Feedback-topology FET compressor. More...

#include <fet_compressor.hpp>

Public Types
using	Ratio = FetRatio

Public Member Functions
	FetCompressorT ()
void	prepare (double sample_rate)
	Designs the resampler prototype, solves the divider's series resistance, and configures the transformer shelf.
void	reset ()
	Zeroes every state buffer.
void	set_input_gain_db (double db)
	The only lever into gain reduction — there is no threshold control.
void	set_output_gain_db (double db)
	Makeup gain, applied after the divider and before the transformer.
void	set_ratio (Ratio r)
Ratio	ratio () const
void	set_attack_us (double us)
void	set_release_ms (double ms)
void	set_knee_db (double db)
	Soft-knee width around the fixed reference, in the DETECTOR's domain.
void	set_transformer_amount (double pct01)
	0..1 depth of the output transformer's low-frequency tilt.
void	set_mix (double pct01)
	0..1 parallel blend.
int	latency_samples () const noexcept
	Exact, parameter independent. See the file doc block.
double	sample_rate () const noexcept
double	oversampled_rate () const noexcept
double	static_curve_db (double level_db) const
	The static characteristic's output for a level in dB, per Giannoulis et al.
double	gain_computer_db (double level_db) const
	Gain reduction the static curve asks for, as a POSITIVE magnitude in dB.
double	measured_static_curve_db (double input_db) const
	The static curve a user actually measures, input dB → output dB: the closed-loop fixed point ŷ = x̂ − bias − D(ŷ), solved in closed form on each of the three branches.
double	measured_gain_reduction_db (double input_db) const
	Measured gain reduction in dB (≥ 0) at a steady input level.
double	loop_slope () const
	The static curve's slope in the detector's domain, B = 1 − 1/R — the loop's only remaining gain term once the divider law is inverted.
double	measured_ratio () const
	The measured input/output ratio: 1 + B = 2 − 1/R, never R.
double	measured_knee_db () const
	The knee width a user measures at the INPUT, in dB.
double	nominal_ratio () const
	The nominal ratio feeding the gain computer (20 in all-buttons-in).
double	effective_knee_db () const
	Knee width in force, after the all-buttons-in widening.
double	bias_shift_db () const
	Detector bias shift in force (non-zero only in all-buttons-in).
double	coloration_depth () const
	Coloration depth ceiling in force: α_max, or α_max·kAbiAlphaExtraMult clamped to kColorationAlphaCeiling in all-buttons-in.
double	attack_coefficient () const
	Detector retain coefficients, exp(−1/(τ·fs_os)).
double	release_coefficient () const
double	gain_reduction_db () const
	Smoothed gain reduction in dB (≥ 0, a magnitude) as of the last oversampled step — the detector itself, exposed for metering and so tests can measure it directly instead of inferring it from audio that crosses zero every half cycle.
GainReduction	gain_reduction () const noexcept
double	control_voltage () const
	The control voltage currently driving the divider, 0..1.
double	divider_conductance (double c) const
	Shunt conductance at control voltage c — the quantity that is linear in control voltage, per Shichman & Hodges.
double	divider_small_signal_gain (double c) const
	Small-signal divider gain at fixed c: R_fet/(R_s + R_fet) with the rest-state trim folded in, so c = 0 is exactly unity and c = 1 is exactly −kGainReductionMaxDb.
double	divider_gain (double c, double v) const
	Instantaneous divider gain, including the coloration term.
double	coloration_multiplier (double c, double v) const
	1 + α(c)·v, the conductance multiplier.
double	coloration_multiplier_bound () const
	The bound on \|1 + α(c)·v\|.
double	control_for_reduction_db (double db) const
	The exact inverse of divider_small_signal_gain in the dB domain: the control voltage that makes the divider apply db dB of attenuation.
bool	divider_supremum_is_provable () const
double	resampler_peak_gain_bound () const
	‖h_up‖₁ · ‖h_down‖₁ for the shipped taps — a true ℓ∞ bound on the resampling pair's peak gain for ANY bounded input, not a step-response measurement.
double	worst_case_gain () const
	Closed-form ℓ∞ bound on \|y/x\| for the whole block at the current settings (series law 8).
SampleType	process (SampleType input)
	One mono sample.
void	process_block (SampleType *io, int n)

