How DFA α1 works
Your heart doesn't beat at a perfectly steady rhythm — the tiny variations between beats carry information. Detrended Fluctuation Analysis (DFA) is a math trick that measures how "self-similar" those variations are across different timescales. The result is a single number, α1. Around 0.75 marks your aerobic threshold — the boundary where breathing starts getting noticeably harder; lab folks call this VT1. Above 0.75 you're in the easy zone, below it you're working. cellTrainer reads RR intervals from your heart-rate strap live, runs the math every few beats, and shows α1 in real time with a color cue. No lab visit needed.
How ectopic detection works
Sometimes a heartbeat fires earlier than expected — an "ectopic" beat. The heart usually compensates with a slightly longer pause right after, then carries on. cellTrainer watches the time between beats from your HR strap and looks for exactly that pattern: one or more early beats followed by a pause. When it spots one, it counts the event (single, couplet for two in a row, run for three or more) and rates severity by how long the compensatory pause was. The Heart Rhythm tile shows each beat as a vertical bar — gray for normal, red for premature, blue for the pause. Usually harmless during exercise, but useful to notice when a cluster shows up.
How HRV works (RMSSD vs SDNN)
Heart-rate variability measures how much the time between heartbeats varies — and your autonomic nervous system leaves fingerprints there. cellTrainer computes two flavours from the same RR stream over a rolling 60-second window. RMSSD looks at how much each beat differs from the next one — a short-term, beat-to-beat metric closely tied to your parasympathetic ("rest and digest") side. It's the single best number for recovery and overall fatigue. SDNN is the standard deviation of all RR intervals in the window — the total variability, blending short- and longer-term influences. Both shrink as intensity rises and your sympathetic system takes over. Watching both side by side gives you a fuller picture than either alone.
How SmO2 (muscle oxygen) works
NIRS sensors shine near-infrared light through your skin into the muscle below — typically a quad or calf. The amount of light that bounces back tells the sensor how much oxygen your hemoglobin is still carrying. The result is SmO2 in percent: around 80 % at rest, dropping into the 30–50 range during hard intervals as the muscle pulls oxygen out faster than blood can deliver. cellTrainer also extracts O2Hb (oxygen-loaded hemoglobin) and HHb (oxygen-stripped hemoglobin) from the same data — three signals from one sensor, useful for separating "the muscle is consuming oxygen" from "blood flow is changing". Supported sensors: Train.Red FYER 1.0 and 2.0.
How CORE body temperature works
Most apps treat "temperature" as ambient. The greenTEG CORE2 sensor measures *body* heat. Worn on the sternum, it combines a heat-flux measurement (how fast your body sheds energy through the skin) with skin-temp and a physiological model to estimate core body temperature (CBT) without an ingestible pill — typically 36.5–38.5 °C at rest, climbing to 39+ °C during hard heat-stress sessions. cellTrainer also derives the Heat Strain Index (HSI) on a 0–10 scale: 0–3 is comfortable, 4–6 means you're working with the heat, 7+ is genuine thermal load. CBT is what counts for adaptation — air-temp comfort lies. The same BLE link forwards your heart rate from a paired strap so the sensor doubles as an HR bridge.
How calories and fat/carb split work
cellTrainer estimates energy expenditure tick by tick, picking the best available source. First choice: a power meter (watts ÷ 25 % mechanical efficiency = kcal/sec). Otherwise the ACSM running formula from speed and incline. Otherwise a heart-rate based formula using your weight, age and sex (Keytel et al.). The fat and carb grams come from a substrate split keyed to your current DFA α1 zone: easy zone burns mostly fat (~70 % fat / 30 % carbs), aerobic 50/50, threshold flips to ~70 % carbs, high intensity ~90 %. Numbers are approximations — useful for fueling decisions ("did I just torch 60 g of carbs?") and trends across sessions, but not lab-grade.
How the AI coach summary works
After a session ends, cellTrainer can optionally ship an aggregated snapshot of the workout to OpenRouter — never the raw tick-by-tick stream, always a compact ~2 KB JSON: workout plan vs. actual, zone time, intensity factor, NP/TSS, DFA α1 distribution, HRV aggregates, ectopic events, and (if a CORE2 was connected) heat-strain stats. The model returns 3–5 sentences in your locale's language: did you hit the target, what physiological effect was achieved, what stood out. You provide your own OpenRouter API key and pick any model — Claude, GPT, Gemini, smaller open-source models, your call. The request goes directly from your device to OpenRouter; cellTrainer's server never sees your training data. The report is saved into the .ctd archive and the .xlsx export so it stays with the session forever.
