Welcome to Yamaha Synth! Thanks for the questions.

The FM-X provides the possibility to use non-sinusoidal oscillators: are known waveforms available?

The short answer is, yes. The FM-X engine of the Montage features Spectral Forms which are best heard to appreciate the difference. A Sine wave is the definition of the Fundamental containing no harmonics (overtones). Harmonics are whole integer multiples of the fundamental frequency. In DX-style FM you build increasing complex wave shapes by the interaction of sine waves. When the pitch of one Operator is uses to modify another is your basic FM. When the pitch ratio of one is 1:1 (equal) to the other, you get a result that is all the harmonics. When that ratio is 2:1 Modifier to Carrier, the result is every other (only the odd numbered) harmonic.

All harmonic we identify as a Sawtooth Wave
Every other harmonic we identify as a Square Wave.

FM-X gives an option to select "All" harmonics (sawtooth like), just "Odd" harmonics (square like) and spectral shapes with a resonant peak (emphasis) ... These non-Sine wave sources can be further strengthened via a parameter called the Spectral Skirt which influences the timbre of the resulting non-Sine wave source. You are provided 6 alternate wave sources of differing harmonic content (source waves) - which allows extremely complex waves from any single Operator... And there are 8 Operators per FM-X Part!

An "All" (harmonic) Spectral Form with a Spectral Skirt set to maximum (wide) will give you a Sawtooth Wave.
An "Odd" (harmonic) Spectral Form with a Spectral Skirt set to maximum (wide) will give you a Square Wave
A "Res" (resonant harmonic) Spectral Form provides increasing higher frequency emphasis and a selection of complex synth sound sources.

These Forms in conjunction with the Skirt parameter generate the same type of energy with a single operator as a Feedback loop on itself would (except with a lot more control)... The classic single Feedback Loops found in the 88 algorithms ensure compatibility with legacy DX/TX FM sounds. But building sounds from scratch with FM-X you can start with a complex wave.

Significantly the resulting array of harmonic complexity as a starting point far exceeds the selection of a handful of wave types in your traditional analog synth. And once you start interacting this singular wave source by modifying or allowing it to be modified by another Operator, you begin to see there is truly an infinite range of wave sources per FM-X Part.

We will be posting some basic Fm tutorials - once you master selecting the Wave shapes building sounds (on purpose) becomes very doable in FM programming. In the 1980's FM was seen as difficult. Back then the DX7 introduced menu driven interface. Using a computer was still new, navigating parameters and numbers in a screen was completely New. Now with everything from games to cell phones, navigating an operating system is literally child's play. Add to this the knobs and sliders that help adjust FM levels and settings... This time around FM will be a lot easier. The reputation of being difficult or hard to use will be greatly reduced. Time is always the best teacher. Everyone is very used to navigating and accessing parameters (this was completely foreign for most in 1983)

It used to take two Operators to create a complex wave shape. And we'll cover that relationship - because it reinforces what you know about the harmonic series and how it is generated.

Then we'll dig into starting with more complex waveforms from a single Operator.

It seems that for every operator can also edit the waveform, using a few parameters: it is true?

Very true.

It has confirmed that each FM-x voice has a multi-band filter with its envelope?

First, there are no Voices in the Montage, what you know as a Voice is now one of eight PARTs that can simultaneously be under Keyboard Control, a PART can be AWM2 or it can be built from an 8 Operator FM-X. Each Operator has a source wave, and an Amplitude Envelope Generator which controls its output over time, and each Operator is either a Modulator (influencer) or a Carrier (outputs sound). Once the Operators have interacted to generate a complex output (sound), that sound can run through a Montage Filter (LPF, HPF, BPF, BEF, Dual band Filters, Etc., etc., etc.) Effects can be added, controllers assigned. Filtering, Effects and Controllers are assigned on the PART COMMON level of editing.

last question: for each operator can provide feedback as to reface the DX?

Explained above.. With the "Spectral Form" and "Skirt" parameters you doing what the reface DX does when it feeds back an Operator on itself.

Hope that helps.

Thank you:D Byeee from Italy...and I hope to buy and play montage soon 😀

The syntesis on FM-X has some similarities to FS1R?

The FM-X has "some similarities" to all Yamaha DX/TX engines. And yes, there are some to the FS1R, but there is no Voice compatibility with FORMANT SHAPING (FS1R) Voices. So far, Voices from DX7, DX7mkII, TX802, TX816 have been announced as compatible (a conversion application) is in the works.

