Spectral parameters

Shared by Triggered Power and Triggered Coherence, on top of the shared capture parameters, unless a table says otherwise.

Everything under ANALYSIS is locked during acquisition, in both plugins: it all reshapes the accumulators, so changing any of it discards what has accumulated.

Estimator

Name	Label	Type	Default	Values	Scope	Locked
mode	Mode	categorical	Spectrogram	Spectrogram, Spectrum	PROCESSOR	yes

Time-resolved spectrogram, or one spectrum per trial window. Every other group's applicability depends on it, so it is first and never greyed.

Frequency axis

Applies in both modes, both plugins.

Name	Label	Type	Default	Range	Step	Locked
freq_min	Freq min	float, Hz	2	0.1 – 1000	0.5	yes
freq_max	Freq max	float, Hz	200	1 – 10000	5	yes
num_freqs	Num freqs	int	60	4 – 512	—	yes
freq_spacing	Spacing	categorical	Log	Linear, Log	—	yes

num_freqs is the number of frequencies in the spectrogram. Cost is linear in it. freq_spacing defaults to Log, which is what you want for a wide range with Morlet wavelets.

Spectrogram — Morlet

Applies in Spectrogram mode with tf_method = Morlet.

Name	Label	Type	Default	Range	Step	Locked
tf_method	TF method	categorical	Morlet	Morlet, Hann STFT	—	yes
n_cycles_low	Cycles low	float	3	1 – 20	0.5	yes
n_cycles_high	Cycles high	float	10	1 – 40	0.5	yes

The number of cycles ramps linearly from n_cycles_low at the first frequency to n_cycles_high at the last, so low frequencies get time resolution and high frequencies get frequency resolution. Fewer cycles means a shorter wavelet: better time resolution, worse frequency resolution.

The ramp runs across the grid index, not across frequency, so a log-spaced grid gets an even progression.

Spectrogram — Hann STFT

Applies in Spectrogram mode with tf_method = Hann STFT.

Name	Label	Type	Default	Range	Step	Locked
stft_window_ms	STFT window	float, ms	256	16 – 4000	16	yes
stft_hop_ms	STFT hop	float, ms	25	1 – 1000	1	yes

A fixed sliding window rather than a frequency-dependent one: stft_window_ms long, stepped by stft_hop_ms. The right choice when you want one time resolution across the whole range. A hop much smaller than the window buys smoothness, not information.

Spectrum — line

Applies in Spectrum mode.

Name	Label	Type	Default	Range	Step	Locked
line_method	Line method	categorical	Multitaper	Multitaper, Hann	—	yes
nw	NW	float	3	1 – 10	0.5	yes
n_tapers	Tapers	int	5	1 – 19	—	yes

nw is the time-bandwidth product for the DPSS tapers. n_tapers is how many of the resulting sequences are used — 2·NW − 1 is the usual choice, hence the defaults of NW 3 and 5 tapers. More tapers means less variance and worse frequency resolution; more than 2·NW − 1 mostly adds bias.

Hann is the single-taper fallback.

Triggered Power only

Per-trial storage — ANALYSIS

Name	Label	Type	Default	Range	Locked
max_trials	Max trials	int	50	1 – 500	yes

Per-trial spectra retained in Spectrum mode, for the individual lines the display draws behind the average. Note the different default and range from Triggered Average's max_trials.

Baseline — canvas

Display-time, and not locked: applied when the display reads the accumulators.

Name	Label	Type	Default	Range / values	Step
baseline_mode	Baseline	categorical	None	None, dB change, Percent change, Z-score	—
baseline_start_ms	Base start	float, ms	-500	-10000 – 10000	10
baseline_end_ms	Base end	float, ms	0	-10000 – 10000	10

Both bounds are relative to the trigger, so the default window is the pre-trigger stretch −500 ms to 0 ms. Z-score uses the across-trial spread of the baseline.

In Spectrum mode this does change what is estimated

It splits the trial window. It stays on the canvas because that is where it is used.

Whitening — canvas

Display-time, and not locked. 1/f removal.

Name	Label	Type	Default	Range / values	Step
whitening_mode	Whiten	categorical	None	None, Fixed exponent, Fitted 1/f	—
whitening_exponent	Exponent	float	1.0	0 – 4	0.1
whitening_overlay	Show 1/f	categorical	Off	Off, On	—

• Fixed exponent multiplies by f^exponent. Cheap and predictable when the exponent is known.
• Fitted 1/f fits the aperiodic background log10 P = offset − exponent · log10 f and divides it out, also reporting the fitted exponent.
• whitening_exponent is read only by Fixed exponent.
• whitening_overlay plots the un-whitened spectrum with the aperiodic background drawn over it, Spectrum mode only. It is the tuning view for the exponent.

Triggered Coherence only

Coherence controls — ANALYSIS

Name	Label	Type	Default	Values	Locked
shift_predictor	Shift pred	categorical	On	Off, On	yes

Also accumulate the trial-shifted null: channel A of each trial against channel B of the previous one.

Defaults to on. Without it there is no way to tell coherence driven by a shared evoked response or a common reference from a real interaction. It is under ANALYSIS because it accumulates a second estimate as trials arrive.

Display — canvas

Display-time, and not locked.

Name	Label	Type	Default	Range / values
coherence_display	Show	categorical	Coherence	Coherence, Phase, Shift predictor, PPC
smooth_time_bins	Smooth t	int	0	0 – 32
smooth_freq_bins	Smooth f	int	0	0 – 32

Phase is the phase of the coherency; PPC is pairwise phase consistency, which is unbiased by trial count.

The two smoothing controls pool neighbouring time-frequency bins into the cross-spectrum sums. Wavelets give one estimate per trial, so with few trials the coherence estimate is badly biased upwards; pooling buys degrees of freedom at the cost of resolution, and is the main stabiliser in Spectrogram mode.

There is no whitening control here: coherence is a normalised ratio, so a per-frequency gain cancels exactly.

Canvas display controls

Not registered parameters, in either plugin.

Control	Values
Colour map	Viridis, Magma, Diverging, Greyscale
Columns	grid layout
Panel height	grid layout

Diverging is blue-white-red, for signed data such as dB change from baseline.

Sessions

Neither spectral plugin writes sessions; the canvases carry no SAVE / LOAD buttons. See Saved sessions.