A first filter

The same 300 Hz high-pass, four ways. Data is always (channels, samples).

Python

import numpy as np
import dspk

x = np.random.default_rng(0).standard_normal((4, 30_000)).astype(np.float32)
spikes = dspk.ops.filter.iir(x, 30_000.0, "highpass", 300.0)
spikes.shape   # (4, 30000)

Rust

use dspk::core::compute::ComputeTarget;
use dspk::ops::core::buffer;
use dspk::ops::{Pipeline, PipelineStage};

let client = ComputeTarget::from_env()?.client()?;
let (channels, samples, fs) = (4, 30_000, 30_000.0);
let x = vec![0.0_f32; channels * samples];

let mut pipeline = Pipeline::new();
pipeline.add(PipelineStage::highpass(300.0));
pipeline.validate(fs)?;

let input = buffer::upload(&client, &x);
let output = pipeline.execute::<f32>(&client, &input, channels, samples, fs)?;
let y: Vec<f32> = buffer::download(&client, output);

Pipeline spec

A spec works on files. Save it as highpass.toml and run dspk run highpass.toml.

dspk_pipeline = 1
edges = [["raw", "hp"], ["hp", "out"]]

[nodes.raw]
op = "read"
path = "recording.bin"

[nodes.hp]
op = "iir"
band = "highpass"
cutoff = [300.0]

[nodes.out]
op = "write"
path = "filtered.zarr"
format = "zarr"

App

1. Open a recording with File → Open.
2. In the Inspector on the right, under processing, use Add op → Filter → IIR and pick the preset Spikes · high-pass 300 Hz.
3. The view previews the result on the device, and updates as you change the values.

That’s the shape of every operation: data in, the same shape out, on the device.

Where next:

• Concepts explains what happens underneath.
• Operations lists everything you can run.