Types of Synthesis

[BluMunk]

As a keys player who has been slowly edging his way into synthesis, I run across a number of different 'types' of synthesis that I don't have a good understanding of.

 

I get (I think) that subtractive synthesis is called such because you start with a signal source, and then use filters to 'subtract' frequencies.

 

Is that basically right? What about other things like 'wavetable'? 'granular'? 'fm'? Others?

[Fulc]

subtractive synthesis is called such because you start with a signal source, and then use filters to 'subtract' frequencies.

 

Is that basically right?

Yes. Modular synthesis is predominantly subtractive in nature, but you can kludge a crude form of FM out of it as well. (Read my well-meaning tutorial here. )

 

What about other things like 'wavetable'? 'granular'? 'fm'? Others?

Wavetable synthesis has always stymied me. I know that the PPG (later Waldorf) did it, but for the life of me I couldn't tell you how. Will someone please step in and straighten us out?

 

FM is short for frequency modulation. In this context it's one oscillator (frequency in the audio range) modulating the frequency of another (frequency also in the audio range). The first oscillator is the modulator and the one being modulated is the carrier. The DX7 had six of these oscillators configured in 32 different arrangements (algorithms).

 

Granular synthesis takes a grain (a small portion) of a waveform and derives its timbres from that. Off the top of my head I don't know of any commercial synths that use it, but that's not to say they don't exist. You could download Csound and play with it to get a feel for granular, but the learning curve is rather steep.

[Sven Golly]

This is where Wikipedia really comes in handy; saves everyone from a whole lot of typing.

 

Check out the Wikipedia entry on Sound Synthesis for an overview of what's out there.

 

The entry on Wavetable Synthesis should hopefully clear up the confusion as well.

[Bob Rinker]

Originally posted by BluMunk:

As a keys player who has been slowly edging his way into synthesis, I run across a number of different 'types' of synthesis that I don't have a good understanding of.

 

I get (I think) that subtractive synthesis is called such because you start with a signal source, and then use filters to 'subtract' frequencies.

 

Is that basically right? What about other things like 'wavetable'? 'granular'? 'fm'? Others?

Some of the answers are covered in the resources listed in this recent thread . At least it's a start.

 

Yes, you are correct - subtractive synthesis starts with waveforms that are rich in harmonics and filters some out. The more "pointy" or sharp the waveform shape, in general, the more harmonics it contains - a sawtooth or pulse has more harmonics than a triangle or sine wave. Additive synthesis is the opposite - adding a bunch of sine waves together to form a tone. A Hammond organ with drawbars is an additive process. The study of waveforms and frequencies is called Fourier analysis.

 

Originally, analog circuits (oscillators) were used to create the basic sounds. These circuits have a distinctive sound of their own, ala Moog, but weren't very good at reproducing the sounds of other instruments. So, somebody got the idea to record a waveform of the instrument (the "sample"), digitize it, and store the waveform as a bunch of numbers - a wavetable. A computer could then play back these numbers through a Digital-to-Analog (D-to-A) converter and reproduce the original sound. If the playback was at a different rate than the original the pitch would change, so the same wavetable could be used for many notes - except that the real instrument typically has different harmonic nuances for different notes. A compromise is to use one sample for, say, every three notes.

 

FM (frequency modulation) synthesis involves the changing (modulating) the frequency of one oscillator with another - a sort-of multiplication of one waveform by another. This has the effect of creating additional frequencies other than the originals. In particular, this method can produce frequencies that are not harmonically related to the originals, which produce "clangy" or bell-like sounds. The frequencies you will get from any two waveforms isn't always obvious, so some really wild sounds can result.

 

Granular - hmmm, don't know. That term might refer to the sampling rate used for wavetable synthesis.

 

I guess I've told you all I know about it.

 

- Bob

[BluMunk]

Originally posted by Sven Golly:

This is where Wikipedia really comes in handy; saves everyone from a whole lot of typing.

 

Check out the Wikipedia entry on Sound Synthesis for an overview of what's out there.

 

The entry on Wavetable Synthesis should hopefully clear up the confusion as well.

 

Actually, that's where I went first, but wikipedia doesn't seem to want to load for me here at work today. And you know, I'm one of those impatient kids who needs things now!

 

Thanks for the links to the direct articles though.

 

 

fulcrum, that looks like a great resource on first glance. When I have time this weekend I'll give it a more thourough(sp?) going-over.

[soundscape]

Originally posted by Sven Golly:

Check out the Wikipedia entry on Sound Synthesis for an overview of what's out there.

