pulp::signal::OuWalkT< SampleType > Class Template Reference

Ornstein-Uhlenbeck walk (Uhlenbeck & Ornstein, Phys. More...

#include <rng.hpp>

Public Member Functions
void	prepare (double update_rate_hz)
void	set_theta (double theta_per_second)
void	set_sigma (double sigma)
void	set_mu (double mu)
void	seed (std::uint32_t s)
void	set_seed (std::uint32_t s)
void	set_correlation_time (double seconds, double sample_rate)
void	reset ()
void	reset_value ()
	Reset to the mean without disturbing the stream position.
SampleType	next ()
SampleType	current () const
SampleType	value () const
double	theta () const
double	sigma () const
double	stationary_std () const
	Closed-form stationary standard deviation, for tests and for callers sizing a modulation depth from a target excursion.

Static Public Attributes
static constexpr SampleType	kClamp = SampleType{1.5}
	A Gaussian walk has unbounded tails; downstream consumers (pitch ratios, delay lengths) do not.
static constexpr double	kMinTheta = 1e-6
static constexpr double	kMaxTheta = 1.0
static constexpr double	kDefaultTheta = 0.05

Detailed Description

template<typename SampleType = float>
class pulp::signal::OuWalkT< SampleType >

Ornstein-Uhlenbeck walk (Uhlenbeck & Ornstein, Phys.

Rev. 36, 1930):

y[n+1] = mu + e^(-theta*dt) * (y[n] - mu)
         + sigma * sqrt(1 - e^(-2*theta*dt)) * g

This is the exact discrete transition, not Euler-Maruyama: both the decay and innovation scale are solved for the configured update interval.

A random walk with a restoring force. Unlike filtered white noise it has a defined mean it returns to, and unlike an LFO it never repeats. The stationary standard deviation is the caller-facing sigma, independent of update rate and theta.

RT contract: prepare() recomputes dt and the noise scale; next(), reset(), and accessors allocate nothing.

USE: the one primitive behind tape wow, filter-cutoff drift, and feedback-network delay-length walks. Anywhere the answer to "why does this sound static?" is "because nothing is moving that the listener did not ask for". theta is how hard it is pulled home (small = long slow excursions), sigma is how far it gets. Output is clamped to +/-kClamp so a large sigma cannot hand a downstream pitch or delay control an absurd value.

Member Function Documentation

current()

template<typename SampleType = float>

SampleType pulp::signal::OuWalkT< SampleType >::current ( ) const

inline

next()

template<typename SampleType = float>

SampleType pulp::signal::OuWalkT< SampleType >::next ( )

inline

prepare()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::prepare ( double update_rate_hz )

inline

Parameters

update_rate_hz

Rate at which next() will be called — the audio sample rate for per-sample use, or the control rate for a decimated one. The walk's statistics are rate-independent.

reset()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::reset ( )

inline

reset_value()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::reset_value ( )

inline

Reset to the mean without disturbing the stream position.

seed()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::seed ( std::uint32_t s )

inline

set_correlation_time()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::set_correlation_time ( double seconds, double sample_rate )

inline

set_mu()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::set_mu ( double mu )

inline

set_seed()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::set_seed ( std::uint32_t s )

inline

set_sigma()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::set_sigma ( double sigma )

inline

set_theta()

template<typename SampleType = float>

void pulp::signal::OuWalkT< SampleType >::set_theta ( double theta_per_second )

inline

sigma()

template<typename SampleType = float>

double pulp::signal::OuWalkT< SampleType >::sigma ( ) const

inline

stationary_std()

template<typename SampleType = float>

double pulp::signal::OuWalkT< SampleType >::stationary_std ( ) const

inline

Closed-form stationary standard deviation, for tests and for callers sizing a modulation depth from a target excursion.

theta()

template<typename SampleType = float>

double pulp::signal::OuWalkT< SampleType >::theta ( ) const

inline

value()

template<typename SampleType = float>

SampleType pulp::signal::OuWalkT< SampleType >::value ( ) const

inline

Member Data Documentation

kClamp

template<typename SampleType = float>

SampleType pulp::signal::OuWalkT< SampleType >::kClamp = SampleType{1.5}

staticconstexpr

A Gaussian walk has unbounded tails; downstream consumers (pitch ratios, delay lengths) do not.

1.5 sits far enough out that clipping is rare at sane settings but the bound still exists.

kDefaultTheta

template<typename SampleType = float>

double pulp::signal::OuWalkT< SampleType >::kDefaultTheta = 0.05

staticconstexpr

kMaxTheta

template<typename SampleType = float>

double pulp::signal::OuWalkT< SampleType >::kMaxTheta = 1.0

staticconstexpr

kMinTheta

template<typename SampleType = float>

double pulp::signal::OuWalkT< SampleType >::kMinTheta = 1e-6

staticconstexpr

---

The documentation for this class was generated from the following file:

• /home/runner/work/pulp/pulp/core/signal/include/pulp/signal/rng.hpp