AFMG - cvut.cz

AFMG AHNERT FEISTEL MEDIA GROUP

Brno 19.10.2011

Acoustic Simulation and Sound System Design

Wolfgang Ahnert

W. Ahnert www.AFMG.eu 2

Overview

• Short review of development of acoustic simulation

• Software EASE

• Speaker, wall and other data

• Extension to lower frequencies

• Intelligibility calculation

• Measurement tools

www.AFMG.eu

Development in Designing

Paths to sure planing: • Roman/ Greek time/ middle age: knowledge based on experiance and

first trial and error reports, i.e. rom. architect Vitruv • since 18. century: investigations, i.e. Chladni or at 1875 Lord Rayleigh,

Prof. Helmholtz • since 1900: roomacoustic basics, Prof. Sabine • by 1935: measurement in models and „Auralisation“ in physical

models, Prof. Spandöck München, Prof. Reichardt Dresden • since 1965: computer model investigations, Prof. Krokstad Trondheim,

afterwards a lot of similar works • since 1995: Auralisation by means of computer models is available

generally

1988, CATT-Acoustic, by Dalenbeck/Sweden

Version 1, now version 8.6

1991, ODEON, by Naylor&Rindel/Denmark

Version 1, now version 10.0

1994, RAMSETE, Faria/Italy,

Version 1, now version 2.xx

1998/2001, CAESAR, 1998 by Vorländer/Schmitz/Aachen/Germany

Version 0.12, 2001 vers. 0.20

2001, EASE, by Ahnert/Feistel/Berlin/Germany

Version 4.0, Autumn 2009 EASE 4.3

Room Acoustics Programs

EASE 4.3. ….5.0

Modules for Editing and Calculation

SpeakerLab

MaterialLab MicrophoneLab

InterfaceLab

Block diagram

SpeakerLab

GLL for different speaker types Point sources Arrays Cluster

GLL for natural sources Human sources Music instruments

Import and export routines

1000 Hz

Point Source presentation

Magnitude and Phase balloon

wrapped

unwrapped

5000 Hz

Point Source presentation

Magnitude and Phase balloon unwrapped

wrapped

Loudspeaker data

Import routine for Wav-files

Identification of the rotation point in data sheets:

Standard AES56-2008

Loudspeaker data base - Cluster

A Cluster calculation routine is a far field approach

Using the measured or simulated balloon data of a single device you calculate the cluster by means of:

power summation

complex integration

all with different angle and frequency band resolution

View cluster shows the components of the cluster

Different cluster calculation

Already in EASE 3.0, Only run time phase consideration

Arrays with In-line Arrangement of Radiators

where n = number of individual loudspeakers d = spacing of the individual loudspeakers = radiation angle = wave length of sound l = (n-1) d = length of the loudspeaker line

Each of the individual loudspeakers radiates the sound spherically and the sound waves get favorably superposed in the far field, whereas the effect of the individual loudspeaker prevails in the near field. For the far field the equation above was given already by STENZEL 1927, 1939 and OLSON 1947 for the angular directivity ratio , the so-called polars.

STENZEL, H.: Über die Richtwirkung von Schallstrahlern, Elektrische Nachrichten-Technik, Band 4 (1927), Seite 239 STENZEL, H.: Leitfaden zur Berechnung von Schallvorgängen, Verlag von Julius Springer, Berlin 1939 OLSON, H.F.: Elements of Acoustical Engineering, D. van Nostrand Company, 2nd edition, New York 1947

Classical columns (Loudspeaker Line, Sound Column)

sinsin

Speaker type Cluster

already in EASE 3.0

Cluster calculations

1 speaker

2 speaker

3 speaker

4 speaker

5 speaker

6 speaker

Arrays with In-line Arrangement of Radiators Many manufactures like L-Acoustics, JBL, Electro-Voice, Nexo, Adamson or Meyer Sound produce line arrays, here as an example Vertec/JBL:

In EASE4.0 it was no longer possible to represent these systems by means of simple point sources with directional irradiation. Thus an algorithm has to be employed that is capable of calculating the attainable sound level according to the array structure as well as the distances and frequencies involved. To this effect there are special product-specific DLLs available that are called up by the respective main program.

