Multispectral PPG
Why 1050 nm vs 940 nm in wearable PPG
Most wrist wearables that go beyond green light settle on a red plus 940 nm near-infrared pair. SciFold adds a fourth optical channel at 1050 nm. This article explains what that extra wavelength actually buys you, why it only becomes useful paired against 940 nm, and what it deliberately does not claim to measure.
The starting point: why green alone is fragile
Green light near 530 nm is the workhorse of consumer PPG. It is strongly absorbed by hemoglobin and penetrates only shallowly, which makes it excellent at tracking the superficial capillary pulse for heart rate. But that same shallowness makes it easy to fool. A change in green amplitude can come from a real physiological shift, or from strap pressure, motion, ambient light leakage, skin temperature, or contact quality. With one wavelength, those causes are hard to tell apart because you only have one view of the tissue.
Adding wavelengths does not magically remove ambiguity. What it does is create a richer optical space in which different causes leave different fingerprints across channels. The value is in comparison, not in any single channel measuring a clinical variable on its own.
What 940 nm is good for: a stable reference
The 940 nm channel is a familiar near-infrared reference in wearable optics for a practical reason: it sits near an isosbestic region for hemoglobin variants, and it is largely insensitive to tissue water. In plain terms, it responds mostly to blood volume and vascular behavior and relatively little to hydration-related optical change. That stability is exactly what makes it valuable — not as the interesting signal, but as the yardstick you measure other channels against. It also penetrates more deeply than green, giving a view of vascular behavior below the capillary surface.
What 1050 nm adds: a deeper, water-sensitive contrast
The 1050 nm channel is chosen for a different property. It sits on the rising edge of a local water absorption band, and it penetrates more deeply still than 940 nm. That combination means its optical behavior carries information about tissue-water-related change that the 940 nm reference is comparatively blind to. On its own, a 1050 nm amplitude is not interpretable — it is influenced by vascular tone, tissue composition, temperature, local perfusion, contact pressure, and motion, all at once. The point is never to read 1050 nm in isolation.
The real move: a differential, not a single reading
The reason to run 1050 nm and 940 nm together is that they respond differently to the thing you care about (tissue-water-related change) but similarly to many of the things you do not (blood-volume change, motion, contact drift, pigmentation). When two channels share the same nuisance influences, comparing them lets much of that common-mode noise cancel out.
Each channel is split into its pulsatile (AC) and quasi-static (DC) parts, and a normalized modulation index — AC divided by DC — is computed per wavelength. Taking the ratio of the 1050 nm index to the 940 nm index yields a differential feature that emphasizes what is specific to the deeper, water-sensitive channel while suppressing what both channels see in common. SciFold treats the result as a Fluid Trend Signal: a relative, baseline-referenced feature, not an absolute number.
What this is not
This is the part that matters most for honesty and for credibility with serious reviewers. A 1050 nm channel on a wrist does not measure hydration percentage, total body water, glucose, or blood pressure. Any of those claims would outrun what the physics of a wrist-worn optical sensor can support. The differential described above is a trend feature — it tells you whether a water-sensitive optical signal is drifting relative to a person's own recent baseline, after motion, temperature, and contact context have been accounted for. It does not output a clinical value, and it is not a diagnosis.
That is why SciFold's public vocabulary stays deliberately narrow: fluid trend, perfusion trend, and signal confidence, surfaced as a simple state such as Normal, Watch, Elevated, or High. The realistic claim — that four wavelengths give a better-conditioned signal space for baseline-relative trend detection — is stronger than an impossible one, because it survives scrutiny.
Why the pairing is a platform decision, not a spec-sheet flourish
Extra wavelengths only earn their place if the fusion logic uses them. The 1050 nm-versus-940 nm differential is meaningful precisely because it is evaluated alongside perfusion, temperature, contact pressure, and motion, and gated by a confidence layer that suppresses outputs when the signal cannot be trusted. A channel added for marketing novelty and read in isolation would add cost without adding information. A channel added as a differential reference, inside a confidence-gated trend engine, is an architecture choice — and that is the distinction SciFold is built around.