stella.util
Class SineError

java.lang.Object
  stella.util.SineError

All Implemented Interfaces:

Multidimensional

public class SineError
extends Object
implements Multidimensional

A class that models a time-dependend shift according to an overlay of sine functions. A PID is then used to correct for this shift, and as a test, the difference between the PID response and the true function is feed into the PID. For optimal performance, the frequencies should be plain integers.
To allow optimization of the PID with a simplex algorithm, this class also implements the Multidimensional interface that returns a measure of the goodness-of-leveling of the PID.

Nested Class Summary

static class	SineError.Parse A command-line frontend to the sine-error class.
private static class	SineError.Relaxation A class that simulates the reaction of the system to an offset.

Field Summary

static int	ABSOLUTESUM The indicator for absoulte error-sum quality measure.
private double[]	amplitude The coefficients of the input function, amplitude.
private double	deadtime The deadtime until the system starts reacting, in PID-t units.
private double[]	frequency The coefficients of the input function, frequency.
private double	lag Latency until the system has reached the new output, in PID units.
private int	maxrelevant The maximum number of offsets allowed.
private double	noise The coefficient for thw gaussian noise.
private static double	OVERREACT If the PID starts to oscilate, the correction will be higher as.
private double[]	phase The coefficients of the input function, phase.
private PIDControl	pid The PIDController handling the error function.
private double	pidfreq The frequency at which the PID is queried.
private boolean	pionly If true, only optimize PI parameters, keep D fixed.
private int	qualitymeasure The quality measure, 0-3.
private int	relaxfreq The period at which offsets are generated in PID-periods.
private static int	RESOLUTION The number of steps skipped for error determination.
static int	SQUARESUM The indicator for error square-sum quality measure.
static int	TABSOLUTESUM The indicator for absolute error times t sum quality measure.
private static double	TINY A small number to prevent division-by-zero in oscillating test.
static int	TSQUARESUM The indicator for error squared times t sum quality measure.
private Random	white The random number generator.

Constructor Summary

SineError(double[] a, double[] f, double[] phi)
Constructs a new sine error function specifiying the coefficients of the error function.

SineError(double[] a, double[] f, double[] phi, double random)
Constructs a new sine error function specifiying the coefficients of the error function.

Method Summary

int	dimension() Return three for the three PID parameters.
double	evaluate(VectorG pidvals) Evaluate the PID.
double	getErrorFunction(double t) Calculates the error function at the given time.
double	getMinimumFrequency() Returns the time step of the simulation.
double	runPID(PrintWriter out) Runs an entire loop using the specified PID and the underlying error function.
void	setDelay(double newdeadtime) Sets the deadtime time.
void	setLatencySpeed(double newlag) Sets the lag time.
void	setMaxRelaxations(int relaxcount) Sets the number of maximal concurrent active relaxation processes.
void	setPID(PIDControl handle) Sets the PIDControler that should level the error function.
void	setPIDFrequency(double f) Sets the PID frequency.
void	setPIOnly(boolean onlypi) Sets the desired optimization strategy.
void	setQualityMeasure(int newqual) Sets the quality measurement.
void	setRelaxationPeriod(int relaxperiod) Set the frequency of the PID corrections as a multiple of the PID internal period.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

SQUARESUM

public static final int SQUARESUM

The indicator for error square-sum quality measure.

See Also:

Constant Field Values

ABSOLUTESUM

public static final int ABSOLUTESUM

The indicator for absoulte error-sum quality measure.

See Also:

Constant Field Values

TSQUARESUM

public static final int TSQUARESUM

The indicator for error squared times t sum quality measure.

See Also:

Constant Field Values

TABSOLUTESUM

public static final int TABSOLUTESUM

The indicator for absolute error times t sum quality measure.