Speaker type Line Array

Renkus-Heinz AimWare

Speaker type Line Array

EASE Focus

Software

Iconyx 8

Electronically steerable column loudspeaker by Renkus-Heinz

DURAN Audio Intellivox 2c

Arrays with In-line Arrangement of Radiators Digitally controlled line arrays

A way of reducing the frequency dependence of the directional characteristics and directivity of sound lines (the figures below illustrate such an unwanted directional effect in three-dimensional representation) consists of supplying the sound signal, with different phases and levels, to the individual loudspeakers in an array.

1000Hz 2000Hz 4000Hz

Line Array Simulation

Line Array Simulation: Direct SPL on faces

New Data Format

Until now: Point Sources, Cluster Dynamic Loudspeaker Libraries DLL and now: Generic Loudspeaker Library GLL

What„s that?

Generic Loudspeaker Library GLL

Why GLL – Another loudspeaker data format? History:

EASE 1.x : 15° and 1/1 Octave half sphere data, magnitude only EASE 2.x : 10° and 1/1 Octave half sphere data, magnitude only EASE 3.x : 5° and 1/3 Octave full sphere data, magnitude only EASE 4.0/4.1: 5° and 1/3 Octave full sphere data,

complex data + DLL modeling capabilities

Problem: Fixed data tables, no user configurability Many constraints due to data reduction or interpolation to fit

tabular format Data resolution not adapted to modeling goals

GLL – Motivation

Typical Problems with Tabular Data: Active or switchable passive multi-way loudspeakers Stack of 2 two-way loudspeakers Column loudspeakers, tapered or digitally steered Touring line arrays Cluster systems

Modeling Requirements: Data resolution > 1/3rd Octave Add coherence information, phase Mechanical configurability (rigging) Electronical configurability (filters) Better integration with manufacturer

control software

Create a GLL What is needed to create a GLL?

Data for individually controlled transducers must be measured individually

For each point source of the model: IR/FR balloon data in sufficient angular resolution (EASERA, TEF, MLSSA, MF, CLIO, LMS, etc.)

Sensitivity is calculated automatically from calibrated on-axis response

Maximum voltage over the rated bandwidth

Impedance data (optional)

=> Stored as a GSS file

Create a GLL Generic Sound Source (GSS) Data:

Data is stored in its native format using high-resolution storage and compression algorithms Rather than enforcing data provision in a fixed format, data points required for prediction are interpolated from available data points

– Preferably complex data, error ranges as determined by rotation point and critical frequency

– New definition of power handling

– Flexible import functions

Create a GLL

What is needed to create a GLL?

Combine sources in a box:

3D location of all transducers / source groups

Define input matrix (external inputs -> acoustic outputs)

Available filters (matrix nodes)

Case drawing

GLL Box Type

Sources

Filter Groups

Input Configurations

LF LF XO

Active System

LF LF XO

Passive System

Design and Analysis Tool Design Tool EASE SpeakerLab:

Driver placement and orientation

Crossover design

Filter development for steered columns

Validation of mechanical design

=> Directivity Prediction

Design and Analysis Tool Analysis Tool EASE SpeakerLab :

For measured data and predicted data

Validation and evaluation over angle and frequency

Variety of graphs: Frequency Response, Balloons, Polars , Beamwidth, Hor./Ver. Mapping, Directivity, etc.

MicrophoneLab

• GLL for different microphone types • Omnidirectional microphones • Directed Microphones • Arrays

• GLL for natural receivers • Human ears • Dummy head • HRTF database

• Import and export routines

38 www.AFMG.eu

Microphone database

First approach to collect microphone data for EASE

MaterialLab

• Interface to SoundFlow • Angle-dependent absorption and scattering values • Interface to Reflex • Data Storage in compiled form • Import of text and picture files • Contact information to web sites • Providing import routines for manufactures to

create material data • Import routine for *.mat files

40 www.AFMG.eu

Calculation of the Acoustic Properties of Multi-Layer Walls

Calculation: • Impedance describes the acoustic

properties:

• Z = p0/v0

• Calculation of transmission:

• t = |pN/p0|²

• Boundary conditions for back size:

• Rigid wall: Sound velocity is 0

• Air: Characteristic impedance of air

• Multi-layer structures are calculated layer by layer: (pn,vn)->(pn-1,vn-1)

• Different computational models can be applied to each layer

pn-1 pn

vn-1 vn

vN = 1/ZAir pN = 1

pN = 1 vN = 0

Rigid Wall

41 www.AFMG.eu

Transmission Loss: Simulation vs. Measurement

100 1000 10000

Frequency / Hz

Transmission Loss [dB]Porous Material placed between two plates

Journal of Sound and Vibration (1992) 155(1), 125-132

SoundFlow

LAURICKS et. al.