FS1R -for those who may have missed it- was a 16 Operator FM rack synth (circa 1998 or 99). 8 Operators (Voiced) made musical tones, 8 Operators (unVoiced) made articulate noises... That's how they got that thing to talk and sing! As far as I know there is not any direct Voice compatibility with FS1R.

Hi Phil,

I hope you saw my message giving you all congrats on this awesome synth! I know you've heard me complain loudly, so I need to COMPLIMENT LOUDLY!

I can't wait to read more about the FM programming (yes, I'm a FM geek!). Maybe a parameter "tree" or flow chart? I realize that may still be in development 😉
This sounds (couldn't resist the pun) really intriguing to me and I want to support it.

Again, GREAT JOB, and I look forward to reading anything you can tell us about parameters within the FM-X.

Steve

also, is there an "EG Bias" available at the operator level that can be assigned to a controller?

thanks!

Thanks for the question,

EG Bias was how, in the original DX-FM engine, you were able to additionally control the Amplitude Envelope Generator within each Operator via the Mod Wheel, Aftertouch, Foot Control or Breath Control. Normally a Program is activated by the Note-On key velocity, that determines the amplitude, and affects the resulting envelope loudness contour. The higher the key velocity the closer the output moved toward the main channel volume setting. EG Bias is a method to add to or replace this function with a physical controller.

Since in FM, the AEG is encased within each Operator - this differs from the analog synth paradigm where all oscillators likely share a single Amplifier Envelope Generator - individual control of the response of Operators allows multi-dimensional control within a sound. For FM, this means you can change harmonic content by manipulating 'modulator' output level independent of the 'carrier' output level.

When the FM engine is encased in a host synth, that host has access to the FM engine at very critical points. The FM-X engine can be accessed with the architecture of Montage. You can 'bias' control of Operator EG on a per Operator basis, from a long list of potential control sources (40 in all), including all the usual suspects (Mod Wheel, BC, FC, AT) and new possibilities like a Motion Sequence, the Envelope of another signal, even an external signal... so an external audio signal can be used as a control source to an FM-X Operator.

Phil,

Thanks for the response & answer- that is awesome, WOW, the real-time control on this axe is going to be amazing....

Can't wait to play around with a Montage eith my own hands. Super excited!

By the way, if one limited oneself to only Voiced Operators not set to "frmt" (formant) form and didn't touch any formant sequencing, the FS1R's "pure FM" functionality does seem quite similar, with respect to number of operators, number of algorithms, choices in spectral form (sine, all, odd, res), and skirt.

Hi Phil,

More questions about "even an external signal... so an external audio signal can be used as a control source to an FM-X Operator."

Does this mean that I could take a signal from a microphone (or say, an aux send from a mixer of a mic source) and use that level to effect the output level of an operator(s)? Could I also use that modulation source to also effect the pitch of an operator? Would the "operator" only respond to the input level, or could it also respond to the pitch or even content of that audio source? (I know you know where I am going with this)

Thanks again,

Steve

Bad Mister wrote:

FS1R -for those who may have missed it- was a 16 Operator FM rack synth (circa 1998 or 99). 8 Operators (Voiced) made musical tones, 8 Operators (unVoiced) made articulate noises... That's how they got that thing to talk and sing! As far as I know there is not any direct Voice compatibility with FS1R.

By the way, I own a FS1R and wonder why Yamaha didn't implemented its techology in full in the Montage. All in all the formant capability would have been very cool in a AWM2+FM-X synthesizer.
Do I have to think it will be implemented in a next instrument that will substitute the Montage?

I've mentioned the same thing. It would be "nice" if when introducing a sound engine that the sound engine would be the "best of breed" example. Montage elected to scale back some features - maybe usability/complexity/cost/etc. If usability/complexity was the main driver - it'd be great to get more creative how to place advanced features "deeper" users still get the "stream-lined" interface as today while also having the flexibility to go deeper and invoke advanced features if they were so bold.

Given the gate count to implement FM is very small - as technology has shrunk transistors many fold since the FS1R - cost in terms of FR4 and silicon is not really a factor. Development time maybe (TTM/schedule - man hours/cost). Who knows.

