PhasefinchSIGNAL LABORATORY / 006REPRODUCIBLE EXPERIMENTS · LOCAL IN YOUR BROWSER
Keep the signal.
Question the filter.
One experiment. Two ways to see it.
Change the conditions, hold the seed, and discover what smoothing leaves behind.
Synthetic data · 1-second records · No account requiredTime domain
AMPLITUDE / AUFrequency domain
AMPLITUDE / AUReady. Run an experiment or save this configuration.
AN OPEN MODEL
A smoother curve
is not always a better answer.
A moving average suppresses noise and can also flatten the signal you care about. Compare both domains and the error against the known clean reference.
Inspect the method +
The signal is a unit sine wave plus seeded uniform noise in [−A, A]. A 32-bit linear congruential generator makes every sample reproducible. Records contain one second of samples; the last sample occurs one interval before 1 s.
The centered moving average uses an odd window. At the boundaries it uses only available samples and renormalizes the divisor. No phase delay is added in the interior; amplitude can be attenuated.
The spectrum is a one-sided DFT with a rectangular window and 1 Hz bins. Amplitude is doubled except at DC and Nyquist. RMSE compares sampled traces with the known clean sampled sine. Signals at or above Nyquist cannot be uniquely reconstructed.
PROPOSED APPLICATION CREDITS
Room for deeper experiments.
Phasefinch credits are planned for advanced filter recipes, batch experiments, and shared experiment notebooks.
These services and token activation are not implemented. The local lab is free to use. No token has been issued.
Proposed: compare a bank of filter recipes and publish a reproducible notebook with collaborators.