42 www.AFMG.eu

Absorption Coefficient: Simulation vs. Measurement

100 1000 10000

Frequency / Hz

Absorption Coefficient Microperforated Panel

J. Acoust. Soc. Am. 104 (5), Nov. 1998

SoundFlow

Dah-You Maa

43 www.AFMG.eu

Software AFMG SoundFlow

Features:

• Entry of layers and thicknesses

• Definition of „raw materials“

• Calculation of transmission, absorption, reflection

• Result output as report document or pictures

• Direct export to EASE

44 www.AFMG.eu

Future Improvements in SoundFlow

• Improve the model for thick plates

• Add different connection types between layers

• Rigid connection (sandwich plates, laminated)

• Rigid line/point connection (studs, structural bridges)

• Viscous

• Conical holes

• Eigenfrequencies on finite sized walls

Boundary Element Method

• Solves differential equations on the boundary of a domain

• Useful for solving scattering problems • Can be coupled to FEM to solve semi-open rooms • Advantages:

• Fast & exact for “small” problems • Can be used for open domains

• Disadvantages: • Scales in contract to FEM • Needs fundamental solution

Calculation of Scattering by a Structured Surface using BEM

Steps: • Structured surface • Tessellation into segments • Incident plane wave • Calculation of sound

pressure at surface • Solving the system of

linear equations • Resulting sound pressure • Calculation of reflected

wave front

pin[0] pin[1] pin[n] …

pout[0] pout[1] pout[n] …

BEM M pout = pin

47 www.AFMG.eu

Quantify scattering

Scattering Coefficient Autocorrelation diffusion coefficient

Normalized diffusion coefficient Correlation Scattering Coefficient Correlation with response of flat panel

S = 0.7

S = 0.15

D = 0.85

D = 0.35

Calculation of Scattering Coefficient

Calculation:

• Comparison with flat panel

• Calculation of spatial distribution using BEM

• Comparison of distributions

• Calculation of correlation function

• -> scattering coefficient

2kHz 2kHz BEM BEM

49 www.AFMG.eu

Calculation of Diffusion Coefficient Calculation:

• BEM – Calculation of spatial distribution

• Autocorrelation

• -> diffusion coefficient

• Comparison with flat panel -> normalized diffusion coefficient

• Also smooth surfaces may yield high diffusion coefficients at low frequencies

• Normalization to suppress this edge effect

• Scattering coefficient compares with geometrical reflection only

2kHz BEM 2kHz BEM

Autocorrelation Autocorrelation

100 Hz

800 Hz

100 Hz

800 Hz

100 Hz

800 Hz

100 Hz

800 Hz

100 Hz

800 Hz

50 www.AFMG.eu

Scattering Coefficients: Simulation vs. Measurement

Comparison:

• Measurement according to ISO 17497-1

• Simulation of „Correlation Scattering Coefficient“ according to Mommertz

• Match is quite good, esp. for use in EASE

100 1000 10000

Frequency / Hz

Measurement

AFMG Reflex

51 www.AFMG.eu

Measurement in cooperation with RPG Diffusors Inc.

Scattering Coefficients: Simulation vs. Measurement

Comparison:

• Measurement according to ISO 17497-1

• Simulation of „Correlation Scattering Coefficient“ according to Mommertz

• Match is quite good, esp. for use in EASE

100 1000 10000

Frequency / Hz

Measurement

AFMG Reflex

52 www.AFMG.eu

Measurement in cooperation with RPG Diffusors Inc.

