Crate ppsd_rs 

Seismic PPSD (Probabilistic Power Spectral Density) computation.

A pure-Rust implementation of the McNamara & Buland (2004) PPSD algorithm, matching ObsPy’s PPSD output within 0.5 dB tolerance. Uses stationxml_rs::Response directly for instrument response evaluation — no intermediate types needed.

Algorithm

The PPSD pipeline for each time segment:

1. Welch’s method — overlap-averaged periodogram with cosine taper and linear detrend
2. Instrument response removal — evaluates PAZ, FIR, and Coefficients stages from stationxml_rs::Response, then divides PSD by |H(f)|²
3. Velocity→acceleration correction — multiplies by ω² = (2πf)²
4. dB conversion — 10 × log₁₀(PSD)
5. Period binning — averages dB values into octave-spaced period bins

Quick Start

use ppsd_rs::process_segment;

// Load inventory (FDSN StationXML or SC3ML — auto-detected)
let inv = stationxml_rs::read_from_file("IA.JAGI.xml").unwrap();
let channel = &inv.networks[0].stations[0].channels[0];
let response = channel.response.as_ref().unwrap();
let sample_rate = channel.sample_rate;

// Compute PSD for one segment
let result = process_segment(
    &samples, sample_rate, nfft, nlap,
    response, &psd_periods, &bin_left, &bin_right,
).unwrap();

Response Evaluation

eval_response walks all stages in a stationxml_rs::Response and computes the combined |H(f)|²:

Stage type	Type	Method
Sensor	stationxml_rs::PolesZeros	Laplace transfer function × gain
ADC	stationxml_rs::Coefficients (Digital)	DTFT of numerators × gain
FIR filter	stationxml_rs::FIR	DTFT with DC normalization × gain
Gain-only	(none of above)	gain²

FIR coefficients are DC-normalized to match evalresp behavior. stationxml_rs::Symmetry expansion is handled automatically (None = all coefficients, Even = mirror, Odd = mirror with negation).

Enums

PpsdError

Errors that can occur during PSD computation.

Functions

cosine_taper

Cosine taper (Tukey window) matching ObsPy’s cosine_taper(npts, p).

eval_response

Evaluate full instrument response |H(f)|² at given frequencies.

fft_power

Compute one-sided power spectrum via real FFT.

konno_ohmachi_smooth

Konno-Ohmachi log-frequency spectral smoothing.

period_bin_average

Bin PSD values into period bins by averaging (ObsPy style).

process_segment

Process one PPSD segment, replicating ObsPy’s PPSD.__process exactly.

setup_period_binning

Compute period binning edges matching ObsPy PPSD._setup_period_binning.

welch_psd

Compute PSD using Welch’s method matching ObsPy/matplotlib’s mlab.psd.

Type Aliases

Result

Result type for ppsd operations.