit sampling resolution. This worsens signal quality and noise performance, but you'll find that many low-end reverb units do in fact utilise 12-bit sampling, or less. Second, you can opt for a machine with a slower built-in computer and consequently reduced bandwidth - though in this case, you should look for at least 10kHz bandwidth for serious work so that you can create reverb brighter than you'd find in nature if you want to.
If the machine you're interested in does offer a decent bandwidth, chances are its designers will have ditched all unnecessary demands on the computer's time, such as programmability, fancy parameter adjustment and flashy displays. Alternatively, they may have kept all these but tried to simulate reverb using fewer building blocks, giving a less sophisticated reverb algorithm. Let's look at the consequences of these two main alternatives.
First off, if you decide to keep all the programmable functions and compromise the reverb algorithms to keep the cost down, reverb density is likely to suffer. Instead of a smooth decay, the widely-spaced coarse reflections give a grainy texture to the reverb, and the treatment of percussion sounds can produce a sound like ripping cloth as the individual reflections appear.
Using an insufficient number of building blocks or poorly designed algorithms can also lead to unnatural colouration of the sound, which usually takes on a ringing or metallic characteristic. You may also notice that the final decay of the reverb isn't too smooth.
On vocals these problems may not be too noticeable, but on percussion, their effect is quite a bit less pleasing - more than likely, your ears and brain won't be convinced by what's being fed to them, and you'll perceive the result as crude and artificial.
Now, this trashy, metallic sound is used to good effect by producers and remix engineers on modern dance records, but it's worth remembering that while a good reverb unit can be made to give you trashy sounds, a poor one can never give you good, natural sounds.
The other approach - limiting the flexibility of the effect by offering only a set number of presets - gives the system's computer a chance to concentrate only on a few specific tasks, so the preset treatments stand a good chance of being high-quality.
On the other hand, losing programmability prevents you from utilising reverb treatments that are subtlely different from everybody else's - and that, in this era of preset digital synch sounds and factory samples, could be important. And if you're a studio owner and some clients ask if their reverb can have "just a little bit more" of something, you could well be stuck.
Then again, it wasn't so long ago that top studios used plate reverbs, where the only things you could alter were the overall damping and the EQ. Nobody complained. A reverb offering as few as a dozen presets could give you a set of treatments ranging from a tight live room to a massive hall, in small enough steps so as not to leave you wanting a sound you couldn't get. And you can always add further EQ or experiment with gates to alter the decay shape.
Remember that if the basic sounds are good, you'll be able to live with them more easily than you would with a machine that lets you program 10,000 permutations of reverb, all of which sound unnatural.
So how do you know what is good digital reverb and what isn't? Simply, use your head. Reverb can't be judged on spec alone, and things like naturalness and stereo spread are totally subjective. Choosing a digital reverb is more akin to wine tasting, and the more models you listen to, the more differences you learn to recognise. Rather than using cheese to take the taste away, try washing your ears out with a burst of white noise between models.
If you listen to any good reverb in a darkened room, you should be able to visualise the environment in which the music is being played. A snare drum or handclap sound from a drum machine will soon show up ringing, coarseness and other vices that prevent the reverb from sounding natural.
Gated and reverse sounds are so much a part of modern recording that you'd be hard-pushed to find a current model of digital reverb that didn't offer them. But the variation in quality of these effects is incredibly wide.
A gated reverb should be dense, solid and exciting... and I've heard some truly awful ones that sound more like dried peas being dropped on a steel plate than true gated reverb.
Reverse reverb is less commonplace but still important. It should be clean and capable of leaving vocals intelligible, and of giving the impression of a sound played in reverse with the start of the sound clearly audible at the end. Of course the sound isn't really played backwards - it's just an electronic conjuring trick. But if it doesn't sound authentic, there's no point using it.
YOU CAN HAVE a digital reverb unit that is flexible and that sounds good - if you're prepared to pay for it. But if you're looking at the very bottom end of the market, you're going to have to lose out in one area or another. The most natural-sounding budget reverbs I've used have offered only a range of presets adequate for live use and for small studio applications.
As a general rule, you need to place sound quality above all else if you're going to be processing drum sounds in the studio. For vocals and keyboards, especially live, you'll probably get away with something that sounds a little less sophisticated.
If you go for an all-singing, all-dancing unit at a bargain price, listen carefully for the sonic problems mentioned above. And don't be fooled by the infinite number of parameter variations you can program; if the basic sound isn't good, there's nothing you can do about it, no matter how many permutations you try.
Remember too that within a mix, the difference between one reverb program and the next one may appear negligible, even though they may sound totally different when tested in isolation with just a single snare-drum beat.
Don't be conned by long reverb times, either; you're unlikely to use anything longer than five seconds in normal music work, and the most widely-used settings are shorter than two seconds.
Alternatively, you may be tempted by a unit that offers other treatments such as delay, chorus and flanging. This is fine if the reverb is good to begin with (some units show compromises in all areas), but bear in mind that unless your chosen system allows you access to several effects simultaneously, you'll be stuck if you need to use reverb and flanging together at a gig, say. In the studio this limitation isn't so serious: you can use one effect when you're recording and another when mixing.
Finally, get what you need, not what you think you want. Close your eyes and let your ears decide.
Feature by Paul White
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!