Software AFMG Reflex

Features:

• Entry of geometry as a cross section

• Calculation of spatial response

• Calculation of coefficients

• Result output as report document or pictures

• Direct export to EASE

53 www.AFMG.eu

Reflection at structured surface

Outlook for AFMG Reflex

• In development • Better handling of small element • Higher order base functions

• Possible extensions • Import/Export of model data • Spline elements • Adding non-rigid BCs • Extending to 2 dimensions

InterfaceLab

Import/Export from and to AutoCad Import/Export from and to SketchUp Import/Export from and to BIM Simple routines to create Models

3D Modules Prototypes Import from other CAD programs

Laser scanning Holographic approach

Interface in EASE 4.3

Support for DXF2000 and Sketchup formats

Import and Export

Up to Google SKP 7.0

Modeling Techniques

Computational

Modelling

Wave-Based

Modelling

Ray-Based

Modelling

Statistical

Modelling

Difference

Methods

Element

Methods

Tracing

Mirror

Image SEA

FEM BEM

Room acoustical computer simulation

Simulation algorithm

CAD model + material data

Objective single number quantities

Realtime convolution

Impulse response calculations

Ray Tracing with counting balloon

Image Model method

Image sources

wall 2

wall 1

ii nnnnnnnnn SSSSS ...... 21121211...

• Geometrical construction

• Audibility test

• Very expensive for high reflection orders

Energy

source

Ray tracing

ray source

wall reflection

detection

r/ c > tmax ?

last ray ?

detector

source

distance law: ray density

Hybrid models

• „Forward audibility test of image sources“

• Rays (cones, ...) hitting a receiver can be addressed to audible image sources

• Dialects: Tracing of cones, triangular beams, pyramids, ..

• Higher Order possible but some mirror images are missed

• Parameter : spatial resolution calculation time

detector

(image) source

distance law: analytically t = rIS/c, E ~ 1/rIS2

Echogram

• Recording of counts and energy in time histograms

tray = rray/c

Fast calculation routines

• Fast Calculation of Impulse Responses

• Sophisticated Hybrid Algorithms

• Use of scattering coefficients

• Use of diffraction effects

• Fast intern calculation routines

• Multi-Thread approaches

• Network calculation

Intersection Algorithms Overview:

- In raytracing for a given ray vector often the next

intersecting surface (triangle) must be found

- Brute force approach, e.g. linear search through all triangles for the nearest intersection point, is very time-consuming

Intersection Algorithms

Previous Algorithm in EASE: - Log-based Hierarchival bounding volumes (HBV)

method

- From Outdoor 3D rendering engine

- Hierarchy of multiple Spheres and Boxes

- a particular child-volume is only tested if the parent was hit

- the resulting computation scaling with the number of triangles is approximately logarithmic

Intersection Algorithms

New Algorithm in v4.2:

- Log-based space partitioning

- Indoor-optimized

- Uniform Single-Level Grid (Each triangle associated with a grid cell is then

tested for intersection with the ray)

- Enables Vector Processing

(Given a ray specified by an origin and direction

vector, a fast grid traversal algorithm computes

the next grid cell intersected by the ray)

Intersection Algorithms Result: Algorithm optimized for acoustic models can be up to 5 times faster (standard

Pentium 4)

Multi-thread Mode in EASE 4.3

Multi-thread Mode

Node Server

Master Server

Simulations

SoftwareJob DB

Job ControllerComputation

Module

Job Controller Job Manager

SSH Deamon

Job StarterSSH Deamon

secure

connection

via the

internet

execute the

computation

module

control the

computation module

via a loopback

connection

start & control

simulation processes

via the local network

Network File System

Storage for

simulation data

and resultsjob control via

loopback

connection

LANInternet

Server Cluster

Job Management

Server Cluster

Calculation Time Improvement

0 1 2 3 4 5 6 7 8 9

Number of Nodes

Multi-Node Calculations

Railway Station

Cathedral

Stadium

Airport model with 1952 loudspeakers

Server Cluster Call

Energy time curve EDT, T10,

T20, T30

C80, C50

LF, LFC

TS, Echo

AlCons, STI, RaSTI

Level etc.