Static Public Attributes
static constexpr int	kOversamplingFactor = 4
	Oversampling factor.
static constexpr int	kResamplerTaps = 65
	Prototype length.
static constexpr double	kResamplerKaiserBeta = 8.0
	Kaiser shape parameter of the prototype.
static constexpr double	kResamplerCutoffFraction = 0.96
	Prototype cutoff, as a fraction of the BASE rate's Nyquist.
static constexpr int	kLatencySamples
	The whole module's latency, in base-rate samples.
static constexpr double	kFetOnResistance = 120.0
	FET fully on — the maximum gain reduction the divider can express.
static constexpr double	kFetOffResistance = 1.0e6
	FET fully off — negligible gain reduction at rest.
static constexpr double	kGainReductionMaxDb = 40.0
	The control-voltage span the divider law is normalised against: the attenuation the divider hits exactly at c = 1.
static constexpr double	kColorationAlphaMax = 0.08
	Coloration depth ceiling.
static constexpr double	kColorationAlphaCeiling = 0.16
	Hard ceiling on the effective coloration depth, whatever the ABI multiplier is re-tuned to.
static constexpr double	kColorationHeadroomClamp = 1.2
	Headroom clamp on the instantaneous sample fed to the coloration term.
static constexpr double	kAbiRatio = 20.0
	ABI reuses the 20:1 static curve rather than inventing a fifth ratio.
static constexpr double	kAbiKneeWidenMult = 2.5
	Widened knee — the documented "curve chaos" softness.
static constexpr double	kAbiBiasShiftDb = 1.5
	Detector bias-point shift, added directly to the gain-reduction magnitude before smoothing.
static constexpr double	kAbiAttackScale = 0.6
	Documented faster reaction.
static constexpr double	kAbiReleaseScale = 0.75
	[design parameter] default 0.75, range 0.5 .. 0.9.
static constexpr double	kAbiAlphaExtraMult = 1.4
	Raises the coloration ceiling — the documented increased distortion.
static constexpr double	kTransformerCornerHz = 60.0
	Corner of the output transformer's low-frequency shelf.
static constexpr double	kTransformerTiltDbMax = 3.0
	Tilt depth at transformer_amount = 1.
static constexpr double	kTransformerQ = 0.70710678118654752
	Butterworth-flat, the shape with no shelf overshoot — so the stage's supremum magnitude is exactly 1 and the gain bound stays clean.
static constexpr double	kThresholdDbfs = -18.0
	The fixed internal reference.
static constexpr double	kLevelEpsilon = dynamics::kLevelEpsilon
	Level-conversion floor, guarding log10(0).
static constexpr double	kInputGainDbMin = -20.0
static constexpr double	kInputGainDbMax = 40.0
static constexpr double	kOutputGainDbMin = -20.0
static constexpr double	kOutputGainDbMax = 20.0
static constexpr double	kAttackUsMin = 20.0
static constexpr double	kAttackUsMax = 800.0
static constexpr double	kReleaseMsMin = 50.0
static constexpr double	kReleaseMsMax = 1100.0
static constexpr double	kKneeDbMin = 0.0
static constexpr double	kKneeDbMax = 6.0
static constexpr double	kDividerGainSupremum = 1.0
	The divider's supremum gain over the whole (c, v) grid.
static constexpr double	kTransformerGainSupremum = 1.0
	The transformer stage's supremum magnitude.

Detailed Description

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

Feedback-topology FET compressor.

Member Typedef Documentation

◆ Ratio

template<typename SampleType = float>

using pulp::signal::FetCompressorT< SampleType >::Ratio = FetRatio

Constructor & Destructor Documentation

◆ FetCompressorT()

template<typename SampleType = float>

pulp::signal::FetCompressorT< SampleType >::FetCompressorT ( )

inline

Member Function Documentation

◆ attack_coefficient()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::attack_coefficient ( ) const

inline

Detector retain coefficients, exp(−1/(τ·fs_os)).

Exposed so a test can predict the closed-loop pole from the SHIPPED values rather than restating them.

◆ bias_shift_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::bias_shift_db ( ) const

inline

Detector bias shift in force (non-zero only in all-buttons-in).

◆ coloration_depth()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::coloration_depth ( ) const

inline

Coloration depth ceiling in force: α_max, or α_max·kAbiAlphaExtraMult clamped to kColorationAlphaCeiling in all-buttons-in.

◆ coloration_multiplier()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::coloration_multiplier	(	double	c,
		double	v ) const

inline

1 + α(c)·v, the conductance multiplier.

