openEMS/FDTD/excitation.cpp

267 lines
6.3 KiB
C++

/*
* Copyright (C) 2010 Thorsten Liebig (Thorsten.Liebig@gmx.de)
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "tools/array_ops.h"
#include "tools/useful.h"
#include <iostream>
#include <fstream>
#include "fparser.hh"
#include "tinyxml.h"
#include "excitation.h"
Excitation::Excitation()
{
Signal_volt = 0;
Signal_curr = 0;
m_Excit_Type = -1;
dT = 0;
m_nyquistTS = 0;
}
Excitation::~Excitation()
{
delete[] Signal_volt;
delete[] Signal_curr;
}
void Excitation::Reset( double timestep )
{
delete[] Signal_volt;
Signal_volt = 0;
delete[] Signal_curr;
Signal_curr = 0;
m_Excit_Type = -1;
dT = timestep;
m_nyquistTS = 0;
}
bool Excitation::setupExcitation( TiXmlElement* Excite, unsigned int maxTS )
{
if (!Excite)
{
cerr << "Excitation::setupExcitation: Error, can't read openEMS excitation settings... " << endl;
return false;
}
if (dT<=0)
{
cerr << "Excitation::setupExcitation: Error, invalid timestep... " << endl;
return false;
}
double f0=0;
double fc=0;
Excite->QueryIntAttribute("Type",&m_Excit_Type);
switch (m_Excit_Type)
{
case 0:
Excite->QueryDoubleAttribute("f0",&f0);
Excite->QueryDoubleAttribute("fc",&fc);
CalcGaussianPulsExcitation(f0,fc,maxTS);
break;
case 1:
Excite->QueryDoubleAttribute("f0",&f0);
CalcSinusExcitation(f0,maxTS);
break;
case 2:
CalcDiracPulsExcitation();
break;
case 3:
CalcStepExcitation();
break;
case 10:
Excite->QueryDoubleAttribute("f0",&f0);
CalcCustomExcitation(f0,maxTS,Excite->Attribute("Function"));
break;
default:
cerr << "Excitation::setupExcitation: Unknown excitation type: \"" << m_Excit_Type<< "\" !!" << endl;
m_Excit_Type = -1;
return false;
}
if (GetNyquistNum() == 0)
{
cerr << "Excitation::setupExcitation: Unknown error... excitation setup failed!!" << endl;
return false;
}
return true;
}
unsigned int Excitation::GetMaxExcitationTimestep() const
{
FDTD_FLOAT maxAmp=0;
unsigned int maxStep=0;
for (unsigned int n=1; n<Length+1; ++n)
{
if (fabs(Signal_volt[n])>maxAmp)
{
maxAmp = fabs(Signal_volt[n]);
maxStep = n;
}
}
return maxStep;
}
void Excitation::CalcGaussianPulsExcitation(double f0, double fc, int nTS)
{
if (dT==0) return;
Length = (unsigned int)(2.0 * 9.0/(2.0*PI*fc) / dT);
if (Length>(unsigned int)nTS)
{
cerr << "Operator::CalcGaussianPulsExcitation: Requested excitation pusle would be " << Length << " timesteps or " << Length * dT << " s long. Cutting to max number of timesteps!" << endl;
Length=(unsigned int)nTS;
}
delete[] Signal_volt;
delete[] Signal_curr;
Signal_volt = new FDTD_FLOAT[Length+1];
Signal_curr = new FDTD_FLOAT[Length+1];
Signal_volt[0]=0.0;
Signal_curr[0]=0.0;
for (unsigned int n=1; n<Length+1; ++n)
{
double t = (n-1)*dT;
Signal_volt[n] = cos(2.0*PI*f0*(t-9.0/(2.0*PI*fc)))*exp(-1*pow(2.0*PI*fc*t/3.0-3,2));
t += 0.5*dT;
Signal_curr[n] = cos(2.0*PI*f0*(t-9.0/(2.0*PI*fc)))*exp(-1*pow(2.0*PI*fc*t/3.0-3,2));