Extension to low frequencies

• Theoretical Background

• Schroeder frequency

• Modal sound field

• Transmission in modal sound field

• Simulation and Measurement

• Finite Element Method - FEM

• Boundary conditions

• Comparison of measurement and simulation results

• Optimizing absorber locations

• Summary to FEM calculation

Schroeder Frequency

Resonances in closed volumes

Modal density ~ f²

Diffuse sound field above Schroeder Frequency

=> Marks transition between wave and geometrical acoustics

Geometrical Acoustics

ustics

Tfs 2000

Modal Sound Field

Modes result from room geometry

Complex spatial resonance patterns in 3D

103 Hz

137 Hz

Sound Transmission in Modal Field

Isolated mode creates max-min pair

Transfer function is symmetric regarding receiver / source

Mode at 40 Hz

Mode at 75 Hz

Sound Transmission in Modal Field

Damping reduces the quality (Q) of the mode

Transfer function is smoother

Exciting undamped Modes

24,5Hz 34,5Hz

Point Source at 25Hz Point Source at 30Hz

Exciting damped Modes

Point source in undamped room

Point source in damped room

Listening Room

Volume 49 m³

RT ~0.15 s

Schroeder-Frequency

fs = 110 Hz

Simulation of Modal Sound Field Modeling wave acoustics using FEM

3D volume mesh of room

Solving the Helmholtz equation with boundary conditions

Boundary Conditions for Simulations

• Geometric acoustics

• Absorption

• Scattering

• Wave based acoustics

• impedance

p 2 v 2

1 2 3 Air Air Z0 = c Z0 = c

Modal Distribution

Measurement versus Simulation

Left Center

Comparison:

• Good qualitative reproduction of transfer function

• Simulation results show critical frequencies

Simulation

Measurement

Simulation

Measurement

Simulation

Measurement

Damping of Selected Modes • TF shows notch at ~90 Hz - how to smooth it?

• => Dampen corresponding mode

• Modes form between hard surfaces, pressure max on the surface

• => Replace hard surface material by absorber to suppress pressure max

• => Mode is attenuated

Measurement

Simulation

~ 90 Hz

Damping of Selected Modes

Measurement Simulation

Without Absorber With Absorber

Summary to FEM Calculation

• FEM modeling for low frequency range

• Determining complex boundary conditions can be difficult

• Qualitative results easily possible

• Optimize modal sound field regarding

• Loudspeaker placement

• Absorber type and placement

Speech Intelligibility

STI according to IEC 60268-16 (2003)

• Background

• Impulse response and modulation functions

• Implementation

• Eyring/Sabine model

• Noise levels

• Signal masking

• Criticism

Background:

Measurement via Modulation or IR: 14x7 MTF Values => STI

Typical MTF, STI and Scale:

STI = 0.590

> 0.75 Excellent

0.6 - 0.75 Good

0.45 - 0.6 Fair

0.3 - 0.45 Poor

< 0.3 Bad

Implementation:

• Eyring/Sabine model

• AURA, Raytracing

Direct Sound Early Reflections / Secondary Sources

Eyring RT/Tail

Direct Sound Early Reflections / Secondary Sources

Reflections

Implementation:

• Signal/Noise Ratio

-50 -40 -30 -20 -10 0 10 20 30 40 50

STI as a Function of S/N (Broadband)

Implementation:

• Signal Masking

45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120

Level in dB(A)

STI as a Function of Broadband Level

A-Weighted

f1 f2 = 2f1

Criticism:

• Stepped masking function

• Being revised for new version

0 500 1000 1500 2000

Delay [ms]

0 500 1000 S

Delay [ms]

Criticism:

• Echo situation: delay dependency

2 Pulses 3 Pulses

Speech Intelligibility • Criticism:

• Linearity of frequency response

• Bandwidth of analog measurements

500 Hz 1000 Hz 2000 Hz 4000 Hz

STIPa signal spectrum

Criticism:

• Binaural?

Latest development in modern measurement tools

• Software 1 (EASERA)

• Scale Model Measurements

• Absorption Coefficient Measurements

• Transmission Loss Measurements

• Software 2 (EASERA SysTune)

• SSA Filter

• Mobile Devices

• Hardware

• AD/DA transducer AUBION X8

Overview

Measurement Software EASERA

Graphic user interface to start a measurement

Measurement Module – Measuring Methods EASERA supports all common Measuring Methods

Simple

Reference

Dual-Channel-FFT

Dual-Channel-TDS

Comparison

External Stimulus

Playback

External

Post Processing - Displays EASERA Post Processing Module, Displays:

• Measures like: EDT, T10, T20, T30, C50, D50, C80, CT, MTF, MTI, STI, STIPa, RaSTI, ALCons, DSPL, TSPL, L50, L80, D/R, G, ST, IACC, LF/LFC (e.g. ISO 3382)

• Statistical Calculation: RMS, Noise Level, Crest Factor, S/N

• Time Domain Displays like: Impulse Response, Step Response , ETC, Schröder, Echogram

• Frequency Domain Displays like: Power Spectrum, Phase, Real & Imaginary Part, Group Delay

