tensr.fitness

Notes Arc E — Tools and trust Post 13 of 14

Your $30 scale is not lying. But it's not omniscient.

No sensor required

A WH-C06 BLE crane scale costs about $30 on Amazon. It samples at 8 Hz — eight force readings per second.

That sample rate gives you a perfectly honest time-under-tension number. It gives you a perfectly honest impulse number. It gives you a perfectly honest set TUT total.

It also gives you a wildly dishonest rate of force development number, if anyone tries to compute it. And a misleading tempo reading on slow eccentrics.

Same sensor. Same data. Same software. Some metrics survive the under-sampling; some don’t. Knowing which is the entire game.

What sensor honesty actually is

Every metric tensr surfaces has a relationship to sample rate:

Metric class8 Hz (C06)80 Hz (Tindeq, C100 retrofit)Why
Force samples / impulse / TUTIntegrated; under-sampling smooths out
Peak force / average forceAggregate over many samples
Rep segmentation⚠️Edge detection blurs at 8 Hz
Tempo (eccentric/concentric duration)⚠️125 ms per sample misleads on slow tempos
Effective rep count⚠️Inherits rep-segmentation noise
RFD (0–200 ms slope)Slope needs sub-200 ms resolution
Time to peak (TTP) ms-precisionms-precision impossible at 125 ms/sample
Bilateral timing offset (ms)Same as TTP
Form-shape variance⚠️Curve shape gets pixelated
Bilateral symmetry (impulse-based)Integrated; rate-tolerant

This isn’t a hardware sales chart. It’s the contract: the app will show you metrics where the data supports them, and hide them — not approximate them — where it doesn’t.

Why it matters

Most fitness data lies politely. A wrist heart-rate that’s actually a wrist accelerometer plus a confident algorithm. A “calorie burn” estimate built on assumptions about your metabolism it has no way to know. A “VO2max” reading inferred from heart rate variability.

The lies are useful, sometimes. They give you a number where measuring the real thing would require lab equipment. The honesty problem is that the user can’t see which numbers are measured and which are inferred.

tensr commits to a different stance: when the sensor can’t support a metric, the metric isn’t shown. When the sensor can support it with caveats, the caveat is on screen next to the number — not buried in documentation.

The user-facing rule: if it’s on screen, it’s real, at this sample rate. Not “real in general.” Not “real if you upgrade.” Real, on the sensor you’re using right now.

What to track together

The sensor-health metrics are best read as one “is my data trustworthy?” view, distinct from the training metrics.

MetricWhat it tells you
Effective sample rateWhat rate is the sensor actually delivering, vs its spec?
Signal noise floorStandard deviation during a 10-second rest — your equipment’s signature
DriftHas the zero baseline moved across the session?
Connection quality% of expected samples received
Sensor latencyInter-sensor offset (relevant for two-sensor work)
Battery levelPractical concern for long sessions

Pair this with the per-metric honesty matrix. The matrix tells you which numbers to trust; the device-health view tells you whether your sensor is performing to its spec on this specific session.

What gear it needs

N/A — this post is about gear.

But here’s the practical takeaway for shopping: what you’re paying for in the higher-rate sensors is the honesty matrix, not the metrics themselves. Most lifters don’t need RFD, sub-second tempo precision, or ms-resolution timing offset. The ones who do — strength athletes, return-to-play rehab, elite power sport — find the upgrade pays itself back instantly.

For everyone else: the C06 is a perfectly good entry point. Honest TUT. Honest impulse. Honest set decay. Honest weekly volume. Honest bilateral symmetry on impulse. That’s most of the catalog.

What to do tomorrow

Pair your sensor for the first time. Let it sit on the bench for 10 seconds while you adjust your headphones. Look at the noise floor on screen.

That number is your equipment’s signature — the standard deviation of force readings when nothing is touching the sensor. A clean sensor reads under ~1 N noise floor. A noisier one reads 3–5 N. A failing one reads 10+ N.

Now check the effective sample rate against the device spec. A C06 should report ~8 Hz; if it’s 4 or 5, your Bluetooth connection is choppy and the metrics inherit that noise.

Two checks. 30 seconds. Now you know what your data looks like at rest, and what to expect during work.

A specific note on shopping: if a sensor’s marketing claims aren’t backed by a published sample rate spec, treat the missing number as a feature absence. “Bluetooth force scale” with no sample rate listed is exactly the kind of detail that does more for you to know than not know.

The C06 is honest about what it is. So is the Tindeq. Most other things in this category aren’t, yet.


What this looks like in tensr.fitness. Open the app, pair a sensor, and the metrics in this post are on the screen the moment you start a set.

A note on the data. Every force sample you record stays on your device unless you opt into sync. The file format is open — SQLite, CSV, NDJSON, all readable with any tool. More on that in the FAQ.

hardware