This generation you do get some new FM features vs. the past - namely in how it integrates with effects and motion control (more flexibility in general). Next gen I'd like to see continued pushing for new stuff (not necessarily just throwback like an FS1R clone) -- but more features of the past (the best of the past) PLUS more new forward-thinking features.

Jason wrote:
Given the gate count to implement FM is very small - as technology has shrunk transistors many fold since the FS1R - cost in terms of FR4 and silicon is not really a factor.

The FM-X in the Montage is certainly not implemented using dedicated logic gates. It is generated using a SWP70 digital signal processor (DSP) which has been programmed to do the FM synthesis. There is an article about the architecture which describes this in detail: http://sandsoftwaresound.net/montage-hardware-platform/
So I guess Yamaha actually might be able to implement the FS1-R architecture even in the current Montage - if they chose to. I am not sure if the DSP would have enough processing power. I would guess, yes, but no way of for me to know. I certainly hope that there will be tons more software updates, hopefully adding tons of new features. Even for the current Montage.

Prelude:

I'm going back adding this line because maybe you misunderstood. "Discrete" logic gates may be what you thought I was conveying - no. Discrete is like a part that's one transistor (or one something - one resistor, one cap). This isn't what I said nor meant. Lets ignore the analog portion of a mixed signal IC - the digital portion of an ASIC is made up of transistors. Traditional construction is with metal layers, doped regions, etc. There's newer tech. out there which adds to the mix. When you hear process specs of "XX-nanometer" - this is the size of a single transistor. We keep squeezing more on a die (the number before "nm" or nano-meter keeps getting smaller).

-----

The SWP70 is the very definition of dedicated logic gates. It's not field programmable. Transistors are fairly cheap. ASIC programs are not necessarily cheap. But the difference between an FS1R-feature-set capable DSP and an FM-X-feature-set capable DSP is very little all things equal.

They may have the "stuff" in the existing silicon. Might not. Point was that the cost of having it vs. not having it is fairly marginal in terms of the atomic-level hardware (gates). At one point it would have cost a fortune to stuff more logic gates to arrive at one design vs. another. Today you have to throw away logic because transistor geometries are so small - your package ends up becoming a limit more times than not. Since there's little up-side in bonding out for FS1R features vs. not having them (it's mostly just part of the signal pipeline - not another pipe necessarily) - there's not a huge need for a package change.

I've read the SSS pages at one point. My comments were much more general than architecture. Throwing architecture details doesn't change my original comment.

The reason why I mentioned that the raw cost of the "sand" part isn't that much of a factor is that cost is often touted as rationale why things do not exist. This may be true from a testing and program perspective. Most of the upside can be amortized and is not what seems to be alluded to when cost adders are discussed from Yamaha's feedback.

I threw out some other possibilities of cost adders - but, as mentioned, they are not "hard" costs. Meaning they do not add to every Montage the same amount like a gold keybed would. The costs add like R&D costs does. You may for it - it's capped (over) - and the more you sell, the more you slice up that cost. If you sell an infinite number of units, then this cost adds a limit of zero cents per unit sold. Certainly, sales will be less than infinite.

-----

... and although "cost" is thrown out as an easy answer of why things are the way they are - there could be other more "real" reasons why an FS1R engine isn't there. Ignore usability/etc - since as I mentioned - great design can make difficult things easy or just "hide" difficult stuff under a layer so you don't have to deal with it - Steve Job mentioning that the file system is something that normal consumers won't have to fool with but will just be an app for power users (came up recently in news about iPhone leaks and such). At any rate - could have been that the ASICs available for TTM (to make schedule) had to re-use stuff before Montage was much of a concept. So despite what may have been better - what was available at the time decided the features just so a product could get out the door. It can take a long time to qualify a new ASIC - and even longer if you run into snags. So you may have to "milk" older tech for a while until your new "core" catches up to your marketing bullet point aspirations. Lots of reasons other than this as well.

Doesn't quite matter which of the "conspiracy theories" is right. Customer feedback should be clearly expressed - and some gathering of customer feedback should be done (as is already done - some debate if it's "good enough" - not going there) in order to shape future products.

I don't offer some of this information in order to generate the correct conspiracy theory. I'm just trying to combat some of the uni-dimensional thinking that usually ends with a rant of how Yamaha should have done something different in the past. Better to focus on the present (program your keyboard to do what you want to make music - using how it is today) with an eye towards the future.