Bachus wrote:SA2 voices, use all kinds of moddeling to reshape the samples, this moddeling technollogy is hugely based on the mathematical models used to influence the real time soundsource of the VL engine, yet in the SA2 (did you see a 2 there) they indeed are used to reshape the sample.. they are not the same moddels as the VL engine, but are based uppon them..
You still don't understand the "physical modelling" part.
Articulation Element Modeling (AEM) in SA2 Voices is still based on Advanced Wave Memory (AWM) sample-based synthesis - which is essentially subtractive synthesis model of modular analog synthesizers, where the simple sound oscillator is replaced by digital samples stored in memory and processed with interpolator/pitch shift blocks.
AEM is just a collection of clever tricks that use additional oscillators (multisamples and Element parameters) to simulate real-world effects like legato and glissando.
It's not really "physical modelling" - as in a mathematical model of some real-world object or circuit which generates the sound of the oscillator.
More here:
http://sandsoftwaresound.net/sa-and-sa2-yamahas-words/
To recap.
1. Physical modelling
VL engine uses a computational mathematical model to of a self-oscillating physical body to
generate the sound. The method is called "digital waveguide" - it provides an approximate solution to the travelling wave equation of a vibrating string or horn. Since ideal strings can be considered one-dimensional (1D) solid body objects, the oscillation can be approximated with a superposition of waves that travel back and forth along its single dimension - hence the "waveguide". In practice, the waveguide model is implemented with bidirectional (two cross-faded channels) digital delay line with an attenuator (filter) between the channels - hence "digital" (i.e. sampled into digital domain). The wave impedance of the attenuator is the essential part which defines the resulting sound.
Basically this works similar to vocal reverb with a very strong feedback: 1) you sing to the microphone (on the VL: mathematically generate the sine wave and additional harmonics based on controller input), 2) the reverberator unit or reverberating room creates self-oscillating echoes that reinforce your own voice (on the VL: digital delay line adds feedback to the initial sound, making it self-oscillate) 3) when you stop singing, the echo slowly fades (on the VL: the sound stops (or slowly fades out) as you stop activating the controller).
https://ccrma.stanford.edu/~jos/pasp/pasp.html
https://ccrma.stanford.edu/~jos/pasp/In ... odels.html
https://ccrma.stanford.edu/~jos/pasp/Di ... odels.html
https://ccrma.stanford.edu/~jos/pasp/Vi ... nstruments
https://ccrma.stanford.edu/realsimple/travelingwaves/
2. Sound generator/oscillator
All this happens in the
oscillator (Voltage Controlled Oscillator in classic analog synthesizer) - the part of the synthesis engine that
generates the sound at the pitch (voltage) specified by the physical controller (keyboard).
In analog subtractive synthesis, the oscillator generates simple sine, triangle and square waves. In analog formant synthesis (vocoder), the oscillator is a sum of several sinusoudal waves.
In digital synthesis, the oscillator can be a spectral model (vocoder, FM synthesis, sine+noise+transients), a physical model (virtual analog, waveguides, filter banks, or physics simulation), or sample-based subtractive synthesis engine (with pitch-shift interpolation of digital samples).
3. Subtractive synthesis
In subtractive synthesis the oscillator only generates sound waves at specified pitch. All additional signal processing - i.e. filtering - is performed
outside of the oscillator. This includes sample-based subtractive synthesis.
These processing blocks that make analogue subtractive synthesis:
1) VCA (Voltage Controlled Amplifier) and AEG (Amplitude Envelope Generator) - modify sound volume according to keypress velocity, amplitute envelope (ADSR - Attack Decay Sustain Release) and/or LFO;
2) VCF (Voltage Controlled Filter) and FEG (Filter Envelope Generator) - a low-pass filter controlled by keypress velocity and/or filter envelope generator, filters out harmonic content generated by the oscillator. Can also have realtime physical controls for cutoff frequency, resonance, and envelope.
3) LFO (Low Frequency Oscillator) - provides additional control source(s) for VCA/VCF.
