Take the 2-minute tour ×
Musical Practice & Performance Stack Exchange is a question and answer site for musicians, students, and enthusiasts. It's 100% free, no registration required.

Is is possible to recreate any sound you hear in a synth? Think of it like ear training, but with a synth. Being able to deconstruct a sound with your ear, know what type of waveform is being used, with whatever envelope and LFO's are on the sound.

I'm talking about complex sounds. The stuff you hear in Transformers for example. Is it possible for someone to recreate a sound just from hearing it?

I've always wondered this. I'm good with synthesis design and understand the concepts of synth design, but what I can't do is tell what waveforms are being used in a sound. I find it hard to distinguish them.

If it is possible are there any tools that can help deconstruct a sound? Or are there any books, or training material on training your ears to be able to do so?

share|improve this question
add comment

5 Answers

up vote 16 down vote accepted

This is a difficult question to answer, because you say you already understand synthesis, and that you're good at it - so it seems you should already know what synthesisers are capable of, and how to make them do it.

So you'll know there are many different kinds of synthesiser, and they can be combined - there's nothing to stop you from controlling an FM synth with an analogue LFO, then munging the output with an analogue filter.

Most synthesisers are limited to a certain range of sounds. You can't make a minimoog sound like a DX7, or like a real grand piano, or like Michael Jackson singing, or like a complex sound effect in Transformers.

However, any sound can be represented by series of numbers (clearly, since that's what's on a CD), and if you were clever enough, you could draw a waveform that plays back as the sound you want.

In reality, sound designers build up complex sounds by layering elements from multiple sources. A sound effect in Transformers could be created by combining a sample of someone hitting a real piece of metal, with a white noise-based whoosh from an analogue synth, a digital reverb, and all kinds of other elements.

One real example is the distinctive sound of Tie Fighters in Star Wars. Famously, this sound effect is made my combining a recording of an elephant call, with a recording of a speeding car on a wet road.

As far as training your ear, the first step is to recognise the elements. Know what a squaretooth wave sounds like compared to a sin wave. Know what resonance and filtering achieve. Knowing what the basic elements sound like, is the first step towards spotting how they combine.

This isn't much different to, for example, knowing what a bass guitar sounds like, and knowing what a guitar sounds like, so that when you hear the two playing in unison, you recognise it as two instruments.

One difference between designing synth patches for music, and designing sound effects, is that sound effects are usually "one shot" affairs. Synth patches intended for making music need to sound good at different pitches. They need to respond in certain ways to louder or quieter playing, pitch-bends, legato.

In the world of film sound effects, traditional "foley" techniques are still used, and there's not much reason to replace them with digital alternatives. If dubbing walking effects, by thrusting shoes on sticks into a tray of gravel, works, why do anything different? I have no doubt that Transformers used, along with electronic sound effects, traditional foley artists. This shows that, even if it's possible to create any sound with a synthesiser, it's often easier to do it a different way.

share|improve this answer
    
What an excellent answer! Thank you. –  Torra Jun 6 '13 at 17:17
add comment

There are really two questions here. First, can a given synthesis algorithm represent all possible audio signals? Second, more related to what you are asking, how would you try to mimic a specific sound using a specific synthesizer?

To answer the first question, there are synthesis models (such as additive synthesis) that can represent any signal, but there are also many other synthesis models that are popular in synths that are not able to represent all conceivable signals. That doesn't necessarily mean that you cannot come close if you try to imitate the sound. Thinking of the problem as ear training, as you say, is very much to the point.

What waveforms are used in a sound is not the right question to ask. The sense of hearing is not very sensitive to phase relations across partials, which influence the waveform a lot. It is the amplitude spectrum that you must get right. And the amplitude envelope, and several other factors that you may add in an ad hoc fashion.

For some practical hints of sound design, take a look at Andy Hunt's book.

share|improve this answer
add comment

Timbre - http://en.wikipedia.org/wiki/Timbre

Best way to recreate a timbre in the real world is to consider two things:

1) The ADSR Envelope

This is the different between a short, sharp sound like a drum hit versus a more gradually fading in sound such as a bowed violin. Adjust the ADSR / Amplitude Envelope of your synth sound.

2) The type of waveform used

For example the closest thing to a brass type sound is usually a triangle wave, other sounds have sort of 'close-ish' wave-form types that you can tinker with.

As for the complex sound effects from movies like Transformers, they are usually very complex and can often be made from several different synths put together. You just have to use your ear to pick up things like amplitude and pitch modulations as well as different effects like bitcrushers, distortion, as well as time-based effects like chorus, phaser etc.

I find Native Instruments synths very useful and versitile - especially Absynth - for what you're talking about because it allows you to have an infinitely changeable envelope for just about any part of the sound that you need. I'm a big fan of Massive as well but there are all sorts of different synths that you can use to experiment with.

You will probably never be able to re-create a complex transformer-like sound effect EXACTLY as you hear it but you can get in the same ballpark with enough time and concentration.

Best of luck.

share|improve this answer
add comment

Today, many keyboard instruments work by having a processor assemble every individual sample of the output waveform, and many of today's processors are sufficiently powerful that they can affort to run many thousands of instructions for each sample they output. This gives the designers and firmware programmers (the people who design the software which runs on the keyboards' processors) considerable flexibility in how sounds should be generated.

Historically, synthesizers were not nearly so versatile. The synthesizers would have a relatively small set of configurable modules which could be interconnected in a relatively limited number of ways. Sometimes the interconnections were handled by physically connecting cables, sometimes by physically moving switches, and sometimes by switchable analog routing circuits. On a few synthesizers, waveforms were generated digitally using some special-purpose digital circuitry. Although it was possible for common processors of the time to run barely fast enough to generate audio waveforms in real time, a typical processor might have be able to run about 30 instructions between samples; digital circuitry which was hard-wired to perform certain appropriate calculations faster could be cheaper than a general-purpose processor which could manage similar speed.

While some vintage synthesizers will bear considerable similarity to each other, there were many different approaches to sound generation used by different companies. Some synthesizers worked by generating one of a small number of electronic waveforms for each pitch and then shaping them using filters and other such elements. Others worked by producing and summing together multiple waveforms for each note. Some like the Yamaha DX series generated samples digitally using a lot of digital alchemy in a process called FM synthesis. Synthesizers that work by build or combine waveforms use various approaches will generally have certain types of sounds that they can produce, but adjusting the parameters of filters and other such elements will generally yield relatively predictable effects (e.g. reducing the cutoff frequency of a low-pass filter will make a sound less "bright"). Synthesizers that use FM synthesis can produce a larger variety of sounds, but in many cases there is no intuitive relationship among the parameters.

Some vintage synthesizers can sound very nice, and they can be fun to play with [in both the "toy around with" and "perform" senses of the word]. One should not, however, expect to be able to pick up any synthesizer and expect to be able to produce any desired sound with it.

share|improve this answer
add comment

There are plenty of tools to help deconstruct a sound.

You can view the waveform on an oscilloscope (old analog or sampling digital) using many different time scales, zoomed in to see the waveform, or with a longer time scale to view the general amplitude envelope.

Spectrum analyzers and audio spectrographs will show you the frequency components and overtone ratios making up any pitched sound, and the frequency distribution for unpitched and semi-pitched noise.

Most computer sound editors (Audacity, et.al.) will do both of the above. Then you can try to set up your synth to construct a similar enough looking waveform and/or spectrum.

You can chop up a sound and feed the tiny segments to a granular synthesizer to rebuild it in all kinds of different way, and see what granules are most important to giving a sound it's distinctive fingerprint.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.