"A rumor was propagated by unknown people at Yamaha (or dealers) that the Japanese units would "blow up" upon being plugged into 1210V AC outlets in the U.S."

 

1210V, eh? Gotta love Wikipedia.

[Sven Golly]

Originally posted by soundscape:

1210V, eh? Gotta love Wikipedia.

You could always edit and fix what is obviously a simple typo, rather than merely bashing an otherwise good resource...

 

To each their own, though.

[soundscape]

Originally posted by Sven Golly:

You could always edit and fix what is obviously a simple typo, rather than merely bashing an otherwise good resource...

Hey, if I wanted to "bash" Wikipedia, and I'm not sure I want to "bash" anything, I'd write a well-reasoned argument about its pros and cons, and believe me, I plenty to say about it (and my concern isn't with a few silly typographical errors and bits of extraneous information.) So I wasn't meaning to say that I thought it was a bad resource in this case. (I often go use it to get initial information on a subject.)

 

But anyway, I find it funny that sometimes entirely irrelevant material appears in the middle of an article, and perhaps I have a really bad sense of humour, but I thought the idea of plugging a DX7 into a 1210V AC outlet and watching it blow up was amusing. That's all.

[JeffLearman]

Note that subtractive techniques can (and often are) used in combination with the other techniques. Any synth with a filter envelope has subtractive synthesis elements. But generally, the term is only used when we start with a simple (square, pulse, triangle, or sawtooth) signal that's rich in harmonics and filter stuff out to get what we get.

 

Another significant type is additive synthesis, which is used by Fairlight, and to some extent, drawbar organs (in a very simplistic way, of course).

 

With additive synthesis, the sound designer analytically breaks down the desired sound into its harmonic component sine waves, and how that harmonic content varies over time when a note is played. On synthesys, the resulting waveform is built by combining sine waves together to make the sound.

 

Stevie Wonder used 'em a lot, though probably not any more. I would expect to see this as a popular form of softsynth since it's rather easily done on today's hardware, but for some reason it doesn't seem to be. Or perhaps I'm just not tuned in.

[Finale]

It's a lot easier to define a goal sound and achieve a close result with subtractive synthesis than with additive synthesis or FM, because you don't need to be a genius in mathematics and in Fourier series to know where you're going.

[Andre Lower]

This might come as a wild post of mine, but I've always wondered about what I'm gonna ask now: Would it be possible/useful to employ some device that would work as an oscilloscope in the audio range to help figure out the harmonic content of a given synth sound?

 

First of all I confess I've never seen an oscilloscope used for audio analysis (perhaps it has a different name, perhaps a regular oscilloscope can be used for that, ...), and I am aware that most sounds would be as hard to "decompose" using said "viewable" device as they are when handled by ear only, the main cause being that it is hard to tell exactly how a given number of original, simple waveforms ended up yielding the "final product" (either the one you hear or see on my hypothetical audio oscilloscope).

 

But that sure would help to point out the eventual differences between a patch that "almost nail" a desired sound and the targeted sound itself, for the waveform differences would point out what is "missing" in the patch at hand.

 

Do I digress too much ? Is anyone still with me ?

[soundscape]

Originally posted by Andre Lower:

This might come as a wild post of mine, but I've always wondered about what I'm gonna ask now: Would it be possible/useful to employ some device that would work as an oscilloscope in the audio range to help figure out the harmonic content of a given synth sound?

Certainly--it's called a spectrum analyser--it uses a Fourier Transform to convert the time domain information to the frequency domain. (Amplitude against frequency instead amplitude against time.)

 

Of course you'd need to look at different parts of the signal (attack vs. sustain portions, etc.)

 

But that sure would help to point out the eventual differences between a patch that "almost nail" a desired sound and the targeted sound itself, for the waveform differences would point out what is "missing" in the patch at hand.

As I understand, some of the problems really boil down to intrinsic differences between synths, so even if you can use oscilloscopes and spectrum analysers to see what's going on, you still might not be able to get the right sound.

