See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/305649333

Computer simulation of audio circuits with vacuum tubes

Conference Paper · May 2016

DOI: 10.1109/SCM.2016.7519700

CITATIONS READS

6 1,539

3 authors:

Timur I. Karimov Denis Butusov

Petersburg State Electrotechnical University Saint Petersburg Electrotechnical University "LETI"
63 PUBLICATIONS   303 CITATIONS    141 PUBLICATIONS   1,084 CITATIONS   

SEE PROFILE SEE PROFILE

Artur I. Karimov
Petersburg State Electrotechnical University
73 PUBLICATIONS   391 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Chaos based sensors design View project

Discrete & Continuous Chaotic Systems Synchronization Problem View project

All content following this page was uploaded by Denis Butusov on 14 October 2017.

The user has requested enhancement of the downloaded file.

 Computer Simulation of Audio Circuits With
Vacuum Tubes
T.I. Karimov, D.N. Butusov, A.I. Karimov
Computer-Aided Design Department
Saint-Petersburg State Electrotechnical University "LETI"
Saint-Petersburg, Russia
e-mail: [email protected]

Abstract—Vacuum tubes are still widespread in modern audio Almost complete overview of guitar amplifier simulation
circuits. The replacement of vacuum tubes with their digital techniques is given in [2]. According to the paper, all
counterparts is an important and nontrivial task. In this paper a simulation methods could be divided on two main branches:
brief overview of methods for analog audio circuits digital black-box approach (it can be called ‘imitation’) and white-
simulation is given. Some techniques to improve the accuracy up box approach (‘modeling’).
to 1% of analog counterpart are proposed, as well as methods to
reduce simulation time up to 5 times comparing to standard Determining idea of the methods from the first branch is to
simulation. find an explicit mathematical formulation of distortion effect
(‘waveshaping’). The main advantage of waveshaping is its
Keywords—digital audio effects; circuit simulation; triode explicitness resulting in comparatively fast computations. But
identification; hardware-targeted method simplified mathematical expressions have no physical
background and hardly lead to realistic sound. Moreover,
I. INTRODUCTION resulting system is memoryless, so sophisticated enhancement
is required to make dynamic response more natural.
Development of computer technologies makes it possible
to replace analog devices with their digital counterparts or Computer simulation is the second branch of emulation
computer programs. This is especially important in a case of techniques, and it does not suffer from most of disadvantages
special hardware that produces different musical effects. Thus noted above. But high accuracy of computations demands high
a musician gets all advantages of digital electronic against computational costs, because underlying circuit equations are
analog: great flexibility in tuning, portability, small size and nonlinear and stiff.
little power consumption.
One computational-effective approach is called DK-
The only weak point is a quality of simulation that often is method (‘D’ stands for ‘discrete’, ‘K’ for ‘Kirchhoff’), it is
supposed to be as unsatisfactory. Some nonlinear effects make based on the idea to solve differential and algebraic parts of
real challenge to engineers because they are produced by circuit equations separately [3]. DK-method implies that all
electronic components with complex multifactor behavior. energy-storing components like inductors and capacitors are
Examples of such analog devices are distorting guitar first discretized with certain numerical method. But it is
amplifiers or Moog ladder filter. possible for only few integration methods.
In this paper practical aspects of guitar preamplifier The last simulation approach is the direct solution of
simulation are considered. In section II we make brief differential-algebraic equations (DAE). That makes possible to
overview of simulation techniques and point their straights implement any method solving DAEs. This approach is time
and weaknesses. In section III we present the case study — consumptive but the most precise as well.
tube preamp simulation. Section IV contains the comparison
of simulation results and analog prototype circuit III. CASE STUDY: TUBE PREAMPLIFIER SIMULATION
measurement; some corrections are applied to the model.
Finally, in Section V, conclusion and discussion are given. For computer simulation we replace the triode with
equivalent circuit of two voltage-controlled current sources
connected as shown at figure 1. Capacitor Cgp represents
II. DIGITAL EMULATION TECHNIQUES FOR DISTORTION EFFECT parasitic capacitance between grid and plate of triode. Triode’s
Electric guitars are mostly used with sound modulators current-voltage curve can be approximated with power
such as overdriving and distorting devices. Constructively the function, but several more complex models were developed
last can be implemented as preamplifiers of dedicated guitar for computer circuit simulation.
amplifiers or effect pedals. Tube amplifiers are considered to
giving the most ‘pleasant’ sound and it is almost unreachable
with analog semiconductor devices [1]. For above-stated
reasons digital emulation of vacuum tube amplifiers is one of
the most challenging and promising tasks in the field of digital
audio systems engineering.

