Why do FM patches so often sound like bells (NOOB ALERT!)

Post your patches here! Put the format(s) in the subject.

Moderators: Joe., lisa, luketeaford, Kent

Post Reply
ulrichburke
Learning to Wiggle
Posts: 9
Joined: Sun Dec 22, 2013 4:54 am

Why do FM patches so often sound like bells (NOOB ALERT!)

Post by ulrichburke » Tue Jan 21, 2020 12:29 pm

Dear Everyone.

If this is in the wrong place tell me, or move it!

OK. I'm trying to wrap my head around synths and mi brane urtz, so be gentle with the stupidity of this question.

I've just downloaded DEXED which is a legit. free Frequency Modulation synth that has a ton of patches. Now from my own understanding, obviously incorrectly, I thought EVERYTHING was Frequency Modulation. When you hear an orchestra on television, you're hearing one carrier wave frequency being modulated (the 'boooop' on a dead channel!) When you use the phone you're modulating a carrier wave. Ditto with FM radio.

So I thought the patches would be rip-snortingly perfect instruments of all descriptions, modulating the synth's frequencies in the same way a rock band with top-range instruments modulates the TV carrier wave, or an orchestra modulates the FM radio frequencies. "Gonna have the Best Sounds Ever!" I thought.

Er - no. What I've got is alien bells, alien ants, alien creatures from Martian forests, a few undistinguished Oriental cymbally things all repeated over and over again. Where's my thinking wrong? I've tried programming the thing myself - not that I understand it, but I twiddled virtual knobs a lot - and I'm beginning to think it's not possible to construct perfect instrument sounds using it. Not just I can't do it cos I don't understand it - which I don't, not really - but that it can't be done. And I don't understand why.

Why can't someone just play an instrument into the synth - which would modulate the frequences, no? - and digitally store the patterns thus created, so they could be called back up to modulate the frequencies again, thus giving you a collection of perfect instruments and the ability to store any sound you played into it, just by storing the modulation patterns? Where's my thinking wrong?

Yours very puzzledly - go gentle with the answers and feel free to modulate the airwaves with hysterical laughter at my lack of knowledge -

Chris.

Be Sandy?
Veteran Wiggler
Posts: 509
Joined: Sun Apr 08, 2018 12:04 pm

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by Be Sandy? » Tue Jan 21, 2020 1:19 pm

FM isn't really "everything" as you describe. While radio may be FM, the sounds that make the music played on one may well not be. To use your orchestra analogy - FM would be trumpet one injecting its sound directly into the tubes of trumpet two to modulate its sound, not the combined sound of the whole orchestra playing together.

Here's a short primer on FM from soundonsound.
https://www.soundonsound.com/techniques ... modulation

FM makes great sounds but I've never really thought of it as producing "instrument perfect" sounds. It depends how you program a synth obviously but it does excel in my opinion in percussive and bass type sounds. And aliens!

Two things to experiment with:

1) Linear FM vs Exponential FM

2) Vary the depth of the modulating signal.

If you're looking for accurate emulations of other instruments using synthesis I'd look into physical modelling.
https://en.wikipedia.org/wiki/Physical_ ... _synthesis

If you're not overly bothered about the synthesis part then there are hundreds of sample libraries out there that are literally the sound of the very instruments you may wish to use.

User avatar
Yes Powder
Powder that makes you say "Yes"
Posts: 1344
Joined: Thu Sep 03, 2015 2:23 am
Location: Albany, New York
Contact:

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by Yes Powder » Tue Jan 21, 2020 2:46 pm