Bounded by coloration_multiplier_bound().

◆ coloration_multiplier_bound()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::coloration_multiplier_bound ( ) const

inline

The bound on |1 + α(c)·v|.

This is the quantity the source spec's registry row calls worst_case_gain; it is NOT one — see the file doc block. Bounds the divider's gain from BELOW.

◆ control_for_reduction_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::control_for_reduction_db ( double db ) const

inline

The exact inverse of divider_small_signal_gain in the dB domain: the control voltage that makes the divider apply db dB of attenuation.

Closed form — one pow, no iteration, which is what makes the loop unity-compensated without an implicit solve in process().

◆ control_voltage()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::control_voltage ( ) const

inline

The control voltage currently driving the divider, 0..1.

◆ divider_conductance()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::divider_conductance ( double c ) const

inline

Shunt conductance at control voltage c — the quantity that is linear in control voltage, per Shichman & Hodges.

◆ divider_gain()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::divider_gain	(	double	c,
		double	v ) const

inline

Instantaneous divider gain, including the coloration term.

v is the divider-referred audio sample; it is clamped to ±kColorationHeadroomClamp before use.

◆ divider_small_signal_gain()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::divider_small_signal_gain ( double c ) const

inline

Small-signal divider gain at fixed c: R_fet/(R_s + R_fet) with the rest-state trim folded in, so c = 0 is exactly unity and c = 1 is exactly −kGainReductionMaxDb.

The coloration term is a DEVIATION from this, never folded into it, so the documented small-signal gain stays exact and auditable independent of drive level (series law 1).

◆ divider_supremum_is_provable()

template<typename SampleType = float>

bool pulp::signal::FetCompressorT< SampleType >::divider_supremum_is_provable ( ) const

inline

◆ effective_knee_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::effective_knee_db ( ) const

inline

Knee width in force, after the all-buttons-in widening.

◆ gain_computer_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::gain_computer_db ( double level_db ) const

inline

Gain reduction the static curve asks for, as a POSITIVE magnitude in dB.

◆ gain_reduction()

template<typename SampleType = float>

GainReduction pulp::signal::FetCompressorT< SampleType >::gain_reduction ( ) const

inlinenoexcept

◆ gain_reduction_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::gain_reduction_db ( ) const

inline

Smoothed gain reduction in dB (≥ 0, a magnitude) as of the last oversampled step — the detector itself, exposed for metering and so tests can measure it directly instead of inferring it from audio that crosses zero every half cycle.

◆ latency_samples()

template<typename SampleType = float>

int pulp::signal::FetCompressorT< SampleType >::latency_samples ( ) const

inlinenoexcept

Exact, parameter independent. See the file doc block.

◆ loop_slope()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::loop_slope ( ) const

inline

The static curve's slope in the detector's domain, B = 1 − 1/R — the loop's only remaining gain term once the divider law is inverted.

◆ measured_gain_reduction_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::measured_gain_reduction_db ( double input_db ) const

inline

Measured gain reduction in dB (≥ 0) at a steady input level.

◆ measured_knee_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::measured_knee_db ( ) const

inline

The knee width a user measures at the INPUT, in dB.

Wider than the one they set, by the loop's own re-mapping.

◆ measured_ratio()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::measured_ratio ( ) const

inline

The measured input/output ratio: 1 + B = 2 − 1/R, never R.

◆ measured_static_curve_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::measured_static_curve_db ( double input_db ) const

inline

The static curve a user actually measures, input dB → output dB: the closed-loop fixed point ŷ = x̂ − bias − D(ŷ), solved in closed form on each of the three branches.

No iteration.

◆ nominal_ratio()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::nominal_ratio ( ) const

inline

The nominal ratio feeding the gain computer (20 in all-buttons-in).

◆ oversampled_rate()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::oversampled_rate ( ) const

inlinenoexcept

◆ prepare()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::prepare ( double sample_rate )

inline

Designs the resampler prototype, solves the divider's series resistance, and configures the transformer shelf.

Allocates nothing (every buffer is a fixed-size array), but is the expensive call and belongs off the audio thread.

◆ process()

template<typename SampleType = float>

SampleType pulp::signal::FetCompressorT< SampleType >::process ( SampleType input )

inline

One mono sample.

Stereo composes two instances driven from a caller-computed shared detector signal.

