Development documentation · 0.2.3 30c2f8ba · Intro examples tested with 0.2.3 · Version details · Known limitations

TimeSeries: Basics#

Create a voltage signal, inspect its metadata, select an interval, and save a plot. Prerequisites: the Quickstart. No optional analysis packages or external data are needed. Allow about 15 minutes to read and run the lesson.

1. Create a signal#

This is the same 40 Hz tone and seeded sensor noise used in the Quickstart. A TimeSeries keeps sample values together with their unit, start time, and cadence.

from gwexpy.noise.wave import gaussian, sine

settings = dict(duration=16, sample_rate=512, t0=0, unit="V")
signal = sine(frequency=40, **settings) + gaussian(std=0.3, seed=10, **settings)
signal.name = "Sensor A"

2. Inspect values and metadata#

value contains an array of sample values; unit, t0, and sample_rate explain what those numbers mean. dt is the time between samples. Keep these attributes when exporting values to another tool.

print(signal[:5])
print("Unit:", signal.unit)
print("Start:", signal.t0)
print("Sample rate:", signal.sample_rate)
print("Cadence:", signal.dt)
assert signal.size == 16 * 512

TimeSeries([-0.33100153,  0.25388934,  0.59692804,  1.07527748,
             0.84930531],
           unit: V,
           t0: 0.0 s,
           dt: 0.001953125 s,
           name: Sensor A,
           channel: None)
Unit: V
Start: 0.0 s
Sample rate: 512.0 Hz
Cadence: 0.001953125 s

3. Select an interval#

The interval starts at 2 seconds and ends just before 4 seconds. Assign the returned series to a new name so the full input remains available. For a TimeSeriesDict, use channels.copy().crop(...) to preserve the collection.

segment = signal.crop(2, 4, copy=True)
assert segment.size == 2 * 512
assert signal.size == 16 * 512

4. Plot and save#

A short interval makes individual oscillations visible. The plot labels the sample unit, and savefig writes a figure that can be used in a log or slide.

plot = signal.crop(2, 2.1, copy=True).plot()
plot.savefig("timeseries.png")
../_images/cde7fa97b34f551245930f63e905687f077c8bc81887de17c2d39958b166e17e.png

5. Estimate an ASD#

Use the full 16-second record for averaging. Two-second Hann-windowed segments with one-second overlap give 0.5 Hz frequency bins. ASD is expressed in V per square root Hz; it is not the amplitude of an individual sine wave.

spectrum = signal.asd(fftlength=2, overlap=1, window="hann", method="welch")
plot = spectrum.plot(xlim=(1, 256))
plot.savefig("timeseries-asd.png")
assert spectrum.unit.is_equivalent(signal.unit / signal.sample_rate.unit**0.5)
../_images/f6f2e51e78507ad6c6fb2a34b566b79a71451d28c50a8f157c197fd5757d25eb.png

Next lessons#