You're confusing FM Radio with FM Synthesis.
Going back to your orchestra analogy, what you hear is not a carrier wave being modulated by the orchestra, because by the time the signal reaches the audio amplifier of your television it has already been demodulated by a circuit in the TV.
Also, the "booop" of a dead channel is not that carrier wave; it's just a standard 1KHz test-tone. A 1KHz sine-wave is not sufficient to modulate full-range audio if the intent is for eventual demodulation and reproduction. The purpose of modulation in a radio transmitter is to get the signal's frequency up into the RF range so it can be wirelessly transmitted, which just so happens to have the side effect of changing the spectrum of the signal. An actual broadcast modulator will first modulate the signal with a carrier well into ultrasonic territory, and then use one method or another to multiply that frequency up to RF ranges. When the signal is picked up by your TV it is demodulated, leaving you only with the sound of the orchestra. If you were to listen to the audio before being demodulated (if your hearing could extend to those ranges) it would sound very strange indeed.
FM Synthesis as it is known to the synth world all stems from John Chowning's work in the field, in which complex sounds are created by running pure sine waves of different frequencies and amplitudes into each other in a similar way that an FM transmitter would, except the carrier waves generally stay within the audio range, and there is no demodulation afterwards because the more complex spectrum is the intended result, rather than a simply a middle-step in a bigger process.
ulrichburke wrote:
Tue Jan 21, 2020 12:29 pm
Why can't someone just play an instrument into the synth - which would modulate the frequences, no? - and digitally store the patterns thus created, so they could be called back up to modulate the frequencies again, thus giving you a collection of perfect instruments and the ability to store any sound you played into it, just by storing the modulation patterns? Where's my thinking wrong?
Easier said than done. There have been attempts over the years to do something like this in one form or another (Quasar M8, Technos Axcel, Yamaha VL1, Hartmann Neuron, to name a few) but it turns out that generating a convincing and dynamic sound to mimic an acoustic instrument requires some heavy duty modeling of the specific instrument you're trying to recreate- down to the shape of the instrument, the materials it's made of, and how the simple(?) act of playing it may affect the parts of that instrument.
As Be Sandy? said, it sounds like physical modeling is what you're looking for. Look up a company called Audio Modeling. It's really fascinating stuff.

User avatar
orangehexagon
Super Deluxe Wiggler
Posts: 1163
Joined: Wed Jul 24, 2013 5:30 pm
Location: CHICAGO

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by orangehexagon » Tue Jan 21, 2020 3:03 pm

There's a lot to unpack in your question, but I just wanted to respond to the last part of your post regarding storing data from a recorded instrument and playing it back as a synthesizer... I've often found that the 'uncanny valley' effect of this is ultimately off-putting. If we're talking about making music here, i've often found that synthesizing a sound that is close to an acoustic instrument, but not too close actually has a much better emotional response (obv subjective). I feel like this is why most physical modelling sound cheesy.

User avatar
orangehexagon
Super Deluxe Wiggler
Posts: 1163
Joined: Wed Jul 24, 2013 5:30 pm
Location: CHICAGO

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by orangehexagon » Tue Jan 21, 2020 3:07 pm

Regarding FM patches all sounding like bells... that might just be the situation with the DEXED presets?

I've been using Ableton's Operator and the Digitone a lot lately (both FM synths) and I've been making some amazing bass drums, thick bass, hi-hats, and snares. Some of the sounds I've been making sound really close to the acoustic thing and some are futuristic AF.

The basic Sine modulating Sine played mid to high octave usually does start off sounding like a bell-ish thing. Experimentation is monumental!

User avatar
commodorejohn
Ultra Wiggler
Posts: 914
Joined: Fri May 03, 2013 4:19 pm
Location: Placerville, CA

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by commodorejohn » Tue Jan 21, 2020 3:31 pm

I don't think anybody's going to top Yes Powder's breakdown of the broadcast process, but just to condense it down to a short summary: in FM radio, the sound you hear is the modulator signal, after the receiver has reconstructed it from interpreting the frequency-modulated carrier signal using the "true" un-modulated carrier frequency (which is known because that's what you tune the radio to) as a reference.

In FM synthesis, the same core process (or something very similar) is used for a completely different purpose: to create complex and interesting timbres that would be difficult to duplicate with more conventional subtractive synthesis methods. The Sound on Sound article linked in one of the replies is the single best summation/explanation I've seen of how this actually works; go read it, whatever you do.

As for the bells! The bells! it's a notable tendency with FM patching. I'm not up on my crazy signal math enough to say what the exact basis for this would be (if any,) but something about the overtone series-es that audio-rate FM produces just lend themselves really well to clangy, chimey, metallic type sounds - and this leads to noticeable selection bias. People tend to make lots of bell type sounds with FM synths because FM synths tend to be really good at them. They can do other things if you know what you're doing and work at it, but it's dead easy to create bell sounds with FM even when you're just starting out.
Computers: Amiga 1200, DEC VAXStation 4000/60, DEC MicroPDP-11/73
Synthesizers: Roland JX-10/SH-09/MT-32/D-50, Yamaha DX7/V50/TX7/TG33/FB-01, Korg MS-20 Mini/ARP Odyssey/DW-8000, Ensoniq SQ-80

"'Legacy code' often differs from its suggested alternative by actually working and scaling." - Bjarne Stroustrup

Marizu
Common Wiggler
Posts: 59
Joined: Tue Sep 27, 2016 5:12 pm

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by Marizu » Wed Jan 29, 2020 5:17 am