◆ process_block()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::process_block	(	SampleType *	io,
		int	n )

inline

◆ ratio()

template<typename SampleType = float>

Ratio pulp::signal::FetCompressorT< SampleType >::ratio ( ) const

inline

◆ release_coefficient()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::release_coefficient ( ) const

inline

◆ resampler_peak_gain_bound()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::resampler_peak_gain_bound ( ) const

inline

‖h_up‖₁ · ‖h_down‖₁ for the shipped taps — a true ℓ∞ bound on the resampling pair's peak gain for ANY bounded input, not a step-response measurement.

Conservative by design: the measured overshoot on a full-scale step is 1.045.

◆ reset()

template<typename SampleType = float>

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

inline

Zeroes every state buffer.

Never allocates; a zero-initialised instance after prepare() is already in this state.

◆ sample_rate()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::sample_rate ( ) const

inlinenoexcept

◆ set_attack_us()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::set_attack_us ( double us )

inline

◆ set_input_gain_db()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::set_input_gain_db ( double db )

inline

The only lever into gain reduction — there is no threshold control.

◆ set_knee_db()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::set_knee_db ( double db )

inline

Soft-knee width around the fixed reference, in the DETECTOR's domain.

The width a user measures at the input is wider — measured_knee_db().

◆ set_mix()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::set_mix ( double pct01 )

inline

0..1 parallel blend.

The dry path is taken PRE input gain and delayed by latency_samples(), so the blend is not a comb filter.

◆ set_output_gain_db()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::set_output_gain_db ( double db )

inline

Makeup gain, applied after the divider and before the transformer.

◆ set_ratio()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::set_ratio ( Ratio r )

inline

◆ set_release_ms()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::set_release_ms ( double ms )

inline

◆ set_transformer_amount()

template<typename SampleType = float>

void pulp::signal::FetCompressorT< SampleType >::set_transformer_amount ( double pct01 )

inline

0..1 depth of the output transformer's low-frequency tilt.

◆ static_curve_db()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::static_curve_db ( double level_db ) const

inline

The static characteristic's output for a level in dB, per Giannoulis et al.

— evaluated in the DETECTOR's domain, which for this topology is the compressor's own OUTPUT. This is NOT the input/output curve a user measures; measured_static_curve_db() is.

◆ worst_case_gain()

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::worst_case_gain ( ) const

inline

Closed-form ℓ∞ bound on |y/x| for the whole block at the current settings (series law 8).

Every factor is either exactly 1 with a proof (the divider, the cut-only transformer shelf) or a shipped gain.

Member Data Documentation

◆ kAbiAlphaExtraMult

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kAbiAlphaExtraMult = 1.4

staticconstexpr

Raises the coloration ceiling — the documented increased distortion.

[design parameter] default 1.4, range 1.2 .. 2.0.

◆ kAbiAttackScale

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kAbiAttackScale = 0.6

staticconstexpr

Documented faster reaction.

[design parameter] default 0.6, range 0.4 .. 0.8.

◆ kAbiBiasShiftDb

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kAbiBiasShiftDb = 1.5

staticconstexpr

Detector bias-point shift, added directly to the gain-reduction magnitude before smoothing.

Note this is added unconditionally, so ABI carries a small standing reduction even below the reference — which is what a bias shift means and what the mode is documented to do. [design parameter] default 1.5 dB, range 0.5 .. 3.0 dB.

◆ kAbiKneeWidenMult

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kAbiKneeWidenMult = 2.5

staticconstexpr

Widened knee — the documented "curve chaos" softness.

[design parameter] default 2.5, range 1.5 .. 4.0.

◆ kAbiRatio

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kAbiRatio = 20.0

staticconstexpr

ABI reuses the 20:1 static curve rather than inventing a fifth ratio.

◆ kAbiReleaseScale

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kAbiReleaseScale = 0.75

staticconstexpr

[design parameter] default 0.75, range 0.5 .. 0.9.

◆ kAttackUsMax

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kAttackUsMax = 800.0

staticconstexpr

◆ kAttackUsMin

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kAttackUsMin = 20.0

staticconstexpr

◆ kColorationAlphaCeiling

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kColorationAlphaCeiling = 0.16

staticconstexpr

Hard ceiling on the effective coloration depth, whatever the ABI multiplier is re-tuned to.

[design parameter] default 0.16, fixed.

◆ kColorationAlphaMax

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kColorationAlphaMax = 0.08

staticconstexpr

Coloration depth ceiling.

