Physiology Data Integrity & Analysis-Readiness Workbench for ECG & RR Time Series
Engineered for human-performance researchers, sports-science engineers, academic physiology labs, and HRV algorithm developers working with chest-strap RR intervals (Polar H10 .csv / .fit) and raw single-lead ECG waveforms (.edf / .csv). CardioLab eliminates the Black-Box Smoothing Hazard by pairing immutable raw baselines (SHA-256) with a synchronized Raw / Processed / Overlay comparator and a 6-event human-in-the-loop Beat Ledger.
NON-MEDICAL SYNTHETIC WORKBENCH: This page renders deterministic mathematical waveforms (10s window, 101 samples) and a 6-event Beat Ledger (BL-01..BL-06). It is not a medical device, does not process patient data, and must never be used for clinical evaluation.
Non-medical notice: Research preview / in-development concept only. All records and waveforms shown on this page are synthetic and must not be used for clinical evaluation.
Interactive browser workbench available below · Zero account required
Primary Job-To-Be-Done
Inspect raw vs detrended ECG/RR waveforms (Δ a.u. = raw − processed) and audit every motion spike or timing perturbation in a traceable 6-event Beat Ledger before computing RMSSD, SDNN, or pNN50.
Key Capabilities
- Solves the Black-Box Smoothing Hazard vs. Manual QC Friction: preserves immutable raw sample arrays (raw_au with SHA-256 source hashing) alongside detrended series (processed_au) so RMSSD, SDNN, and pNN50 are never silently distorted by opaque consumer interpolation.
- Synchronized 3-Mode Signal Comparator (Raw, Processed, Overlay) computing exact point-by-point amplitude deltas (Δ a.u. = raw − processed, e.g., +0.420 a.u. at BL-03 t = 4.2s) across 101 deterministic samples and 3 selectable time windows.
- Traceable 6-Event Human-in-the-Loop Beat Ledger (BL-01..BL-06) & Analysis-Ready Package: logs observed_shape, proposed_action, and review_status (normal_reference, inspect_before_exclusion, candidate_detrend, verify_beat_boundary) for reproducible Python/R and AI Interpretation Pack handoffs.
Privacy & Local Data Boundary
All waveforms and ledger rows are generated from a deterministic built-in mathematical array. The website requests no Web Bluetooth permissions and accepts no biometric uploads.
01. INTERACTIVE ARCHITECTURE SIMULATOR (SAMPLE FIXTURE)
Inspect the transformation and boundary handling live
This interactive workbench runs locally in your browser on built-in synthetic fixtures.
CardioLab Synthetic Signal Workbench & Traceable Beat Ledger
Stage 1 · Immutable Raw Baseline (`SHA-256`)
Preserves original raw_au samples with SHA-256 source hashing so black-box smoothing never overwrites raw chest-strap RR or ECG traces.
Stage 2 · Prevents Unflagged `RMSSD` / `SDNN` / `pNN50` Distortion
Audits motion spikes (BL-03 +0.420 a.u.) and timing perturbations (BL-05) in a 6-event human-in-the-loop Beat Ledger (BL-01..BL-06).
Stage 3 · Analysis-Ready Package & AI Interpretation Pack
Emits detrended processed_au paired with manifest.json and the Beat Ledger receipt for reproducible Python/R or LLM workflows.
raw_au
1.362 a.u.
processed_au
1.435 a.u.
Δ a.u. (raw − proc)
-0.073 a.u.