978-1-4673-8919-8/16/$31.00 ©2016 IEEE

 There are few approaches to improve performance. For
example, the whole circuit can be divided into blocks (so
called block-wise decomposition) with 2 triodes in each block.
That makes it possible to reduce significantly simulation time
though some properties of the system can be lost.
Another feature is that cathode follower can be eliminated
in model because it is required just for impedance conversion
and has constant response to audible frequencies. In our case,
circuit of Orange DH-15 preamp with 32 nodes could be
a) b) decomposed into 3 blocks with 12, 13 and 17 nodes
Fig. 1. Common cathode amplification stage and triode’s model respectively. Thus simulation runs 4-5 times faster. This time
can be reduced even more with assumption that first triode
The most developed of them are Koren’s model [4] and stage works in linear mode.
Dempwolf’s model [5]. The last model’s formulae have
discontinuous derivatives and thus can be considered as the Indeed, guitar signal has rather small amplitude, typically
most efficient computationally. up to 0.5 V when all strings are stroke, while valuable
nonlinear distortion starts at 1 V. Figure 3 demonstrates full
6
x 10 and decomposed models spectra comparison. General shape of
6 the spectrum is strongly preserved, even when 1-st stage is
Serial evaluations of f(t,x) per sec

Lobatto IIIC replaced with linear filter (2-nd order low pass IIR filter was
5 Euler used with fitted coefficients).
Euler(2) serial 50
4
Euler(2) parallel
45
3 Radau IA
40
2
35
Amplitude Spectrum, dB

1 30

0 25
1 0 -1
10 10 10 20
RMS error, V
15
Decomposition
Fig. 2. Comparison of numerical methods for triode stage simulation problem. 10 Decomposition, 1st stage as filter
Euler-2 is an extrapolation implicit Euler method of order 2
Full schematic
5
Input Signal
In our study the different numerical methods were
0
compared for a single stage: implicit Euler’s method in 10
2
10
3
10
4

traditional and extrapolated versions, implicit midpoint rule, Frequency, Hz

trapezoidal rule, Lobatto IIIC method, Radau IA method and
others. It was discovered that method should be L-stable to Fig. 3. Different model implementations spectra
show good performance on the triode stage simulation
problem. Figure 2 represents the simulation results, giving the
IV. MODEL VERIFICATION AND IMPROVEMENT
comparison of studied methods for triode simulation problem.
Four points for each method represents performance for 1, 2, Computational experiments demonstrated dramatical
4, 8 times oversampling (from left to right) assuming base influence of triode current-voltage curve approximation to
frequency of 44.1 kHz. simulation results. At the same time tubes of similar
designation (e.g. 12AX7) have variety of parameters up to
Performance tests on real hardware systems showed that 20% [5]. We found that precise simulation requires individual
parallel extrapolation methods are mostly suitable for fitting of particular triode parameters. It can be performed by
platforms where parallelization do not causes many additional measuring triode curves. The promising approach is parameter
costs (e.g. FPGA). For modern multicore CPUs, however, fitting according to time response
solver parallelism gives no benefits. It can be considered that
most stable methods should be preferred, e.g. implicit Euler. Using this technique, difference between simulation results
and measured signal is minimized by triode model parameters
As the practical circuit, preamp of Orange DH-15 “Dark tuning.
Terror” guitar amplifier was taken and its model was
simulated. This preamp consists of 4 common cathode triode Figure 4 illustrates the last attitude. For this example
stages, tone stack and cathode follower. Simplification of such voltages at nodes #6 and #7 were taken. Simulation with
models is very important, especially for real-time applications. parameters obtained from precise measurement of grid current
still gives inaccurate solution (fig. 4a).
 д
20 20

10 10
Node voltage, V

Node voltage, V
0 0

-10 -10
#6 - measurement #6 - measurement
-20 #7 - measurement #7 - measurement
-20
#6 - model #6 - model
#7 - model #7 - model
-30 -30
0 0.005 0.01 0.015 0.02 0.025 0 0.005 0.01 0.015 0.02 0.025
Time, sec Time, sec

a) b)
Fig. 4. Comparison between two parameter sets: initial, obtained from static voltage-current triode curve approximation, and optimized, based on transient analysis
data fitting

Gradient descent optimization method leads to another

solution that much better fits measurement data (fig. 4b). More ACKNOWLEDGEMENT
precise fitting should take into account variability of This paper was supported by the Fund for Assistance to
amplification coefficient µ of tube and other nonlinear Small Innovative Enterprises of Russian Federation under
phenomena. project 5160GU1/2014 from 19.02.2015.
Nevertheless, simulation of the full guitar sound-
modulating chain (preamp plus speaker) shows good REFERENCES
consistency with measurement. [1] E. Barbour, The Cool Sound of Tubes, IEEE Spectrum 35(8), 1998.
[2] J. Pakarinen and D. T. Yeh, A review of digital techniques for modeling
V. CONCLUSION AND DISCUSSION vacuum-tube guitar amplifiers, Computer Music Journal, vol. 33, no. 2,
pp. 85–100, 2009.
In this paper brief overview of analog audio circuits [3] G. Borin, G. De Poli, and D. Rocchesso, Elimination of delay-free loops
simulation techniques is given. The case study of vacuum tube in discrete-time models of nonlinear acoustic systems, IEEE
preamplifier was investigated. It was shown that the most Transactions on Speech and Audio Processing, vol. 8, no. 5, pp. 597–
precise simulation method is the solution of DAEs of the 604, 2000.
system. To reduce complexity block-wise decomposition was [4] I. Cohen and T. Helie, Simulation of a guitar ampli er stage for several
triode models: examination of some relevant phenomena and choice of
used. Triode curve approximation is the most important part of adapted numerical schemes, 127th Convention of Audio Engineering
computer model, and precise simulation requires individual Society, pp. 1–9, 2009.
fitting of all tubes. We used modified Dempwolf’s model with [5] K. Dempwolf and U. Zölzer, A Physically-Motivated Triode Model for
static amplification coefficient. Further improvement of model Circuit Simulations, 14-th Int. Conference on Digital Audio Effects
requires changing of dynamic as well as defining formulae for (DAFx-11), pp. 257–264, 2011.
this changes.

View publication stats