FM synthesis: uhhh what?

[ectogemia]

So I've read half a dozen articles on FM synthesis, and I understand the physics and math of it just fine. Unfortunately, that theoretical knowledge doesn't quite translate to producing a target sound quite like it does for most other synthesis methods. I've been practicing with it over the past week or so, and while I'm certainly getting better at coaxing cool things out of it, I feel like I'm hitting a wall.

It's probably worth saying that I've been doing all of my FM synthesis out of Zebra2.5 which only has a very basic FM synth within it -- no crazy FM mod matrix in the conventional sense. Could that be what's holding me back?

Is anyone aware of any tips, tricks, tutorials, common approaches, etc. to producing different types of sounds out of an FM synth? Perhaps some advice on how to wrap my mind around the mod matrix in more complex FM synths and how to use it predictably and effectively?

[zircon]

http://www.sonicstate.com/articles/article.cfm?id=148

Watch this first.

[ectogemia]

http://www.sonicstate.com/articles/article.cfm?id=148

Watch this first.

Hah, I just watched that on Friday. It was so good, I took notes.

It absolutely helped, and I've even been able to apply some of the harmonic principles he talked about, but I still can't conceive of how to apply that to an FM mod matrix.

[Gario]

That video... shit, man, that helps me out. I didn't even know I needed help with that, yet.

[zircon]

Hah, I just watched that on Friday. It was so good, I took notes.

It absolutely helped, and I've even been able to apply some of the harmonic principles he talked about, but I still can't conceive of how to apply that to an FM mod matrix.

You just take the same concept and apply it to more layers, that's all. So instead of a sine wave (Op A) modulating another sine (Op , you have an already modulated signal (Op A -> Op C) modulating another already modulated signal (Op B -> Op D). There's no reason to dive into complexity if you don't have a need for it though. Personally, I don't see the need to use more than fairly basic FM synthesis. It's hard to program very complex FM sounds because there are so many bad results. Why complicate life if you don't have to? What are you trying to do that you can't do with simple FM?

[ectogemia]

you have an already modulated signal (Op A -> Op C) modulating another already modulated signal (Op B -> Op D).

That's honestly mostly what this thread was about. I reached a wall because I was doing Op A --> Op B kinds of synthesis, sometimes with a little Op B --> Op C, but I never really did mix two different carriers' signals. I guess what I'm stabbing for is if there are any predictable ways to manipulate FM at the level you described. It's just such an abstract form of synthesis that I was hoping for a few guidelines to go by instead of having to really get into it all on my own.

Why complicate life if you don't have to?

Because when I finally produced some FM patches that didn't suck, it got me really excited about using FM as a source for a ton of new sounds. I suppose I just don't know at this point if I'm complicating things to the point of diminishing returns or not.

What are you trying to do that you can't do with simple FM?

Hell if I know I just started with FM, so I'm not sure what the possibilities are. Browsing through Sytrus presets blew my mind, and perhaps it's because I'm not entirely familiar with Sytrus's interface, but some of those impressive sounds seemed to come from pretty complex-looking FM matrices. At least that's my novice estimation. Maybe it was just creative effects processing of simple FM modulation? Maybe one or both?

[zircon]

No, you're right that some of the Sytrus patches are outstanding. I guess what I'm saying is that I don't know of any shortcuts or common matrix setups, so a lot of it is massive trial & error and simply studying existing patches. Try disabling knobs one at a time (right click in Sytrus) and see how that affects the sound, for example. I really can't claim to be a master of FM synthesis myself and usually only do custom stuff from scratch if I'm making a basic, more aggressive bass or lead. The rest of the time I use existing patches with alterations.

[liquid wind]

virt's presentation on FM synthesis is also pretty beast, just note that some of the internet links he mentions don't exist anymore

[Nase]

i suppose it's pretty much the same as learning subtractive synthesis, just a bit more complex and counter-intuitive. but you still learn by trying out shit, reverse engineering stuff, and learning from others about effective ways of creating certain types of sound.

in the subtractive (pre-wired) world there's usually only so much you can do, and you quickly gain the analytic skills to mentally deconstruct sounds, hearing the different modules at interplay.

not having bothered much with FM myself, i'm having trouble hearing what's going on beyond the most basic 2-OP sounds. that'd surely change though if i decided to spend some serious time with an FM synth.

you'll just have to get at it yourself and see if it's worth the effort. i've watched most of the links posted here i think, but in the end, they didn't really raise my level of enthusiasm for doing FM synthesis ^^

[ectogemia]

virt's presentation on FM synthesis is also pretty beast, just note that some of the internet links he mentions don't exist anymore

Ooooo, very nice. This is definitely what I was talking about. Mega bonus points for virt being the speaker.

Well, if virt is saying that most of the great patches are produced by random tweaking, I believe it. It seems like there's not a lot of rhyme or reason to FM just based on the sideband generation process's nature. Oh, well, it's still fun to screw around with all the knobbies

Anyone else have any other resources or tips?