α(c) = kColorationAlphaMax · c, so the distortion scales with how hard the FET is being driven. Honest gap: no citable literature gives a coloration-depth constant for any historical unit — those numbers are measured and proprietary, not published. Original engineering, sized to sit inside the documented qualitative envelope. [design parameter] default 0.08, range 0.0 .. 0.2.

◆ kColorationHeadroomClamp

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kColorationHeadroomClamp = 1.2

staticconstexpr

Headroom clamp on the instantaneous sample fed to the coloration term.

Bounds the correction so it can neither invert nor blow up the divider law. [design parameter] default 1.2, range 1.0 .. 1.5.

◆ kDividerGainSupremum

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kDividerGainSupremum = 1.0

staticconstexpr

The divider's supremum gain over the whole (c, v) grid.

Exactly 1: the stage is a passive attenuator and the coloration term cannot lift it, because Δg·(1 − α·v_clamp) ≫ α·v_clamp·g_min at the shipped constants. divider_supremum_is_provable() checks that inequality numerically so a re-tune of the design parameters cannot silently break the claim.

◆ kFetOffResistance

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kFetOffResistance = 1.0e6

staticconstexpr

FET fully off — negligible gain reduction at rest.

[design parameter] default 1 MΩ, range 200 kΩ .. 5 MΩ.

◆ kFetOnResistance

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kFetOnResistance = 120.0

staticconstexpr

FET fully on — the maximum gain reduction the divider can express.

[design parameter] default 120 Ω, range 50 .. 500 Ω.

◆ kGainReductionMaxDb

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kGainReductionMaxDb = 40.0

staticconstexpr

The control-voltage span the divider law is normalised against: the attenuation the divider hits exactly at c = 1.

[design parameter] default 40 dB, range 20 .. 60 dB.

◆ kInputGainDbMax

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kInputGainDbMax = 40.0

staticconstexpr

◆ kInputGainDbMin

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kInputGainDbMin = -20.0

staticconstexpr

◆ kKneeDbMax

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kKneeDbMax = 6.0

staticconstexpr

◆ kKneeDbMin

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kKneeDbMin = 0.0

staticconstexpr

◆ kLatencySamples

template<typename SampleType = float>

int pulp::signal::FetCompressorT< SampleType >::kLatencySamples

staticconstexpr

Initial value:

=

2 * ((kResamplerTaps - 1) / 2) / kOversamplingFactor

pulp::signal::FetCompressorT::kOversamplingFactor

static constexpr int kOversamplingFactor

Oversampling factor.

Definition fet_compressor.hpp:269

pulp::signal::FetCompressorT::kResamplerTaps

static constexpr int kResamplerTaps

Prototype length.

Definition fet_compressor.hpp:275

The whole module's latency, in base-rate samples.

Exact, parameter independent, computed from the geometry above.

◆ kLevelEpsilon

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kLevelEpsilon = dynamics::kLevelEpsilon

staticconstexpr

Level-conversion floor, guarding log10(0).

Declared in dynamics_core.hpp alongside the conversions that consume it, and re-exported here so the call sites and the catalog keep reading one name.

◆ kOutputGainDbMax

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kOutputGainDbMax = 20.0

staticconstexpr

◆ kOutputGainDbMin

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kOutputGainDbMin = -20.0

staticconstexpr

◆ kOversamplingFactor

template<typename SampleType = float>

int pulp::signal::FetCompressorT< SampleType >::kOversamplingFactor = 4

staticconstexpr

Oversampling factor.

Chosen, not defaulted — see the file doc block. Fixed at build time because changing it changes latency_samples(). [design parameter] default 4, range {2, 4, 8}.

◆ kReleaseMsMax

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kReleaseMsMax = 1100.0

staticconstexpr

◆ kReleaseMsMin

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kReleaseMsMin = 50.0

staticconstexpr

◆ kResamplerCutoffFraction

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kResamplerCutoffFraction = 0.96

staticconstexpr

Prototype cutoff, as a fraction of the BASE rate's Nyquist.

Lower buys alias rejection and costs passband; 0.96 is the balance measured in the file doc block. [design parameter] default 0.96, range 0.85 .. 1.0.

◆ kResamplerKaiserBeta

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kResamplerKaiserBeta = 8.0

staticconstexpr

Kaiser shape parameter of the prototype.

House policy value. [design parameter] default 8.0, range 6.0 .. 12.0.

