What a Real Desaturation Curve Looks Like — and What Doesn't
Minute three of a five-minute bike interval, and the SmO2 trace on the coach's tablet does something nobody planned for: it drops from 54% to 31% across two samples, sits there for a beat, then jumps back to 58% before resuming its slow physiological slide. Power output never wavered. Heart rate kept climbing along its normal curve. Nothing about the athlete's effort explains a 23-point swing that resolves itself in under two seconds — muscle tissue oxygenation simply doesn't move that fast, because the blood flow kinetics that drive it are bounded by vascular response time, not by however abruptly a chart auto-scales. What just got recorded is a motion artifact, not a desaturation event, and until you can tell the two apart on sight, every SmO2 number pulled from that session is a guess wearing a decimal point.
What a Real Desaturation Curve Looks Like — and What Doesn't
Muscle deoxygenation in response to a step change in effort follows blood flow and oxygen extraction kinetics, not an on/off switch. Exercise physiology work on muscle deoxy-hemoglobin kinetics has repeatedly shown a mono-exponential rise or fall with a time constant typically landing somewhere in the 10-to-20-second range at exercise onset, sometimes longer during recovery. That's the shape a trustworthy SmO2 trace should have: a curve that bends, not a cliff that drops. A genuine desaturation event during a hard interval looks like a smooth decline over roughly ten to thirty seconds, tracking the rise in effort, then a smooth recovery curve of similar or slightly longer duration once the effort eases.
An artifact breaks that shape in a specific way — it changes faster than the tissue physically can. If two consecutive samples, typically a second or two apart depending on the device's sample rate, show a swing of more than roughly 10 to 15 percentage points with no corresponding change in power, pace, or cadence, that's not muscle chemistry. It's optical or mechanical interference at the sensor-skin interface. The practical rule worth writing on a whiteboard: real SmO2 changes are rate-limited by blood flow; artifacts are not rate-limited by anything, because they aren't measuring blood flow at all in that moment.
Three Different Failures That All Look Like a Spike
Coaches troubleshooting a noisy SmO2 trace tend to lump every weird number under a single bad-sensor label, but three distinct mechanisms produce three distinct signatures, and mixing them up leads to the wrong fix. Sensor slip is an optical failure. Adipose tissue thickness is a signal-quality ceiling set before the session even starts. Mechanical or pressure artifact is a movement-timed distortion layered on top of a real signal.
| Root cause | What triggers it | Signature on the raw trace | What actually fixes it |
|---|---|---|---|
| Sensor slip / probe lift-off | Explosive or multiplanar movement, sweat loosening adhesive, a strap that doesn't conform to the muscle belly | A sudden jump toward an extreme (near 0% or pinned near 99%) within one or two samples, uncorrelated with effort, often right after a jump, cut, or change of direction | Retape the sensor housing itself, not just the strap; add a light-blocking sleeve; re-check tension between sets, not only at the start |
| Adipose tissue thickness (ATT) | Present at placement, not a one-time event — worsens with any drift of the probe over the fat layer | A compressed dynamic range through the whole session: small, sluggish movements even during a genuinely hard effort, or low-amplitude noisy oscillation rather than a clean curve | Measure ATT with a skinfold caliper at the exact placement site before testing; apply the device's ATT-correction setting if one exists; where ATT is high, read trend direction rather than absolute percentage |
| Mechanical / pressure artifact | Strap tension changing as the muscle belly swells and contracts under dynamic load — a pedal stroke, a running stride, a grip over the sensor site | A rhythmic wobble locked to movement cadence rather than to effort, or a spike that repeats at the same phase of every cycle | Loosen strap tension slightly; move placement away from the point of maximum girth change; if the oscillation persists, relocate to a site with less mechanical deformation |
Adipose tissue thickness deserves particular respect because it's not intermittent the way the other two are — it sets a ceiling on signal quality for that athlete on that muscle, every single session, regardless of how carefully the sensor is taped. Van Beekvelt and colleagues (2001), in a foundational study on NIRS performance in skeletal muscle published in the Journal of Applied Physiology, quantified this directly: NIRS-derived signal amplitude fell by roughly 4% for every additional millimeter of adipose tissue thickness over the muscle, and the paper's working recommendation was to treat sites with more than about 15mm of subcutaneous fat as unreliable for absolute NIRS values without a correction factor. The limitation worth flagging is that the study measured this relationship at rest with controlled arterial occlusion on a small, relatively homogeneous sample — it establishes the mechanism cleanly but wasn't designed to capture how ATT interacts with movement artifact during dynamic, high-intensity exercise.
Confirming It's an Artifact Before You Throw the Rep Out
Before deleting a suspicious sample or writing off a whole session, run it through four checks. First, cross-reference the timestamp against power, pace, or cadence data from the same session — a real desaturation event tracks effort; an artifact usually doesn't. Second, if the device exposes a raw signal-quality or light-intensity channel underneath the processed SmO2 number, check it; a sudden drop in signal quality at the same timestamp as the SmO2 spike is close to a confirmed diagnosis of sensor slip. Third, note whether the timing lines up with a specific movement event — a jump landing, a hard change of direction, a max-effort grip — rather than with a change in physiological demand. Fourth, apply the rate-of-change test from the previous section: does the swing exceed what a 10-to-20-second kinetic time constant could physically produce between those two samples?
