There's a nice series on Substractive Synthesis on this very site:
https://www.yamahasynth.com/learn/synth-programming/subtractive-synthesis-101-part-one-the-basics

TLDR: any synth that has a Filter in its sound path is a Subtractive synth. Samplers, Wavetable synths, Virtual analog, Analog are all subtractive synths.

The basic sound path of a subtractive synthesis being:
Wave -> Filter -> Amplifier

The only difference is in the "Wave" and the features it has.

In the AWM2 engine, the Wave part are Samples (recorded sounds). There are various features in samplers, like triggering different samples based on various criteria, changing the phase, starting at various places in the samples etc.

In a Wavetable synth, the "Wave" part is playing Wavetables, again with various features available.

In an Analog or Virtual Analog synth, the "Wave" part is an actual Oscillator that creates the wave on the fly. This means that various characteristics of the Wave can be modulated in real time, like Pulse Width, the shape of the wave, phase etc.

The AWM2 engine has a lot of samples that are recordings of classic analog waveforms like Sine and Square and Saw, but they're just that: recordings. You can emulate quite well a subset of what a virtual analog synth does with those, but you can't have that animation given by a real oscillator.

Nice reply, dude. 🙂

The "modeling" part of "analog modeling" implies a mathematical model for a synthesizer, typically composed of wave generators (e.g., oscillator), filter, envelope generator, VCA, LFOs, effects, etc. Some underlying computational engine executes code based on the model, producing a digital audio stream typically passed into digital-to-analog conversion.

In the effects world, a digital reverb is, effectively, a mathematical model of old-school echo chambers, springs, and so forth.

Yamaha very likely have a tight functional specification for the AWM2 pipeline, describing specific filter characteristics, wave generation, and so on. This is the starting point for Yamaha's hardware engineers. They probably have a VHDL or Verilog description that drives hardware synthesis (the equivalent of a compiler except it compiles hardware circuits). They don't crank out an AWM2 implementation willy-nilly. 🙂

Analog modeling synthesis could be implemented in software (firmware) targeted for a general purpose digital signal processor (DSP) or other computational engine. This gets people's hearts racing because the mathematical model could describe one of their favorite blasts from the past -- Minimoog, 2600, CS-80, whatever -- or something totally new. The mathematics is not limited to the well-known, well-worn AWM2 model.

That's my take, anyway -- pj

Music tech blog: sandsoftwaresound.net

Being a newbie in the synth department, is there something the analog modelling can do that AWM2 can't?

Actually each of them can do things that the other can't.

Generally speaking the differences between them boil down to:

1. what you start with
2. what elements you can control and the degree to which you can control them
3. what elements you can add or remove

The above replies explained the basics.

Although both AWM2 and analog start with 'waves' the content of those waves is typically vastly different. A recorded wave is a sample of real-world sound as received by one or more microphones.

Recorded sounds don't travel in a straight line from the source to the microphone. They bound off walls, chairs, ceilings, lamps, smoke molecules, etc. And they bounce and reflect multiple times before reaching the same microphones in many different phases.

You don't have ANY control over those incidental changes when you start with a recording.

On the other hand you can't realistically ADD all of those incidental things in an Analog model.

Even the waves you start with for Analog might have elements you can't control such as the number, phase and strength of each individual harmonic in a provided 'Sawtooth' wave.

The choice between methods often comes down to WHAT type of sound you are trying to produce and the quality of the starting wave samples that you have available.

If you are trying to produce a real-world sound like a piano it makes sense to start with a piano rather than build from scratch. You don't HAVE to do that but you will get closer to your goal quicker.

That is because there are high quality piano samples available already.

If you are trying to produce the sound of a muffled explosion it will be hard to find an appropriate high quality recording that you can start with.

Summarizing what I first said above:

1. what bulding blocks (waves) do you have available to use
2. what tools do you have thata work with those building blocks
3. how much time & effort do you want to expend

There is no 'one size fits all' solution

Here's a collection of my thoughts regards MODX/MODX+ capability as an emulation of a Classic Analog Subtractive Synth, in respect of the call for a "VA" type Synth Engine.