Again, physical modelling has
none of these functional blocks. Instead of shaping/filtering the harmonic content generated by a simple oscillator according to key velocity, LFOs or EGs, it generate the required sound in the oscillator itself - though it's typically a very complex oscillator which takes a lot of parameters, realtime controls, and even dedicated processing units, and it has very high computational complexity which limits the maximum polyphony.
4. Sample-based subtractive synthesis
Digital sample-based synthesis is functionally similar to analog subtractive synthesis, but it has a significant advantage - much larger parameter memory, which allows multitimbral operation and increased polyphony essentially "for free".
The principal processing blocks are the functional analogs of analog subtractive synthesis:
1a) Oscillator/AEG - in sample-based subtractive synthesis, the oscillator needs to interpolate between samples to generate a pitch shift from the sample frequency (specified at sampling time) to the target frequency (pitch) defined by key number. The amplitude envelope can have additional control points, such as Decay1/Hold/Decay2, which can also have both time and amplitude.
2a) Filter/FEG - digital sample-based synths have many types of filters including resonant (self-oscillating) filters, and can also have additional envelope parameters.
3a) LFO - digital synths can have multiple LFOs routed to each other to generate complex modulations.
5. Patch parameter storage
Because every patch (program, Voice) parameter can be in local memory, a 16 note (voice) polyphonic digital sampling synth - which is capable of reading back 16 separate sample memory blocks and having 16 interpolator/pitch shift units - can be at the same time 16-part multi-timbral, i.e. can play back 16 separate programs at once, since it holds the data for all of them; sending these to the tone generator chip is instant.
Recent digital sample-based synthesizers have even more parameter memory - thanks to hundreds of megabytes of local RAM - and enormous amounts of flash-based or RAM-based sample memory (up to several gigabytes). This makes possible 1) lots of velocity layers - not just multisamples, where key banks have separate samples for a range of notes and keypress velocities, but also separate oscillator settings - Elements of each Voice in Yamaha's language, which also include above-mentioned AEG, filter and FEG, LFO, and effect processor routings.
This means that 128 voices (Notes) of polyphony allow quite complex programs (Voices) constructed from several oscillators (Elements) sounding
all at once, and it won't have an adverse effect on polyphony.
6. Articulation Element Modelling
Now to Super Articulation 2 Voices. The specs idicate that Tyros3/4/5 and related Clavinovas are still based on Advanced Wave Memory (AWM) tone generator, i.e. the familiar sample-based subtractive synthesis engine described above, there are samples (stored in flash memory) played back by AWM tone generator chip.
SA2 programs (Voices) are really made from additional samples for playback effects, each with separate oscillator settings (Elements in Yamaha language). There are probably a few additional Elements, or some additional parameters in the existing Elements, which are triggered when the OS detects legato, staccato, bending, breathing, etc. Yamaha calls it "Articulation Element Modelling" (AEM).
AEM is not physical modelling, as in "mathematical model of a real-world physical object, used to generate the sound in the oscillator". The sound source in AWM tone generator still comes from pre-recorded and pre-processed multisamples stored in permanent memory, with velocity and note ranges and dedicated AEG/FEG/Filter/LFO/FX settings that make an Element, where a collection of these multisample/settings (Elements) makes up a program (Voice).
In Yamaha's own words:
http://ca.yamaha.com/en/products/musica ... s/cvp-509/
- Articulation Element Modeling (AEM) Technology
AEM is Yamaha’s proprietary tone generation technology. AEM selects, in real time, the optimal sampled data for the piece being performed, recreating the smooth, natural sounds of acoustical instruments. AEM is used in some of the Super Articulation Voices.
http://usa.yamaha.com/news_events/piano ... f-realism/
- Super Articulation 2, created with Articulation Element Modeling, is capable of reproducing natural legato and staccato expression, vibrato, pitch bend and glistening glissando passages. The feature builds on Yamaha's legendary MegaVoice technology, which added an element of humanistic playing techniques to recorded instrument samples. For the guitar, these include muted string sounds, finger slides, fret noises and harmonics. Wind instrument voices include the ability to add strategically placed breaths that sound like a person is performing, from passages played on clarinets and saxophones to the bent notes of a harmonica.