• Averaging, Integrating, Smoothing, Differentiating, Resampling

• Tabular View

• Overlay View

• Units like: Volt, Pa, FS, %, W, Ws, dBu, dBV, dBm, dBp, dBSPL, dBFS

• Waterfall

• Spectrogram

View&Calc Processing

Post Processing - Editing

www.sda.de

EASERA Editing Functions:

• Editing with User-defined Filters and Windows

• Editing by Data Averaging, Integrating, Smoothing, Differentiating, Resampling, Multiplication, Summation, Division, Subtraction, Power, Interpolation

Scale Model Measurements

EASERA Scale Model Measurements: - In air up to 192 kHz sample rate: Scale of 1:20

corresponds to 9.6 kHz sample rate => upper limit is 4 kHz octave

- Air attenuation is significant and must be compensated - Requires very low noise floor, since ΔL ~ t - Rescale by changing sample rate according to scale factor

Air Compensation

Change of Scaling Factor

Results of Scale Model Measurements

EASERA Scale Model Measurements:

170 ms at 192 kHz

3.4 s at 9,6 kHz 2.3 s

Live Sound Measurements

Live Sound Measurements Typical setup when using speech or music

signals:

Spectrally Selective Accumulation SSA • Speech or music signal: • Dedicated measurement signals

often not possible to use • Music and/or speech available during

rehearsal, less disturbing • „Non-ideal“ measurement signals

• Irregular in frequency • Irregular in time • No advance knowledge

• => Spectrally Selective Accumulation*:

• Extracts valid data • Rejects noise, interference,

perturbations

Classic Music: Non-Continuous in Frequency, Preferred Tones

Speech: Non-Continuous in Time and Frequency

*Patent pending

Spectrally Selective Accumulation SSA

Spectrally Selective Accumulation SSA Signal Threshold Filter (Action 1)

• Data only used when critical S/N reached

• Excludes unexcited frequencies from subsequent processing

• Threshold spectrum can be measured or entered

• For reference and signal spectrum

Rejected

Spectrally Selective Accumulation SSA Excursion Filter (Action 2):

Compares new transfer function measurement with existing

Deviating values outside tolerance are discarded/suppressed

Provides high immunity against temporary perturbations

Rejected

Coherence Filter (Action 3):

• Coherence: Linear relationship between input I and output O

Coherent Incoherent Rejected

C → 1 for <O/I>² ≈ <O²>/<I²>

C → 0 for <O/I> → 0

C → const for noise floor > 0

Spectrally Selective Accumulation SSA

SysTune Web Interface

SysTune @ iPad:

- SysTune plug-in with web interface

- Make measurements with SysTune from mobile device

- Control SysTune remotely

- For any platform with a browser

How it works: Laptop SysTune

Soundcard

Mobile Device

Wireless Microphone

Sound System

Output

Control

Reference Input

Supported Functions:

- Start / Stop Analysis

- Spectrum, IR, TF

- Frequency Resolution

- Capture Overlays

- Rename/Remove

Overlays

- Show/Hide Overlays

AUBION X.8

Professional Soundcard:

- Fully digital and calibrated

- 8 Inputs

- Integrated with SysTune and EASERA

- Ethernet-based

- Daisy-chaining possible

AUBION X.8 Features: - 8 Inputs (4 mic/line, 4 line), 2+2 outputs - 2x Ethernet with switch, SPDIF I/O optional - 10 Hz – 20 kHz, Dynamic Range 110 dB - THD < -100 dB, Crosstalk Attenuation > 110 dB - ASIO drivers

AFMG - cvut.cz

Documents

Transcript of AFMG - cvut.cz

GLL (Generic Loudspeaker Library) - AFMG

ŠTĚPÁNKOVÁ - cvut.cz

Program - cvut.cz

F3 - cvut.cz

WINTERSTEINER - cvut.cz

OMELKA - cvut.cz

AXONOMETRIE - cvut.cz

Polykardiografie - cvut.cz

ZAHRADNÍČKOVÁ - cvut.cz

HRUDKA - cvut.cz

Diferenciál - cvut.cz

AFMG SysTune Developed by AFMG Ahnert Feistel Media Group ...

SemvložtezadáníVašípráce. - cvut.cz

MATAR - cvut.cz

Preprocesor - cvut.cz

VÝKRESY - cvut.cz

02 - cvut.cz

AFMG SoundFlow

PETR - cvut.cz

Foundations - cvut.cz