OVERVIEW
What it does
Iconic is a DSP technology that simulates physical systems in the digital domain. In its current form, Iconic simulates the sound of a vacuum-tube guitar amplifier by processing a non-linear dynamical model which characterizes the amplifier’s electronic circuit.
The entire Iconic modeler was developed in-house.
How it works
An amplifier’s circuit is divided into smaller circuits called stages. A set of simultaneous equations is used to represent the circuit of an amplifier stage. All the sets of circuit equations representing each stage constitute the mathematical model of the amplifier. Iconic employs an efficient Newton algorithm to solve all sets of simultaneous equations that represent a non-linear dynamical system; in this case, a vacuum-tube guitar amplifier. Here, the dynamical parts of the amplifier are the energy-storage components, i.e. the capacitors and inductors, while the non-linear parts are the vacuum tubes and the magnetization saturation of the transformer core
The circuit equations are derived beforehand using Modified Nodal Analysis and then are organized into a vector. This vector, along with its Jacobian matrix and the circuit’s nodal voltages (also a vector), is passed to the Newton solver, which calculates a new set of nodal voltages. This process occurs at every sample of the input signal from the guitar. One of the nodal voltages represents the amplifier’s input. Another one of the nodal voltages represents the amplifier’s output. The input and output voltage signals are routed to the host’s respective sound device ports
Innovations
While algorithms based on Newton’s method are nothing new (it’s a 300-year old method), Iconic does incorporate the following in-house technologies which increase processing efficiency and which make real-time operation possible:
A new algorithm for directly evaluating multivariate non-linear functions.
A new look-up table algorithm for evaluating multivariate non-linear functions. (Look-up tables are not used in the protoypes.)
An approximation of the long-tail pair phase inverter circuit that results in the reduction of its associated circuit matrix.
Prototypes
Using the Iconic modeler, amplifier models based on the following circuits were created and tested:
Marshall® 2204, a.k.a the “Master Volume”, a.k.a. the JCM 800
Fender® AB763, a.k.a. the Blackface circuit
Soldano® SLO100 with modified interstage coupling capacitances
For each amplifier, the following were modeled:
Triode preamp stages
Phase inverter
Tone stack
Power Amp (two pentodes)
Preamp tube Miller capacitance
Preamp interstage capacitor coupling
Power amp sag
Negative feedback loop between power amp and phase inverter stages (enables the use of Presence control)
Power amp output transformer based on the linear Gyrator-Capacitor model (a non-linear model is available, but it is not used in the prototypes)
The Marshall-based and Soldano-based Iconic models use an approximated phase-inverter circuit, which behaves in a similar manner to a long-tail phase inverter (single-ended input, differential output) but results in a faster run time. The Fender-based Iconic model uses the full long-tail phase inverter circuit. Look-up tables are not used in any prototype.
The prototypes of the above-mentioned amplifier models were embedded in single-threaded, 32-bit VST plugins. The plugins were created using the Audio Plugin Generator (APG) application from dlab®.
The above image shows the Iconic Marshall-based amplifier model VST plugin running in its standalone host. The plugin .DLL files can be used in any VST host or DAW.
Real-Time Performance
The performance and quality of the amplifier models were evaluated by examining the direct output from each amplifier stage and by convolving the power amp output signal with 3rd-party guitar speaker cabinet IRs. The evaluations included signal analysis and playing guitar through the Iconic amplifier models.
The Iconic amp models were run on a laptop with a 4th-generation Intel® i7 processor (4 physical cores, 2 threads/core, 2.4 GHz).
As a Matlab® function in Simulink®, each entire Iconic amplifier model ran in real time at 4x the input sampling frequency of 44.1 kHz with a buffer latency of 0.36 ms (16 samples).
As a single-threaded, 32-bit VST plugin, each entire Iconic amplifier model ran in real time at 2x the input sampling frequency of 44.1 kHz with a buffer latency of 0.36 ms (16 samples).
If implemented for multi-core processing, therefore, the Iconic amplifier models have the potential to run in real time at even higher multiples of the sampling frequency. Running at higher multiples of the sampling frequency reduces aliasing caused by high-gain distortion, thus resulting in higher sound quality. Also, a multi-core implementation of Iconic can enable the use of the non-linear output transformer model and the full phase inverter circuit in the amplifier models currently using its approximation circuit.
LISTEN TO THE RESULTS
MUSIC PRODUCED WITH ICONIC
The above-mentioned Iconic amplifier models were used in all rhythm and lead guitar tracks on three songs from the Agan Rock EP. (See the EP’s Track Credits for song details and listen to each one on the web page). Iconic amplifier models were also used in the rhythm guitar tracks on one song from the Electric Fez: Project A’ EP. (See also the EP’s Track Credits.) In these songs, the guitar parts were processed in real time at 2x the input sampling frequency.
No distortion or overdrive effects of any kind were used in creating the sound of the guitar tracks processed with Iconic. The distortion and overdrive sounds in these tracks were derived solely from the Iconic amplifier models. All tracks were processed in real time.
Applications
The technologies developed for Iconic are not limited to just guitar amplifier modeling. They can be easily adapted to model any linear or non-linear dynamical system – some of them also in real time. Some applications include the following:
In-Car Entertainment (In-Vehicle Infotainment): hi-fi amplifier simulation, sound enhancement and effects
AI: Optimization algorithms for supervised neural networks
Mechanical systems: vehicles, guidance and control systems, machines, engines
Structures: bridges, towers, skyscrapers, castles made of sand
Medicine: neuron clusters, vascular fluid flow, biological processes
Econometrics: currency, stock and commodities markets, trading patterns, trends
The call
We are currently seeking interested parties for collaboration and funding in order to develop, market and sell software and hardware products that will implement the Iconic modeler. Those interested can refer to the Contact page.