This is a really interesting question. I think that it may be uesful to look at it from the other side.
"Why does subtractive synthesis not usually sound like bells?"
Traditional subtractive synths start with waveforms like sawtooth and square wave. These base waveforms are actually clusters of harmonics that are related to the root pitch. They are basically consonant with the root pitch. This means that if you start to shape the waveforms with a filter, you are still working with a collection of sounds that are harmonically related.
A bell sound is interesting because it starts with an inpulse of dissonant tones and then settles down in to a fundamental root tone. In FM synthesis, it is easy to generate the dissonant frequencies that are common in bell sounds. In a DX model, each operator consists of a tone generator, an envelope and a vca (okay, it's not voltage controlled). This makes it easy to generate the dissonant strike tones that we associate with bells. If the strike operator is short and the main operator is longer, then you will get dissonance foillowed by consonance, which is the bell type of sonic signature.
FM is massively rewarding. Understanding the 'algorithms' (patching) is key to being able to make the sounds that you intend. There are plenty of great tutorials on Youtube. I recommend the PowerDX ones.
In our Eurorack world, we have amazing modules like Akemie's Castle which allow us to feel our way through this process, rather than think our way through it. You still benefit from understanding the types of sound that you are likely to get from the different operator configurations.

I'm not sure whether it is theoretically possible to construct 'perfect insurument sounds' using FM. You need to appreciate that many wind and bowed instruments modulate their sound massively over time, so you would probably need to constantly modify your synthesis. I have seen incredibly powerful additive synthesis (not the same as FM) that seems to be able to take sonic signatures and manipulate them. I think that the Rossum Panharmonium does this. I know that FL Studio's Harmor plug in does it. My Waldorf NW1 does it, too, but it has low resolution analysis, so whilst it sounds charming, it doesn't really sound accurate. This isn't traditional sampling as the samples that are taken are analysed and then resynthesised from sine waves.

User avatar
oberdada
Common Wiggler
Posts: 219
Joined: Tue Nov 01, 2016 12:06 pm

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by oberdada » Fri Jan 31, 2020 7:23 pm

Bell sounds are inharmonic, and FM with an inharmonic carrier to modulator ratio produces inharmonic spectra. Then the amplitude envelope needs to be decaying, and if the modulation index has a similar decay it will simulate the dying out of high partials before lower partials. However, the way the modulation index corresponds to a spectral profile through those esoteric Bessel functions is a sound signature of FM that cannot be found in acoustic sounds. (Feedback FM is better behaved in this respect.)

If you restrict the c:m ratio to integers or simple rationals like 1/2 or 3/2 you will get harmonic sounds. Ring modulation and frequency shifting can also be used to produce inharmonic spectra and are often used along with FM to make all sorts of bell sounds.

There have been attempts to model at least harmonic tones of some acoustic instruments using FM and automated search of the parameter space by genetic algorithms. Andy Horner has written a few papers on the subject.

User avatar
galanter2
Common Wiggler
Posts: 127
Joined: Mon Aug 04, 2014 8:32 pm
Location: Texas A&M University
Contact:

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by galanter2 » Fri Jan 31, 2020 10:31 pm

The above is probably the best of the answers. To supplement that at an even more basic level.

A vibrating string, or organ pipe, or (for the most part) horn, etc. gets its timbre, its unique sound, because it vibrates at frequencies that are simple multiples of the root note. This creates a rich harmonic timbre.

Bells, however, vibrate at frequencies that are not simple multiples of the root note. Because those frequencies are not harmonic, i.e. are inharmonic, it creates a metallic inharmonic timbre. (It may even be difficult to identify what the root note is because some of the harmonics will be *below* the apparent root.)

A traditional analog synth uses oscillator waveforms like triangle, square, and sawtooth. Those waveforms are more like a vibrating string or organ pipe, and tend towards rich harmonic timbres like strings, organs, horns, etc.

An FM synth uses frequency modulated waveforms. FM creates frequencies that are (usually) not harmonic, and are more like the inharmonic timbres of a bell.

User avatar
commodorejohn
Ultra Wiggler
Posts: 914
Joined: Fri May 03, 2013 4:19 pm
Location: Placerville, CA

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by commodorejohn » Fri Jan 31, 2020 10:39 pm

FM can create harmonic overtone serieses just as easily as it can create inharmonic ones, so that's not entirely true. But it's much easier to create an inharmonic series with FM than with conventional simple-oscillator subtractive synthesis. (Actually, I'm kind of curious - is it even possible for a simple cyclic waveform to contain inharmonic overtones? Seems like it wouldn't be...)
Computers: Amiga 1200, DEC VAXStation 4000/60, DEC MicroPDP-11/73
Synthesizers: Roland JX-10/SH-09/MT-32/D-50, Yamaha DX7/V50/TX7/TG33/FB-01, Korg MS-20 Mini/ARP Odyssey/DW-8000, Ensoniq SQ-80