◆ kResamplerTaps

template<typename SampleType = float>

int pulp::signal::FetCompressorT< SampleType >::kResamplerTaps = 65

staticconstexpr

Prototype length.

Odd, and (kResamplerTaps − 1)/2 is divisible by kOversamplingFactor, which is what makes the base-rate group delay an exact integer rather than something to round. [design parameter] default 65, range 33 .. 257 (of the form 8k+1).

◆ kThresholdDbfs

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kThresholdDbfs = -18.0

staticconstexpr

The fixed internal reference.

Configurable only here, never at runtime: the absence of a threshold control is the point of the design. [design parameter] default −18 dBFS, range −24 .. −12 dBFS.

◆ kTransformerCornerHz

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kTransformerCornerHz = 60.0

staticconstexpr

Corner of the output transformer's low-frequency shelf.

A transformer loses level below its primary-inductance corner, so the shelf CUTS — which is also what keeps it out of the worst-case gain bound. [design parameter] default 60 Hz, range 20 .. 200 Hz.

◆ kTransformerGainSupremum

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kTransformerGainSupremum = 1.0

staticconstexpr

The transformer stage's supremum magnitude.

A low shelf with a non-positive gain and Q ≤ 1/√2 is monotone, so its peak is its high-frequency asymptote, exactly 1.

◆ kTransformerQ

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kTransformerQ = 0.70710678118654752

staticconstexpr

Butterworth-flat, the shape with no shelf overshoot — so the stage's supremum magnitude is exactly 1 and the gain bound stays clean.

[design parameter] default 0.70710678, range 0.3 .. 0.7071.

◆ kTransformerTiltDbMax

template<typename SampleType = float>

double pulp::signal::FetCompressorT< SampleType >::kTransformerTiltDbMax = 3.0

staticconstexpr

Tilt depth at transformer_amount = 1.

[design parameter] default 3 dB, range 0 .. 6 dB.

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

/home/runner/work/pulp/pulp/core/signal/include/pulp/signal/fet_compressor.hpp

Pulp 1.88.0 Cross-platform audio plugin framework

Public Types

Public Member Functions

Static Public Attributes

Detailed Description

Member Typedef Documentation

◆ Ratio

Constructor & Destructor Documentation

◆ FetCompressorT()

Member Function Documentation

◆ attack_coefficient()

◆ bias_shift_db()

◆ coloration_depth()

◆ coloration_multiplier()

◆ coloration_multiplier_bound()

◆ control_for_reduction_db()

◆ control_voltage()

◆ divider_conductance()

◆ divider_gain()

◆ divider_small_signal_gain()

◆ divider_supremum_is_provable()

◆ effective_knee_db()

◆ gain_computer_db()

◆ gain_reduction()

◆ gain_reduction_db()

◆ latency_samples()

◆ loop_slope()

◆ measured_gain_reduction_db()

◆ measured_knee_db()

◆ measured_ratio()

◆ measured_static_curve_db()

◆ nominal_ratio()

◆ oversampled_rate()

◆ prepare()

◆ process()

◆ process_block()

◆ ratio()

◆ release_coefficient()

◆ resampler_peak_gain_bound()

◆ reset()

◆ sample_rate()

◆ set_attack_us()

◆ set_input_gain_db()

◆ set_knee_db()

◆ set_mix()

◆ set_output_gain_db()

◆ set_ratio()

◆ set_release_ms()

◆ set_transformer_amount()

◆ static_curve_db()

◆ worst_case_gain()

Member Data Documentation

◆ kAbiAlphaExtraMult

◆ kAbiAttackScale

◆ kAbiBiasShiftDb

◆ kAbiKneeWidenMult

◆ kAbiRatio

◆ kAbiReleaseScale

◆ kAttackUsMax

◆ kAttackUsMin

◆ kColorationAlphaCeiling

◆ kColorationAlphaMax

◆ kColorationHeadroomClamp

◆ kDividerGainSupremum

◆ kFetOffResistance

◆ kFetOnResistance

◆ kGainReductionMaxDb

◆ kInputGainDbMax

◆ kInputGainDbMin

◆ kKneeDbMax

◆ kKneeDbMin

◆ kLatencySamples

◆ kLevelEpsilon

◆ kOutputGainDbMax

◆ kOutputGainDbMin

◆ kOversamplingFactor

◆ kReleaseMsMax

◆ kReleaseMsMin

◆ kResamplerCutoffFraction

◆ kResamplerKaiserBeta

◆ kResamplerTaps