Some MSF FM Questions

[dmbaer]

I've been playing around with FM in SMF and have a list of questions.

1. The merger modules offers a couple of flavors of FM but no PM (Phase Modulation) as far as I can tell. To the extent I understand it, these should work identically, at least for sine waves. But one synth I have, Rob Papen Blue 2, they sound very different indeed (but maybe that's just a Blue 2 thing). In any case, is PM planned as an addition to the merger module?

2. Many implementations of FM (or PM) offer the ability for an operator to modulate itself. Moreover, the DX7 offered two algorithms in which the lowest operator in a stack can modulate the highest operator in the same stack. I see no way to duplicate this with the MSF FM implementation. Am I missing something?

3. This one is the biggie: If I implement a simple two operator stack using pure sine waves, 100% wet in the merger module and 100% depth in the merger, around middle C (MIDI note 60), I get more or less an expected results. Around MIDI note 80 and above and the results get really ugly and unusable. This is supposed to duplicate a two-operator stack with the modulator depth at 100% and both operators at the same pitch (ratio 1:1) - nothing fancy in the least. I played with upping the oversampling with no substantial improvement. I should maybe have included a preset that demonstrates this, but it takes less than a minute to set this up. Any thoughts?

4. A simple question: Is there a difference in modulating the depth in the merger module (in FM mode) and modulation the side chain input to the merger module in an identical fashion instead? I'm specifically trying to understand what to do with the case of two parallel modulators (with two different envelopes) modulating another operator. This is a very common thing in FM presets - I'm not trying to do something off the beaten track.

[Chandlerhimself]

I’m sure Vojtech can answer these better than I can, but Maybe I can answer some of it.

1. I beleive FM is phase modulation and FM(abs) is actually frequency modulation. I could be wrong about this, but I think this is the case. Most synths that are said the be FM are actually PM, so I’m guessing that is what is being done here.

2. I hope in the future a dedicated FM module is created and it has the self modulation feature. As it is now IMO wouldn’t it just be easier to use a new OSC with the same settings as the first. The last part about the lower operators modulaing the higher, I don’t quite understand.

3. I’ll test that and see if it happens on my system. It sounds like aliasing, but oversampling should take care of it. Maybe its a bug.

4. I’d like to kkow this too, but I’d guess the volume of the side chain and the depth will work in the same way.


I hope in the future there is a FM module with a matrix for this type of thing. I like the way it is now too, but for more complex patches something like sytrus or FM8 would be easier.

[MeldaProduction]

Ok so :

1 & 2) FM here is actually sort of simulated - you cannot do FM synthesis without sort of doing the oscillation internally. Once you have 2 inputs, where you have no idea what the signals are, it's not the synthesis anymore, and more like the original frequency modulation itself. And when you think about it, phase modulation and frequency modulation are kind of the same thing, in a way you'd need different shapes.
As Chandler said, I'll probably create some dedicated FM matrix synth when time allows. I'd need some inspiration though, since I'm not really a FM fan . Any ideas? I can see Sytrus and FM8 now, so I'll check these when time allows.

3) I tried (from what I understood) and seems fine here. Please send me some preset. Note that unless you are using sine waves or use Low quality, the oscillators internally produce alias-free curves, which may look very different from what you'd be expecting, so it may be worth checking if disabling these features for the modulation sidechain wouldn't help. I think the switches are not available yet, next update, today/tomorrow.

4) I believe these should pretty much work the same.

[jmg8]

For the FM inspiration please take a look at Arturia DX7.

For me I would like to see in the new FM module the ability to change algorithms. It is a really great way to change the sound up and gives unexpected and often great sounding results.
Here is some info on DX7 algorithms:

[Whywhy]

I like the ableton operator, simple matrix or that :

[dmbaer]

The last part about the lower operators modulaing the higher, I don’t quite understand.

Go here:

http://blog.dubspot.com/propellerhead-p ... nthesizer/

scroll down a bit and take a look at algorithms 4 and 6.

[dmbaer]

I'll probably create some dedicated FM matrix synth when time allows. I'd need some inspiration though, since I'm not really a FM fan . Any ideas?

I *definitely* have some ideas, but I want to give it some serious thought before spouting off. But my approach does not involve a matrix - you've already got one in MSF, right? I'll post my thoughts/suggestions in a day or two.

[Whywhy]

Perhaps, an fm oscillator inspired by the wavetable module, without fix pm algorithms:

I think, 4 op's are enough- for my need-

4 Melda osc, represented like the individuals waves osc in the wt module,
in each osc's, 4 volume output box, with a modulator per voices value.
Like that, we could be able to make complex routing with feedback.

Edit : I forget, each osc, operators, should have a dedicated Oct, semitone, cents, phase setting.

My question, is about the transform per voices, it will be available for each osc's or for the output 4op's osc ?

[dmbaer]

I tried (from what I understood) and seems fine here. Please send me some preset.

Here's the preset. Actually, it's that preset on some steroids - it's a reasonably useful FM behavior learning tool and works nicely from the Easy view.

