Why PMX exists: a note on sensor transduction, resolution, and provenance.
Plenty of online discussion compares fitness sensors — "the Polar H10 is more accurate than the Garmin optical sensor" — without any grounding in what these devices actually measure or how. A number on a screen isn't automatically a measurement; a measurement implies knowing something about the transduction chain that produced it, and where it can go wrong. This is a short note on that chain for heart-rate sensing specifically, and why PMX is built the way it is as a result.
A chest strap like the Polar H10 measures the heart's electrical depolarization directly, via ECG electrodes. The QRS complex — the sharp spike marking each heartbeat — has a very high rate of voltage change, so a peak-detection algorithm has a narrow, well-defined event to timestamp. That's why chest straps can report beat-to-beat timing (R-R intervals) with genuine precision.
A wrist-based optical sensor like Garmin's Elevate does not measure the electrical event. It measures photoplethysmography (PPG): an LED illuminates the skin and a photodiode detects changes in reflected light as blood volume in the capillary bed changes with each pulse. That pulse arrives at the wrist after a real physiological delay (pulse transit time, roughly 100–250ms, itself variable with blood pressure and vascular tone) and is shaped by the arterial wall's mechanical compliance and by light scattering in tissue. Every stage of that chain — electrical to mechanical to pressure wave to blood volume to light absorption to photodiode current — is an additional smoothing/averaging step. PPG is inherently lower-bandwidth for beat-to-beat timing than ECG, independent of motion artifact, which is a separate problem stacked on top.
The ANT+ Heart Rate Profile transmits beat timing in units of 1/1024 second — about 0.977ms of quantization resolution. That's a real, specified number for what a chest strap like the H10 reports over the wire. But it raises the question this whole note is really about.
A protocol field quantized to 1/1024s tells you the grid the timestamp is reported on. It does not tell you how precisely the underlying peak-detection algorithm located the actual QRS event on that grid — those are two different claims, and conflating them is a basic measurement-science mistake. The physical argument above is what actually justifies trusting ECG-derived timing more than PPG-derived timing for this purpose: not because one device reports more decimal places, but because the QRS spike is a sharper physical event to timestamp than a PPG upstroke ever can be.
This matters most for anything computed from beat-to-beat intervals rather than just average heart rate. True heart-rate variability metrics (RMSSD, SDNN, and similar) assume the interval sequence reflects the heart's actual electrical timing. PPG-derived interval sequences carry pulse-transit-time variability as an additional, physiologically real confound — which is why the literature sometimes calls the PPG-derived quantity pulse-rate variability (PRV) rather than HRV at all. It isn't just a noisier version of the same measurement; for this purpose, it's arguably measuring something different.
None of the above means PPG is "bad" — for a real-time "am I over my training threshold" display, its smoothing is close to irrelevant, maybe even desirable. The point is that which sensor is right depends on what you're about to do with the number, and you can't make that call if you don't know which sensor produced it, or if a platform has already silently fused multiple sources into one opaque field.
That's the actual point of PMX: not declaring a winner between sensors, but refusing to hide the provenance. A PMX-bridged session can record pmx_H10_bpm alongside a watch's native hr field in the same .fit file — so the choice of which stream is canonical for a given analysis is yours to make, deliberately, rather than a platform's firmware fusion logic making it for you, invisibly.