}
SetNyquistNum( CalcNyquistNum(f0+fc,dT) );
}
void Excitation::CalcDiracPulsExcitation()
{
if (dT==0) return;
Length = 1;
// cerr << "Operator::CalcDiracPulsExcitation: Length of the excite signal: " << ExciteLength << " timesteps" << endl;
delete[] Signal_volt;
delete[] Signal_curr;
Signal_volt = new FDTD_FLOAT[Length+1];
Signal_curr = new FDTD_FLOAT[Length+1];
Signal_volt[0]=0.0;
Signal_volt[1]=1.0;
Signal_curr[0]=0.0;
Signal_curr[1]=1.0;
SetNyquistNum( 1 );
}
void Excitation::CalcStepExcitation()
{
if (dT==0) return;
Length = 1;
delete[] Signal_volt;
delete[] Signal_curr;
Signal_volt = new FDTD_FLOAT[Length+1];
Signal_curr = new FDTD_FLOAT[Length+1];
Signal_volt[0]=1.0;
Signal_volt[1]=1.0;
Signal_curr[0]=1.0;
Signal_curr[1]=1.0;
SetNyquistNum( 1 );
}
void Excitation::CalcCustomExcitation(double f0, int nTS, string signal)
{
if (dT==0) return;
if (nTS<=0) return;
Length = (unsigned int)(nTS);
// cerr << "Operator::CalcSinusExcitation: Length of the excite signal: " << ExciteLength << " timesteps" << endl;
delete[] Signal_volt;
delete[] Signal_curr;
Signal_volt = new FDTD_FLOAT[Length+1];
Signal_curr = new FDTD_FLOAT[Length+1];
Signal_volt[0]=0.0;
Signal_curr[0]=0.0;
FunctionParser fParse;
fParse.AddConstant("pi", 3.14159265358979323846);
fParse.AddConstant("e", 2.71828182845904523536);
fParse.Parse(signal,"t");
if (fParse.GetParseErrorType()!=FunctionParser::FP_NO_ERROR)
{
cerr << "Operator::CalcCustomExcitation: Function Parser error: " << fParse.ErrorMsg() << endl;
exit(1);
}
double vars[1];
for (unsigned int n=1; n<Length+1; ++n)
{
vars[0] = (n-1)*dT;
Signal_volt[n] = fParse.Eval(vars);
vars[0] += 0.5*dT;
Signal_curr[n] = fParse.Eval(vars);
}
SetNyquistNum( CalcNyquistNum(f0,dT) );
}
void Excitation::CalcSinusExcitation(double f0, int nTS)
{
if (dT==0) return;
if (nTS<=0) return;
Length = (unsigned int)(nTS);
// cerr << "Operator::CalcSinusExcitation: Length of the excite signal: " << ExciteLength << " timesteps" << endl;
delete[] Signal_volt;
delete[] Signal_curr;
Signal_volt = new FDTD_FLOAT[Length+1];
Signal_curr = new FDTD_FLOAT[Length+1];
Signal_volt[0]=0.0;
Signal_curr[0]=0.0;
for (unsigned int n=1; n<Length+1; ++n)
{
double t = (n-1)*dT;
Signal_volt[n] = sin(2.0*PI*f0*t);
t += 0.5*dT;
Signal_curr[n] = sin(2.0*PI*f0*t);
}
SetNyquistNum( CalcNyquistNum(f0,dT) );
}
void Excitation::DumpVoltageExcite(string filename)
{
ofstream file;
file.open( filename.c_str() );
if (file.fail())
return;
for (unsigned int n=1; n<Length+1; ++n)
file << (n-1)*dT << "\t" << Signal_volt[n] << "\n";
file.close();
}
void Excitation::DumpCurrentExcite(string filename)
{
ofstream file;
file.open( filename.c_str() );
if (file.fail())
return;
for (unsigned int n=1; n<Length+1; ++n)
file << (n-1)*dT + 0.5*dT << "\t" << Signal_curr[n] << "\n";
file.close();
}