EEG Spectrum

CPU side (per frame). The node resolves the Spectrum JSON, sorts electrode keys into anatomical front-to-back order (Fp, F, C, P, T, O — handling A_/B_/C_ prefixes), and reads the frequency resolution from the FFT EEG payload (_freqRes, falling back to a cached value or 0.244 Hz). It converts Freq Min/Freq Max into a visible bin window, clamps each PSD slice to [0, 1e6], and applies exponential smoothing with vDSP: smoothed = smoothed × α + new × (1 − α), where α is the Settings-panel Smoothing value. The smoothed bins for all electrodes are packed into one reusable shared-storage MTLBuffer. In Single mode, only the selected electrode's slice is uploaded.

Auto-scaling. If Power Max is greater than 0 it is used directly. If it is 0 or negative, the node auto-scales: an EMA tracker follows the global maximum bin with 20% headroom (autoMax = autoMax × 0.95 + globalMax × 1.2 × 0.05), floored at 0.001. This is what lets a plot stay readable as signal strength changes.

GPU side. A fullscreen quad over the plot area draws band zones, grid, and — in Line/Filled style — the spectrum itself, evaluated as a Catmull-Rom spline (4 subdivisions per bin, 3 bin spans searched) against a point-to-segment SDF for gap-free, aspect-corrected anti-aliased curves. In Bars style a second instanced pass draws one quad per bin per electrode (80% fill, 20% gap). Output is premultiplied alpha; edge softness comes from Grafit's global AA multiplier.

Coordinates and outputs. X/Y position the plot's center in Grafit coordinates (origin at viewport center, shorter viewport edge maps to [-1, 1]); Width/Height set its extent. With labels enabled a 12% left margin and 8% right margin are reserved, and traces render only in between. The Image output feeds a Canvas node — nothing appears on screen without one — and setting Enable to false clears the texture.

Power scale

Display modes (Settings panel)

Render styles (Settings panel)

Band zones

Band ranges are not hardcoded — they come from the _bandRanges payload attached by FFT EEG, so whatever bands you configure upstream are what get drawn. The first five are shaded as background zones at 8% alpha, and a colored boundary line marks each band's lower edge (the first band's edge is skipped). Colors are fixed per band name:

Grid and labels

Frequency grid lines use a nice-number step (…0.1, 0.2, 0.5, 1, 2, 5, 10…) targeting about 8 minor divisions across the range, with major lines at 5× the minor step; horizontal power lines are fixed at 0, 25, 50, 75, 100% of the plot height. Labels are AppKit text overlays: a left-side power ruler (up to 12 ticks), right-side electrode legends colored by hue, a unit indicator above the plot (µV² or dB), and frequency labels beneath every grid line (up to 50), with major-line labels drawn brighter.

Data outputs

• Peak Frequencies — per electrode, a nested object with peakFreq (Hz of the strongest visible bin) and peakPower. Note the plural key name and the nested structure; both the outer object and the band object are ordered anatomically.
• Bands — per electrode, the mean power inside each configured band, keyed by band name in the order FFT EEG supplied.
• Width / Height — pass through the resolved Width and Height inputs as floats, handy for aligning other generators to the same box.

Interactions worth knowing: Freq Min is forced below Freq Max (if you invert them the node repairs the range), and the frequency window together with the FFT's bin resolution determines visibleBins — a narrow window over a high-resolution FFT yields fat, well-separated bars, while a wide window makes Bars style dense enough that Line or Filled reads better. Power Max and the Power Scale setting jointly define the vertical mapping, and the left ruler always relabels itself to match. Smoothing (Settings) trades responsiveness for stability and applies before every other calculation, including the peak and band outputs.

• BrainBrick Device Input → FFT EEG → EEG Spectrum → Canvas — the standard live PSD display for a neurofeedback session
• EEG Spectrum (Bands) → JSON Array Get → Remap → Canvas Opacity — let alpha-band power fade a visual layer in and out
• EEG Spectrum (Peak Frequencies) → JSON Array Get → Remap — drive a color, rotation, or particle rate from each electrode's dominant frequency
• EEG Spectrum in Stacked mode with a tall Height — compare spectral profiles across all electrodes at a glance for montage checks
• EEG Spectrum in Single mode with the Electrode picker — focus a large hero plot on one site (e.g. an occipital channel for alpha work)
• BrainBrick Device Input → EEG Plot → Canvas on one layer and EEG Spectrum → Canvas on another — a research dashboard showing time and frequency domains together
• EEG Spectrum (Bands) → Smooth → threshold logic — trigger scene changes when a band crosses a level, using the same smoothed data the plot shows

• Generators are invisible without a Canvas node on Image; Canvas owns layer order (Layer 1 = back) and blend mode.
• Set Power Max to 0 whenever signal strength varies between sessions — auto-scaling keeps the plot readable without manual tuning.
• Band zones and boundary lines mirror whatever ranges FFT EEG is configured with, so change the bands there, not here.
• Bars style at a wide frequency range renders one instance per bin per electrode; Line or Filled is both cheaper to read and usually cheaper to draw at high bin counts.
• Turn Show Label off for production visuals — it also removes the side margins, letting the spectrum use the full plot width.
• Anti-aliasing follows Grafit's global AA multiplier; there is no per-node AA quality control.
• Electrode keys are kept in anatomical front-to-back order (Fp, F, C, P, T, O) through ordered JSON, so Peak Frequencies and Bands iterate in the same order as the legend on screen.