observed_shape
High-frequency motion spike (+1.64 a.u. raw / +0.92 a.u. detrended)
proposed_action
Flag interval for researcher inspection; do not auto-delete
Cursor: t = 4.2s | raw_au = 1.362 | processed_au = 1.435 | Δ a.u. = 0.073
Beat Ledger — Traceable Processing Record (example only / not a patient reading)
Selected event: BL-03 (Δ a.u. = -0.073)| event_id | t_seconds | observed_shape | proposed_action | reason | source | review_status |
|---|---|---|---|---|---|---|
| BL-01 | 1.0s | Synthetic nominal R-peak (+1.18 a.u.) | Retain fiducial timestamp | Matches nominal 1.0s synthetic cycle period with low baseline drift. | synthetic | normal_reference |
| BL-02 | 2.0s | Synthetic nominal R-peak (+1.15 a.u.) | Retain fiducial timestamp | Consistent synthetic RR interval (1.00s). | synthetic | normal_reference |
| BL-03 | 4.2s | High-frequency motion spike (+1.64 a.u. raw / +0.92 a.u. detrended) | Flag interval for researcher inspection; do not auto-delete | Simulated electrode motion burst superimposed on baseline wander between 3.8s and 4.5s. Illustrative detrend reduces wander while preserving event marker. | synthetic | artifact_flagged |
| BL-04 | 5.0s | Post-disturbance recovery peak (+1.12 a.u.) | Mark preceding RR interval as uncertain | Interval from 4.2s disturbance cannot be treated as a clean physiological beat. | synthetic | review_required |
| BL-05 | 7.6s | Premature timing perturbation (+1.31 a.u. at 0.60s coupling) | Preserve raw morphology in Beat Ledger for manual review | Synthetic timing anomaly injected at t=7.6s. Unreviewed smoothing could obscure timing offset. | synthetic | review_required |
| BL-06 | 8.8s | Compensatory synthetic pause (+1.16 a.u. after 1.20s) | Retain with linked annotation to BL-05 | Demonstrates traceable pairing between premature synthetic event and subsequent interval. | synthetic | normal_reference |
Static All-States Reference Summary (No-JS & Audit View) — Time vs. Raw & Processed Normalized Amplitude + 6 Beat Ledger Events
| t_seconds | raw_au (a.u.) | processed_au (a.u.) | delta_au (Δ a.u.) | flagged_event_id |
|---|---|---|---|---|
| 0.0s | -0.080 | -0.080 | 0.000 | — |
| 0.2s | -0.010 | -0.080 | 0.070 | — |
| 0.4s | 0.055 | -0.080 | 0.135 | — |
| 0.6s | 0.112 | -0.080 | 0.192 | — |
| 0.8s | 0.156 | -0.080 | 0.236 | — |
| 1.0s | 1.386 | 1.120 | 0.266 | BL-01 |
| 1.2s | 0.199 | -0.080 | 0.279 | — |
| 1.4s | 0.195 | -0.080 | 0.275 | — |
| 1.6s | 0.173 | -0.080 | 0.253 | — |
| 1.8s | 0.136 | -0.080 | 0.216 | — |
| 2.0s | 1.285 | 1.120 | 0.165 | BL-02 |
| 2.2s | 0.023 | -0.080 | 0.103 | — |
| 2.4s | -0.045 | -0.080 | 0.035 | — |
| 2.6s | -0.115 | -0.080 | -0.035 | — |
| 2.8s | -0.183 | -0.080 | -0.103 | — |
| 3.0s | 0.955 | 1.120 | -0.165 | — |
| 3.2s | -0.296 | -0.080 | -0.216 | — |
| 3.4s | -0.333 | -0.080 | -0.253 | — |
| 3.6s | -0.355 | -0.080 | -0.275 | — |
| 3.8s | -0.370 | -0.083 | -0.287 | — |
| 4.0s | -0.487 | -0.115 | -0.372 | — |
| 4.2s | 1.362 | 1.435 | -0.073 | BL-03 |
| 4.4s | -0.452 | -0.125 | -0.327 | — |
| 4.6s | -0.215 | -0.080 | -0.135 | — |
| 4.8s | -0.150 | -0.080 | -0.070 | — |
| 5.0s | 1.120 | 1.120 | 0.000 | BL-04 |
| 5.2s | -0.010 | -0.080 | 0.070 | — |
| 5.4s | 0.055 | -0.080 | 0.135 | — |
| 5.6s | 0.112 | -0.080 | 0.192 | — |
| 5.8s | 0.156 | -0.080 | 0.236 | — |
| 6.0s | 1.386 | 1.120 | 0.266 | — |
| 6.2s | 0.199 | -0.080 | 0.279 | — |
| 6.4s | 0.195 | -0.080 | 0.275 | — |
| 6.6s | 0.173 | -0.080 | 0.253 | — |
| 6.8s | 0.136 | -0.080 | 0.216 | — |