In any case, using this preset with the default settings should exactly duplicate the simple case I originally described: sine modulating sine at 1:1 ratio, 100% modulation depth. However, you might wish to check that I've set up the merger module correctly. That could be the problem.

Here 'tis:

Code: Select all

$eNrdWVl32joQfs+v0HH61hKQd+cAPVkg4d4QXEy2RwMKuDG244VAfv0dSzYYmy1N0-aWPMQzGmn20Weofp1NbDQlfmC5To3DRxUOEWfgDi1nVOOi8LGkcl-rB9W24UbOsGkOQtefByQMYT1Ah7CDQ4cXtts37aBn9hl961oDWM-QdjQhNa5yxEsYq6osCrjCc+jMtW3TC8iwa43GoW46xAYh4Ee+T5xQ9wloap3XuGYbXZl9Dl3ctAzrlfBwrKKoX1SZX-AaZjCP+RVN-aKpEocuv914gTnxbOoJnHrVNYAIQ+LH8hhkeazk2XAEL2ocapuzq2bn1HYHTzG7xgmgqh3ZoaWbvjkhIB4szKfH36YxVFRFk1WuXjWivm5HI8tJwlavMtJYDd9aVQaxySAkQ7YDFHwRuHR-kD6gZmTb12Yc2k4wsMAaUIPokcFTy6Fm5Zc5dE-5D6CH22DR4Y1jBa7TcMy+TTJebjAWPB2bHkFGr6Hr3YbR6MUayvVqefX0BWO7-fyK-fFjJwohLRt8wX+cL23ijwjNQ8vxorDlDMkMU42UNkJi2jixgPkppI90LeMrO4pDEIHMSSWcyaK80fNz4oXj5NGf3xEWxMO2O4STmz55jqDT52jiDiMIJ9Tu+pCUM76WFxVYjEeh3A8yHXD-vuKvLIu-vF33PahtOFNcr-Z8MnFtd1M+qRCV5feR5ZmssI+swGTFfWRFJivtIysxWXkfWZnJKvvIKkxW3UdWBVk274FARhSEZpwizO3cvNgFJ0DOMboizigu0DjjOWkqwe+UEHZKiDslpJ0S8k4JZaeEul3ilkAHWOEcNa5v0xXMli5sa0hYTDOLKl3rWRNyZU0smClw06FHyw+gyWVBVgQN86IIE8I2U5YkKGJFxHRj7iar5xgVdE4eTWDdmna05uK+8N3Iq3HJvU-v4MbQgqZns+Xa9SemDc29QrGJxrAAxMAj6cUZ62Cip3PmA-JSS+K+T5-piO2azggkRQ4t+OzgxBhUQqmK1HhafTkH8x7jvMd7OxnjDzij676wgbbOZ2IfwVBzgk1+68QfAOAJtnjbcgCkhVRFzvFF7QS0SEpIdy0nRBg9LENwVKmoa8KA82GQC4kXBGERija7KeKL9w0ZvzOnH5Jv4hCfAoYSyqIHII0kEPT-1jqQ8wFQNtbBbuf3qIQ4FsiYuG44dkgQ-KawZPRviY2Sj42KTk3n6QxeBOKJwlEqVlqpcYblkCUDMwYyniPTz-ABXDPdZIh6vgUu2JlVYWUX6oLgclHMbO0COMjtlTLLhvnC-UAWGeiIF5IkwNHBSt6+RdbgKQEnmzPXh33IgV0rOWvMwi4JaF3k+xqWPRdsIgnenS16D94i8umlJILXluQEaFBJlJOPkml4gZeE9MPVP13eek8dTdNxT2ufn7T88azf1Us9DV9+9-7xG6Xr1zb9E19Lo8-30vP0sufoYv9k-txshlFQAsyVGvJDbq11Q83UX-3TybaPflIqMs9f9Jfl+oqJuVpW87WsvW+mdaACp6TYv3HkRbhjf7ibs+lXpLd0t26Fg3HMTk1bVIiUMY6X1nS6VrgNKz916hkEIIrrkF8879KIZNRvGXe4iAnyoAC-8yakL2QfHIP05RMeE3XbfC4AACxmh7yyMuTFY7wy4-ksHX-3kqWFmOZXBjg+FlYmNj4Wl7Qc0ypXxB9LuNmNX1LfNrfPjtuIbts+vBOgsxyYHzjTeWnDNAS8v+xZPtu0AkyC+qegPeannmq+wogmYyPqfr6807XT64fx6UNn+LKYj+J3NZ2PI30+mp39+zK4i0aleSMQrJvmSH5Vax8yzjFeO8-F84w96bzuLOe5eP7Zf6YfP1i-nrM3X8RioYjVrTD2zPR9i-zJIBa-CcTiwu2GtY0AaJfzfzKIxW8Hsbhwt-GVvwDGbsri-wnEYu3vQLF85ZfBWL6AUXj8nqm2Ccb+MuCK3wdc+QJ+4fmfOOh+A27Fb8atPF9gBfWD6skgtKaE-YSGskQ6IlZ-C0mpnk9YqcchY1-rnkH1j1x-3grJpOU8umBC8gNdcRfzjVqZ1Rnb43sxEiPxF6D0G8jki8c7axiO2-Wq4Ub+gCCQf7RsWEp+tYtPSkXK7dS58tqfI+sH-wHOokKH

[dmbaer]

Any ideas? I can see Sytrus and FM8 now, so I'll check these when time allows.