It's long. Read or don't. It's an answer to all those posters in other forums who criticise and belittle the MODX for NOT having a "VA Engine".

1) Popular Opinion
A common detraction in Synth forums is that the MODX is "weak" because it doesn't have a Virtual Analogue (VA) Synth Engine.

I often read claims that the MODX shines for "Real Instrument" sounds, but is poor for classic "Analog Synth" sounds. This is usually accompanied by claims that "other" comparably priced Synthesisers are "better". These comments often read as disingenuous "praise" for the MODX, and it's the frequent disingenuity that actually makes my skin crawl.

"He tried really hard, and I was very impressed by the amount of effort he put in, but unfortunately he never quite made the grade. If we need someone to call in for support when work loads are high we will certainly consider getting in touch".

That is Disingenuity.

The proof of the pudding, is in the eating. If the MODX is putting out sounds that you cannot distinguish from a real Minimoog, Prophet 5, OBX, Jupiter 8, Juno 60 etc... then how you arrived at those sounds is immaterial.

The raison d'etre of any Synthesiser is to "Synthesise Sound". The MODX is exceptionally competent from every conceivable angle in this respect. Therefore at the very least, it is by definition, a Sound Synthesiser. Credit should be given where it is due.

2) Control
The hidden agenda behind these kinds of comments is not lack of capability on the MODX's part, but lack of capability on the User's part. It's a fairly typical human reaction to reject anything that is not familiar. Therefore, they make the assumption that the unfamiliar "AWM2 Engine" is less competent than a familiar "VA Engine" as found on Nords, Fantoms et al.

This assumption is fallacious.

The only advantage a VA Engine presents over the AWM2 Engine is familiarity. A VA Engine provides a well established, 60 year old recipe.

Therefore Users do not need to encumber themselves with the reading of manuals, or the learning of anything new. If they have prior experience with classic Subtractive Synths, or "VA's", they assume they can get started quickly.

3) Ease of Use
There is a presumption that "Sound Design" with a Classic Subtractive synth is "Easy". I would disagree.

It is easy to randomly twist a few Knobs and be happy with the result...
if the end result is a generic "Synthy sounding Sound". To some people, this seems to be the defining characteristic of, and all that, a "Synth" needs to be.

My experience tells me that Sound Design, even on a "familiar" Control layout, is actually quite difficult and complex. Experienced synthists spend a lot more time designing and scrutinising their design, than they are given credit for. That's why their designs sound great, and how they make it look simple.

For a start, it requires that you have a deep working knowledge of Subtractive Synthesis, both in operation, but also in fundamental sound design and sound synthesis.

To illustrate, you have a limited selection of Oscillators and Oscillator Waveforms. Which of those waveforms should you use to create (Synthesise) a sound? What should be their relative Pitch and Level? What Filter Characteristics need to be applied? What time-based shaping needs to applied (Modulation, Envelopes, LFO's)?

These are the same considerations that anyone needs to work through, regardless of Synth Engine/Type.

These are considerations you must process with a "VA". Unless you want "quick, easy, random", and can live with the result.

Its no different with "AWM2". Except you are not limited to 1, 2 or 3 Oscillators, or 4 or 5 waveforms.

Based on that alone, AWM2 is more powerful, and more capable, as a "Sound Synthesiser" than a "VA" could ever hope to be.

4) Knob Per Function
Carrying on with the previous logic, a VA Synth Engine, would really only be at an advantage if it carried the same amount of Physical Controls you would find on a Classic Analogue Subtractive Synth.

If those physical controls are not immediately presented, and instead become "Menu Parameters" then any perceived advantage is lost. The "ease of use" argument disappears. Now, on a menu-driven level playing field the MODX AWM2 Engine is at a clear advantage. You can do more. A lot more.

5) But AWM2 cannot do...

a) Oscillator Sync
b) Pulse Width Modulation (PWM)

No it can't. But it can emulate them, in the small number of cases where those Subtractive Synth functions were part of the sound. Just to be clear, not all Subtractive Synths provided Oscillator Sync or PWM... some did, some didn't. For those that didn't... it didn't prevent them from becoming "classics".

