A quick hot fix

• Randomize button: respects baseline & eligible rows, safe during sessions
• Session saving: study folder + Study_S###R##[_v##].mat
• Plot saving: same scheme + collision-safe PNGs
• Blinding: saved metadata retained but kept out of filenames
• Minor UI polish & stability fixes

• Blinded/Unblinded modes — choose how to run experiments.
• Feedback UI refactored to a state machine for smoother behavior.
• Channel mismatch alert to prevent misconfigured device/LSL setups.
• New probeset and normalization tweaks for more stable values.
• Session data now saves transfer runs.
• Fixes: bad-channel deselection, sham device issues (incl. missing channel).
• Docs & configs: README was refreshed, and device models were updated.

🚀 What’s New
• Save transfer trials & raw feedback
– Now each exported sample has both raw and normalized feedback values.

• Record run type
– Every sample now carries a "feedback" vs. "transfer" flag in the saved .mat, making downstream analysis simpler.

🐞 Bug Fixes
• Unidentified short-channel crash
– Guard against missing SS channels in session.pushSample—we only shift SSwindow fields that actually exist, preventing “Unrecognized field” errors.

Fixed

• Sham device raw‑760 nm count corrected from 51 to 52 (added devch 51)

New

• Immediate sample pulling on OPEN (restores v1.1 behavior)

Fixed

• 0 Hz sample‐rate display
• Delay (“DELAY”) drift

• Sham protocol now works.
• SS channel count displays correctly
• The UI now accurately detects and shows the number of short‑separation channels you’ve selected—no more off‑by‑one glitches.
• All‑channels selection UI bugs ironed out
• Various edge‑case issues in the channel‑selection tables have been smoothed over, so selecting or deselecting channels is rock‑solid.
• Device configurations fixed

[v1.2.5] – 2025-06-11

Added

• ShamGauss protocol
  • Gaussian-noise sham algorithm fully implemented.
  • process(...) signature updated to accept the new 11 arguments (marker, srate, idx, data, SSdata, windownum, window, SSwindow, isfullwindow, prevfeedback, prevmarker) and delegate to the original Gauss logic unchanged.

Changed

• Protocol list
  • Removed the old “Gauss” entry; renamed/refactored it to ShamGauss.m with no adapter layer required.
  • MovAvg remains as the active real‐time NF algorithm, while “ShamGauss” serves solely as the control.

Release v1.2.4 – Robust Channel Selection & Improved UI Sync
Highlights:

• Intelligent Channel Selection:
  Channels from the device’s JSON channel_map are now only pre-selected if the user hasn’t already made a selection. This ensures default selections are helpful but never overwrite manual user choices.

• Manual Overrides are Persistent:
  Users can deselect or reselect channels at any time—these choices persist throughout the session, even after UI refresh. The deselection part happens blindly without knowing which channels are selected.

• UI and Logic Always in Sync:
  The interface accurately reflects the channels being processed for both NF (long) and SS (short) roles, providing full transparency and reliability.

• Improved Accessibility:
  The “Bad Channels” (Deselect) button is now only enabled after the LSL stream is open, preventing invalid actions.

• Streamlined Layout:
  “Bad Channels” controls have been moved directly under “Channels” in the settings for a more intuitive workflow.

• An optional channel JSON parameter: Load your long- and short-channel definitions directly from a JSON map.

• Long and short channels validation: Checks that the map has both long_channels and short_channels and that every ID is in range.

• No overlaps or duplicates: Detects if a channel appears in both long and short lists or if any ID is repeated.

• User-friendly errors: Pops up clear, one-line alerts (e.g., “Field ‘short_channels.HbO’ contains duplicate IDs: [1]”).