For anybody wanting to get their hands on the best FM tutorial ever (IMO), get this Kindle mini-book:

https://www.amazon.com/How-Make-Noise-F ... B008H7CEQG

It costs all of three bucks (USD).

I assume this is just extracted from the full book How to Make a Noise by Simon Cann. I reread it recently and all kinds of lights popped on in my head. It demystifies how to effectively use FM better than anything else I've ever encountered.

[jmg8]

The whole book is available for free here.
https://noisesculpture.com/how-to-make- ... ogramming/

[Chandlerhimself]

I think an approach similar to Sytrus, FM8 or Rhino will work the best. I think FM8s matrix is the best because you can easily see the values by looking at the matrix and you can tell where the feedback is going. It also allows all the “algorithms” from the DX7 while also allowing new ones not available in older hardware. The matrix isn’t the easiest thing to work with, but all the other alternatives seem much worse. Maybe there is an easier way that nobody has come with, but IMO its not worth it trying to re-invent the wheel.

I’d also like to see ratios and hz used instead of octaves and cents. Most FM synths and tutorials use ratios, so it will be less confusing if the same terms are used. I’m thinking something that looks like the FM8 matrix with 6 pannels under it for the 6 operators. Each pannel has a waveform, ratio, hz and level control that are modulatable. The only other thing I can think of is a unison control. This would IMO combine all the good points of sytrus and fm8 and make it easy to program.

[MeldaProduction]

Well, we have quite a few contenders here

[dmbaer]

This is my proposal for a clean FM solution in MSF. Let me first establish my bias with respect to FM. I’m not a big fan of traditional (DX7-like) FM sounds on their own – they are not all that interesting. To me the real value comes in using an FM sound as a component of a richer sound that includes sample playback or analog-like timbres. This is quite ably demonstrated by Rob Papen Blue II, which has some incredibly elegant sounds that incorporate FM but that don’t sound at all like they were made with a DX7 approach. Rhino, for all its age, is another synth that doesn’t do too badly in this regard.

So, this suggestion assumes the FM capability will often be used in larger contexts. The other assumption is that unusual applications of FM, however far off the beaten track, can already be accomplished courtesy the merger module, so an MSF FM module does not have to cover all the possibilities. My suggestion is for a straightforward FM module that does DX7-ish sound creation only. For a great many situations, that will be that’s desired.

As a model, take a look at how this was implemented in UVI Falcon. You can find a brief description, including a picture of the available algorithms, here (scroll about 1/3 down):

http://soundbytesmag.net/falcon-synth-o ... e-up-look/

Something like this is what I’d like to see in an MSF FM oscillator module. I’d also limit the MSF FM module to just four (or fewer) operators, but I wouldn’t bother with all the algorithms in the Falcon version. Rather than the final one in Falcon (two parallel stacks of two operators), for example, you could just use two parallel MSF FM modules each with a two-operator stack. The ninth one also is totally unnecessary. What I’d do instead is to go back to the original DX7 list and first eliminate all the algorithms with a single carrier and five modulators on top. Those are rarely useful for anything. Then pair down the list further by separating parallel stacks in the algorithms. E.g., I would hope to include the right half of DX7 algorithm 4, the one with the full stack feedback (see the link above for a picture), but there would be no need for the left-hand stack since a second FM module with a three-operator stack (and with feedback level set to zero) could be used instead.

The MSF FM module would have multiple algorithms, but each one need not have fully four operators. Again, the DX7 algorithm 4 in the MSF version, for example, could just have the right stack of three operators. To fully duplicate DX7 algorithm 4 in MSF, you just use two FM modules with the three-high stack MSF FM algorithm. However, we’d also want to see the right hand stack of algorithm 3, also just three operators, but with the feedback configured differently. Again, see the link above for a picture of algorithm 3.

The MSF FM module could, I suppose, offer wave forms other than sine, but that’s not all that useful in most cases. Certainly the full Melda oscillator capability would be serious overkill. With FM, you can’t stray very far from sine waves before the results quickly become unmusical. However, a few alternates like a triangle and a filtered square wave might come in handy from time to time. One other possibility might be for the (single) carrier operator to use a sidechain-supplied signal as that which is modulated.

There would not need to be a lot of modulation targets in this module. Level (for each operator) would be the big one. Pitch/frequency/ration/whatever would be the other.

So, there you have it. I want a solution somewhat like that in Falcon, but with the differences as outlined above. If this idea gains any traction, I’d be happy to propose the set of algorithms to constitute the algorithm capability in an MSF FM module.

[Fourstepper]

I like what chandler said