Crum, O'Connor, Van Loo, Valckx, and Stannard (2017), writing in the European Journal of Sport Science, tested the reliability of a commercial NIRS device (Moxy) across an incremental cycling test to exhaustion in trained cyclists. Reliability was solid at low-to-moderate workloads, with good agreement between repeated trials, but it degraded noticeably at the highest-intensity stages near peak effort — the point in the test where pedaling force, postural shift, and strap-muscle interaction all increase simultaneously with true desaturation. The authors couldn't fully separate mechanical noise from a genuinely faster physiological response at maximal effort, and that's the honest limitation: at the exact intensity where coaches most want a clean number, motion and physiology are hardest to pull apart, which is precisely why the four-check process above matters more at threshold and above than it does during an easy warm-up.
Building a Confidence Band Before You Trust the Number
- Measure adipose tissue thickness first. A skinfold caliper at the exact sensor site, doubled and halved per the standard skinfold convention, gives a field-usable proxy for ATT. Treat under 10mm as a reliable placement, 10–15mm as usable with caution, and anything beyond 15mm as trend-only data per athlete.
- Run a vascular occlusion test to anchor the physiological range. With the athlete resting and the sensor secured, inflate a cuff proximal to the sensor to a suprasystolic pressure — commonly cited protocols use roughly 250–300mmHg — and hold for three to five minutes or until the SmO2 trace flattens at its minimum. Release the cuff and record the reactive hyperemia peak. That minimum-to-maximum span becomes the individual's physiological 0-to-100% anchor for the session, rather than assuming every athlete's raw sensor scale means the same thing.
- Collect a clean, near-motionless baseline before adding intensity. A short seated or standing hold at low effort with minimal limb movement establishes what the trace looks like without mechanical interference, so later spikes have something honest to be compared against.
- Re-tape after warm-up, not just before it. Sweat loosens adhesive within the first ten to fifteen minutes of most sessions. A strap that felt secure in the locker room is a common source of mid-session slip.
- Log ATT alongside every profile. An athlete's subcutaneous fat changes across a season with body composition shifts; a placement that was reliable in pre-season testing can drift into the caution zone months later without anyone re-checking.
Mistakes That Keep the Noise Coming Back
Taping once at the start and never re-checking. The single most common cause of mid-session slip: a strap secured before warm-up, never revisited once sweat and movement have had twenty minutes to work on the adhesive.
Treating every spike as a fatigue signal. A coach who reads a sudden SmO2 drop as proof the athlete just hit the wall and adjusts the training plan on that basis, without checking whether power or pace moved at all, is coaching off a sensor artifact rather than off the athlete.
Comparing raw percentages across athletes with different adipose tissue thickness. A lean sprinter and a heavier lineman taped identically will show meaningfully different dynamic ranges for reasons that have nothing to do with training status. Compare trend and rate of change within an athlete, not absolute numbers between athletes, unless ATT has been measured and matched.
Placing the sensor at the point of maximum girth change. Vastus lateralis placement too close to where the muscle belly bulges hardest under load invites the pressure-artifact pattern described above. A few centimeters of repositioning toward a flatter section of the muscle often resolves a rhythmic wobble that retaping alone never fixed.
Frequently asked questions
01My SmO2 reading jumped from 45% to 90% in one second during a sprint — is that possible physiologically?+
02How much adipose tissue thickness is too much for reliable NIRS readings?+
03Can a vascular occlusion test be done in a field setting without lab equipment?+
04Why does reliability seem to get worse specifically at the hardest part of a workout?+
05Should I just discard any reading that spikes and moves on?+
Related Articles
Blood Flow Restriction (BFR) Training Complete Guide
Complete BFR training guide: occlusion pressure, rep schemes, hypertrophy mechanisms, rehab protocols, and how to monitor velocity during low-load BFR.
IMU Sensor Placement: Best Practices for Reliable Movement Data
Where you strap an IMU changes the number it reports. A field guide to lumbar, sternal, wrist, and shank mounts, with the errors that quietly wreck data.
HRV-Based Training Recovery Guide: Autonomic Monitoring for Strength Athletes
HRV drops but you still feel fine — noise or fatigue? See the three-zone protocol and weekly templates that turn readings into real training calls.
IMU Drift and Sensor Fusion: Why Your Data Wanders and How It's Fixed
A bar velocity reading that creeps upward for no reason isn't a broken app. It's uncorrected IMU drift, and a 90-second calibration habit controls it.
IMU Gyroscope Saturation in Rotational Sports: Why the Data Flat-Lines and How to Fix It
A flat 2,000°/s ceiling on a pitcher's shoulder trace isn't a bug — it's gyroscope saturation. Here's how to spot clipped data and set the right range.
AC Joint Separation Return-to-Contact Benchmarks: Shoulder Stability Tests Before Tackling
An AC joint separation that stopped hurting isn't the same as one that can absorb a tackle. See the stability tests that predict contact tolerance.
Adductor Squeeze Return Readiness: Symmetry Criteria After Groin Strain
Adductor squeeze return criteria: use symmetry percent and pain threshold, not a borrowed force number, to time your return after groin strain.
Anaerobic Speed Reserve: How to Calculate and Use It
Calculate anaerobic speed reserve from MAS and max sprint speed, then use the number to profile athletes and pick the right speed or aerobic training bias.
Measure performance with lab-grade accuracy