[Nase]

Well, if virt is saying that most of the great patches are produced by random tweaking, I believe it. It seems like there's not a lot of rhyme or reason to FM just based on the sideband generation process's nature.

while that's good advice, it's probably as helpful to remember some 'template' sounds you can quickly patch together, then go on from there.

you could try replicating some signature fm sounds first, like typical bell- or brasslike timbres. knowing how to quickly dial up these sounds should be a nice stepping stone to doing more advanced stuff on their basis.

FM is great at making sounds that are somewhere in the grey area between harmonic and inharmonic. inharmonic is easy enough to do as you have surely experienced

it's worth knowing how to keep your sounds 'clean' when you want to.

[Rozovian]

There's a limit to how inharmonic you can make it if you can only use integer multiples of the note pitch. It's there to keep you from making interesting, slightly detuned fm sounds and terrible, terrible noise.

You'll find that the cleanest results come from powers of two, and that integers lead to better results than non-integers. Once you know why, you'll understand fm synthesis a lot better. Start thinking about integer multiples, waveforms, and harmonics.

[lazygecko]

Where's the second part of that Sonicstate video? Navigating the website is a mess.

[Yoozer]

Deconstruct the sound you want in components - e.g. initial attack and sustain sound, or low velocity timbre/high velocity timbre. Then, use groups of two or three operators as building blocks for these sounds.

Don't bother making anything you could do in a fraction of the time on a subtractive machine unless you are making a soundset or want to see how clever you can be - that includes making Sega Genesis tracks.

Gecko - I don't think part two was ever made.

[SnappleMan]

I don't think the piece by piece deconstruction method works with FM synthesis that well. Not for me at least... It took me a while of just playing with it to finally understand how to get the sounds I want from it, and I don't even fully understand the math behind it. I just really made it a point to learn the mod matrix inside and out, and everything just started making all kinds of logical sense to me.

[Fishy]

The maths behind FM are not too complicated. FM is basically a super vibrato where the rate of vibrato is linked by (what can be a very simple) equation to the frequency of the note you're trying to play.

So you have a fundamental sine wave frequency oscillator that produces a sine tone. Then you have another oscillator that varies the frequency of your first tone like vibrato. At a modulation frequency of below the audible spectrum, (a vibrato of 20Hz or so) it sounds like anything from slight to very silly vibrato. Above a vibrato frequency of 20Hz though, strange things start to happen. Instead of vibrato you start to get additional frequencies related to the fundamental appearing that change the timbre instead of the pitch. What they are depends on the relationship between the fundamental frequency, and the modulation frequency. This relationship is called the modulation index.

So if you have a modulation index of 2, that means the mod frequency is always twice that of the fundamental. So when you play a note at 440Hz A on your keyboard, that would be being varied with a vibrato of 880Hz, which is ridiculous, but sounds awesome. In fact if you go through and try out all the basic mod indexes of 1, 2 3.... up to 10, you've basically heard all the sounds a Sega Genesis makes cause of it's simple FM synths .

Just try and keep all your modulation indexes integer numbers and the sounds will always be fairly harmonic. FM tends to produce spectra that only contain frequencies that are multiples of the fundamental, rather then standard harmonic series components. That's why the majority of first order FM sounds are similar in timbre to bells and chimes.

Obviously you can muck things about with FM8 and such by starting with a square wave instead of a sine wave. Or modulating the mod frequency with another oscillator. Try building a simple uber vibrato in Reactor if you have it. It really helps you understand whats happening.

[ectogemia]

Thanks, fish. Very useful info.

I had surmised most of that after seeing the first half of virt's seminar linked earlier in the thread, but I hadn't heard about the modulation index. I've just been doing FM by detuning the modulator by even octaves which maintains the integer ratio, but I guess I've been missing some potential timbres in doing that.

Do some FM synthesizers come with tweakable modulation index parameters or do they tend to expect the user to tweak the frequency of the modulator by hand?

[Fishy]

Well the makers of FM synths probably just want you to muck about with it and see what fun sounds you can get, and figure out how it works like that, rather then why it works. As soon as you start blathering on about maths in the manual only people like me will care.

The only reason I know this is cause I programmed an FM synth for funsies in Max MSP.

[Rozovian]

The best way to learn it, as Fishy already alluded to, is to create one of your own. I've screwed with FM a bit in Reaktor, and it's really a lot quicker to learn that way than by reading and screwing on invisibly connected knobs in an unfamiliar interface.

[Fishy]

honestly it's super simple to build one.

1. Oscillator of frequency a

2. Oscillator of frequency a mutiplied by selectable modulation index (any integer between 1 and 20 are all useful harmonic sounds)

Oscillator 2 modulates Oscillator 1 like a regular vibrato. Instant Sega Genesis.