| 7.0s | 1.285 | 1.120 | 0.165 | — |
| 7.2s | 0.023 | -0.080 | 0.103 | — |
| 7.4s | -0.045 | -0.080 | 0.035 | — |
| 7.6s | 1.185 | 1.220 | -0.035 | BL-05 |
| 7.8s | -0.183 | -0.080 | -0.103 | — |
| 8.0s | -0.245 | -0.080 | -0.165 | — |
| 8.2s | -0.296 | -0.080 | -0.216 | — |
| 8.4s | -0.333 | -0.080 | -0.253 | — |
| 8.6s | -0.355 | -0.080 | -0.275 | — |
| 8.8s | 0.841 | 1.120 | -0.279 | BL-06 |
| 9.0s | -0.346 | -0.080 | -0.266 | — |
| 9.2s | -0.316 | -0.080 | -0.236 | — |
| 9.4s | -0.272 | -0.080 | -0.192 | — |
| 9.6s | -0.215 | -0.080 | -0.135 | — |
| 9.8s | 1.050 | 1.120 | -0.070 | — |
| 10.0s | -0.080 | -0.080 | 0.000 | — |
02. PROBLEM & INTENDED OUTCOME
The Black-Box Smoothing Hazard: Why Unflagged Artifacts Distort RMSSD & SDNN
A single unflagged motion spike (e.g., strap slip at t = 4.2s) or sensor timing dropout in a chest-strap RR or single-lead ECG recording can inflate or depress time-domain HRV metrics (RMSSD, SDNN, pNN50). Consumer wellness apps hide this behind aggressive black-box smoothing that overwrites raw samples, while legacy academic tools (Kubios, LabChart) require tedious manual editing without emitting a machine-readable transformation ledger.
Engineered deliverable goal
CardioLab enforces a 3-stage Physiology Data Integrity architecture: (1) Immutable Raw Signal Ingestion with SHA-256 hashing, (2) Synchronized Raw / Processed / Overlay inspection with exact Δ a.u. readouts and a 6-event Beat Ledger (BL-01..BL-06), and (3) an Analysis-Ready Package & AI Interpretation Pack (manifest.json + Beat Ledger CSV/JSON) so downstream Python/R scripts know every modified beat.
- 01.Prevents RMSSD / SDNN / pNN50 distortion caused by unflagged motion spikes or black-box smoothing
- 02.Synchronized Raw, Processed, and Overlay comparator with exact Δ a.u. (raw − processed) readouts
- 03.6-event Human-in-the-Loop Beat Ledger (BL-01..BL-06) + Analysis-Ready Package for reproducible research
03 & 04. ARCHITECTURE, PERMISSIONS & KNOWN LIMITATIONS
Inspect Technical Pipeline, Permissions & Known Limitations
03 & 04. ARCHITECTURE, PERMISSIONS & KNOWN LIMITATIONS
Inspect Technical Pipeline, Permissions & Known Limitations
How the pipeline handles structure and unknown states
- Stage 01
Stage 1 · Immutable Raw Signal & Hardware Artifact Ingestion (SHA-256)
Preserves the original chest-strap RR (.csv / .fit) or single-lead ECG (.edf / .csv) sample array (raw_au) with a cryptographic SHA-256 source hash so baseline wander and motion spikes are never destructively overwritten.
Boundary note: Example only; 10.0-second synthetic mathematical waveform (101 samples) on this website. - Stage 02
Stage 2 · Synchronized Raw / Processed / Overlay Comparator & 6-Event Beat Ledger
Computes detrended processed_au alongside raw_au, displaying exact amplitude deltas (Δ a.u. = raw − processed) across BL-01..BL-06 with explicit human-in-the-loop review statuses (inspect_before_exclusion, candidate_detrend, verify_beat_boundary).
Boundary note: Demonstrates how a +0.420 a.u. motion spike (BL-03 at 4.2s) and timing perturbation (BL-05 at 7.1s) are audited before RMSSD/SDNN calculation. - Stage 03
Stage 3 · Analysis-Ready Package & AI Interpretation Pack (manifest.json)
Exports the detrended series paired with a machine-readable manifest.json and Beat Ledger receipt so downstream Python/R pipelines or LLM analysis sessions know exactly which beats were modified or excluded.
Boundary note: Requires human researcher inspection; strictly non-medical research output.