"'Legacy code' often differs from its suggested alternative by actually working and scaling." - Bjarne Stroustrup

User avatar
galanter2
Common Wiggler
Posts: 127
Joined: Mon Aug 04, 2014 8:32 pm
Location: Texas A&M University
Contact:

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by galanter2 » Fri Jan 31, 2020 11:05 pm

Like I said: "FM creates frequencies that are (usually) not harmonic,"

User avatar
oberdada
Common Wiggler
Posts: 219
Joined: Tue Nov 01, 2016 12:06 pm

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by oberdada » Sat Feb 01, 2020 3:28 pm

commodorejohn wrote:
Fri Jan 31, 2020 10:39 pm
(Actually, I'm kind of curious - is it even possible for a simple cyclic waveform to contain inharmonic overtones? Seems like it wouldn't be...)
Simple answer: No. Then again, it depends on how you define "inharmonic". For the mathematically minded, there is a straightforward definition of harmonic spectra as having partials related by ratios of two integers. For inharmonic partials the ratios need to take irrational values.

From a more perceptual point of view, you could take a few very high harmonic overtones of an extremely low fundamental frequency, perhaps in the subaudio range. These partials may not sound like they are harmonically related although the waveform will repeat cyclically at the period corresponding to the fundamental.

User avatar
commodorejohn
Ultra Wiggler
Posts: 914
Joined: Fri May 03, 2013 4:19 pm
Location: Placerville, CA

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by commodorejohn » Sat Feb 01, 2020 9:01 pm

Yeah, that's about what I figured.
Computers: Amiga 1200, DEC VAXStation 4000/60, DEC MicroPDP-11/73
Synthesizers: Roland JX-10/SH-09/MT-32/D-50, Yamaha DX7/V50/TX7/TG33/FB-01, Korg MS-20 Mini/ARP Odyssey/DW-8000, Ensoniq SQ-80

"'Legacy code' often differs from its suggested alternative by actually working and scaling." - Bjarne Stroustrup

User avatar
W23
1-Post Wiggler
Posts: 1
Joined: Mon Nov 25, 2019 3:30 pm
Location: Houston

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by W23 » Wed Feb 05, 2020 12:28 pm

One of the best things about Dexed is that there are tens of thousands of DX7 presets freely available. I can’t remember how many presets come with Dexed itself, but it’s in the thousands. Find some presets that you like, and have a real close look at them. What are the carrier and modulator frequencies used? What algorithm is used, and in what order does it assign the carriers and modulators? Where do the LFOs figure in...for vibrato, or a swell? You’ll start seeing relationships pretty quickly.

I’m definitely not a sound design expert. I still spend a lot of time just fumbling around and just find it easier to let the synth (I mostly use FM8) lead me down an interesting path, but paying attention to frequency relationships helps a bunch. Aim for even number operator frequencies and multiples of frequencies. Odd number frequencies produce inharmonic results, but are quite useful with short envelopes for bell tones and even transient sounds like bass strings and electric piano tines. It also helps to break down the components of a single sound to the transient, the body, and what happens as the sound fades out. Often, I just click through the various available algos until I run across something interesting as well. FM8 even has envelope presets that are quite complex, but envelopes are pretty easy to draw in Dexed

User avatar
oberdada
Common Wiggler
Posts: 219
Joined: Tue Nov 01, 2016 12:06 pm

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by oberdada » Thu Mar 26, 2020 3:27 pm

I just discovered a phase modulation version of exponential FM which it is quite different. Its partials don't shift with the modulation index. But I'm not aware of any implementations of it, nor have I seen it mentioned anywhere.

You can read about it here on the bottom of the page.

Xomrys
Learning to Wiggle
Posts: 26
Joined: Wed Feb 26, 2020 12:54 am

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by Xomrys » Fri Mar 27, 2020 1:00 am

Bells have sparse spectra, which is what you get when you FM random whole number ratios (4:1 etc)

Try lots of 1:1 and 2:1 ratios also feedback FM, also octave doubling (two independent operator networks all exactly one octave apart), etc. you can get closer spectrally to typical saw, triangle waveforms etc.

You could also learn to enjoy the bells...

User avatar
Tun
Common Wiggler
Posts: 116
Joined: Tue Apr 09, 2019 8:56 pm
Location: Thailand

Re: Why do FM patches so often sound like bells (NOOB ALERT!)

Post by Tun » Fri Mar 27, 2020 1:18 am

The questions have already been answered. OP, have a listen to this if you haven’t already.


Post Reply

Return to “Synthesis Techniques”