Also, let's not forget, the MODX has a fully fledged FM-X Engine, and Timbres approaching those "Sync" and "PWM" sounds can be achieved in FM... albeit with more complex design (a time burden).

For those not familiar with AWM2, the MODX has multiple "Sync" and "PWM" Samples which are used to recreate those sounds. The end result is fine. More samples and variety are available in 3rd Party Libraries.

It's a case of throwing away the baby with the bathwater. Pointing out a paint scratch on an otherwise beautiful looking sports car.

"But, but, but, it can't do...." . Nearly every critical post in every forum. Don't be misled. Consider all the things it CAN do, over and above a VA.

Analog Modelling sounds GOOD!

Often, they move through shapes really well, between forms, because everything is modelled. And, quite often, just about everything can be modulated. Some of the newer ones (last dozen or so years) offer some kind of morphing of shapes for some or all four key ingredients: oscillators, low frequency oscillators and envelopes - and the ability to use each of these to modify and/or modulate each of the others, and to modulate the form of the filter, too.

And most are massively easier to create and edit sounds within than the byzantine menu diving of the MODX.

Probably best of all, in terms of quick and easy features, is unison spreads (in most analogue modelled synths) being quick to create, configure and then modulate.

[quotePost id=120571]

Analog Modelling sounds GOOD!

[/quotePost]

Therefore you are implying that AWM2 sounds bad. It doesn't. Far from it. The end results (audio) are excellent.

You also tabled another common misleading argument... that VA (Virtual Analog Subtractive Synth Model) is simple, quick and easy to create/edit.

This would be the case if the Synth itself was a "VA", for example Modal Cobalt8, Roland GAIA, Access Virus TI, Waldorf Kyra etc.
That is where the Synth's control surface is laid out "knob per function" like a "real" Analogue Subtractive Synth.

But the case in point here is a "VA Engine". The Algorithm inside a multi-timbral keyboard that serves to produce "Subtractive Synth" sounds. All controls, with maybe a couple of exceptions, are Menu Based. Therefore, there is no advantage over the current menu-based version of AWM2.

Flipping the coin, if the AWM2 Engine was placed into dedicated hardware, providing Knob per Function control, the "VA is easier to use" argument disappears.

The only argument left in this vein is that VA is more common and familiar. Accepted. But this does not make anything that "isn't VA" bad. That's just opinion born of plain old bigotry.

Everything "sounds good".
And things said about VA like how much motion and modulation can be embedded in it, are doubly true for FM.

My problem is not with VA, which is great (just like any other synthesis form, btw).
My problem is
1. In general, with the downright deranged obsession the industry has with this particular synthesis method, which has been beaten to death already while not being in any way more interstingly sounding than others (quite the opposite); and
2. In particular, with targeting the one workhorse synth line that actually offers a particular and original synthesis method in an inflated sea of VA synths, and pretending said synth is somehow worse because it's not just another VA synth. In fact, it's much, much better because of that.
So fuck VA, it's already way too much of it on the market.
And unfortunately this whole campaign means that Yamaha is bowing to the mob and will grow this already saturated slice, instead of concentrating on FM or other more original synthesis approaches.

Analogue Modelling also sounds LUSH!

In the list of things it can do that AWM2 can't, or doesn't immediately and innately provide:

Analogue modelling provides oscillation dynamics not possible from samples, and often not readily, easily or actually possible with FM, either.

There's a beautiful, welcoming, familiarity to the jumping off points of analogue modelling, inherent to the world's experiences with sounds and songs, before it can then go in just about any direction, including the more modern and extreme.

Analogue modelling can create FM, PM and AM, and be extended to both build and use wavetables, and to sweep through wave shapes, whilst animating and phasing the pulse widths, for all the different locations waves are used.

It's so a perfect blend of sanguine and melancholy, familiar and modern, in most all its flavourings, that it'll never age out.

And it's particularly responsive to warming effects and filtering, and filtering and warming effects.

But, perhaps most importantly, most implementations permit easily ensuring any oscillation starts on max, lowest or zero values as desired, which is especially important control for the initiations of lower frequency sounds like bass and kicks.