See Also:

Constant Field Values

amplitude

private double[] amplitude

The coefficients of the input function, amplitude.

frequency

private double[] frequency

The coefficients of the input function, frequency.

phase

private double[] phase

The coefficients of the input function, phase.

noise

private double noise

The coefficient for thw gaussian noise.

white

private Random white

The random number generator.

pid

private PIDControl pid

The PIDController handling the error function.

pidfreq

private double pidfreq

The frequency at which the PID is queried.

relaxfreq

private int relaxfreq

The period at which offsets are generated in PID-periods.

maxrelevant

private int maxrelevant

The maximum number of offsets allowed.

deadtime

private double deadtime

The deadtime until the system starts reacting, in PID-t units.

lag

private double lag

Latency until the system has reached the new output, in PID units.

qualitymeasure

private int qualitymeasure

The quality measure, 0-3.

pionly

private boolean pionly

If true, only optimize PI parameters, keep D fixed.

RESOLUTION

private static final int RESOLUTION

The number of steps skipped for error determination.

See Also:

Constant Field Values

OVERREACT

private static final double OVERREACT

If the PID starts to oscilate, the correction will be higher as.

See Also:

Constant Field Values

TINY

private static final double TINY

A small number to prevent division-by-zero in oscillating test.

See Also:

Constant Field Values

Constructor Detail

SineError

public SineError(double[] a,
                 double[] f,
                 double[] phi)

Constructs a new sine error function specifiying the coefficients of the error function. No white noise.

SineError

public SineError(double[] a,
                 double[] f,
                 double[] phi,
                 double random)

Constructs a new sine error function specifiying the coefficients of the error function. White noise is added.

Method Detail

setPID

public void setPID(PIDControl handle)

Sets the PIDControler that should level the error function.

setPIDFrequency

public void setPIDFrequency(double f)

Sets the PID frequency.

setMaxRelaxations

public void setMaxRelaxations(int relaxcount)

Sets the number of maximal concurrent active relaxation processes.

setRelaxationPeriod

public void setRelaxationPeriod(int relaxperiod)

Set the frequency of the PID corrections as a multiple of the PID internal period. If set to e.g. 5, only every fifth PID-loop will add an offset process.

setQualityMeasure

public void setQualityMeasure(int newqual)

Sets the quality measurement. Defines how the quality of the PID is measured. The values allowed are:

• SQUARESUM:
• ABSOLUTESUM:
• TSQUARESUM:
• TABSOLUTESUM:

All of the quality measurements are obtained only after the intial transient response, defined by the number of steps given in #RESPONSE.

setDelay

public void setDelay(double newdeadtime)

Sets the deadtime time. This is the time it takes the system to start reacting to a new signal. The time unit is the inverse of the PID frequency.

setLatencySpeed

public void setLatencySpeed(double newlag)

Sets the lag time. This is the time it takes the system to fully adjust to the new output value. The time unit is the inverse of the PID frequency.

setPIOnly

public void setPIOnly(boolean onlypi)

Sets the desired optimization strategy. If true, only P and I are changed.

getErrorFunction

public double getErrorFunction(double t)

Calculates the error function at the given time. The function is evaluated as

       f = Σiaiċsin(2πfi*t+φi
       

getMinimumFrequency

public double getMinimumFrequency()

Returns the time step of the simulation. This is the time that has to be used to produce a repetition in the error-function after t=1. Calculated correctly only if the frequencies are plain integers.

runPID

public double runPID(PrintWriter out)

Runs an entire loop using the specified PID and the underlying error function. If the PrintWriter supplied is not null, the result of each step are writtn within, The procedure is as follows.

• We start at time t=0 and increase the time in steps returned by #getTimeStep until we reach t >= 1.
• We start at t=0 calculating the error function and plugging it into the PID with #addError.
• At any time ti>0, we retrieve the PID response with #getCorrection and feed the difference of the error function and the correction into the PID as the new error.
• At any step, the absolute value of the difference between the error function and the PID is calculated and added to the offset.
• At t=1, the offset is returned.

dimension

public int dimension()

Return three for the three PID parameters.

Specified by:

dimension in interface Multidimensional

evaluate

public double evaluate(VectorG pidvals)

Evaluate the PID. This runs the PID and returns the offset. This offset, in turn, should be minimized. The parameters in the vector argument are interpreted as the proportional, integral, and differential parameter of the PID, in this order.

Specified by:

evaluate in interface Multidimensional