 

Here's some interesting stuff: ("Why a Moog Sounds Like a Moog"):

 

http://matrixsynth.blogspot.com/2006/01/why-moog-sounds-like-moog.html

[Dan South]

Originally posted by Bob Rinker:

Originally, analog circuits (oscillators) were used to create the basic sounds. These circuits have a distinctive sound of their own, ala Moog, but weren't very good at reproducing the sounds of other instruments. So, somebody got the idea to record a waveform of the instrument (the "sample"), digitize it, and store the waveform as a bunch of numbers - a wavetable. A computer could then play back these numbers through a Digital-to-Analog (D-to-A) converter and reproduce the original sound. If the playback was at a different rate than the original the pitch would change, so the same wavetable could be used for many notes - except that the real instrument typically has different harmonic nuances for different notes. A compromise is to use one sample for, say, every three notes.

This is essentially correct, but I'd like to clarify the terminology.

 

Sample-based synthesis uses the same technology as your CD player. An acoustic source (piano, guitar, sax, etc.) is mic'd to create an analog electrical signal. A device called an A-to-D converter "samples" that signal many times per second (about 44,100 times per second in the case of a compact disc). By sampling, I mean that it measures the amplitude of the signal at that instant.

 

These waveforms are stored for playback. They may be compressed, and one sample may be used for several playback notes (saves memory over sampling each note). And for some instruments, like piano, each note is sampled several times for different volume levels (pp,p,mp,mf,f,ff,etc.).

 

Anyway, the other gentleman already described this process to a degree, but this is not what most people consider to be "wavetable synthesis." This is sample playback synthesis, because when you press a note in sample playback synthesis, it starts playing one or more samples and holds them to the end of the note.

 

Wavetable synthesis differs as follows. The sample(s) that you play at the beginning of the note might not be the same samples that you are playing as the note sustains. Here's an illustration.

 

Sample playback synthesis

 

If you play a note that contains a trumpet sample and an organ sample, both samples play until the end of the note. Their volumes can be controlled by envelopes, so you might not hear them all the way through, but they're still being player, even if their volume is zero.

 

Wavetable synthesis

 

You press a note and hear the attack of a trumpet, which then fades (or jumps abruptly) to a sample of an organ, which can then fade or jump to something else (flute, snare drums, distorted guitar, hi hat, etc.) You can even set this wave sequence to loop over time, i.e. run through the list over and over.

 

Some people consider sample playback synthesis and wavetable synthesis to be the same, and that's not inaccurate. The words can be used interchangeably. But wavetable synthesis has a more specific meaning, the one that I've described above. You'll have to judge from context which meaning someone is using at any given time.

[Odyssian]

I've often noticed references to "analog synthesis" and "modular synthesis" ("modular synthesis" is even listed as a topic link in another message forum).

 

I'm pretty sure that analog and modular are not synthesis types, but are descriptions of the types of circuit design (analog) and a type of synthesizer (modular).

 

The copied text below from the Wiki doesn't list modular or analog as synthesis types, so not only am I probably correct, I'm also very anal..

 

===================

 

Synthesizer basics

 

There are two major kinds of synthesizers, analog and digital.

 

There are also many different kinds of synthesis methods, each applicable to both analog and digital synthesizers. These techniques tend to be mathematically related, especially frequency modulation and phase modulation.

 

* Subtractive synthesis

* Additive synthesis

* Granular synthesis

* Wavetable synthesis

* Frequency modulation synthesis

* Phase distortion synthesis

* Physical modelling synthesis

* Digital sampling

* Subharmonic synthesis

[Bob Rinker]

Originally posted by Odyssian:

I'm pretty sure that analog and modular are not synthesis types, but are descriptions of the types of circuit design (analog) and a type of synthesizer (modular).

 

You are probably correct - these are "implementations" rather than actual synthesis types. However, they do imply characteristics that these technologies represent. "Analog" implies (mostly) subtractive synthesis using analog circuit-based oscs and filters which are less precise and stable than their digital counterparts, but that's what gives them their character. "Modular" implies all of the above, plus the signal path is not predetermined - it pretty much must be patched together from scratch. As a result, modular synths provide the most versatility, although it probably takes half an hour just to get s single sound out. I realize these are over-generalizations and over-simplifications, but hopefully you get the point.

 

- Bob

[soundscape]

Originally posted by Bob Rinker:

If the playback was at a different rate than the original the pitch would change, so the same wavetable could be used for many notes - except that the real instrument typically has different harmonic nuances for different notes. A compromise is to use one sample for, say, every three notes.

As I understand, one of the main problems is that many instruments have 'fixed pitch' elements (formants) relating to their physical characteristics. Roland's Variphrase technology, I believe, is designed to overcome some of these problems when pitch shifting.

 

Of course there are other problems, Yamaha's 'Sound Board Simulation FX' (e.g. on the S90ES) attempts to overcome one of them...