Explicit scope and verification requirements
| Area | Why accessed | On-device handling | Verification note |
|---|---|---|---|
| Synthetic waveform dataset | Render interactive SVG trace and Beat Ledger table. | Computed deterministically from static constants in browser memory. | Verified by unit tests on raw vs. processed sample arrays. |
| Web Bluetooth / hardware sensors | Not accessed by this website. | Zero device or sensor permissions requested. | Hardware support belongs to future desktop R&D scope (gate O-10). |
Known Limitations
- In development R&D concept; strictly non-medical and not for clinical evaluation or patient monitoring.
- Waveforms and Beat Ledger entries on this website are 100% synthetic mathematical fixtures.
- Real Bluetooth sensor compatibility (such as Polar H10 BLE PMD) and clinical DSP accuracy are not claimed or verified.
05. FREQUENTLY ASKED QUESTIONS
Common questions about CardioLab Workstation
Is CardioLab Workstation medical software?
No. No clinical use, diagnostic capability, or regulatory approval is claimed. All waveforms and Beat Ledger records shown here are synthetic examples.
Can I connect a Polar H10 sensor to this website now?
No. This website does not request Bluetooth permissions or connect to hardware sensors, and desktop sensor support is not yet verified.
Is the signal filter clinically validated?
No. The workbench demonstrates an illustrative mathematical transformation so you can inspect how raw and processed traces are compared alongside a Beat Ledger.
Need production AI document exports today? Noxport AI Exporter is Available Now.
While this R&D workbench is in preview/development, our flagship Noxport AI Exporter (v3.0.4.226) is shipping today with 6 Basic Formats Free Forever and 4 Pro Formats (Word .docx, Vector .pdf, Pro .md, Professional Pack .zip).
NOXPORT LABS R&D PORTFOLIO · RELEASE NOTIFICATIONS
Follow R&D workbenches in the Noxport Labs portfolio
Inspect interactive synthetic workbenches in your browser or choose specific R&D projects below to receive release notifications.
AnswerBell AI
Monitors long-running deep-research streams across a 3000ms DOM quiet window so you know when a thread is ready for Noxport AI Exporter compilation.
YouTube → Gemini Gem
Converts long technical lectures into millisecond-anchored Markdown & JSONL corpora with explicit missing-caption gap flags (seg-005 unavailable).
Vitalis Health Studio
Demonstrates a local read-only SQLite MCP policy boundary that allows bounded 28-day aggregates while blocking raw record dumps with JSON-RPC -32001.
Explore CardioLab Workstation architecture & documentation
Inspect our demonstration architecture documentation or explore another project in the Noxport Labs portfolio.
Div I · Available
Noxport AI Exporter
Turn 50–150+ turn ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek, Copilot, and Qwen sessions into clean, editable Word (.docx), searchable vector PDF, Pro Markdown, and .zip bundles locally in your browser. Engineered for staff/principal engineers, independent technical consultants, AI solution architects, and computational researchers — with zero cloud uploads, no account required, and zero silent truncation.
Div I · In development
AnswerBell AI
AnswerBell AI eliminates the "Babysitting Tax" and context-switch fatigue when running multi-minute reasoning models (o3-pro, Deep Research, Claude Extended Thinking) across 10–30 open browser tabs. Unlike naive notification extensions that trigger false-positive chimes on a single Stop-button flicker, AnswerBell reconciles multi-signal DOM evidence over a 3000ms quiet window to distinguish true completion from mid-run tool pauses.
Div II · Preview
YouTube → Gemini Gem
Engineered for technical researchers, PhD students, and investigative analysts synthesizing multi-video lecture series, conference keynotes, and technical interviews into grounded corpora for Google Gemini Gems and NotebookLM. Unlike 1-click YouTube AI summarizers that emit unverifiable prose, YouTube → Gemini Gem preserves a 1:1 raw_lines audit trail, millisecond timecodes, and fail-closed missingness flags.
Div III · Preview
Vitalis Health Studio
Engineered for quantified-self engineers, privacy-conscious longevity researchers, and AI builders who own multi-year archives from Apple Health (export.xml), Garmin (.fit), Oura, and Whoop. Vitalis resolves the dilemma between Vendor Truncation (mobile apps locking your data inside opaque 7-day readiness scores) and Cloud AI Exfiltration (pasting raw multi-year CSV/XML dumps into cloud LLMs).