Sines of reality
The new thing in software is copying the old. E&T filters out the bad from the good in synths and finds out how to remake the Fender sound.
When it launched in 1983, Roland's silver TB-303 Bassline did not sell well. The Japanese company badly misread its market. The company conceived the Bassline as an accompaniment machine for keyboardists and guitarists, promoting it alongside the TR-808 drum machine. The only problem was that the TB-303 did not sound much like a bass guitar and was extremely tedious to program.
But Nathaniel Pierre Jones (aka DJ Pierre) picked one up a few years later in Chicago. His initial hope was to to emulate the techniques of Ron Hardy, who worked as a DJ in a local nightclub, and who used a drum machine to mix extra beats into the records he played. But Pierre also tweaked the knobs of the TB-303 so that it sounded less like a bass and more like a sound-effects machine, and in so doing provided a new addition to his arsenal. Soon, Pierre-inspired tracks began to supplant the regular diet of disco played in Chicago and moved outwards to spread around the world.
Acid house and the TB-303
The incomprehensible user interface of the TB-303 encouraged producers to tweak knobs and press button almost at random: the result was the strangled squawk of acid house, which erupted into the UK charts in the late 1980s. While the government passed laws against public gatherings playing repetitive beats, the TB-303 became the one synth that any budding dance producer wanted. By 1994, Norman Cook, (aka Fatboy Slim), summed it up with 'Everybody Needs a 303'. Although the popularity of acid house as a genre has long since evaporated, the TB-303 is still a sought-after synth - selling for more than three times its original price - and has been joined by other old analogue hardware in desirability among musicians and collectors.
Angus Hewlett, the boss of London-based music-software company FXpansion, says: "The analogue revival has been going on longer than analogue was ever pronounced dead. The DX7 came out in 1983 and the analogue revival really started around 1991. So, it was dead for less than ten years and going strong for the best part of 20."
The scarcity of vintage analogue hardware and the high component cost of effective replicas has spawned a mini-industry dedicated to emulating these classics in software, with others choosing to remake the sound of old mixing desks and famous guitar effects (see 'Fender in the box', p34). But it has been a learning process.
"Going back to 2003/04, we had virtual analogue synths that would respond quite nicely in the middle of the musical range. But, when you push them, they don't behave the same way as hardware. They don't do 'nasty' the way analogue does," says Hewlett.
There is even heated debate over what the hardware itself does. The architecture of the TB-303 is one such battleground.
The core of the acid house sound is the TB-303's low-pass filter. With the cutoff frequency and resonance turned up high, the sound veers between angry wasp and robotic gurgle, with a rawness that other synths, such as the MiniMoog, do not have.
Tim Stinchcombe, a test engineer at UK-based Heber over the past ten years has dug into the design of analogue-synth circuits since buying a Doepfer modular system. The Moog filter acquired the name 'ladder filter' because of the shape of its circuit. Each pole of the filter is a matched pair of transistors that, presented on a schematic, are stacked to look like a ladder.
Many synth-makers and hobbyists have since replicated the filter in hardware - and, with less success, in software - but at the start of the 1980s, Moog's patent was still in force.
One advantage of the transistor ladder filter, according to Stinchcombe is that each pole is effectively isolated from its neighbours, which greatly simplifies the mathematical analysis. "It shows the genius of Robert Moog that he had an intuitive feel for how to do filters," he says.
The TB-303 filter is quite different. At first glance, it looks very similar. It even has the same number of ladder 'rungs' as the Moog, something that intrigued Stinchcombe as Roland touted the design as a three-pole, 18dB filter rather than a four-pole, 24dB circuit. In place of the transistors of the Moog structure, Roland inserted diodes.
"As soon as you put in the diodes, you lose the isolation and it becomes much harder to analyse. You end up with a much more inelegant transfer function at the end," says Stinchcombe.
Stinchcombe insists that the filter is, at its heart, a 24dB design, which some Roland fans find hard to accept. "They don't want to hear otherwise because Roland called it an 18dB," he says.
In reality, once you add in the effects of all the components in the circuit, the Roland filter is far more complicated than a Moog-type structure. It has as many as ten poles, some in odd places, although some are effectively cancelled out.
"Decoupling capacitors can have a considerable effect on the circuit," says Stinchcombe.
Hewlett says a lot of engineer knowledge is preserved in the original circuit boards. "They were very inventive. They would spend a lot of time on each design. And a lot of it was done to sound good."
Synth circuits such as the square-wave oscillator in the TB-303 used extreme short-cuts. The oscillator is little more than a simple comparator fed by the sawtooth generator, but it does generate a hollow-sounding square wave that works well as a bass sound.
"There is no single, hard and fast set of rules for what sounds good and sounds bad. But, by and large, the sort of inaccuracy you get from the discrete digital domain does not tend to sound good. Mistakes in the voicing behaviour that might come from a naïve digital implementation will tend to sound unmusical," says Hewlett.
For its recently launced DCAM Synth Squad software, which is designed to capture the character of old analogue hardware, Hewlett and co-lead developer Andrew Simper had to go back to the circuits as the most reliable way to derive the behaviour of the synths, particularly in more complex parts such as the filters and envelope generators.
"With DCAM, we build the circuit in a simulator. We build the schematic from the service manual or other sources or by reverse engineering the circuit board," Hewlett explains. "We analyse the characteristics and build approximations to the components within that. Then we listen to it and tune it even further."
Hewlett adds: "The idea behind Synth Squad is to get the overall sound of an analogue synth. We want to be able to dial in a range of sounds rather than to get a specific one-to-one thing," says Hewlett.
For their software, AudioRealism in Sweden and Polish group D16 have focused squarely on the Roland silver boxes. For D16, circuit modelling is one key element of the process.
"There's no simple recipe," explains Sebastian Bachlinski of D16. "Certainly, we analyse each circuit and the interactions between circuits. The internal structure has a direct impact on the sound but we never do 100 per cent emulation of electronic circuits. After an analysis we try to 'forget' about the circuit construction and do everything that's necessary to make our plug-ins sound the same as the original but in the digital signal processing domain.
"This translation from the analogue world to the digital is crucial. At this stage everything can be won or everything can be lost. Almost every time it demands the design of new algorithms to model a specific nuance."
Oscillators can often be modelled quite simply as waveform generators, says Bachlinski. "But implementing envelopes is quite different. In this case it is necessary to have a solid theoretical ground in the analysis of electronic circuits."
When building the prototype, which Bachlinski estimates consumes 30 per cent of the time of the overall job, about half the time is spent on analysis with the rest split between algorithm implementation and tweaking and tuning.
To try to demonstrate the accuracy of its model, D16 posts spectral plots against the results produced by a real TB-303. Does it make sense to try to get so close to the real thing? Once an instrument has been recorded and a battery of studio effects added, will people notice?
The audience might not hear a difference but the musicians might. Hewlett says the feel of the instrument is vitally important. "If it doesn't respond to their playing properly, then they aren't going to like it," he says.
The sound of analogue is not likely to go away fast despite attempts to introduce novel forms of synthesis.
"The reason synth music is stuck with oscillators passed through resonant low-pass filters is because there is a lot of musicality to be won that way. There is something that seems to work. Musicians aren't as individual as they often like to make out," says Hewlett.